Mouse Medial-Prefrontal Cortex Involvement in Trace Fear ... · Mouse Medial-Prefrontal Cortex Involvement in Trace Fear Memory during Wakefulness and Sleep Hendrik William Steenland
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Mouse Medial-Prefrontal Cortex Involvement in Trace Fear Memory during Wakefulness and Sleep
by
Hendrik William Steenland
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
occurs in response to predators, when there is no option for escape from the situation (Blanchard
& Blanchard, 2008). Moreover, freezing appears to be a most effective behavioural strategy if
the prey has not already been spotted by the predator (Eilam, 2005). The most common way to
induce freezing behaviour, under experimental situations, is to present noxious painful stimuli
(typically to the feet). Models of fear conditioning in mice and rats have been instrumental in
uncovering the molecular biology and basic circuitry controlling fear behaviour.
1.2.2 Fear conditioning
Pavlovian fear conditioning has been used as a basic paradigm in the examination of fear
memory. In a typical fear-conditioning experiment, rodents are presented with an emotionally
neutral conditioning stimulus (CS), such as a tone, which is paired in time with an aversive foot
shock (US) (Fig. 1.1A). With repeated pairings of the CS and US, the presentation of the CS
alone begins to elicit defensive behaviours such as freezing (Blanchard & Blanchard, 1969b;
Fanselow & Bolles, 1979). This presentation of the tone continually updates the animal to the
impending danger. This type of conditioning is thought to rely predominantly on the amygdala
(LeDoux, 2000; LeDoux, 2003). By contrast, the trace fear conditioning paradigm involves the
association of the CS and US; however these stimuli are separated by a time interval (Fig. 1.1B).
4
Figure 1.1 Cued delay and trace fear conditioning Conditioning and testing protocols for cued delay and trace fear memory. During conditioning, both a shock and tone are repeated for up to 10 trials, while fear behaviour (e.g. freezing) is monitored. The major difference between delay (A) and trace fear conditioning (B) is that; under delay conditioning the tone co-terminates with the shock stimuli, while under trace conditioning the tone terminates before the onset of the shock. Thus, under trace fear conditioning, the animal must make an association between tone and shock over a time (or trace) interval. Fear memory is tested similarly in both protocols, with the repeated presentation of the tone by itself. Provided the animal has made an association between the shock and the tone, the animal will demonstrate defensive behaviors such as freezing. The delayed fear paradigm is a typical method to examine the function of the amygdala fear circuit, while trace fear conditioning is thought to engage extra-amygdaloid structures necessary to bridge the trace interval (figure by H. Steenland, refs in text). Thus, trace fear conditioning generates a representation (a fear memory) which is maintained in
expectation of a shock in the absence of the CS. Despite the seemingly simple distinction
between these two paradigms, the trace fear paradigm must engage additional brain structures
which enable the bridging of the temporal gap between the US and CS, such as the anterior
cingulate cortex (ACC) (Han et al., 2003), prelimbic cortex (Baeg et al., 2001; Knight et al.,
5
2004; Runyan & Dash, 2004; Runyan et al., 2004) and the hippocampus (McEchron et al., 1998;
Burman et al., 2006). Moreover, the trace fear memory paradigm may be representative of
everyday events in which individuals anticipate negative future outcomes. Trace fear
conditioning will be extensively reviewed and diagrammed in chapter 2. However, for now, the
examination of basic fear circuits is most important.
1.3 Basic fear circuitry
This section examines the basic circuits of fear conditioning, from the initial point of
processing the tone and shock stimuli to the point of gernerating motor output (i.e. freezing
behaviour). Figure 1.2 is a model of the basic fear circuits and will serve as a network on which
to build more complicated circuits related to trace fear conditioning (Chapter 2).
1.3.1 Fear expression circuits (central amygdala)
The basic circuit of fear behaviour has been extensively studied and has centered on the
amygdala. The amygdala is a small almond shaped structure residing in the temporal lobe. The
amygdala can be subdivided into 3 nuclei which include the lateral nucleus, basolateral nucleus
and the central nucleus (LeDoux, 2000; LeDoux, 2003) (Figure 1.2). Neurons of the lateral
nucleus receive auditory (CS) and somatosensory inputs from the thalamic and cortical
processing regions (Turner & Herkenham, 1991; Mascagni et al., 1993; Romanski & LeDoux,
1993). The lateral amygdala connects with the central amygdala both directly and indirectly via a
6
projection through the basolateral nucleus (Pitkanen et al., 1997; Pare & Smith, 1998).
Accordingly, the central amygdala is thought to be involved in controlling behaviours related to
fear expression (LeDoux, 2000; LeDoux, 2003). For example, the central nucleus of the
amygdala projects to the lateral hypothalamus to control blood pressure (LeDoux et al., 1988),
paraventricular nucleus of the hypothalamus to influence adrenocorticotropic hormone secretion
(Beaulieu et al., 1986; Gray et al., 1989), the periaqueductal gray to control conditioned freezing
(LeDoux et al., 1988; Bellgowan & Helmstetter, 1996; De Oca et al., 1998) and the nucleus
Figure 1.2 Basic fear circuits Representation of the fundamental systems involved in controlling fear behavior. The input stage represents the collection of information from pain and auditory modalities during fear conditioning. These systems transduce this information to be processed in the spinal cord, brainstem, thalamus and auditory cortex, and eventually converge in the lateral nucleus of the amygdala (LA). The central nucleus of the amygdala (CEA) then directs the main outputs controlling fear including: autonomic, behavioral and arousal systems. This thesis is focused on freezing behavior and so the projection from the periaqueductal gray (PAG) to the spinal cord is highlighted. From the spinal cord, the motor out-flow can control muscles of the body (e.g. neck) to express freezing. Note, higher order systems may play a role in fear conditioning, but will be reviewed in the context of trace fear conditioning and memory in chapter 2 (figure 2.2). TE, temporal, BLA basolateral nucleus of amygdala, CEA, central nucleus of amygdala, DMN, dorsal motor nucleus, VTA, ventral tegmental area, LC, locus coeruleus, LDT, laterodorsal tegmentum, ACTH, adrenocorticotropic hormone, VPL, ventroposterio-lateral, MDvc, mediodorsal vental caudal, MG, medial geniculate nucleus. (figure by H. Steenland, refs in text)
7
reticularis pontis caudalis to control conditioned startle responses (Hitchcock & Davis, 1986;
Rosen & Davis, 1990; Rosen et al., 1991). The central nucleus of the amygdala is also thought to
control cortical arousal (LeDoux, 2000; LeDoux, 2003) through connections with the locus
Fibiger, 1992) and the dopaminergic ventral tegmental area (Kaufling et al., 2009).
1.3.2 Reception circuits (lateral amygdala)
Classical fear conditioning has two basic sensory components which generally include a
tone (CS) and shock (US) stimuli, occurring in close temporal proximity. For an organism to
associate these two modalities it could be speculated that there must be points of convergence in
the brain where these two systems overlap, and this appears to be the case. Tone-related
information is thought to reach the amygdala via the medial geniculate nucleus (in the thalamus)
terminating in the lateral amygdala (Clugnet & LeDoux, 1990). Neurons in the lateral amygdala
respond 12-25msec after the presentation of clicks or broadband white noise with frequency
preferences around 10-12kHz (Bordi & LeDoux, 1992; Bordi et al., 1993). It was also shown
that the dorsal aspect of the lateral amygdala receives converging input from both pain-related
information and tone-related information (Romanski et al., 1993). Unit recording studies have
shown that learning-related activity in the thalamic pathway occurs earlier than that in the
auditory cortex (Quirk et al., 1995; Quirk et al., 1997). Indeed it has been shown that complete
ablation of the auditory cortex has no effect on cued fear conditioning; however if the medial
geniculate nucleus of the thalamus is ablated, cued fear conditioning is impaired (Romanski &
LeDoux, 1992).
8
1.3.3 Contextual integration (basolateral nucleus of the amygdala)
The basal nucleus of the amygdala receives large projections from the hippocampus
(Pikkarainen et al., 1999). In addition, perirhinal cortical stimulation produces long-term
potentiation (LTP) in the basolateral, but not in the lateral nucleus of the amygdala (Yaniv &
Richter-Levin, 2000). Long-term potentiation (LTP) is the main synaptic model of learning and
memory and will be addressed in chapter 3 (section 3.4). Unit recording from the basolateral
nucleus suggests that it is involved in learning multimodal sensory information (Toyomitsu et al.,
2002). Consistently, a recent study demonstrated that contextual conditioning, but not auditory
cued fear conditioning, is disrupted by pre-treatment of ibotenate-induced damage to the
basolateral amygdala (Onishi & Xavier, 2010). The results indicate that the basolateral amygdala
is not involved in fear expression, but rather in integrating information related to context (Onishi
& Xavier, 2010).
1.3.4 Electrophysiological evidence of plasticity in the lateral amygdala
As demonstrated above, the lateral amygdala represents a point of convergence of pain
and auditory information, and therefore may be expected to have changes associated with fear
learning and be modifiable though manipulations of molecular biology. Neuronal and field
potential recordings have shown that the tone-evoked neural responses in the lateral amygdala
are potentiated with fear conditioning to shock (Quirk et al., 1995; Quirk et al., 1997). An
alternative way to examine this pathway is to by-pass the painful peripheral input and directly
stimulate the thalamus in association with tone input. Tetanic stimulation of this thalamic
pathway enhanced evoked-potentials elicited from auditory stimulation (Clugnet & LeDoux,
1990; Rogan & LeDoux, 1995; Rogan et al., 1997).
9
Understanding the network interactions responsible for retrieval of fear memory is a
challenging pursuit. Seidenbecher et al. (2003) examined how fear memory retrieval might be
possible between the amygdala and hippocampus. These investigators recorded local field
potentials from both the hippocampus and the lateral amygdala, and examined the spontaneous
activity which occurred during cued fear memory retrieval. Interestingly, during fear behaviours,
the amygdala and hippocampus exhibited 4-5Hz theta frequency brain waves (Seidenbecher et
al., 2003). Most importantly, these brain waves were synchronized, indicating that the two
regions were either communicating (Seidenbecher et al., 2003) or being driven by a common
oscillator. A follow-up study showed that this 4-5Hz activity occurred during contextual fear
memory retrieval as well, but it only occurred after 24hrs post-conditioning and not 2min, 30min,
2hrs or 30 days after learning (Narayanan et al., 2007). While this study reports that electrode
locations were in the lateral amygdala, it is likely that a substantial degree of the recorded
activity comes from the basolateral amygdala, since this is a terminal zone for the hippocampus
(Pikkarainen et al., 1999).
1.4 The unconditioned stimulus and basic pain transduction circuitry
1.4.1 Nociception
The skin contains a variety of receptors which can respond to damaging mechanical
stress, noxious heat and cold, and changes in local metabolism (e.g. acidity) (Craig, 2003). This
information is transduced at the periphery and terminates on lamina I neurons in the dorsal horn
of the spinal cord (Craig, 2003). Peripheral fibres which carry this information, consist of the Aδ-
and C-fibres (Craig, 2003). In particular, cutaneous Aδ-fibres can respond to pinch and rapidly to
heat >46ºC, or to pinch and slowly to heat >53ºC. Other cutaneous Aδ fibres respond only to heat
(Craig, 2003). C-nociceptors have a similar response profile, responding to heat, pinch or both.
10
Other C-fibres respond to tissue damage and inflammation (Craig, 2003). Aδ-fibres activation is
associated with the early components of pain (e.g. sharp pain) while C-fibre tend to produce
sustained response to pain (1-2sec later), which may produce a burning sensation. Typically,
shock stimuli are used in fear conditioning. The skin is quite resistant to shock stimuli since its
resistance is about 100 kOhms (Fish & Geddes, 2009). The skin also acts as a capacitor, in that
rapidly changing voltage is the most effective means to stimulate the skin (Fish & Geddes,
2009). Shock stimuli would presumably activate c-fibers and Aδ-fibres, though there seems to be
a lack of information on this topic.
1.4.2 Spinothalamilic projections
Aδ- and C-fibres terminate in lamina I of the spinal cord in a somatotopic manner (Craig,
2003). Based on experiments of cutaneous stimulation, lamina I neurons which project through
the spinothalamic tract can be distinguished, based on their response to either sharp or burning
pain (Andrew & Craig, 2002; Craig & Andrew, 2002). Lamina V neurons receive input from
collateral Aδ-fibres. These neurons in this region are considered to respond to a wide variety of
modalities including noxious and innocuous stimulation (Craig, 2003). Collectively lamina I and
lamina V-VII both supply the spinothalamic tract. The lamina I spinothalamic tract projects
extensively to the posterior part of the ventromedial nucleus of the thalamus (Craig, 2004) where
neuron responses mirroring the activity lamina I of the spinal cord have been detected (Davis et
al., 1999; Craig, 2003), and to the ventral caudal part of the medial dorsal nucleus (MDvc)
(Craig, 2003). By contrast, parts of the spinothalamic tract which originate from layer V
terminate in the posterior part of the ventromedial nucleus (VMpo), ventral posterior inferior
nucleus (VPI) to supply the basal ganglia, motor and parietal areas (Craig, 2003). As discussed
above, the spinothalamic tract region of the thalamus is also connected to the lateral amygdala
11
(Ledoux et al., 1987; Turner & Herkenham, 1991). Indeed some of the cells in the lateral
amygdala respond to nociceptive stimulation (Romanski et al., 1993). Nociceptive information
from the spinal cord and the parabrachial nucleus may also be carried to the central nucleus of
the amygdala directly (Bernard & Besson, 1990; Cliffer et al., 1991).
1.4.3 Spinothalamo-cortical connections
Different parts of the thalamus project to distinct cortical regions to process distinct
aspects of pain. For example, the MDvc thalamus projects to the anterior cingulate cortex, where
the affective component of the pain is processed (Devinsky et al., 1995; Rainville et al., 1997;
Hofbauer et al., 2001). By contrast, the VPI and VMpo thalamic region projects to the
somatosensory SII and area 3a of SI (Bushnell et al., 1999; Willis et al., 2002; Craig, 2003;
Duncan & Albanese, 2003). However there is still debate over the organization of these thalamic
nuclei (Willis et al., 2002). The SI cortex is somatotopically organized and has been implicated
in both the localization of pain and the discrimination of pain (Bushnell et al., 1999; Willis et al.,
2002; Craig, 2003; Duncan & Albanese, 2003). It is also likely involved in top-down regulation
of pain. For example, direct glutamatergic activation of the SI region in rats activates descending
corticofugal inputs which potentiate noxious-evoked responses in the thalamus (Monconduit et
al., 2006). A direct comparison of the ACC with that of SI showed that the ACC neurons were
less responsive and had longer latencies than that of the SI to laser-heat stimuli (Kuo & Yen,
2005). The insula is thought to be innervated by projections from the thalamic VPI (Craig, 2003).
The insula cortex is a viscerosensory region which processes cardiovascular, cardiopulmonary,
gastrointestinal and gustatory information (Cechetto & Chen, 1990; Allen et al., 1991; Zhang &
Oppenheimer, 1997; Jasmin et al., 2004).
12
1.5 The conditioning stimulus and the auditory transduction circuitry
During fear conditioning, a tone is played in association with a shock. This information
must be transduced from the ear to relevant areas involved in fear memory for the association of
the shock and the tone to be made. As has been pointed out, the lateral amygdala receives
converging input from both pain and sound modalities from the thalamus, possibly to make fear
memory associations; however, the auditory cortex is also activated during fear conditioning.
Thus, this section serves as a basic review of the auditory system and how it may relate to fear
conditioning
1.5.1 Auditory cortex and cued-fear conditioning
It has been shown that lesions to the medial geniculate projection zone in the thalamus
block auditory fear conditioning; however, lesions to the auditory cortex for which the medial
geniculate projects, do not impair auditory fear conditioning (LeDoux et al., 1984). This finding
suggested that the medial geniculate has a variety of subcortical targets. Indeed, the medial
geniculate was found to connect with the amygdala and inferior colliculus (LeDoux et al., 1985).
The primary auditory cortex, TE1 (temporal lobe1), was not found to originally project to the
amygdala (LeDoux et al., 1991), but anterograde labelling from the ventral TE1, TE2 and caudal
TE3 regions of the auditory cortex do show a connection to the lateral amygdala (LeDoux et al.,
1991; Romanski & LeDoux, 1993). However, complete destruction of the auditory cortex does
not impair auditory fear conditioning (Romanski & LeDoux, 1992). Indeed, during conditioning
the neurons in the auditory cortex take more trials to learn, whereas the learning in the amygdala
happens rapidly (Quirk et al., 1997). Interestingly, a few auditory neurons were found to
anticipate cue stimuli; thus, the auditory system may be more involved in the prediction of future
events related to auditory stimuli (Quirk et al., 1997). To examine the role of the auditory cortical
13
responses in fear conditioning, Armony et al. (1998) removed the amygdala and showed that late
auditory-evoked responses, associated with conditioning in the auditory cortex, were attenuated.
Cued fear conditioning therefore appears to engage the auditory cortex but does not appear to use
the auditory cortex.
1.5.2 Auditory transduction circuits in the brain
The auditory system is not a fundamental topic of this thesis, though it deserves some
brief discussion since the conditioning stimulus used in this study is typically a white noise.
White noise is defined as a random signal with an equal power spectral density in each frequency
band (Wikipedia, 2010). In the inner ear, different positions along the basilar membrane of the
cochlea respond to different frequencies of sound (Kandel et al., 2000). Since white noise has a
range of spectral energies it would be expected that the cochlea would be active along its length.
This activity is mechanically transduced by hair cells and the auditory nerve (cranial nerve VIII)
which sends this information via the lateral lemniscus to the dorsal medulla inferior olive and
finally to the medial geniculate of the thalamus (Kandel et al., 2000).
1.6 Motor output
Freezing behaviour is quite obvious; however whole networks of motor neurons which
would normally activate movement must produce an absence of movement. Since the central
nucleus of the amygdala is essential for fear expression (section 1.3.1) the systems responsible
for behavioural arrest will be examined. Damage to the central amygdala abolishes freezing
behaviour related to conditioning (section 1.3.1) and this region projects prominently to the
periaqueductal central gray (LeDoux et al., 1988; Bellgowan & Helmstetter, 1996; De Oca et al.,
1998).
14
The periaqueductal gray is located in the midbrain. Lesion studies show that removal of
the caudal aspect of the periaqueductal gray blocks conditioned freezing behaviour (Liebman et
al., 1970; LeDoux et al., 1988), while having no impact on arterial pressure related to
conditioning (LeDoux et al., 1988). To understand what motor regions the periaqueductal gray
may innervate, retrograde and anterograde labeling studies have been conducted (Mouton &
Holstege, 1994). It was found that the periaqueductal gray robustly projects to the upper cervical
spinal cord in the lateral parts of lamina V, VII and the dorsal part of lamina VIII. Importantly,
this part of the cervical spinal cord is involved in the control of neck muscle activity; thus, this
pathway may mediate head movements related to freezing behaviour (Mouton & Holstege,
1994). It should be mentioned that the dorsolateral periaqueductal central gray is innervated by
both dorsal and ventral aspects of the ACC, while the prelimbic and infralimbic cortices
innervate both dorso- and ventrolateral divisions of the periaqueductal central gray (Floyd et al.,
2000; Gabbott et al., 2005). The results suggest that higher order control of freezing may be
possible through direct innervation of the periaqueductal gray.
1.7 Comments
This concludes our examination of the basic circuitry involved in auditory fear
conditioning. A few themes appear important with this evaluation. Firstly, fear behaviour
circuitry appears to have parallel systems involving the spinothalamic tract, projecting to the
amygdala, ACC, SI, SII and insula brain regions. Secondly, the amygdala, ACC, infralimbic, and
prelimbic cortex all innervate the periaqueductal central gray, each possibly exerting control over
freezing behaviour. However, it should be mentioned that simple forms of fear conditioning may
prime systems for action without their actual necessity for the task or paradigm at hand. For
15
example, the insula may be primed for activity in parallel with the amygdala during cued delay
fear conditioning, but might not be required for the actual conditioning itself.
1.8 EXPERIMENT SET 1: EMG is an effective measure of fear behaviour
1.8.1 Abstract
Conventional methodology for examining conditioned fear has involved visual-based
technologies such as cameras and direct observation. These methods are excellent screening tools
of fear behaviour (i.e. freezing). However, they do not provide a physiological output, which
could be useful to examine neurophysiological correlates of fear. We found that dorsal neck
electromyography can be used as a method to score fear behaviour and is as effective as visual-
based scoring. Since the electromyogram is a physiological measure it can be used in either the
light or dark to score freezing behaviour, without use of any camera or visual observation. We
also show that electromyogram-based scoring methodologies, in conjunction with an
electroencephalogram, are useful to discriminate fear from sleep.
1.8.2 Rationale
Genetic mouse models are a fundamental tool to examine the molecular basis of
behaviour (Tang et al., 1999; Mogil et al., 2000; Wei et al., 2002; Fanselow & Poulos, 2005; Ko
et al., 2005; Zhao et al., 2005; Steenland et al., 2008a). Conditioned fear is one of the most
widely used methodologies to examine emotional memory in rodents (LeDoux, 2000). Fear
behaviour can be identified as an absence of movement and is known as freezing (Bolles, 1970;
Blanchard & Blanchard, 1972; Fanselow, 1984; LeDoux et al., 1986). However a similar
definition holds for the behavioural description of sleep, which is also seen as a state of
immobility (Siegel, 2005). Thus it is essential that freezing, which consists of a defensive
16
behaviour (LeDoux, 2000; McNaughton & Corr, 2004; Kim & Jung, 2006) be dissociated from
sleeping, which ultimately disengages an animal from the external environment. Confusion over
identifying these two behaviours could become an issue when examining fear memory in genetic
mice which also demonstrate alterations in sleep (Espinosa et al., 2004; Steenland et al., 2008a).
The most common method to examine fear behaviour is through direct visual observation
or through use of motion sensitive cameras (Bolles, 1970; Blanchard & Blanchard, 1972;
Fanselow, 1984; Anagnostaras et al., 2000; Marchand et al., 2003; Espinosa et al., 2004;
Takahashi, 2004; Kopec et al., 2007). Direct visual observation can produce inaccuracies in
measurement due to varying reaction times and attention. Camera systems do not provide a
physiological correlate of behaviour, but infrared cameras are available for recordings in the dark
(Takahashi, 2004). Thus, it would be useful for fear measurement methods to provide both a
physiological correlate and for freezing behaviour to be recorded regardless of the time of day.
Trace-fear conditioning in rodents, involves the association of a neutral conditioned
stimulus (CS; generally a tone) with an aversive unconditioned stimulus (US; shock) over a time
interval (referred to as the trace interval). The acquisition and retention of trace-fear memory
requires the amygdala, hippocampus and prefrontal cortex (McEchron et al., 1998; Buchel et al.,
1999; McEchron & Disterhoft, 1999; McEchron et al., 2000; Crestani et al., 2002; Han et al.,
2003; Knight et al., 2004; Runyan et al., 2004; Chowdhury et al., 2005; Misane et al., 2005;
Bangasser et al., 2006; Carter et al., 2006; Knight et al., 2006; Burman & Gewirtz, 2007).
Methodologies which might permit precise timing of a freezing behaviour, such as
electromyography (EMG), may be helpful for examining the neural correlates which dictate fear.
Additionally, some evidence suggests that contextual information may influence trace-fear
conditioning (McEchron et al., 1998; Quinn et al., 2002). Accordingly, conducting trace-fear
conditioning in the dark would eliminate confounding information of the environmental context.
17
The current study validated a simple, inexpensive, physiological method to examine trace-fear
conditioning in the dark or light, utilizing EMG of the neck.
1.8.3 Materials and Methods
1.8.3A. Animals
Surgery was performed on 6 C57BL6 mice for in vivo electrophysiology studies (8-14
weeks old). Mice were maintained on a 12:12-h light-dark cycle (lights on at 8:00 A.M.) and had
access to food and water ad libitum. Procedures conformed to the recommendations of the
Canadian Council on Animal Care and the University of Toronto Animal Care Committee
approved the protocols.
1.8.3B Surgical preparation
Mice were anesthetized with 1-2% isoflurane which was mixed with oxygen (30%
balanced with nitrogen) and delivered to the mice via nose cone throughout the surgery. All
electrodes were pre-attached to a miniature connector. The abdomen and scalp of mice were
shaved and then cleaned with iodine (Triadine) and alcohol. The skull of the mouse was fixed
into a stereotaxic adapter (502063, WPI, Sarasota, Fl, USA) mounted on a stereotaxic frame
(Kopf Model 962, Tujunga, CA, USA). Three small holes (1.19 mm diameter) were drilled into
the skull for differential frontal-parietal recordings. Electrodes, consisting of a wire attached to a
jeweler’s screw (with contact end ground flat), were fixed into the holes to record EEG
(electroencephalogram) at the following coordinates relative to bregma: frontal cortex (AP 2.2,
ML 1.0), parietal cortex (AP -2.2, ML -2.5) and ground (AP -3.0, ML 3.0). For dorsal neck
muscle recording, left and right nuchal muscles were exposed and Teflon-coated stainless steel
electrodes (part# AS632, Cooner Wire) were sutured (4.0 silk) to each muscle to record neck
18
EMG. All wires and connectors were secured to the skull with dental cement and cyanoacrylate
glue (Krazy glue). Mice were injected, intraoperatively with buprenorphine (0.05 mg/kg, SC) as
an analgesic, and 1.0 ml sterile saline (IP) for hydration. Mice were placed on a warm heating
pad until they showed signs of ambulation, and were permitted to recover ~14 days prior to
recording.
1.8.3C Electrophysiological recordings
On the day of the experiment, a light-weight cable was connected to the assembly on the
mouse head. The signals were routed through a commutator (Crist Instruments, Hagerstown,
MD, USA). The mice were placed in a shock chamber (Med Associates, St. Albans, VT) to
either examine fear conditioning or to record EEG. The shock chamber was situated in a sound
attenuated cubicle (ENV-017M, Med Associates, St. Albans, VT, USA). Electrophysiological
signals were amplified and filtered (Super-Z headstage amplifiers and BMA-400 amplifiers and
filters, CWE Inc., Ardmore, PA, USA) as follows: cortical EEG 1000X at 1-100Hz and neck
EMG 2000X at 100-1000Hz.
1.8.3D Trace-fear conditioning
Trace-fear conditioning was performed in an isolated shock chamber. The CS was an
80dB white noise, delivered for 15s, and the US was a 0.75mA-scrambled foot-shock for 0.5s.
The mouse was acclimated for 60s, and presented with ten CS–trace–US–ITI trials (trace of 30s,
inter-trial interval (ITI) of 210s). For the camera-based method, data were video recorded using
FreezeFrame Video-Based Conditioned Fear System and analyzed by Actimetrics Software
(Coulbourn Instruments, Wilmette, IL). In addition, whenever freezing behaviour was observed
directly, the experimenter delivered a 3V pulse to the analog to digital converter, via depression
19
of a foot pedal. The pulse was recorded with Spike2 software alongside the EMG record.
Freezing was defined as the absence of movement (of at least 1s) with the exception of breathing
for visual and camera-based methods. Since neck EMG is well known to correlate with an
animal’s movements, we tentatively defined EMG-based freezing as an absence of muscle twitch
activity (of at least 1s). All three scoring methodologies were conducted simultaneously to
examine whether or not EMG freezing measures would correlate with visual-based freezing
measures.
1.8.3E Analyses and statistics
For all comparisons, differences were considered significant if the null hypothesis was
rejected at P < 0.05 using a two-tailed test. Repeated-measures ANOVA (RM-ANOVA) was
performed and followed by post-hoc comparisons with the Bonferroni-corrected P to infer
statistical significance for EMG magnitude comparisons. Freezing behaviour was analyzed using
Pearson-product moment correlation to examine the validity of using EMG as a measure for
conditioned freezing behaviour. Analyses were performed using Sigma-Stat (SPSS Inc.,
Chicago, IL, USA). Data were plotted with SigmaPlot (Systat Software, San Jose, CA, USA).
For the sleep study the mice were scored to be in Non-rapid eye movement sleep
(NREM) if the EMG was of low muscle tone and the EEG was characterized by high voltage
slow waves (1-4 Hz range). Wakefulness was characterized by low voltage, high frequency EEG,
and muscle tone associated with movement. EEG was analyzed in 5s epochs for periods of
wakefulness and NREM sleep using Spike2 software. Scripts for EEG analysis (Sudsa-version
2.2) were obtained from Cambridge Electronic Design (CED). Fast-Fourier transform was used
to convert EEG waveforms into total power (μV2), which was binned every 5s for the following
frequency bands: δ1 (0.5-2Hz), δ2 (2-4Hz), θ (4-7.5Hz), β1 (7.5-13Hz), β2 (13-20Hz) and α (20-
20
30Hz). All neck EMG recordings were smoothed (25ms time constant) and rectified (Spike2
software, 1401 interface, CED Ltd., Cambridge, UK) to calculate the area under the curve. For
EMGs during active wakefulness and NREM sleep, the area under the curve was calculated
every 5s. For EMG-based freezing behaviour the magnitude of the EMG was determined for the
length of the freezing bout and normalized to a 5s period for comparison to other behavioural
states.
1.8.4. Results
Figure 1.3A shows the placement of the neck-EMG recording electrodes to examine
freezing behaviour and sleep. Figure 1.3B shows an example of a baseline EMG recording and
how the signal was processed for quantification.
Figure. 1.3 Surgical preparation (A) Left and right neck muscles (arrows). The muscle overlying the neck muscles has been gently pulled back for better visualization. Two loops of exposed wire were made at the end of a Teflon coated stainless steel wire. The loops were then sewn to left and right muscles. (B) Representative neck EMG traces. The upper trace is and EMG recorded between left (L) and right (R) neck muscles. The second trace is the same signal which has been rectified (Rt). The third signal is the same signal rectified and smoothed (Rt/Sm). Upward arrows denote an increase in movement. Arb., is arbitrary units.
Animals were conditioned in the trace-fear paradigm while their fear was simultaneously
recorded via visual observation, freeze frame camera and with neck EMG electrodes. Figure
1.4A shows simultaneous visual scoring with camera and neck recordings during trace-fear
21
Figure. 1.4 Neck EMG recordings can provide measurement of fear behavior (A) Simultaneous recordings from a visual observer, freeze-frame camera, and neck EMG during the traceconditioning paradigm (during 10th training trial). The conditioning protocol is depicted at the top, showing that awas played followed by an interval (trace interval) then by a shock. It was readily observed that the rectifiedsmoothed EMG signal paralleled the camera-based movement detection, with periods of quiescent EMG activittwitches) correlating with periods of inactivity. Under each measurement, the freezing behavior duration is indicsimilar to the visual-based scoring. Visual-based scoring paralleled both the EMG and the camera-based scoring methowever it was obvious that the latencies of the detection were longer (based on observer reaction times). (B) Scatterand regression lines (+/-95% confidence intervals) for the percentage freezing during tone and trace intervals for eachof the ten trace-fear training trials. Comparisons are made for each of the fear conditioning scoring methods. All scmethods were highly correlated with one another. Abbreviations as for figure 1.3. conditioning. The animal was found to develop visually observable freezing behaviour during the
conditioning. The automated camera system also detected this freezing behaviour. Interestingly,
the neck muscle EMG was found to shift from twitch-like muscle activity during visually
observed movements to a non-twitch activity, with a stable level of muscle tone. All methods of
collecting freezing behaviour data appeared to correlate with one another. The percentage of
22
freezing during the tone and during the trace interval was quantified for 10 trials of conditioning.
Figure 1.4B, C, and D, show correlations plotted for the percentages of freezing for all three data
collection methods. Figure 1.4B shows that there was a strong statistically significant positive
correlation between the visual scoring and the camera scoring (r2= 0.82, p<0.001). Figure 1.4C
shows that there was a strong statistically significant positive correlation between the camera
scoring and the neck EMG scoring (r2= 0.88, p<0.0001). Figure 1.4D shows that there was a
strong statistically significant positive correlation between the EMG scoring and the visual
scoring (r2= 0.83, p<0.0001). Based on these findings we conclude that EMG scoring is as
accurate as both the visual method and the camera method. Thus the EMG-based method would
be sufficient to yield measures of fear behaviour during any point in a 24 hour day, whether
lights are on or off.
Fear behaviour is not always easy to discriminate from sleep when visually observing
rodents. Importantly, fear memory is often studied when rodents would naturally be sleeping (i.e.
lights on phase). We therefore examined whether the automated camera system and simple EMG
recording could mistake sleep for fear memory in an animal which was not fear conditioned.
Figure 1.5A shows an example of the automated camera system detecting fear behaviour in an
untrained animal. Consistently, an examination of the EMG suggests an absence of movement.
However, in this experiment, simultaneous EEG was conducted to examine brain wave
signatures associated with sleep. Figure 1.5B shows that the addition of and EEG recording
identifies this state as NREM sleep rather than fear. The EEG is characterized by high voltage
slow waves compared to the wakefulness that follows the sleep. Figure 1.5C and D show
quantified EMG activity and EEG power confirming the behavioural states of the mouse.
23
Figure 1.5 Falsely scoring sleep as fear behaviour (A) Shows a simultaneous recording with a freeze-frame camera while EMG activity was recorded during the light phase, when mice normally sleep. There is a long period of behavioural quiescence, as detected by the automated camera system and also viewed with neck muscle EMG recording. The animal then appears to become active toward the end of the measurement period. Freezing behaviour is scored below each trace. (B) When the EEG activity was observed during this period (shaded boxes in A and B) it was revealed that the mouse was actually sleeping (NREM) as indicated by high voltage slow waves in the EEG (panel I). The animal can be then seen to arouse from sleep (panel II) as revealed by high frequency low amplitude EEG and the increase in EMG activity. Thus, camera and even EMG methodologies may be recording sleep behaviour when they should be recording fear behaviour. (C) Reduced neck EMG is observed during NREM sleep in comparison to arousal from sleep. (D) Quantification of EEG frequency bands during NREM sleep, demonstrating an increase in power of high voltage slow frequencies relative to wakefulness. Abbreviations as for figure 1.3. We next examined whether the magnitude of the EMG would be useful for discriminating
freezing behaviour from both NREM sleep and wakefulness. EMG values were quantified for 5
chronically instrumented animals during EMG-based freezing behaviour, wakefulness and
24
NREM sleep. RM-ANOVA revealed that there was a significant effect of the recorded state on
neck muscle EMG. (F(2,10)=20.17, P=0.001). Figure 1.6B shows that there was a significant
decrease in EMG from wakefulness to freezing behaviour (t=2.90, P=0.032), and from freezing
behaviour to NREM sleep (t=3.04, P=0.013). The finding suggests that the magnitude of the
EMG is sufficient to discriminate each of these states.
Fig. 1.6 Discriminating freezing behaviour from sleep-wake states Grouped data (n=5) showing that each state could be quantitatively discriminated based on muscle tone measurements. Abbreviations as for figure 1.3, * is p<0.05. We therefore suggest that EMG-based freezing behaviour is characterized by neck
muscle activity that is greater in magnitude than NREM sleep but less than that of wakefulness.
In addition, freezing behaviour occurs in the absence of neck muscle twitches and must last for at
least 1 second to be scored.
25
1.8.5 Discussion
The current investigation sought to examine an inexpensive alternative approach for fear
behaviour scoring, to examine mice across the natural light-dark cycle. We have shown that
freezing behaviour can be scored with use of a simple EMG record. We also show that the way
in which freezing is behaviourally defined overlaps with the way that sleep is defined. That is;
immobility with the exception of breathing. Thus EEG recordings in conjunction with EMG
recording can be helpful to discriminate fear behaviour from sleep.
A major advantage of recording neck EMG for freezing behaviour is that it provides
precise timing of when the freezing occurs which is crucial for the investigation of the neural
correlates of trace-fear memory. Moreover, the EMG activity can be collected using the same
analog-digital converter as that for recording EEG or brain cells.
The major drawbacks to the EMG method described here are that it requires surgery and
there is currently no automated method to score the EMG-based freezing activity. Thus, rapid
screening would be problematic. However, this does not preclude using a camera system and
automated scoring method to do initial screening. The examination of the neurophysiology could
then be followed up with the EMG-based methodology to score freezing.
It has been suggested that resting or sleeping, as scored via visual observation, only
occurs after extinction training of at least 20 minutes (Marchand et al., 2003). Preliminary
observations from our laboratory also suggest that this is the case. Accordingly, studying animals
that have altered sleep habits might cause these animals to fall asleep during memory testing or
extinction which would be detected as freezing behaviour. We have shown that visual-based
measures of freezing and EMG measures can mistake sleep for freezing behaviour. However, we
also show that the magnitude of the EMG can help clarify if an animal is sleeping or in a state of
26
fear. Thus it is recommended that EMG and EEG be used to verify the true behavioural state of
an animal during memory testing of fear conditioning paradigms.
27
CHAPTER 2
SYSTEMS CONTROLLING TRACE FEAR BEHAVIOR
28
2.1 Elaborating the fear circuits with trace fear
In contrast to delay or cued fear conditioning, trace fear conditioning involves the
presentation of the CS and US separated by a time interval (Figure 1.1, Chapter 1). As mentioned
in Chapter 1, this may engage more circuits that would be necessary to make an association
between CS and US across the trace fear interval. Chapter 1 demonstrated that neck EMG can be
a useful means to measure freezing behaviour in fear paradigms. Moreover, this technique
permits the investigator to know precisely when the mouse begins or terminates freezing
behaviour. Thus, for the first time electrophysiological brain cell recordings can be conducted in
parallel with a precise measure of the onset of freezing behaviour. The ensuing Chapter will
elaborate upon the circuitry presented in Chapter 1, to include systems involved in trace fear
memory (Figure 2.1). This chapter will conclude with detailed experiments investigating the role
of the ACC in trace fear conditioning.
2.2 Amygdala and trace fear
Since the function of the amygdala has already been reviewed in the previous chapter,
this section will focus on the potential role of the amygdala in trace fear conditioning. Pare and
Collins (2000) conducted a study in which warning signals would predict a shock. In their
paradigm several tones were played which preceded shock stimulus. Field potential and unit
recordings were obtained from the lateral amygdala during conditioning. This paradigm is
somewhat analogous to the trace fear paradigm in that tones predicted a shock over a time
interval. In this study neurons were found to become locked with the tone signals, coincident
with augmented tone-evoked potentials and in a few cases neuron activities began to ramp,
appearing to be related to the expectation of the shock. Cell activity also became more
synchronized, appearing to be correlated with the theta band field potentials (4-7Hz).
29
Collectively, the results indicate that either the lateral amygdala itself or other structures are
capable of maintaining the activation of the central amygdala, even in the absence of tone
stimuli. In a more recent study, conducted in humans using fMRI, it was found that the central
nucleus of the amygdala became active when threatening stimuli (shock) were in close proximity
(Mobbs et al., 2007). However, when the threat was relatively far away, the basolateral amygdala
tended to be more active, suggesting that it may be involved in the assessment of the threat and
possibly escaping it. Thus the lateral and basolateral amygdala appear to be important in the
anticipation of threats, while the central amygdala appears important for imminent threat. Thus,
the inputs that modulate or drive these parts of the amygdala may help elaborate the networks
involved in trace fear conditioning.
2.3 Hippocampus and trace fear
2.3.1 Basic structure of the hippocampus
The medial temporal lobe system has been long implicated in the formation of declarative
memory (Squire & Zola, 1996). Declarative memory is defined by memory of facts and events
(Squire & Zola, 1996). The medial temporal lobe consists of the hippocampal formation and
extra-hippocampal structures, such as the parahippocampal and perirhinal cortices. (Squire &
Zola, 1996; Amaral & Lavenex, 2006). The hippocampal formation is organized into seven
divisions including Cornu-Ammonis regions 1-4 (CA1-4), the dentate gyrus, subiculum and
entorhinal cortex (Squire & Zola, 1996), and has been well implicated in trace fear memory
(McEchron et al., 1998; Quinn et al., 2002; McEchron et al., 2003; Weitemier & Ryabinin,
2004; Misane et al., 2005; Quinn et al., 2008). The entorhinal cortex is the major input to the
hippocampus and also serves as a major interface between the brain and the hippocampus. The
basic circuit involves information flow through the entorhinal cortex to the dentate gyrus, then to
30
CA3, then CA1, and finally through the subiculum where it travels back out through the
entorhinal cortex (Amaral & Lavenex, 2006). A summary of the major connections of the
hippocampal formation is depicted in Figure 2.1 (pg 31). Damage to hippocampus in amnesia
& Squire, 1998). Thus, it was proposed that awareness is required for trace eye-blink
conditioning (Clark & Squire, 1998). Consistently, it was found that fear (as measured by skin
response) was experienced during trace fear acquisition only when the subject was cognitively
aware of the CS which predicted the aversive stimulus (Knight et al., 2006).
2.3.2 Hippocampus and trace fear conditioning
Pharmacological inhibition of the dorsal and ventral hippocampus prior to training impairs trace
fear memory when tested 24 hours after conditioning (Esclassan et al., 2009). Consistently, it
was found that pre-training lesions restricted to the dorsal hippocampus of the rabbit impairs
trace fear cardiac responses (McEchron et al., 2000). By contrast, it was observed that
excitotoxic lesions to the ventral, but not dorsal, hippocampus impair trace fear acquisition
(Yoon & Otto, 2007); however the dorsal and ventral hippocampus were found to be important
for retrieval of trace fear memory 24 hours after conditioning (Yoon & Otto, 2007). The ventral
hippocampus (CA1 and subiculum) is known to project to the basolateral amygdala (Canteras et
al., 1992), and damage to this region disrupts contextual fear conditioning (Maren & Fanselow,
1995). Thus, while the involvement of the hippocampus in trace fear conditioning is
unequivocal, the hippocampal region which is most important cannot be agreed upon.
31
Figure 2.1 Hippocampus and prefrontal circuits Top middle: Areas of interest, including hippocampus, amygdala and prefrontal cortex. Left side: Cross-section through the hippocampus showing the Cornu Ammonis regions (CA1-3), dentate gyrus (DG), and entorhinal cortex (EC). Schematic organization of hippocampal connections is depicted below. Top right: divisions of the prefrontal cortex (each color represents a division). This thesis focuses on regions within the medial division including prelimbic (PrL) and anterior cingulate (labelled as CG1 in the mouse). Bottom right: is a coronal cross-section showing layer by layer efferent and afferent connections to and from relevant structures involved in fear conditioning (center circle). Connections which are common to both the ACC and the prelimbic cortex are colored in black. Connections which belong solely to the prelimbic cortex are in red. Notice that the prelimbic cortex but not the ACC receives input from the CA1 region of the hippocampus. In addition the ACC is only connected with the dorsal aspect for the striatum but the prelimbic cortex is connected with both dorsal and ventral striatum. Also, note that the hippocampus, prelimbic cortex and the ACC all share connections with the basolateral amygdala (BLA). Ifl, infralimbic, AID, agranular insula dorsal, IV, agranular insula ventral, M1, motor cortex,, MD, mediodorsal thalamus, PAG, periaqueductal gray, PrC, precentral, Lat-hypothal, lateral hypothalamus, LO, lateral orbito, VLO, ventral lateral orbito ,VO, ventral orbito. (figure by H. Steenland, refs in text)
The trace fear interval is typically between 15 and 30 seconds, and manipulations of this
time may be useful to determine the minimum interval necessary to engage the hippocampus or
32
other structures. Injection of NMDA receptor antagonist D-APV into the dorsal hippocampus (in
and around CA1) prior to conditioning, blocks trace fear acquisition for time intervals of 15 to 30
seconds, but not intervals less than 15 seconds (Misane et al., 2005). These results suggest that
the dorsal hippocampus plays an important role in bridging the time interval between tone and
shock.
It was hypothesized that the hippocampus maintains temporal contiguity for the tone and
shock stimuli across the trace fear interval (Bangasser et al., 2006). To examine this issue,
Bangasser et al. (2006) first confirmed that lesioning the rat dorsal and ventral hippocampus
impaired trace fear memory 24 hours later. However, if rats are trained with additional tone
stimuli (2sec) coincident with the shock, the lesions did not impair trace fear memory. Thus, the
hippocampus appears to be important for associating two discontiguous events (i.e. events that
occur at different times) (Bangasser et al., 2006). How exactly the hippocampus is involved in
this contiguity is still left up to debate.
It has been shown that CA1 neurons are activated during early conditioning to the CS and
this effect decreases during later conditioning trials (Gilmartin & McEchron, 2005a).
Consistently, the activity of tone-evoked neuron responses in CA1 correlated with the degree of
freezing behaviour during learning in rats (Chen et al., 2009). Functional imaging studies in
humans also demonstrated that the magnitude of activity in the hippocampus is greatest during
early trial of trace fear training, and its activity is predictive over the behavioural association of
the shock and tone (Knight et al., 2004). However, the dentate gyrus has also demonstrated
increased firing for both the US and CS progressively across training trials (Gilmartin &
McEchron, 2005a). Thus, converging evidence indicates that the hippocampus is related to the
US and CS association and the early processing during trace fear acquisition.
33
Knockout of NMDA receptors in CA1 pyramidal neurons is sufficient to slow the
acquisition and impair the retention of trace fear memory (Huerta et al., 2000). It has been shown
that dorsal intra-hippocampal injections of NR2A antagonist (NVP-AAM077), prior to training,
impairs contextual and trace fear acquisition and decreases c-Fos expression (Gao et al., 2009).
By contrast, the NR2B antagonist (RO-25-6981) blocked trace fear conditioning without any
impact on contextual fear (Gao et al., 2009). Cultured primary hippocampal neuron studies
revealed that the NR2A antagonism likely affects the ERK1/2 pathway to impact c-Fos through
MSK1 (Gao et al., 2009). MSK1 is known to be a down-stream effecter of the ERK pathway
(Sindreu et al., 2007). However, activation of the NR2B subunit was not found to influence the
ERK pathway (Gao et al., 2009). Interestingly, a study conducted much earlier (Runyan et al.,
2004) had shown that infusion of an ERK inhibitor into the hippocampus, prior to conditioning,
reduced both contextual fear and trace fear memory 48 hours after trace fear acquisition.
Collectively, the ERK pathways appears important for contextual and trace fear memory,
possibly mediated through activation of NR2A containing NMDA receptors.
2.3.3 Hippocampus and trace fear memory retrieval
Examination of CA1 neurons in rabbits and rats, during trace fear memory recall,
revealed that these neurons fire when animals are anticipating the shock stimuli (McEchron et
al., 2003; Chen et al., 2009). In addition, the activity of these neurons appeared to be related to
the strength of the CS-US association (McEchron et al., 2003). Consistently, neurotoxic lesions
to the dorsal hippocampus demonstrate that it is involved in the storage of recent but not remote
memory (Quinn et al., 2008). Chowdhury et al. (2005) discovered that lesions of the dorsal
hippocampus (mostly in CA1), 1-2 days after conditioning, impaired trace fear memory recall.
34
2.3.4 Gene expression in the hippocampus following trace fear conditioning
One way to explore how trace fear memory differs from cued or delay fear memory is to
look at the expression of genes or proteins following induction of trace fear memory or after
retrieval of this memory. Weitemier and Ryabinin (2004) conducted a study in which they used
immunohistochemistry to examine protein expression in the mouse hippocampus. Interestingly,
they found that c-fos was differentially enhanced in the CA3 and dentate gyrus as a result of
trace fear conditioning. Following memory recall (tested 24 hours after conditioning) zif268 was
elevated in the trace fear condition above that of the delay fear condition in CA3 and the DG. It
is interesting that neither acquisition nor later recall increased expression of Zif268 in the CA1
region of the hippocampus, which has been strongly implicated in trace fear memory (see
above). Putting technical issues aside, this may be explained, by asserting that the integrity of the
hippocampal circuitry is what is important for fear memory acquisition rather than any one
specific region (for example CA1 alone). Also, equally interesting from the perspective of motor
systems, is that during trace-fear acquisition there tended to be more immobility compared to that
of delay fear conditioned animals (Weitemier & Ryabinin, 2004). Thus, the results may simply
be related to differences in motor inactivation, a possibility which appears to be rarely taken into
account in studies of c-fos and Zif268.
A recent study has examined a large variety of genes using RNA microarray analysis of
the mouse hippocampus (unspecified locations in dorsal aspect) following trace-fear
conditioning (Sirri et al., 2009). Thirty minutes after trace fear conditioning there was a rise in:
et al., 2006). The output from the prelimbic cortex to the basolateral amygdala is from layers II
and V and is excitatory (Gabbott et al., 2005; Likhtik et al., 2005) suggesting a degree of
reciprocal laminar distribution. Considerable literature also suggests that the prelimbic and
infralimbic cortices receive monosynaptic input (~15msec lag) from the subiculum/CA1 region
37
of the hippocampus (Swanson, 1981; Ferino et al., 1987; Thierry et al., 2000). However this
direct input appears most prominent from the ventral hippocampus (Swanson, 1981; Ferino et al.,
1987; Thierry et al., 2000). Interestingly, there are reports of phase locking between the
prelimbic neuron activity and the hippocampus theta rhythm during spatial memory tasks (Jones
& Wilson, 2005; Siapas et al., 2005). This phase locking is offset by about 50msec (Jones &
Wilson, 2005; Siapas et al., 2005). There is also evidence that monosynaptic connections to the
prelimbic cortex activate inhibitory interneurons followed by activation of pyramidal neurons
(Gabbott et al., 2002; Tierney et al., 2004), which may be necessary for the generation of theta
(4-7.5Hz) waves (Siapas et al., 2005).
It has been found that there is convergent input from the basolateral amygdala and the
hippocampus in the prelimbic cortex (Ishikawa & Nakamura, 2003). If both inputs are stimulated
at the same time the response in the prelimbic cortex is enhanced; however, if either input is
stimulated before the other, a depressed response to the latter stimulation will occur (Ishikawa &
Nakamura, 2003). The results suggest that synchronizing inputs (for example, local theta field
potentials) from both hippocampus and the amygdala may have an augmenting effect on the
prelimbic cortex.
2.4.2 Prelimbic cortex and trace fear conditioning
Only a few studies have examined the involvement of the prelimbic cortex in trace fear
conditioning. Baeg et al (2001) found that the majority of neurons changed their firing to
coincide with the presentation of the CS for both delay and trace fear (2sec trace interval)
paradigms, and these responses decreased during fear memory extinction. Interestingly, putative
inhibitory interneurons and putative pyramidal neurons showed different patterns of responses.
Inhibitory neurons tended to show transient responses and increased their firing rates following
38
CS presentation; however, pyramidal neurons were observed to show long lasting depressed
activity following the CS (Baeg et al., 2001). Baeg et al. (2001) also report that there was no
obvious relationship between neuron activity and the magnitude of freezing itself. However a
recent study examining delay fear conditions showed that the majority of neurons in the
prelimbic cortex responded in an excitatory way to the CS and that in many cases this activity
was increased coincident with freezing behaviour (Burgos-Robles et al., 2009). These authors
suggest that the prelimbic cortex drives the basolateral nucleus directly to impact the central
nucleus for fear behaviour expression (Burgos-Robles et al., 2009).
Gilmartin and McEchron (2005b) showed that prelimbic neurons have learning-related
increases in activity to the CS and US in a trace fear paradigm (20sec trace interval).
Interestingly, a subset of prelimbic neurons also exhibited increases in activity during the trace
interval. These results suggest that the prelimbic cortex may be important for bridging the time
gap of the trace fear interval, through maintaining the association between tone and shock stimuli
(Gilmartin & McEchron, 2005b). Converging evidence for this concept comes from frontal
imaging studies in humans, in which the frontal operculum and middle frontal gyri were
differentially activated during trace conditioning relative to delay conditioning (Knight et al.,
2004). Gilmartin and McEchron (2005b) also showed that the neuron response of the infralimbic
cortex tends to decrease during the CS, US presentation, possibly implicating an inhibitory
output from the prelimbic to this region.
In a delay fear conditioning paradigm conducted by Burgos-Robles et al. (2009), it was
also found that prelimbic neurons become more synchronous with conditioning and burst more
often. This finding is consistent with recent data showing that theta oscillations recorded from
the prelimbic and infralimbic cortex demonstrate enhanced theta power during a trace (1.5sec)
reward test (Paz et al., 2008). It was also found that neuron activities tend to fire near the
39
negative trough of the field potential. This result indicates that the hippocampus may be driving
prefrontal memory recall (Paz et al., 2008).
2.4.3 Prelimbic cortex and trace fear memory
Runyan et al. (2004) found that immediately after trace fear training, ERK is
phosphorylated. Interestingly, blocking ERK activity prior to conditioning did not influence trace
fear acquisition but did impact memory retention 24 and 72 hours after conditioning. A similar
result was produced when ERK was blocked immediately after trace fear conditioning,
suggesting that some process occurs after trace-fear conditioning to consolidate memory,
possibly in the prelimbic cortex. Runyan and Dash (2004), also blocked dopamine receptors in
the prelimbic cortex prior to trace fear conditioning. Antagonizing dopamine was found to block
the rise in ERK phosphorylation and reduce long-term memory retention (Runyan & Dash,
2004). Consistently, blocking the PI3Kinase pathway in the prefrontal cortex after trace fear
conditioning interferes with remote memory retention, examined 3d and 6d after trace fear
conditioning training (Sui et al., 2008). Finally, neurotoxic lesions to the prelimbic cortex made
after conditioning demonstrate that the prelimbic cortex is involved in the storage of remote
memory (Quinn et al., 2008).
2.4.4 Basic structure of the anterior cingulate cortex
The ACC is an agranular (lacking layer IV) midline cortical area located dorsal to the
prelimbic cortex. The ACC is defined and innervated in layer III by projections originating from
midline and intralaminar nuclei of the medial thalamus (Vogt et al., 1979; Devinsky et al., 1995;
Hsu & Shyu, 1997; Hsu et al., 2000; Shyu et al., 2004). Layer VI of the ACC projects back
extensively to the mediodorsal thalamus (Gabbott et al., 2005). Also, Layer V of the ACC
40
projects prominently to the dorsal but not ventral striatum and no report has ever demonstrated a
direct connection with the hippocampus (to my knowledge), which distinguishes it from the
prelimbic and infralimbic cortices (Figure 2.1). Thus, striatal output and innervation from CA1
may be a helpful method to differentiate functions of ACC from that of the prelimbic cortex. For
example, the ventral striatum is argued to be important in allocating attention resources to cues
which predict danger (Dalley et al., 2004; McNally & Westbrook, 2006) which is consistent with
the mnemonic function of the prelimbic region. By contrast, the dorsal striatum has been
implicated in aversive learning (Delgado et al., 2008), which is consistent with neural responses
of the ACC (see sections 2.5.1 - 2.5.4). Similar to the prelimbic cortex, the ACC projects to the
lateral hypothalamus, periaqueductal gray and the basolateral amygdala (Floyd et al., 2000;
Gabbott et al., 2005). The connection between the ACC and the basolateral amygdala has been
identified as glutamatergic in nature (Bissiere et al., 2008).
2.4.5 Pain responsive neurons in the anterior cingulate cortex
Shock stimuli are a major component of the trace fear paradigm. Not only is information
regarding nociception sent to the amygdala, it is also sent to the ACC (Fig. 2.2 pg 41, ACC is
labelled as CG1 or 2 in mouse). Neurons of the ACC respond predominantly to noxious pinch or
electrical stimulation to the skin; however responses can also be elicited with pressure and
tapping of the skin, and colonic distension (Davis et al., 1994; Ploghaus et al., 1999; Wei et al.,
1999; Wei & Zhuo, 2001; Kung et al., 2003; Yang et al., 2006). In addition, the receptive field of
ACC neurons can respond to stimulation from large portions of either side of the body. Most
ACC neurons in these studies responded with excitation to peripheral stimulation and were
located in layer V in rats and layer III in rabbits (Sikes & Vogt, 1992; Yamamura et al., 1996). In
addition to spike recording conducted in the ACC, extensive evoked-field potential recording has
41
Figure 2.2 Elaborated fear circuits This is an elaboration of the fear circuit presented in Figure1.2, however this circuit now includes hippocampal (top) and prefrontal circuits (left) and the appropriate connections with the thalamus. Notice the added connections between the CG1 (rodent ACC), Prelimbic, CA1 hippocampus and the BLA. In the context of trace fear memory, the prelimbic cortex and CA1 hippocampus may maintain the memory over the trace interval through activation of the BLA. The BLA would then activate of central amygdala to produce freezing behavior. The cingulate cortex may be involved in the affective component of the shock stimuli, or be involved in modulating fear behavior though connection with the periaqueductal gray (PAG). Abbreviations are the same as for Figure 1.2 and 2.1, IFL is infralimbic. (figure by H. Steenland, refs in text)
42
been done in anesthetized rat preparations (Wei & Zhuo, 2001; Sun et al., 2006; Yang et al.,
2006). In these studies it was shown that noxious peripheral stimulation elicited sink currents in
lamina II/III and V of the ACC, reflecting synchronous depolarization of neuron populations.
These currents are mediated through glutamatergic synaptic transmission and could be abolished
by medial thalamic lesions (Yang et al., 2006). While both the rat and the mouse have brains of
comparable scale and shape, it cannot be assumed that the regional boundaries and neuron
response profiles of the mouse brain are the same as that of the rat. Based on the afferent and
efferent connections with the ACC, it is likely be involved in processing incoming sensory input
in layer III and possibly be involved in a motor output via layer V. However a basic description
of ACC neuron responses to peripheral stimulation has never been conducted in the mouse.
2.4.6 Anterior cingulate cortex and trace fear conditioning
Much less work has focused on the role of the ACC in trace fear memory. A study by
Han et al. (2003) demonstrated that lesions to the ACC in mice impairs trace fear conditioning
without affecting delay fear conditioning. In addition, it was reported that c-fos expression was
elevated in the CG1 subregion of the ACC when measured 30 minutes after trace fear
conditioning. Interestingly, ACC lesioned mice demonstrated impairments in trace fear
acquisition as early as the first training trial. The authors attribute their finds to the importance of
the ACC in attention processes. Indeed, distracting stimuli presented during training, are
sufficient to produce deficits in trace fear memory similar to the effect of ACC lesioning (Han et
al., 2003). Consistently, mouse models of Fragile-X syndrome and chronic pain (Complete
Freund's Adjuvant) both demonstrate impaired trace fear conditioning, possibly because both
animal models are impaired in attentional processing (Zhao et al., 2005; Zhao et al., 2006).
43
However, fMRI studies in humans has not demonstrated any significant role for the ACC
in trace fear conditioning that it doesn’t already play in delay conditioning (Buchel et al., 1999;
Knight et al., 2004). Since there is nearly equal evidence for and against the possibility of the
ACC in trace fear memory, its involvement should be considered unresolved. Rather, more basic
recordings need to be done to ascertain the activity of neurons during trace fear conditioning. To
really get an understanding of what the ACC does, the following section will examine research
on the role of the ACC in various learning paradigms and behaviours relevant to trace fear
conditioning.
2.5 Close examination of anterior cingulate cortex
2.5.1 Cued and contextual fear conditioning
Blocking glutamatergic transmission in the ACC is sufficient to impair the acquisition of
Inc) were implanted in the mouse ACC for freely behaving and anesthesia drug delivery. In the
case of freely behaving mice, a guide cannula (with a dummy probe) was implanted previously
on the day of surgery (at least 1 week before the experiment). For these experiments a frontal
parietal EEG electrode was also placed on the skull in addition to EMG electrodes to record from
the neck (see above). For anesthesia experiments the microdialysis probe was lowered into the
ACC on a steep 70º angle (AP 0.6 , ML 0.5, DV 2.3). Careful attention was paid to avoid cortical
surface vasculature. After 30 minutes had elapsed, the reference electrode was lowered in place
near the shaft of the microdialysis probe. The recording electrode was then slowly lowered into
the brain in 5µm steps, parallel to the probe. Mineral oil was applied to the exposed surface of
the brain every hour. Mice were placed on a heating pad (37 ºC) throughout the experiment. The
dialysis probe was perfused (1-3µL/min with ACSF for ~1hr or more following implantation).
The composition of the ACSF was the same as that used for in-vitro recording.
2.7.3H Trace fear conditioning
Trace fear conditioning was performed in an isolated shock chamber (Med Associates, St.
Albans, VT). The conditioned stimulus (CS) was a 80dB white noise, delivered for 15s, and the
unconditioned stimulus (US) was a 0.75mA-scrambled foot-shock for 0.5s (Wu et al., 2008).
Mice were acclimated for 60s, and presented with ten CS–trace–US–ITI trials (trace of 15 sec,
58
inter-trial interval (ITI) of 225s). One day after training, mice were acclimated for 60s and
subjected to ten CS–ITI trials (ITI of 225s) in a novel chamber to test for trace fear memory
(Huerta et al., 2000). Freezing is typically defined as the absence of movement with the
exception of breathing. Conventional trace fear paradigms have been performed with animals
during the day, when animals normally sleep. This was previously necessary because
behavioural scoring and motion detection required light to determine freezing behaviour. Thus,
most studies have been carried out when the animal would naturally be sleeping. To circumvent
this potential confound we utilized neck EMG recording to examine freezing behaviour in the
dark (0.2 lux) when mice are naturally awake (Steenland & Zhuo, 2009). The additional
advantage of recording in the dark with EMG is that it may remove contextual cues. ACC field
potentials and unit recordings were obtained from animals during the paradigm to examine the
role of the ACC in trace fear conditioning.
2.7.3I Histology
At the completion of freely behaving animal experiments, mice were euthanized with
isoflurane and a DC current (100-500uA) was passed between recording electrodes and ground,
to lesion the brain. The brain was then removed and fixed with 10% formalin solution. After at
least a week, the brains were transferred to 30% sucrose, and cut in 40µm coronal sections with a
cryostat (CM 1850; Leica). Sections were mounted and observed under microscope immediately
after sectioning. Some brains were also processed for Nissl staining. Lesion and microdialysis
sites were localized and co-registered with the stereotaxic atlas of the mouse brain (Paxinos &
Franklin, 2000).
2.7.3J Recording
59
Electrophysiological signals were amplified and filtered (Super-Z headstage amplifiers
and BMA-400 amplifiers and filters, CWE Inc., Ardmore, PA, USA) as follows: EEG 1000X at
1-100Hz, EKG 1000-5000X at 50-100Hz, and EMG 2000X at 100-1000Hz. An electronic
blanker (Model, SP1 EKG blanker) was specially modified for use with mice (Modified by
Charles Ward, CWE Inc.) to remove the EKG artifact from diaphragm EMG recording. The
blanker was modified to detect and remove EKG artifacts of heart beats up to 1000 beats/s. The
blanker also provided a rectified diaphragm signal. Since diaphragm recordings included
prominent EKG artifact superimposed on the diaphragm signal, the signal was sent to two
amplifier channels to obtain separate recordings of EKG and diaphragm EMG. Signals were
digitized, smoothed (25ms time constant), rectified when necessary (Spike2 software, 1401
interface, CED Ltd., Cambridge, UK) and recorded on a computer.
Spike recording was done with a custom headstage operational amplifer, built from a
TL064 package in the unity gain follower configuration (Jeantet & Cho, 2003). This is a basic
headstage model that can be used for in-vivo freely behaving animals or for in-vivo anesthetized
work. The headstage has four channels and is powered by a +/-3 V DC split battery supply. The
headstage was interfaced with BNC connectors and sent to a differential amplifier (Brownlee
Precision Instruments, BP-440, San Jose, California, USA). At the level of the differential
amplifier the signals were spliced and sent to two different channels to separate the local field
potential (for anesthesia only) from spike recording. Spike recording was filtered between 300
Hz to 10 kHz with 1000X amplification. Local field potentials were filted between 1-100Hz and
amplified by 1000X. Spike signals were then digitized at 20kH, EEG, field potential and EKG
signals at 400Hz, and EMG signals at 4000Hz with an analog to digital converter (1401
interface, CED Ltd., Cambridge, UK).
60
2.7.3H Analyses
Physiological signals were analyzed in 5s epochs for the periods of wakefulness and sleep
using Spike2 software. Scripts for EEG analysis (Sudsa-version 2.2) were obtained from CED.
Custom scripts were written in our laboratory (by H.W.S.) for quantifying heart rate, breathing,
neck EMGs and spike recording. Fast-Fourier transform was used to convert EEG waveforms
into total power (μV2), binned every 5s, for the following frequency bands: δ1 (0.5-2Hz), δ2 (2-
4Hz), θ (4-7.5Hz), β1 (7.5-13Hz), β2 (13-20Hz) and α (20-30Hz). Neck muscle recording was
quantified by integrating the area under the rectified and smoothed EMG waveform every 5s.
Spike sorting was done off-line using a template matching program provided by Spike2
software. The threshold for analyzing spikes was set at 2x the baseline noise level. The spike
template ranged from 0.5 ms before the spike peak, to 1.5 ms after the peak. After template
generation, principle component analysis was performed and the templates were then revised
based k-means clustering of the spikes. Peristimulation-time histograms were generated with a
bin size of 10ms. Single units from each ensemble were normalized as z scores (Tsai et al.,
2004). In brief, the 5-s period before stimulation was used as a baseline period. The mean firing
rate and standard deviation of the 500 bins in this baseline period were calculated. All bin values
were transformed to Z scores according to the mean and standard deviation of the baseline
period. A 99% confidence level (Z>2.33 or Z<-2.33) was used for identifying if a single unit
responded (Kuo & Yen, 2005). Chi (χ2) test was used to compare different neural populations.
Under all recording conditions several features of extracellular recorded spikes were
recorded, including: firing rate, half-width time and peak to peak time. Interneurons and
pyramidal neurons were discriminated based on two dimensional plots of half amplitude versus
61
peak to peak time. Autocorrelations were performed for each neuron. Bursting cells were
classified as cells whose autocorrelations peaked at 3-6 ms followed by an exponential decay.
Cells were classified as bursting if the, maximum peak on the auto-correlogram between 3 and 6
ms was > or = 50% of the maximum bin value of the first 50ms (Bartho et al., 2004). The
criterion for regular spiking (non-bursting) neurons was that the mode of the interspike-interval
histogram was >35ms. Regular spiking neurons rarely discharged in bursts and the auto-
correlogram showed an exponential rise from time 0 to tens of milliseconds. Cells that don’t
match either criterion were labeled as indeterminate.
2.7.4 Results:
2.7.4A Characterization of spontaneous field potentials in the ACC during trace fear
conditioning
Trace fear conditioning (Figure 2.3A) involves the association of a tone and shock across
a time interval and is thought to engage the ACC (Han et al., 2003). Since the trace fear
conditioning paradigm involves a variety of stimuli (e.g. sound and shock) and motor responding
(freezing behaviour) along with learning, it is an ideal paradigm to examine the function of the
ACC in sensory-motor integration and plasticity. Neck EMG was used as previously described
(Steenland & Zhuo, 2009), to determine when the mouse was freezing (Fig. 2.3B). As expected,
a significant alteration in freezing behaviour during the trace interval was detected during
conditioning (10 training trials) and memory testing (24 hours later) (F(20, 130) = 5.10.67; p <0.001
, one-way ANOVA). Post-hoc analysis revealed that the mice both learned and remembered the
tone and shock association (all t>3.45, all P<0.015, Fig. 2.3C).
Trace fear conditioning was conducted in parallel with local ACC field potential
recording to examine whether changes in spontaneous oscillation state or alteration in tone-
62
Figure 2.3 Trace fear conditioning and measurements of fear (A) In the standard trace fear paradigm, a tone and shock are presented, separated by a time interval (trace interval). The paradigm is repeated until the mouse learns that the tone predicts the shock (as evidenced by freezing during the trace interval). Memory can be tested 24 hours later, with presentation of the tone alone. If the mouse expects to be shocked, it will freeze during and after the presentation of the tone. Highlighted green segments indicate regions of analysis in B. (B) Example of EMG-based freezing in one mouse during the trace fear interval. Each panel during conditioning and memory represents one trace interval trial. Green underline indicates periods scored as freezing. (C) Grouped data showing an increase in EMG-based freezing during both learning and memory compared to baseline (BL). * indicates statistical significance P<0.05, Large green dots correspond to freezing behavior in B. Arb., arbitrary units, BL, baseline, ES, expected shock, ITI, inter-tribal interval, Rt/Sm, rectified and smoothed signal, S, shock. Error bars are +/- SEM. evoked potentials could be detected. It was previously shown that local field potentials, in the 4-
7.5Hz range, can be augmented in the ACC and the prelimbic cortex during reward anticipation
(Tsujimoto et al., 2006; Paz et al., 2008) and in the amygdala and hippocampus during fear
behaviour (Pare & Collins, 2000; Seidenbecher et al., 2003; Narayanan et al., 2007). Examples
63
Figure 2.4 Local field potential changes ACC accompany trace fear conditioning (A) Example Fourier transform computed for one mouse on spontaneous ACC field potentials during each phase of the trace fear paradigm. Comparisons are made directly to baseline (50-60sec of time before presentation of first tone). Clear power enhancements can be seen between 2-4Hz for tone and 4-7.5Hz for trace interval (arrows). A consistent increase was also seen between 13-20Hz. Shock stimuli did not change field potential power. Below each power spectrum is a range of frequencies for which the power spectrum was binned for analysis purposes. (B-G) Shows binned averages of all trials for each power spectral bin. The 2-4Hz power was significantly increased for the tone interval, while the 4-7.5 and 13-20Hz bands were enhanced for the trace fear interval. * indicates statistical significance P<0.05 compared to baseline (BL). of the Fourier transform (prior to frequency binning), for spontaneous field potentials from one
animal during trace fear conditioning, are depicted in figure 2.4A. Data was then binned into
64
frequency ranges for statistical analysis (Fig. 2.4B-G). A significant alteration in 2-4Hz
powerwas detected during trace fear conditioning (F(3,27) = 4.49; p =0.011 , one-way ANOVA).
Post-hoc analysis revealed that 2-4Hz power was increased during the tone condition above that
of baseline (t = 2.70; p =0.035, Fig. 2.4C). This fits with the possibility that relevant sensory
information may come to activate the ACC with conditioning. A significant alteration in 4-7.5Hz
power was also detected during trace fear conditioning (F(3,27) = 5.027; p =0.007, one-way
ANOVA). Post-hoc analysis revealed that 4-7.5Hz power was increased during the trace interval
above that of baseline (t = 2.65; p <0.04, Fig. 2.4D) consistent with findings from other brain
regions (Pare & Collins, 2000; Seidenbecher et al., 2003; Tsujimoto et al., 2006; Narayanan et
al., 2007; Paz et al., 2008). Thus, these oscillations may be related to memory recall of fear
behaviour or attention to impending danger. A significant alteration in 13-20Hz power was also
detected during the trace fear interval (F(3,27) = 6.67; p =0.002 , one-way ANOVA). Post-hoc
analysis revealed that 13-20Hz power was increased during the trace fear interval above that of
baseline (t = 4.08.; p = 0.001, Fig. 2.4F). Interestingly no changes in spontaneous EEG were
detected following the shock stimuli, compared to baseline. However our analysis was restricted
to frequencies below 30Hz and stimulus artifacts precluded observation at the exact point of
stimulation. Thus, shock analysis was carried on the 15 seconds of data which was collected
immediately following the termination of the shock. Moreover, evoked-pain potentials may be a
more relevant measure in this case.
2.7.4B Characterization of tone-evoked and motor-triggered potentials in the ACC during trace
fear conditioning
The prelimbic cortex and lateral amygdala demonstrate enhanced tone-evoked potentials
with conditioning (Pare & Collins, 2000; Mears et al., 2009). Figure 2.5A (top panel) shows an
65
Figure 2.5 Tone-evoked and motor-locked potentials in the ACC (A) Demonstrates that tone-evoked potentials could be detected in the ACC during trace fear conditioning and do not necessarily require a robust motor activation. Upper trace is an example of an average of ten trials during conditioning from one mouse. Middle trace is the field potential average and standard error of the mean for all mice recorded. Notice the peak around 50msec and the trough around 150msec. Lower trace is the rectified and smoothed EMG potential average and standard error of the mean for all animals recorded. (B) Demonstrates EMG triggered potentials during trace fear conditioning which appear to precede motor activation. Upper trace is an example of averaged EMG-triggered field potentials from 100 freezing-movement events. Middle trace is the field potential average and standard error of the mean for all animals recorded. Notice the peak activation around 70msec. Lower trace is the rectified and smoothed EMG potential average and standard error of the mean for all animals recorded. Arb., arbitrary units. AVG, average. Grey rectangle represents Z-scores for 99% confidence. example of an average of 10 tone-evoked potentials from the ACC of one mouse for all tones
during trace fear conditioning. Grouped data is also depicted (Fig. 2.5A, middle panel) in register
with changes in neck EMG (Fig. 2.5A, bottom panel). In 10 of 10 mice, tone-evoked potentials
could be observed with a significant upward deflection between 30 and 50msec, consistent with
other brain regions (Pare & Collins, 2000; Mears et al., 2009). A significant, latent, downward
deflection was also detected in 5 of 10 tone-evoked potentials with an average trough occurring
66
around 150msec. An increase in EMG activity was only detected in 3 of 8 animals in response to
tone stimuli (Fig. 2.5A, lower panel). Thus, these tone-evoked potentials appeared to be related
to a sensory function of the ACC rather than a motor function per-se.
Potentials in the motor cortex and in premotor regions of the brain are commonly
associated with motor action (Donchin et al., 2001; Rickert et al., 2005; Roux et al., 2006).
When animals are trained during trace fear conditioning, they go through many periods of
spontaneous freezing which are eventually followed by spontaneous movement. These periods
occur during inter-trial intervals and during the conditioning trials. The execution of movement
from a point of freezing behaviour may be an attempt to explore the safety of the environment or
even to escape from that environment. We examined local field potentials at a point where
freezing (lasting at least 2 seconds) evolves into movement. To do this, local field potentials
were back-triggered and averaged from the onset of EMG activation. Figure 2.5B (top panel)
shows an example and average back-triggered potential for one animal from the time of freezing
to the time of movement. Grouped data is also depicted (Fig. 2.5B, middle panel) in register with
changes in neck EMG (Fig. 2.5B, bottom panel). In 8 of 10 animals, field potentials could be
observed with a significant upward deflection which peaked between 50-70msec after the
movement. The waveform repolarized even while the EMG remained elevated. Interestingly,
based on the average waveforms of all data, the spontaneous EMG-related potential started to
become elevated ~25msec before motor movement, demonstrating that the ACC potentials may
have a premotor component.
2.7.4C Learning-related spontaneous and evoked-field potentials in the ACC during trace fear
conditioning
67
The association between tone and shock was increased with conditioning as demonstrated
by freezing measurements (Fig. 2.3B, C). Thus, neural correlates of this memory may be
reflected by learning curves of spontaneous and tone-evoked field potentials. The results from
figure 2.4C, D and F demonstrated that 2-4, 4-7.5 and 13-20Hz rhythms change during trace fear
conditioning. However, trial by trial analysis revealed that no significant learning curve existed
for the 2-4Hz frequency for the tone stimulus (F(10, 88) = 1.12; p <0.358 , one-way ANOVA, data
not shown). By contrast, a significant effect for the trace fear interval on 4-7.5 Hz power was
detected (F(10, 88) = 2.10; p <0.032 , one-way ANOVA). Post-hoc analysis revealed that a
significant learning curve for 4-7.5 Hz power (t=3.31, p=0.013, 7th time point, Fig. 2.6A).
Correlation analysis also revealed that the 4-7.5Hz learning curve co-varied with the freezing
learning curve (r2= 0.25; P<0.001). In addition, a significant effect for the trace fear interval on
13-20Hz power was detected (F(10, 88) = 2.25; p <0.022, one-way ANOVA). Post-hoc analysis
revealed that a learning curve occurred for 13-20Hz power (t=3.76, P=0.003, 8th time point, Fig.
2.6B). However, correlation analysis revealed that the 13-20Hz curve did not co-vary with the
freezing learning curve (r2= 0.33; P<0.083). It should be noted that; after the 7-8th conditioning
trial, 4-7.5 and 13-20Hz spontaneous field potential power decreased to the preconditioning
baseline level, while fear behaviour was still elevated. Thus, these potentials only correlate with
freezing behaviour for part of the conditioning. Interestingly the 4-7.5 and 13-20Hz power
learning curves mirrored one another and analysis revealed a robust correlation between them
(r2= 0.62; P< 0.007, Fig. 2.6C). Based on these results, we expect the enhanced cortical 4-7.5 and
13-20Hz rhythms may impact ongoing neural activity patterns during the trace fear interval (for
example; reduction of activity, see Fig. 2.11). While the role of these spontaneous field potentials
is currently speculative, they are not likely to reflect memory recall in this case.
68
Figure 2.6 Tone-evoked ACC potentials predict freezing behaviour (A) Four to 7.5 Hz power recorded during the trace interval shows a significant learning curve, though the effect is not robust and does not completely parallel freezing behaviour (grey line). (B) Thirteen to 20Hz power recorded during the trace interval shows a significant learning curve, though the effect is not robust and does not completely parallel freezing behaviour (grey line). (C) The changes in 4-7.5Hz power was strongly correlated with changes in 13-20Hz power and may be related with the decrease in activity of regular spiking neurons (Fig. 2.11). (D) Example temperature plots of tone evoked potentials during 10 trials of trace fear conditioning. Arrow depicts points of analysis at 50msec (P50) after commencement of tone presentation. (E) Tone-evoked potentials (P50 values) show a significant and robust learning curve throughout conditioning that parallel freezing behaviour (grey line). (F) P50 values and freezing quantities were significantly correlated, suggesting that tone-evoked potentials can predict freezing magnitude during the trace interval, 15sec after the tone-evoked potential occurred.
We previously showed that tone-evoked potentials were augmented by trace fear
conditioning (Fig. 2.5A). To examine whether these potentials demonstrate a learning curve, trial
by trial analysis was conducted. A significant alteration in tone-evoked potentials was detected
during trace fear conditioning (F(9, 80) = 4.27; p <0.001, one-way ANOVA). Post-hoc analysis
revealed that a robust learning curve could be detected for the tone-evoked potentials at the post
stimulus 50msec (P50) time point (all t>3.25, all P<0.015, Fig. 2.6D and E). Correlation analysis
of these learning curves revealed a significant co-variation between fear behaviour and P50
69
values (r2= 0.79; P<0.001, Fig. 2.6F). In contrast to spontaneous field potential power, tone-
evoked potentials remained elevated until the end of the conditioning trials. Collectively, the
results suggest that the association between the shock and tone stimuli gradually produce robust
enhancements of cue-related processing in the ACC, with modest enhancement in 4-7Hz field
potentials power during the trace interval.
2.7.4D Identification of neuron types and properties
The results from the field potential studies implicate the ACC in motor movements, learning the
association between cues and pain, and possibly processing during the trace interval. To examine
the neural correlates of trace fear conditioning, in-vivo freely behaving extracellular spike
recording was conducted (Fig. 2.7A). In addition, to further characterize ACC activity in the
mouse, neuron responses were probed for sensitivity to noxious and innocuous stimuli. Finally,
to parse out sensory neuron responding from motor-related activity, neuron recordings were
conducted in an anesthetized model where motor movements are inhibited (Fig. 2.7B). However,
prior to the analysis of neuron activities in these models, neuron types need to be identified. The
following segment describes the criteria by which neurons were sorted according to pyramidal
Vs non-pyramidal and bursting Vs regular spiking cells.
2.7.4E Identification of putative pyramidal and non-pyramidal neurons
Neurons were sorted (Fig 2.7C.) and identified as putative pyramidal and non-pyramidal,
based on their individual extracellular spike waveforms. A previous report in rats demonstrated
that neurons which have peak to peak values less than 0.5msec, cluster together, and are
considered to be non-pyramidal neurons, while those above this threshold are considered to be
pyramidal neurons (Bartho et al., 2004). Consistent with this idea, peak to peak time provided
70
Figure 2.7 In-vivo spike recording to investigate receptive functions of ACC (A) Multichannel spike recording from the mouse ACC. Manual microdrives can independently position 3 electrodes to improve isolation of single units. The first trace represents a well-isolated neuron. The remaining traces are from two other electrodes in which neurons had not yet been isolated. (B) Multichannel recording in a mouse anesthesia preparation. In this preparation, spike recordings were performed from the ACC simultaneously with the local field potential. Frontal-parietal (F-P) EEG in conjunction with diaphragm (DIA) and electrocardiogram (EKG) were used to determine the health and stability of the preparation. Animals could then be stimulated with peripheral noxious or innocuous stimuli. (C) Multiunit recording from a single electrode (upper trace). Principal component analysis (PCA) was used to sort the multiunit record into individual units (lower three traces). Once sorted, the waveforms can be averaged to yield values of half-width and the peak to peak time. (D) Scatter-plot of peak to peak time and half-width for each individual neuron recorded during wakefulness in freely behaving animals, from sorted multiunit records. Neurons appeared to form clusters above and below 0.5msec peak to peak time. (E) Scatter plot of peak to peak time and half-width for each individual neuron recorded under anesthesia conditions, from sorted multiunit records. Neurons appeared to form clear clusters above and below 0.5msec peak to peak time. However it is interesting to note that a third cluster threshold appears around 0.8msec. Rt/Sm is rectified and smoothed, ch is channel.
71
the most obvious clustering while half-width did not show obvious clustering for both freely
behaving and anesthetized conditions (Fig. 2.7D and E respectively). The division of neuron
clusters appeared around 0.5msec peak to peak time.
To confirm that our “blind” clustering technique was appropriate, we examined whether
clustering could be observed under in-vitro patch-clamp conditions in which the experimenter
was not blinded. In this study, neurons were identified under microscope, based on their
morphology (Fig. 2.8). Neurons were injected with a range of currents to evoke action potentials.
Similar to others (Henze et al., 2000), we found that if the instantaneous slope of the intracellular
signal were plotted relative to time, the intracellular signal could be converted to a waveform that
matches the time course of the extracellular signal (Fig. 2.8A). Pyramidal neurons were clearly
separable from non-pyramidal neurons based on peak to peak times of the converted intracellular
signal (Fig 2.8B). However, the threshold which divided the pyramidal from non-pyramidal
clusters was ~1.6msec which is about 3 times higher than in-vivo conditions (0.5msec) and is
likely attributed to differences in recording conditions (for example 25ºC in-vitro bath
temperature). Since clustering occurred under in-vitro anesthesia and freely behaving conditions,
and is consistent with others (Henze et al., 2000), we adopted the 0.5msec peak to peak time as
an adequate threshold to discriminate pyramidal from non-pyramidal neurons under in-vivo
conditions in the ACC. After neurons had been identified, we found that the relative proportions
of putative pyramidal to non-pyramidal neurons recorded under anesthesia and freely behaving
conditions were similar (Figures 2.10, 2.13 and 2.14), with pyramidal neurons predominating at
78-80% and interneurons composing a mere 20-25%.
72
2.7.4F Identification of bursting and regular spiking cells
Neurons can also be classified based on their degree of bursting behaviour (Connors et
al., 1982; McCormick et al., 1985; Steriade et al., 1993c; Harris et al., 2001; Bartho et al., 2004).
Figure 2.8 Identified pyramidal and interneurons clustering under in-vitro conditions (A) Simultaneous patch clamp and lose patch recording from the same neuron. The middle panel shows superimpositions of the two averaged spike waveforms. Note, the timing of the extracellular waveform peaks do not match that of the intracellular peaks. However, when the instantaneous slope of the intracellular waveform plotted as a function of time, the time course of the intracellular signal matches the loose patch extracellular signal (lower panel). Thus, measurements of half-width and peak to peak time can be calculated from the converted intracellular signals for comparison with in-vivo extracellular data. (B) Scatter plot of peak to peak time and half-width for each individual neuron recorded under in-vitro conditions from visually identified pyramidal and non-pyramidal neurons. All non-pyramidal neurons had peak to peak times that were less than 1.6msec showing that peak to peak time was a very reliable designation for discriminating neurons (data contributed by Dr. XiangYao Li of the Min Zhuo laboratory).
73
Autocorrelation of spike times was used to determine whether neurons exhibit bursting
behaviour (Bartho et al., 2004). Since bursting behaviour may be related to the generation of
oscillations observed during anesthesia (Steriade et al., 1993c) and oscillations were found to
change during trace fear conditioning (Fig. 2.4 and 2.6), we examined whether propensities of
observed bursting and regular spiking neurons would differ across all three in-vivo recording
conditions (anesthesia, general freely-behaving, and trace fear). Examples and data regarding
autocorrelations for bursting, indeterminate, and regular spiking cells for all three recording
conditions are given in figure 2.9B, 2.13A and 2.14A (right hand side). If a neuron is highly
correlated with itself, the activity of the neuron will fall consistently within a time frame of each
spike. If this occurs within 3-6ms, the neuron is classified as a bursting cell. If the neuron
autocorrelation shows a mode greater than the 35ms time point it is classified as regular spiking
(non-bursting). Neurons in which the mode of the autocorrelation fell between 35 and 6ms were
considered indeterminate neurons.
Anesthesia conditions were the most frequent condition in which to record bursting
neurons, with 53.5% of pyramidal bursting neurons compared to general freely behaving
recordings (7.1%, χ2 = 15.36, P <0.001) and trace fear recordings (6.4%, χ2 = 25.68, P <0.001). It
is interesting to note, that while 4-7.5% and 13-30 Hz field potential activity was elevated during
trace fear conditioning, this did not seem to increase the propensity of recorded bursting cells.
It is typical to expect the behaviour of interneurons to be regular spiking; however there
are reports of multipolar bursting interneurons in the cortex (Blatow et al., 2003). Indeed, we
found that 10% putative non-pyramidal neurons recorded under anesthesia displayed bursting
behaviour (Fig. 2.14A). Alternatively, this result could be related to the threshold criteria of the
neuron classifications.
74
Regular spiking pyramidal neurons were much less frequently recorded under anesthesia
conditions (2.8%) than under freely behaving conditions (42.9%, χ2 = 24.39, P <0.001) or during
trace fear experiments (21.3%, χ2 = 8.63, P <0.003). There was a near significant decrease in
regular spiking pyramidal neurons recorded under the trace fear conditions relative to the freely
behaving conditions (χ2 = 3.37, P <0.07). This may be related to an inhibitory action over regular
spiking neurons in the trace fear paradigm (Fig. 2.11, next section). To compare proportions of
non-pyramidal regular spiking neurons, data were pooled from freely behaving and trace fear
experiments (because of low sample size). We found a similar amount of regular spiking non-
pyramidal neurons during anesthesia (20%, 4/20 neurons) when compared to the pooled freely
behaving data (28.5%, 6/21 neurons, χ2 = 0.031, P <0.86).
The results suggest that anesthesia conditions, in which clear oscillations can be seen in
the electroencephalogram and local potentials (Fig. 2.7B), may favor the excitation of pyramidal
bursting cells at the expense of regular spiking pyramidal cells. This may manifest as a change in
the sampling of neuron bursting types without altering the actual ratio of pyramidal to non-
pyramidal neurons.
2.7.4G Unit recordings in the ACC during trace fear conditioning
Lesion of the ACC blocks trace fear conditioning, but not delay fear conditioning (Han et
al., 2003). The ACC has also been implicated in conditioned avoidance and escape responses
(Gabriel et al., 1991a; Gabriel et al., 1991b; Freeman et al., 1996; Kubota et al., 1996; Koyama
et al., 1998; Koyama et al., 2000; Koyama et al., 2001; Johansen & Fields, 2004). However, the
ACC has been demonstrated to have premotor activity with expectation of reward (Shima &
Tanji, 1998; Tsujimoto et al., 2006). Unit recordings of the ACC have never been conducted
during trace fear conditioning. Thus we examined the neural responses of the ACC to the various
75
conditions of the trace fear paradigm and closely examined how neural activity was related to
freezing and movement behaviour. Every neuron was tested in response to 4 tones and 4 shocks,
with each animal being trained for up to 12 trials on day one. In between training trials, animals
were permitted to rest (~1-2 hours), while new cells were manually isolated with the
advancement of electrodes using a microdrive. In addition, 2 animals were trained for 12
additional trials on a second day. Throughout all experiments, neck EMG was monitored for
freezing behaviour.
Figure 2.9 shows an example of a putative pyramidal neuron (2.9A) with indeterminate
bursting properties (2.9B), which has been analyzed in a variety of ways. This particular neuron
increased its activity after shock stimulation (Fig 2.9C); however, the activation of this neuron
also tended to mirror changes in neck muscle movements (Fig. 2.9C, lower panel). Thus we
decided to examine the activity of this neuron more closely. Figure 2.9D shows a single period of
freezing followed by a period of movement in which the neuron’s activity increased,
corresponding with the neck muscle movement (2.9D, lower panel). Figure 2.9E shows grouped
data of 18 such EMG movements in which the spike activity was highly correlated. On average
the neuron ramped its activity, starting 3.2 seconds before the motor movement. The neuron
reached a peak firing rate ~400msec before the peak in EMG activity. A significant correlation
(r2 =0.31, P<0.001) was detected between the EMG amplitudes and the spike rate of this neuron.
Figure 2.10 shows grouped data from all the neurons (represented with colour plots) recorded for
trace fear experiments, analyzed in the same fashion as in figure 2.9. Figure 2.10 shows that the
rate of recording responses did not depend on whether pyramidal neurons or non-pyramidal
neurons were recorded, with 40 of 47 tests (85.1%) yielding responses for pyramidal neurons and
10 of 12 tests (83.3%) for non-pyramidal neurons (χ2 = 0.09, P =0.766). Neurons
76
Figure 2.9 Analysis of ACC neuron during trace fear conditioning (A) Example of spike-sorted putative-pyramidal neuron (B) Auto-correlation demonstrating this neuron is of the indeterminate type.(C) Upper panel shows that this neuron became active prominently after the application of the shock stimuli. Lower panel shows reduction in EMG activity during tone and trace intervals. EMG activity dramatically raises after the shock stimuli, coincident with the neural activation. Thus it appears that this cell not only responds to pain, but may also be involved in motor behavior. (D) Example of single unit activity during freezing behavior and during spontaneous neck muscle activation. (E) Upper 2 panels show trial by trial activations (temperature plot yellow = 11 spikes/bin), and average of this neuron during freezing and spontaneous neck muscle activation. Lower 2 panels show trial by trial activations (temperature plot yellow = 0.39 arbitrary EMG unit) and average of neck EMG. Notice that the neuron develops premotor activation and its peek activity precedes the motor movement by 400msec. (F) Correlation analysis revealed a strong significant relationship between spontaneous motor movement and neuron firing frequency. Green colored bins represent significant responses above baseline (excitatory) based on a Z-score for 99% confidence (grey bar).Yellow colored bins represent significant responses below baseline (inhibitory) based on a Z-score for 99% confidence (grey bar).
77
responded most prominently to: motor movements which terminated freezing behaviour (67.8%),
shock stimuli (62.7%), tone stimuli (47.5%), and finally the trace interval (42.4%). It is easy to
see that there is considerable overlap with neurons that are involved in motor movement and
neurons activated during other phases of the trace fear paradigm (Fig. 2.10, left and right panels).
Out of all 59 neurons studied, only 7 (11.9%) neurons changed their responses to some part of
the trace fear paradigm, without also being related to motor movements. On the converse, a
similar quantity of neurons (9 of 59 =15.3%, χ2 = 0.0723, P =0.79) appeared to be strictly related
to some aspect of trace fear conditioning (tone, trace or shock stimuli).
Examination of inhibitory and excitatory responses shows that regular spiking, pyramidal
neurons were significantly more (6 of 10 = 60%) inhibited during trace fear conditioning as
compared to indeterminate cells (3 of 34 = 8.8%, χ2 = 9.49, P =0.002). Close inspection of these
results shows that this inhibition could occur during the tone, trace or shock intervals. This
finding is consistent with the slowing of the ACC field potentials during the tone and trace
intervals, and the trend toward reduced propensity to record regular spiking pyramidal neurons
during conditioning. We next examined whether pyramidal neurons were generally excited or
generally inhibited compared to non-pyramidal neurons during trace fear conditioning. To
explore this further, the average of each neuron for each conditioning phase was computed
(baseline, tone, trace interval and shock (Figure 2.11). Bursting neurons were excluded from the
analysis to keep the comparison symmetrical. A significant main effect of neuron type (F(1,54) =
4.67; p <0.035 , 2-way repeated measures ANOVA) and conditioning phase was found (F(3,162) =
4.67; p <0.035, pyramidal neurons. The only other interesting observation was that pyramidal
neurons during the trace interval were significantly inhibited compared to baseline (t=2.89,
P=0.026, Bonferroni t-test).
78
Figure 2.10 ACC neurons respond to trace fear conditioning and demonstrate premotor activity The left hand column shows all recorded neurons, their responses to particular phases of the trace fear paradigm, and motor-related activations. Excitatory responses are represented by green boxes while inhibitory responses by yellow. If both excitatory and inhibitory responses happened during a particular phase the box is a yellow-green intermediate. It is easy to see that the vast majority of recorded neurons demonstrate premotor activity. In addition, regular spiking neurons are often inhibited during trace conditioning. Figures to the right show example responses to the different phases of the trace fear paradigm and EMG triggered activations. Each different phase occurs at the time points between the dotted lines. BL is baseline.
79
Taken together, it appears that there was a net depression of pyramidal neuron firing during the
trace fear interval with the majority of the effect likely being carried by fast regular spiking
neurons.
Figure 2.11 Pyramidal neurons are inhibited during the trace interval Putative pyramidal neurons are inhibited selectively during the trace fear interval and tend to have lower firing rates then that of putative non-pyramidal neurons. * indicates statistical significance P<0.05 compared to pyramidal neuron baseline (BL).
80
The propensity of neurons to commence firing before or after motor movements was compared.
To identify motor-related cells, the magnitude of the EMG activity for each time point, was
correlated to the neurons firing rate (as in Fig 2.9E and F). It should be mentioned that not all
neurons which had motor activations as judged by Z-score analysis, had significant correlations
with motor movement. Significant correlations were considered motor-related neurons. Out of
these neurons, a greater amount had premotor behaviour (41of 43 = 95.3%) compared to neurons
which became active during the motor movement itself (2 of 43 = 4.7%, χ2 = 24.0, P <0.001).
The data strongly suggest that the ACC has a prominent premotor function in the mouse.
2.7.4H Motor learning in the ACC during trace fear conditioning
Since the largest proportion of neurons recorded corresponded with premotor activity, we
examined whether neurons gradually develop this activity (Fig 2.12A and B) and whether they
alter their tuning-curves with conditioning (Fig 2.12C and D). Figure 2.12B shows the
percentage of neurons recorded that did and did not develop premotor activity. The proportions
of learning to non-learning neurons were similar (χ2 = 2.67, P =0.263). Neurons may also change
their tuning to motor behaviour. Figure 2.12D shows the percentage of neurons recorded that did
and did not change their tuning to motor movements. The proportion of tuning to non-tuning
neurons were similar (χ2 = 0.016, P =0.992). Collectively the results suggest that a variety of
neuron types in the ACC are capable of developing premotor activity and changing their tuning
to motor behaviour throughout the course of conditioning.
Based on the results from field potential recordings and current observations, the ACC
appears to be strongly related to mediating and modifying premotor output, learning aversive
81
Figure 2.12 Motor plasticity in the ACC during trace fear conditioning. (A) Heat plot (2.0 spikes/bin =red) for an example neuron which developed premotor- related activity across 4 conditioning trials (for the second block of conditioning, trials 4-8).Twenty-four sweeps of data were taken during periods of spontaneous freezing behavior, irrespective of training trial. (B) Percentage of all neurons which developed locking to EMG activity. (C) Heat plot (7.0 spikes/bin =red) for an example neuron which changed its tuning to EMG activity across 4 conditioning trials (for the first block of conditioning). Twenty-four sweeps of data were taken during periods of spontaneous freezing behavior, irrespective of training trial. (D) Percentage of all neurons which changed their tuning to EMG activity.
cue-related information, and responding to pain stimuli. Based on the enhancement of 4-7.5 Hz
rhythms, during the trace interval, and the propensity for reduced firing of regular spiking
neurons during the trace interval, it appears that the ACC is an unlikely candidate to maintain
memory across this interval. However, the current analysis did not permit us to clearly
discriminate neurons involved in pain from neurons involved in motor movement. If ACC
neurons are related to the affective memory of pain, they may drive motor behaviour even when
no pain stimulus is present. The remaining studies and analyses are aimed at teasing apart
82
whether neurons in the ACC can respond to pain stimuli without necessarily having premotor
activity.
2.7.4I Functional characterization of mouse ACC neurons to direct peripheral stimulation in
freely behaving animals
We characterized the response of ACC neurons to both noxious and innocuous peripheral
stimuli. Figure 2.13 shows a summary of all the neuron types recorded under freely behaving
conditions (2.13A, left), with select examples for responsive neurons (2.13A, right). Responses
were detected for noxious pinching of the tail, pinching right and left ear and for innocuous
approaches of a probe, brushing and air puffs to the fur. Pinch stimuli were applied with
tweezers for up to 5 seconds. Fur was brushed on the back of the animal with a probe, in a
direction opposite to the hairs’ growth, for up to 5 seconds. Air puffs were applied via a
pressurized canister. Figure 2.13B shows raw recording traces from different neurons to various
peripheral stimuli. The propensity of recoding responses to peripheral stimulation did not depend
on whether pyramidal or non-pyramidal neurons were recorded, with 31 of 74 tests (41.8%)
yielding responses for pyramidal neurons and 11 of 22 tests (47.6%) for non-pyramidal neurons
(χ2 = 2.47, P <0.12). In addition, it is of interest to note that the propensity to record neurons with
excitatory responses was higher (38 of 95 tests, 40.0%) when compared to inhibitory responses
(3 of 95 tests, 3.2%, χ2 = 35.95, P <0.001). Analysis was conducted to examine whether the ACC
responded differentially to noxious pain Vs innocuous stimuli. The propensity to record
responses did not depend on whether noxious or innocuous stimulation were applied, with 9 of
33 tests (27.3%) yielding responses to noxious stimulation and 30 of 62 tests (48%) for
innocuous stimulation (χ2 = 3.143, P =0.076). Thus, the ACC of the mouse does not seem to have
a preference for responding to noxious nociceptive stimuli. Since neuron responses in freely
83
84
Figure 2.13 Noxious and innocuous responding neurons under freely behaving conditions (A) The left hand column represents the number of neurons recorded and responding (either excitatory or inhibitory) to particular probe stimuli. The figures in the middle column show examples of neural responses to probe stimuli. Green coloured bins represent significant responses above baseline (excitatory) based on a Z-score for 99% confidence (grey rectangle).Yellow colored bins represent significant responses below baseline (inhibitory) based on a Z-score for 99% confidence. Application of stimulus occurs at the time points between the dotted lines (Stim). The right-most columns depict autocorrelations from each of the example neurons. The different colors in the graphs represent the different ranges to classify each type of neuron as bursting (white), indeterminate (yellow) and regular spiking (green). (B) The first trace shows a unit recording from the ACC in which the cell responded to the approach of probe stimuli. In this case the neuron responded after termination of the approach. The same neuron was found to respond to a pinch of the tail (trace two). In this case however, the response lasted the duration of the pinch extending well after the pinch. Trace three shows the response to a brush of the fur. In this case the neuron responded only during the brush stimuli. The fourth trace shows a neuron that was inhibited following a puff of air to the face of the animal. The neuron reduced its activity during the stimulation extending to after the puff had terminated. The final trace shows that when EMG activity was recorded it could be seen to correlate with spike activity. Appro, approach of probe, BL, baseline, LEP, left ear pinch, REP, right ear pinch, PT, pinch tail.
behaving conditions are complicated by motor movement, we next examined responses under
anesthesia.
2.7.4J Functional characterization of mouse ACC neurons to direct peripheral stimulation under
anesthesia
Since nearly every means of peripheral stimulation will result in some form of motor
movement, we also used an anesthesia model in which we could monitor responses to peripheral
stimulation while precluding motor output. A variety of anesthesia experiments have investigated
ACC neurons in rats and rabbits in a similar manner (Sikes & Vogt, 1992; Yamamura et al.,
1996; Hsu & Shyu, 1997; Hsu et al., 2000; Gao et al., 2006; Yang et al., 2006).
Figure 2.14 shows a summary of all the neuron types recorded in the anesthesia
experiments (2.14A, left) with select examples of responsive neurons (2.14B, right). Responses
were detected for pinching of the tail, right and left foot, right and left ear and to innocuous
brushing of the fur. Similar to freely behaving conditions the propensity of recording responses
to peripheral stimulation did not depend on whether pyramidal or non-pyramidal neurons were
recorded, with 52 of 167 tests (31.1%) yielding responses for pyramidal neurons to peripheral
85
stimulation and 29 of 58 tests (50%) for non-pyramidal neurons (χ2 = 2.47, P =0.12). In addition,
it is of interest to note that the propensity to record responding neurons was higher for excitatory
responses (67 of 206 tests, 32.5%) as compared to inhibitory responses (14 of 206 tests, 6.8% χ2
= 26.13, P <0.001).
Analysis was next conducted to examine whether the ACC responded differentially to
noxious pain Vs innocuous brush stimuli. Similar to freely behaving conditions, the rate of
recording responses did not depend on whether noxious or innocuous stimulation were applied,
with 66 of 183 tests (36.1%) yielding responses to noxious pinch stimulation and 14 of 42 tests
(33.3%) for brush stimulation (χ2 = 0.024, P =0.88). These values are in a similar range to that
observed in freely behaving conditions for noxious and innocuous stimuli (27% and 48%
respectively). Finally, the propensity to record neural responses to peripheral stimulation under
anesthesia (81 of 225 tests = 36%) and that of freely behaving conditions (39 of 95 testes = 41%,
χ2 = 0.530, P <0.46) did not yield any difference. This result suggests that the anesthesia
conditions did not dramatically reduce the degree to which units responded to peripheral
stimulation, indicating that the ACC can process sensory input without the necessity of premotor
response. Collectively, these findings confirm that the ACC responds to noxious and innocuous
sensory input in the mouse, consistent with findings in other species (Sikes & Vogt, 1992;
Yamamura et al., 1996; Hsu & Shyu, 1997; Hsu et al., 2000; Gao et al., 2006; Yang et al., 2006).
Moreover, it suggests that the ACC is important as an integrator of sensory/affective input with
premotor output.
86
Figure 2.14 Neurons responding to noxious and innocuous stimuli under anesthesia (A) The left hand column represents the number of neurons recorded and responding (either excitatory or inhibitory) to particular probe stimuli. The figures in the middle column show examples of neural responses to probe stimuli. Green colored bins represent significant responses above baseline (excitatory) based on a Z-score for 99% confidence. Yellow colored bins represent significant responses below baseline (inhibitory) based on a Z-
87
score for 99% confidence. Application of stimulus occurs at the time points between the dotted lines. The right-most columns depict autocorrelations from each of the example neurons. The different colors in the graphs represent the different ranges to classify each type of neuron as bursting (white), indeterminate (yellow) and regular spiking (green). (B) Electrode trajectory and recordings from different depths from the cortex within the ACC. Anesthesia was used in these experiments to reduce motor movements. In these experiments, electrodes were lowered and neurons probed with peripheral stimulation to the body. In this example a pinch stimulus (horizontal bar above each trace) was applied to the tail. While probe stimuli seem to have minor effects for the second and third trace, a robust excitation can be seen in the 5th trace. BL, baseline, CG cingulate, M, motor, LEP is left ear pinch, REP is right ear pinch, Appro is approach of probe, PT is pinch tail.
2.7.4K Direct stimulation of the ACC
Stimulation of the ACC has been known to induce vocalizations (Jurgens & Pratt, 1979;
Jurgens, 2009), fear behaviour (Johansen & Fields, 2004; Tang et al., 2005), and motor
al., 1976; Sinnamon & Galer, 1984). In addition, it has been reported that cingulate seizures in
humans are characterized by paroxysmal motor attacks during sleep, vocalizations, arousals from
sleep and nocturnal paroxysmal dystonia (Schindler et al., 2001; Vetrugno et al., 2005). Eliciting
ictal discharges in a brain region is a useful method to uncover the function of that region. For
example, blockade of GABA type-A receptors in the motor cortex has been found to elicit ictal
activity coinciding with electromyography activity, even under anesthesia (Castro-Alamancos &
Borrell, 1995). Recently it was shown that the potassium channel blocker, 4-amino pyridine (4-
AP), produces ictal activity in an ACC slice preparation (Panuccio et al., 2009). To examine
whether the activation of the ACC would induce emotion-related motor movements, we focally
applied 4-AP (via microdialysis) to the ACC while recording EEG to measure ictal activity, in
freely behaving animals.
In 2 animals microdialysis probes were implanted into the midline ACC. A frontal cortex
screw electrode was placed near the midline above the ACC. An additional reference electrode
was placed over the parietal cortex for differential recording. Thirty minutes after probe
implantation and ACSF perfusion (1-3ul/min), mice showed normal sleep and wakefulness EEG
88
oscillations (Fig. 2.15A). The perfusate was washed for an additional 1-2 hours before
experimental manipulations. When the perfusate was switched to 1-5mM 4-AP, ictal oscillations
appeared in the EEG. Ictal EEG events corresponded closely with EMG movements of the neck
(Fig. 2.15A). Mice were seen to run in circles (Fig. 2.15A, middle panel), vocalize, jump and had
enhanced startle response to sound (clapping of the hands). Mice were also seen to exhibit
behaviour which resembled small, fast forward motions followed by a backward motion of the
same magnitude. Post-ictal behaviour resembled that of freezing, with a sustained absence of
movement. Interestingly, audible vocalizations were often accompanied by changes in high
frequency EEG (Fig. 2.15, right most panel).
To examine the evolution of the seizure activity from the ACC, local field potential and
unit recording were carried out in combination with frontal-parietal recordings under anesthesia
(n=2). Application of 1mM 4-AP to the ACC changed multiunit recordings and local field
potentials along a similar time frame (Fig. 2.15B). Ictal activity was seen in the local field
potentials, but could also invade the more filtered spike recordings (filtered between 300Hz and
20kH). It appeared that the ictal activity was a summation of multiunit activity, since it became
difficult to observe the multiunit activity during ictal spikes (Fig. 2.15B, left to right panel). Ictal
activity was seen to gradually spread from the ACC to the surface EEG recording during the
perfusion. An additional caveat of our findings is that the 4-AP was able to activate brain cells
even in the down state of the frontal parietal oscillation (Fig 2.15B, right-most panel). To our
knowledge there is only one other method to block the down states and that is with intracellular
injection of caesium to block potassium channels (Metherate & Ashe, 1993). Our anesthesia
experiments demonstrate that 4-AP facilitates population spikes. The output of these spikes
appear to produce emotional motor behaviours as seen by the effects of 4-AP in the ACC in
freely behaving animals.
89
Figure 2.15. Ictal activity in the ACC produces circling behavior, vocalizing and synchronization of multiunit activity (A) Left panel: simultaneous recording of EMG and EEG during ACSF perfusion into the ACC. Lower trace is expanded view in which no correlation can be seen between EEG and motor movement. Middle panel: application of 4AP to the ACC generates rhythmic ictal events which are synchronized with movements of the neck. Right panel: mouse vocalizes and EEG changes to a higher frequency. (B) Left panel: simultaneous recording of ACC spikes, local field potentials and frontal parietal (F-P) EEG during ACSF perfusion into the ACC. Middle panel: application of 4AP to the ACC generates rhythmic ictal events in the local field potential which invades multiunit recording trace. Arrows indicate locations of individual spikes and arrowheads indicate areas where many spikes appear to be summated to produce a population spike. Right panel: After abatement of the ictal events neurons appear to fire during the down state of the surface EEG (arrows).
90
2.7.4L Histological verification of electrode sites
Nearly all experiments involved histological verification of electrode cites (Fig. 2.16).
The only experiments in which electrode cites were not verified were the anesthesia experiments
in which the depth of the electrodes or microdialysis probes could be estimated accurately with
the micromanipulator. In other experiments (freely behaving spike, and trace fear, and local field
potential), DC current (100-500uA) was passed between the ground electrode and the recording
electrodes to make a lesion (Fig. 2.16C). Finally, in freely behaving microdialysis experiments, a
sizeable lesion site was made by the microdialysis probe and was used as an estimate of the
electrode and probe locations. Nissl staining was performed for microdialysis experiments, freely
behaving, and trace fear unit recording studies. The majority of electrodes were on target within
the ACC. However, one of the electrode sets in the trace fear experiments was slightly near the
border of M2 and CG1. Moreover, during surgery, electrodes were prepositioned ~500um below
the cortical surface to avoid hitting the motor cortices, so this particular electrode set was an
unlikely cause for concern.
91
Figure 2.16. Verification of electrode positions (A) Locations of the field potential electrodes. Each line represents the deepest location of one tip of the bipolar electrode (tips had a 0.5mm offset). (B) Locations of spike electrodes used in trace fear experiments. Each line represents 2-4 electrodes used in each of four mice. (C) The first panel shows Nissl stain and the effect of passing current across 4 electrodes in the ACC (4 round lesions clustered together). The remaining two panels show the locations of the electrode tracts from 4 separate mice (each line represents 2-4 electrodes). (D) For the anesthesia preparation, the tract was estimated from the angle of brain penetration and the depth to which the electrode was moved. There were two major hot spots for recording, which included the middle and superficial layers. (E) Location of the region where neurons were visually selected for in-vitro slice recording. Most recorded neurons were situated in layer II-III while a select few were in layer I. (F) The first panel shows Nissl stain and the lesion site left by the microdialysis probe (arrows). To the right is the identified location of 2 microdialysis probes for freely behaving experiments.
92
2.7.5 Discussion
The current study examined the input and output relationship of the ACC using the trace
fear conditioning paradigm, peripheral stimulation, pharmacology and an anesthesia model. We
show that tone-evoked potentials, rather than spontaneous field potentials in the ACC, are the
best predictor of freezing behaviour during the trace fear interval. This establishes the ACC as a
major player in making associations between tone and shock stimuli. The recordings of tone-
evoked neuron activity are consistent with this observation.
In contrast to what we expected, putative pyramidal neurons were largely inhibited during
the trace interval, suggesting that the ACC is not involved in bridging the temporal gap between
the shock and the tone. Interestingly, the majority of recorded neurons demonstrated premotor
activity that co-varied with the magnitude of the motor movement. Many of these neurons also
responded to shock stimuli. To tease apart whether the ACC can be active independent of motor
output, neuron proportions were compared between freely behaving conditions and those of
anesthesia. Similar proportions of neurons were found to respond to noxious and innocuous
stimuli under these conditions, even when the animal was not moving during anesthesia. The
results indicate that the ACC can receive input without necessarily being linked to motor output.
Finally, direct stimulation of the ACC, by blocking 4-AP sensitive potassium, produces an array
of emotionally-related behaviour. Based on these findings we suggest that the ACC is one of the
major affective-motor systems of the brain.
2.7.5A Characterization of field potentials in the ACC during trace fear conditioning
Changes in spontaneous field potentials in the 4-7.5 Hz range are know to occur in the
frontal cortex during the expectation of reward stimuli (Tsujimoto et al., 2006; Paz et al., 2008)
and avoidance conditioning (Stark et al., 2007; 2008). Our experiments show that local field
93
potential recordings in the 4-7.5 and 13-20Hz range were increased during trace fear
conditioning and paralleled one another. However this activity did not completely parallel the
learning curve and was not robust. Indeed the slowing of this activity may actually be related to
the overall decrease in pyramidal neuron spike frequency during the trace interval (Fig. 2.11). It
is of interest that 4-7.5Hz in the amygdala and hippocampus has been linked with freezing
behaviour itself (Pare & Collins, 2000; Seidenbecher et al., 2003; Narayanan et al., 2007) and
therefore may represent communication between a numbers of structures involved in fear
behaviour. However, there is no evidence that the ACC receives direct activation from the
hippocampus, which might drive these 4.7.5Hz rhythms. By contrast the prelimbic cortex
receives a direct and converging input from CA1 hippocampus and amygdala (Thierry et al.,
2000; Gabbott et al., 2002; Ishikawa & Nakamura, 2003; Tierney et al., 2004). Indeed there is
evidence that the hippocampus and the prelimbic cortex are involved in bridging the temporal
gap of the trace interval (Pare & Collins, 2000; Baeg et al., 2001; Gilmartin & McEchron,
2005b). So it might be expected that the prelimbic cortex would demonstrate non-decrementing
4-7.5 rhythms which correlate closely with freezing behaviour.
Tone-evoked potentials have been reported to become enhanced with fear conditioning in
the prelimbic cortex (Mears et al., 2009). These results suggest that there is a reduction of
inhibitory gating to the prelimbic cortex. Similarly, visually-evoked signals are enhanced in
aversive conditioning in the human ACC (Buchel et al., 1998). Consistently we find that the
tone-evoked potentials are robustly enhanced with increased conditioning, and also parallel the
behavioural learning curve remarkably well. These potentials are not locked to motor behaviours
(Fig. 2.5A), suggesting that they are related largely to sensory or affective processing. Moreover,
it was commonly observed that neurons were activated with the presentation of the tone stimulus
(Neuron 2 and 10, Fig. 2.10). Thus the ACC seems to be important for the ability of an animal to
94
remain receptive to biologically relevant information (e.g. tone) which signals the future
occurrence of an aversive stimulus (e.g. shock). Whether this functions to focus attention to
impending danger, or retrieve memory of previous aversive events, remains to be discovered.
2.7.5B Unit recordings in the ACC during trace fear conditioning
Lesions of the ACC blocks trace fear conditioning, but not delay fear conditioning (Han
et al., 2003). However, the ACC was shown to be equally active in delay fear conditioning as it
was in trace fear conditioning in humans, and deactivations of the ACC were seen with
conditioning (Knight et al., 2004). Our results show that the activity of ACC neurons is reduced
during the trace interval. These de-activations tended to occur in pyramidal cells (Fig. 2.11) with
the highest percentage occurring in regular spiking (Fig. 2.10) neurons. However, when motor
movements were observed during the tone or trace interval, they tended to be robust and
correlated with neuron firing frequency. Thus, it is possible that reducing the activity of ACC
neurons during periods of freezing may permit these neurons to rebound into a firing mode
associated with motor movements. This would account for why many neurons demonstrate trace
interval activations (usually with motor movements, Fig. 2.10) while the population average
(Fig. 2.11) shows a net decrease in pyramidal activity during these periods. This finding fits
nicely with previous observations that regular spiking neurons of the prelimbic cortex are largely
inhibited during trace fear conditioning (Baeg et al., 2001).
While we found a decrease in ACC neuron activity during the trace interval, other studies
show that c-fos is elevated selectively in the ACC by trace fear (Han et al., 2003). In addition, a
recent report showed that blocking or lesioning the ACC prevents cued-delay fear acquisition
(Bissiere et al., 2008). The major difference between those studies and the current one is that unit
and field potential recording have excellent temporal resolution; while lesions, pharmacology
95
and immediate early gene studies do not. It should be emphasized that the ACC tone-evoked
potentials were augmented in a learning-dependent fashion. This suggests that the ACC is
permitting information flow though its network. Thus, previous studies may have blocked tone-
evoked potentials or measured increases in c-fos related to these potentials. Experiments need to
be conducted to selectively turn off neuron activity during tone-evoked potentials, to examine
whether these potentials have any bearing over behaviour, or rather, are epiphenomena of an
already sensitized network.
The seemingly multimodal behaviour of neurons, overlapping with motor movements and
pain, makes sense if an animal is to avoid pain in the future through motor behaviour, and fits
nicely with recent predictions for the overlap of pain and movement systems (Dum et al., 2009).
However, it is surprising that ACC neuron activity would correlate with motor behaviour even
when the animal is not experiencing pain. One explanation for this is that the ACC retains
information important for escape from painful situations. In the case of chronic pain and visceral
hypersensitivity, where the pain or discomfort is inescapable, the ACC is known to become
sensitized (Gao et al., 2006; Zhao et al., 2006; Toyoda et al., 2009) and spontaneous firing rates
of ACC are reported to have doubled (Gao et al., 2006). Consistently, we found that a significant
number of neurons developed premotor activity and changed their tuning throughout
conditioning, implicating their sensitivity to the conditioning paradigm. Moreover, if the ACC is
involved in an affective (Devinsky et al., 1995; Rainville et al., 1997) or a motivational
component to escape aversive stimuli (Gabriel et al., 1991a; Gabriel et al., 1991b; Freeman et
al., 1996; Kubota et al., 1996; Koyama et al., 1998; Takenouchi et al., 1999; Koyama et al.,
2000; Koyama et al., 2001; Johansen & Fields, 2004; Ding et al., 2008), premotor activation may
reflect these processes, rather than direct motor activations per-se.
96
In the current study it was found that 4-AP induced vocalizations and other motor
behaviours, was consistent with studies of others (Penfield & Welch, 1951; Showers, 1959;
electrically-evoked vocalizations in the ACC can be blocked with glutamatergic antagonism in
the periaqueductal gray (Kyuhou & Gemba, 1999), which is also a site involved in the expression
of fear behaviour (De Oca et al., 1998; Mobbs et al., 2007). Thus, the ACC may be situated a
step or two back from the actual motor execution. In such a case the ACC could be involved in
integrating affect and motor execution and would have the capacity of plastic modification and
tuning.
2.7.5C Functional characterization of mouse ACC neurons to direct peripheral stimulation
The ACC is well characterized in response peripheral noxious and innocuous stimulation,
in humans, monkeys, rabbits, and rats (Sikes & Vogt, 1992; Yamamura et al., 1996; Hsu &
Shyu, 1997; Hutchison et al., 1999; Hsu et al., 2000; Gao et al., 2006; Yang et al., 2006).
However, a similar characterization has never been conducted in mice. We repeated earlier
studies of peripheral noxious and innocuous stimulation. We found that a significant proportion
of neurons were activated in the ACC in response to shock (62.7%), which is comparable to that
recorded with laser-heat stimulation (51%) (Kuo & Yen, 2005). Noxious pinch stimulation to the
body activated 27.3% of neurons in freely behaving animals and 36.1% of neurons under
anesthesia, which is comparable to the 39% of noxious responsive neurons in the ACC of an
anesthetized rat preparation reported previously (Yamamura et al., 1996). Since the number of
responsive neurons recorded under anesthesia is comparable to that of freely behaving
conditions, the premotor activity of the ACC is unlikely to be necessary for motor acts; rather
this activity may be involved in modulating motor output.
97
2.7.5D Neuron identification
One of the major weaknesses of extracellular recording is the ability to identify
interneurons and pyramidal neurons. We tried to circumvent this problem by using a pre-
established grouping/clustering method (Bartho et al., 2004) and attempted to support our results
by using the same method in-vitro (Fig. 2.8). In this analysis it was found that peak to peak
thresholds were much larger in-vitro (1.6msec) as compared to in-vivo conditions (0.5msec).
However all interneurons showed lower peak to peak times as compared to pyramidal neurons
in-vitro. The discrepancy between in-vivo and in-vitro conditions may be related to the
conversion of the intracellular waveform to the extracellular wave-form, or to the lower
temperature of the in-vitro bathing solution. In all recordings conducted, for each set of
experiments, we achieved the same relative number or pyramidal Vs non-pyramidal neurons
suggesting that our sampling of each experiment had reasonable reproducibility. We used burst
analysis (Bartho et al., 2004) to further classify neurons. It is of interest to note that the
propensity to record bursting pyramidal neurons tended to be higher under anesthesia conditions
(53.5%). Indeed, only 2.8% of the recorded pyramidal neurons were regular spiking. This data is
reasonably consistent with a recent in-vitro study from our laboratory in which the lowest
recorded neuron number was from regular spiking neurons (24%) followed by bursting (31%)
and indeterminate (or intermediate) neurons (45%) (Cao et al., 2009). Moreover, McCormick et
al. (1985) demonstrated that bursting behaviour in the cingulate cortex neurons is likely to be
removed if neurons are depolarized sufficiently toward threshold prior to being activated. This
observation is the simplest explanation for the increased sampling of burst-related cells seen
under anesthesia conditions. This would also explain why many bursting cells respond to
98
noxious stimulation under anesthesia, while many regular or indeterminate spiking cells respond
to noxious stimulation in freely behaving conditions.
The ACC is typically referred to either as a motor area (CMA) or as an affective area
(ACC). However, our data suggests that the ACC integrates affective information with premotor
commands in the mouse brain. Premotor commands may be related to vocalization, directing
escape, or safety behaviour. Moreover, the activity of the ACC seems to be tightly linked with
cue information indicating the emotional significance of potentially dangerous situations.
99
CHAPTER 3
SLEEP, SYNAPTIC PLASTICITY AND TRACE FEAR
100
3.1 Overview
The previous chapter focused on how trace fear conditioning influences brain waves and
unit activity in the ACC. The following chapter investigates how fear memory may be related to
genetics and sleep. As such, the mechanisms of sleep and the molecular biology of memory
formation are reviewed. The chapter finishes with a study of the influence of learning and
memory enhancement on sleep.
3.2 Description of sleep
Sleep is defined as a state of immobility with reduced responsiveness to the environment.
However, sleep can be reversed, which distinguishes it from anesthesia and coma (Siegel, 2005).
Another defining characteristic is that the removal of sleep results in the body attempting to
recover sleep (Siegel, 2005). The most common way to measure sleep is through use of cortical
electroencephalogram or electrocorticogram in conjunction with muscle recordings from the
body. Sleep is largely classified into two additional states. The first state is refered to as non-
rapid eye-movement (NREM) sleep or slow wave sleep (SWS), while the other state is referred
to as rapid eye-movement (REM) (Figure 3.1) (Aserinsky & Kleitman, 1953). During
wakefulness, the electrocorticogram, recorded from the surface of the cortex, is composed of
low-amplitude fast-frequency signals (30-60Hz) (Maloney et al., 1997). Electromyography of the
neck also shows that the muscles are generating tone and are involved in motor movements of
the head. By contrast, NREM sleep is characterized by high voltage and predominant slow wave
potentials (0.5-4 Hz), sleep spindles (7-13Hz) and behavioural quiescence, evidenced by a
reduction in postural muscle tone. Rapid eye-movement (REM) sleep is characterized by EEG
very much similar to that of wakefulness; however muscle tone is completely abolished with the
exception of occasional muscle twitches and rapid movements of the eyes. Considerable research
101
literature exists on connecting REM sleep with learning and memory for which several reviews
exist (Smith, 1996; Hobson et al., 1998), however an emerging literature also suggests that slow
wave sleep, learning and memory are connected (Benington & Frank, 2003; Tononi & Cirelli,
2006); The review presented here will focus on systems involved in controlling slow wave sleep
and wakefulness and how these may connect to learning and memory.
Figure 3.1 Defining characteristics of sleep Upper trace: shows EMG activity across each sleep and wakefulness state. EMG gradually declines from one state to the next. Finally, during REM sleep postural muscles develop atonia, with the exception of the occasional muscle twitches. Lower trace: is the frontal-parietal EEG changes that occur across sleep and wakefulness states. The EEG activity during wakefulness is high frequency low voltage. During NREM sleep the voltage changes to high voltage low frequency. Finally, during REM sleep the EEG appears similar to wakefulness.
3.3 Circuits
The following section details the primary circuits involved in the regulation of
wakefulness and slow wave sleep behaviour states. These circuits include a vast array of
ascending neurotransmitter systems impinging on the thalamus and the cortex to maintain
wakefulness. The decrease in the activity of these circuits accompanies a progression into sleep.
It should be mentioned that a variety of descending mechanisms are also involved in the
102
reduction in muscle tone during sleep and are reviewed elsewhere (Chase & Morales, 1990;
Rekling et al., 2000).
3.3.1. Wakefulness promoting systems
There is a variety of arousal systems which project to the thalamus and the cortex. Each
system distributes a specific neurotransmitter and their output varies as a function of the sleep
and wakefulness states. Systems involved in wakefulness will be discussed here, followed by a
description of systems involved in sleep promotion in the next section (3.2). Figure 3.2 is a
depiction of all the circuits, and some of the important neurotransmitters mechanisms discussed
in this and later sections.
A region involved in producing arousal was first discovered through a variety of
responsible for generating arousal were concentrated in the oral pontine and mesencephalic brain
stem reticular formation. They produce arousal through a dorsal thalamic relay to the cortex and
a ventral relay coursing through the hypothalamus and basal forebrain (Siegel, 2004; Jones,
2005).
The dorsal and ventral pathways originate from two cholinergic synthesizing nuclei
known as the lateral dorsal tegmentum (LDT) and the pedunculopontine tegmentum (PPT)
(Jones & Beaudet, 1987). Neurons of the LDT and PPT discharge during waking, decrease their
discharge during NREM sleep and increase their discharge during REM sleep (el Mansari et al.,
1989; Steriade et al., 1990). Thus, the activity of the cholinergic system appears to parallel the
103
Figure 3.2 Sleep circuitry Basic circuits involved in both the generation of wakefulness and NREM sleep. The bottom left corner shows key neurotransmitter systems composed of cells which turn on either during NREM sleep (Sleep ON) or wakefulness (Arousal). During wakefulness many of these regions project to the cortex (top left) releasing neurotransmitters which can depolarize or modulate neural activity. In the thalamus (middle circle) relay neurons are activated by these neurotransmitter systems, this has the effect of inactivating an Ih current and blocking a K-leak current. The net effect is that the thalamus is kept in a tonic mode; whereby information from the periphery can be relayed to the cortex. During NREM sleep, neurons of the inhibitory basal forebrain and ventrolateral preoptic area (VLPO) (bottom left) become active and inhibit a variety of wake-on circuits. This reduces neurotransmitter release to the cortex and the thalamus. Thus the thalamus is disfacilitated and begins to oscillate in a burst more, in parallel with the cortex. The interaction between Ih and IT channels produces burst oscillations in the thalamus (see text for more detail). In addition feed-back inhibition from the reticular nucleus of the thalamus is involved in these thalamic oscillations, producing sleep spindles and delta waves. (figure by H. Steenland, refs in text).
104
changes in global EEG activity. The dorsal pathway activates midline and intralaminar nuclei of
the thalamus (Woolf & Butcher, 1986; Hallanger et al., 1987; Steriade et al., 1990) which in turn
projects diffusely to the cortex (Minciacchi et al., 1986). In the thalamus, acetylcholine activates
nicotinic and muscarinic receptors to enhance cortical activation (Curro Dossi et al., 1991;
McCormick, 1992b; Steriade et al., 1993a). Consistently, stimulation of the LDT or PPT
cholinergic system produces cortical arousal via activation of muscarinic and nicotinic receptors
in the thalamus (Curro Dossi et al., 1991; Steriade et al., 1993a).
The ventral pathway consists of cholinergic neurons which project to the cholinergic
basal forebrain and the histaminergic neurons of the hypothalamus, which in turn project to the
cortex, producing arousal (Siegel, 2004; Jones, 2005). The basal forebrain consists of the
substantia-innominata nucleus basalis complex, diagonal band of Broca and the medial septal
nuclei which are cholinergic containing. Neurons of the substantia-innominata nucleus basalis
complex project prominently to the cortex (Rye et al., 1984). Like the cholinergic neurons of the
brainstem, basal forebrain cholinergic neurons are active during wakefulness, decrease their
activity during NREM sleep and increase their activity during REM sleep (Lee et al., 2005a).
Interestingly, the discharge of these neurons is correlated with gamma (30-60Hz) and theta brain
frequencies (4-8 Hz), while correlating negatively with 1-4Hz frequencies (Lee et al., 2005a).
The locus coeruleus is a pontine structure which forms a diffuse projection system
throughout the forebrain and the thalamus (Lindvall et al., 1974; Jones et al., 1977; Jones &
Moore, 1977; Kromer & Moore, 1980). The neurons of this structure demonstrate state-
dependent activity, with the highest firing rates during wakefulness, with progressively lower
activity during NREM sleep and cessation of firing during REM sleep (Aston-Jones & Bloom,
1981). Pharmacological activation of the locus coeruleus in an anesthetized rat preparation
resulted in pronounced shifts from low frequency high voltage EEG to that of high frequency
105
low voltage EEG, characteristic of wakefulness (Berridge & Foote, 1991). Similarly, electrical
stimulation of the locus coeruleus in cats also produces cortical arousal (Steriade et al., 1993a).
Dorsal and median Raphé nuclei project diffusely to the forebrain and to the thalamus
(Cropper et al., 1984; Vertes, 1991; Vertes et al., 1999). Serotonergic neurons of the dorsal
Raphé nucleus are most active during wakefulness, progressively decrease their discharge during
NREM sleep and further inactivate during REM sleep (Jacobs & Fornal, 1991). Interestingly,
stimulation of the Raphé nuclei induce EEG signs of drowsiness and lesion produces a state of
cortical arousal (Yamamoto et al., 1979); however, contrary findings have also been observed,
with Raphé output promoting arousal (Houdouin et al., 1991).
Blocking histamine receptors is known to produce enhancements in slow wave sleep,
while increasing histamine levels produces arousal (Lin et al., 1988). The tuberomammillary
nucleus of the posterior hypothalamus consists of histamine-containing neurons and is a major
source of diffuse projections to the cortex and the thalamus (Saper, 1985; Panula et al., 1989).
The activity of these neurons is highest during wakefulness, low during NREM sleep and abates
during REM sleep (Brown et al., 2001; Saper et al., 2001). However a more recent study, with
identified juxtacellular labelled histaminergic neurons demonstrated that these neurons are only
active in wakefulness (Takahashi et al., 2006).
The posterior hypothalamus is also the locus of orexinergic neurons (de Lecea et al.,
1998; Sakurai et al., 1998). Orexins stimulate behavioural arousal and cortical activation via
diffuse projections and excitatory influences on the cerebral cortex, lateral dorsal tegmentum,
substantia innominata, thalamocortical, lateral hypothalamic, dorsal and median Raphé, and
locus coeruleus (Peyron et al., 1998; Espana et al., 2005). Orexins are also known to have
excitatory effects on a variety of these arousal regions (Ivanov & Aston-Jones, 2000; Liu et al.,
2002; Kohlmeier et al., 2004; Wu et al., 2004; Xia et al., 2005). Neurons of the orexin system
106
are most active during wakefulness, decrease during NREM sleep and REM sleep (Lee et al.,
2005b; Mileykovskiy et al., 2005).
All of these systems are involved in promoting wakefulness or maintaining cortical
arousal. Thus removal of activity in each of these systems may be responsible for the production
of NREM or REM sleep states. The following section addresses how this may come about.
3.3.2 Sleep promoting systems
Neurons of the non-cholinergic basal forebrain and the hypothalamic preoptic area
discharge at higher rates during NREM sleep (Szymusiak & McGinty, 1989; Szymusiak, 1995;
Szymusiak et al., 1998; Lee et al., 2005a). Interestingly, after lesioning the ventrolateral
preoptica area (VLPO), the degree of neuron loss correlates closely with the loss of NREM sleep
(Lu et al., 2000). In addition, c-fos immunoreactivity is elevated in these regions during recovery
from sleep deprivation, and co-localize with GABAergic neurons (Lu et al., 2000; Modirrousta et
al., 2004). GABAergic VLPO and basal forebrain neurons appear to innervate orexin neurons
(Gritti et al., 1994; Jones, 2005; Sakurai et al., 2005), histaminergic neurons (Jones & Cuello,
1989; Saper et al., 2001), LC neurons (Jones & Cuello, 1989; Sherin et al., 1998; Saper et al.,
2001; Steininger et al., 2001) and the dorsal and median Raphé (Sherin et al., 1998; Steininger et
al., 2001). The VLPO also terminates within the cholinergic basal forebrain, PPT and LDT
(Saper et al., 2001). Thus these GABAergic systems may function to inhibit arousal-related
systems. The interactions of these neural networks have led some to devise models for how
wakefulness may switch to sleep (Saper et al., 2001). However, these models appear to be rather
premature. At this point it is probably safe to postulate that sleep onset occurs through active
inhibitory processes from sleep-active inhibitory neurons onto wake-active neurons. The net
effect may be the deactivation of arousal-related input to the cortex and thalamus. The
107
relationship between these arousal systems and the thalamus is discussed next within this
context.
3.3.3 Thalamocortical system
The thalamus, residing in the middle of the brain, is a key gateway to the cortex. It is
generally organized according to two main neuron types. The first neuron type is referred to as
the relay neuron, which passes information to the cortex (McCormick, 1992b; McCormick &
Bal, 1997). The second type of neuron is referred to as the reticular neuron, and these neurons
form a layer of GABAergic interneurons which inhibit the relay neurons (McCormick, 1992b).
In Chapter 2, regions of the thalamus which receive information from the spinothalamic tract and
subsequently project to the anterior cingulate cortex were reviewed. Here we cover more
extensively the architecture and cellular properties of the thalamus and its role in triggering and
generating electrocortical activity associated with sleep and wakefulness.
Thalamocortical neurons generate repetitive burst discharges during sleep that appear to
ride on a slower depolarizing potential (Hirsch et al., 1983; McCarley et al., 1983). By contrast,
wakefulness is characterized by tonic firing of these neurons (Hirsch et al., 1983; McCarley et
al., 1983). There is a general agreement that when the thalamus switches to this bursting mode
during NREM sleep, information flow through the thalamus is largely blocked (Livingstone &
Hubel, 1981). However, some evidence suggests that information may reach the cortex if
stimulation to the thalamus occurs during the down state of the oscillation, rather than during the
upstate (Watson et al., 2008).
Thalamocortical neurons are endowed with unique properties. These neurons demonstrate
rebound burst properties following intracellular hyperpolarization and tonic firing properties with
stimulation (Livingstone & Hubel, 1981). This rebound property is well accepted to be mediated
108
by low threshold calcium currents IT (Jahnsen & Llinas, 1984a; b; McCormick & Bal, 1997) and
has been extensively reviewed by McCormick (1992b). Specifically, this channel is activated as
membrane potential rises toward -65 mV, in which a burst of action potentials will be produced
and the channel becomes inactivated. To de-inactivate this channel and promote bursting
behaviour, other channels must be present to reduce membrane potential into a target range of
the IT channel. Indeed, the in-activation of a hyperpolarization-activated cationic (Na+ and K+) Ih
current during each burst of spikes produces a hyperpolarizing overshoot. This overshoot then
activates both the IT and Ih , following the burst, thus de-inactivating the IT for another calcium
spike and burst of action potentials. The frequency of the bursting behaviour is in a range which
is consistent with delta waves (0.5-4Hz) (McCormick & Pape, 1990; Dossi et al., 1992;
McCormick & Huguenard, 1992).
The reticular nucleus of the thalamus is an inhibitory network that surrounds the relay
cells of the thalamus (McCormick & Bal, 1997). The reticular cells actively inhibit the relay cells
and neighbouring reticular cells of the thalamus; however the relay corticothalamic neurons
actively excite the reticular cells (McCormick & Bal, 1997). The interactions are thought to be
involved in the generation of sleep spindles during NREM sleep (McCormick & Bal, 1997).
3.3.4 Integrating the thalamocortical system with the arousal system
The ascending arousal system and the burst properties of the thalamus have already been
covered. It is well understood that neurotransmitters from the ascending arousal systems are
responsible for producing states of wakefulness, whilst their withdrawal produces sleep. As seen
in figure 3.2, cholinergic input from the LTD and PPT, norepinephrine from the locus coeruleus,
serotonin from the Raphé nuclei, and histamine from the tuberomammillary nucleus all impinge
upon the thalamus to influence whether it is in a bursting or tonic firing mode (McCormick,
109
1992b). Specifically, the activation of muscarinic receptors by acetylcholine, H1 receptors by
histamine, α1-adrenoceptors by norepinephrine reduce potassium leak currents of relay neurons
and therefore produce a net depolarization (McCormick & Prince, 1987a; b; 1988; McCormick
& Williamson, 1991; McCormick, 1992a; McCormick & von Krosigk, 1992). Thus, removal of
this input then activates K-leak channels and hyperpolarizes relay neurons. The
hyperpolarization permits activation of the IT current, promoting a bursting mode and reinforcing
slow wave oscillations. By contrast, activation of histamine H2 receptors, serotonergic receptors,
and β-adrenergic receptors influence the Ih current in relay neurons (Pape & McCormick, 1989;
McCormick & Williamson, 1991).
3.3.5 Slow corticothalamic oscillations
Experimental data in sleeping animals suggests that delta oscillations and spindles are
organized according to a slow oscillation (<1Hz) originating in the cortex (Steriade et al.,
1993b). Indeed, the slow oscillation has been found to survive brainstem transaction and
thalamic lesions (Steriade et al., 1993c). The hyperpolarization phase of this slow oscillation is
associated with inactivity of both pyramidal and interneurons (Steriade et al., 1993d) while the
depolarization phase of this rhythm is associated with neuron activation, delta (1-4Hz) and
Src (Yu et al., 1997; Salter & Kalia, 2004) and CaMKIV (Kang et al., 2001; Wei et al., 2002;
Wu et al., 2008). The involvement of CaMKIV will be examined in detail in subsequent sections
(section 3.4.4 and 3.4.5). Moreover, if sleep is involved in memory formation, these two
processes should show mechanistic overlaps with respect to molecular biology (section 3.5.2).
However, for now, examining the consequences of LTP induction is worthwhile.
3.4.3 Long-term potentiation expression
The major mechanisms for the expression of LTP involves increasing the
phosphorylation of AMPA receptors, thereby increasing the current thorough the receptor
(Benke et al., 1998; Derkach et al., 1999; Lee et al., 2003) or the insertion of new AMPA
receptors into the postsynaptic membrane (Malenka & Nicoll, 1999; Bredt & Nicoll, 2003).
Examination of LTP protocols readily show that LTP is short-lived (30-60 mins) (Malenka &
Bear, 2004) relative to the memories which can last an entire lifetime. However it has been well
established that the longer forms of LTP require gene transcription and protein synthesis
(Abraham, 2003; Pittenger & Kandel, 2003).
A host of signalling molecules are implicated in the induction of “late phase” LTP and
include CaMKIV, PKA, and MAPK, which are involved in the activation of a key transcription
factor known as Cre-response element binding protein (CREB) (Silva et al., 1998; Abraham,
2003; Pittenger & Kandel, 2003). While a vast literature exists on CaMKII, which is positioned
in the cytosol, much less literature has focussed on CaMKIV which is located in both the cytosol
and the nucleus of neurons (section 3.4.4). Thus, the CaMKIV may be in a unique position for
controlling protein expression linked with LTP (section 3.4.5). Such a protein may be more
113
relevant to long-term changes in the brain. The following section will focus on what is known
about CaMKIV in synaptic plasticity, learning and memory.
3.4.4 CaMKIV molecular biology
CaMKIV is localized throughout the brain with strong expression is found in the
neocortex, hippocampus and cerebellum and slightly lesser expression in the ACC and amygdala
(Nakamura et al., 1995). Within brain cells, CaMKIV is localized within the nuclei of neurons
with lesser expression in the cytosol (Jensen et al., 1991; Nakamura et al., 1995; Bito et al., 1996;
Ibata et al., 2008). Importin-α is the protein which transports CaMKIV from the cytosol to the
nucleus (Kotera et al., 2005) and catalytic activation of CaMKIV is required for this to happen
(Lemrow et al., 2004).
Figure 3.3 summarizes the CaMKIV pathway based on data that is presented in this and
later sections. When calcium enters a neuron it can bind to a calcium sensor known as
calmodulin (Chin & Means, 2000). The activation of calmodulin produces a conformational
change that exposes its hydrophobic residues (Chin & Means, 2000). Exposure of these residues
permits activation of CaMKIV (Means et al., 1997). In addition, CaMKinaseKinase can
phosphorylate the activation loop of CaMKIV at residue Thr-196, increasing its
calcium/calmodulin dependent activity (Kasahara et al., 2001; Tokumitsu et al., 2004). However,
Protein phosphatase2A has been found to de-phosphorylate CaMKIV at Thr-196, to undo the
activity of CaMKinaseKinase (Tokumitsu et al., 1994). Interestingly, CaMKIV can also exhibit
activity in the absence of calcium (Chatila et al., 1996; Tokumitsu et al., 2004) and is capable of
auto-phosphorylation (Okuno et al., 1995; Chatila et al., 1996).
114
Figure 3.3 CaMKIV pathway Top: Depiction of CaMKIV activation pathway leading to DNA transcription. Following DNA transcription, AMPA receptors may be inserted in the plasma membrane to amplify incoming glutamate signaling. Lower left: panel showing the DNA activation machinery with nuclear CaMKIV phosphorylating CREB and CREB binding protein (CBP). Lower right: schematic diagram of CaMKIV domain structure. ER, endoplasmic reticulum, AC1, adenylate cyclase 1,TrKB, tyrosine receptor kinaseB, IP3 inositol-triphosphate, imp importin, TBP, TATA binding protein. BDNF, brain derived neurotrophic factor. (figure by H. Steenland, refs in text).
115
CaMKIV phosphorylates Ser-133 of CREB (Bito et al., 1996; Kasahara et al., 2001) and
CREB binding protein (CBP) (Chawla et al., 1998) and this stimulates gene transcription.
CaMKIV is also known for cross-talk between signalling pathways. For example, CaMKIV can
phosphorylate the calcium/calmodulin binding domain to prevent it from activating the AC1
pathway (Wayman et al., 1996) and can directly phosphorylate the small GTP-binding protein
Rap-1b (Sahyoun et al., 1991). This may lead to the activation of the MAP-kinase pathway,
which as mentioned is important in the hippocampus and prelimbic regions for trace fear
acquisition (Runyan & Dash, 2004; Runyan et al., 2004; Gao et al., 2009).
3.4.5 CaMKIV and plasticity
Induction of LTP in the CA1 hippocampus, with high frequency stimulation, results in
rapid activation of CaMKIV that lasts for up to 30 minutes (Kasahara et al., 2001). This results in
the phosphorylation of CREB, which then produces an increase in c-fos expression (Kasahara et
al., 2001). There is also evidence that CaMKIV transcription regulation impacts brain derived
neurotrophic factor expression (Shieh et al., 1998). Consistently, CaMKIV knockout mice have
dramatically reduced CREB phosphorylation, LTP induction in the hippocampus, and LTD
induction in the cerebellum (Ho et al., 2000). However, CaMKIV knockout mice show no
deficits in spatial memory or locomotor activity (Ho et al., 2000; Ribar et al., 2000). By contrast,
interfering with CaMKIV expression with a dominant negative form of CaMKIV impaired the
hidden platform version of the Morris spatial water maze task (Kang et al., 2001).
Studies involving fear memory have also been conducted with CaMKIV mice (Wei et al.,
2002; Ko et al., 2005; Fukushima et al., 2008; Wu et al., 2008). CaMKIV knockout mice show
contextual and auditory cue fear memory deficits at 1 and 7 days after conditioning, and reduced
elevations in pCREB following conditioning in the ACC, CA1 hippocampus and amygdala (Wei
116
et al., 2002). It was also found that CaMKIV knockout mice have impaired LTP to high
frequency stimulation in ACC (Wei et al., 2002). Interestingly, this study did not reveal any
difference in responsiveness in these mice to inflammatory pain in the formalin test (Wei et al.,
2002), suggesting some selectivity to modalities which aren’t involved in pain.
Over-expression of the dominant negative form of CaMKIV, which inhibits calcium
stimulated CaMKIV only in the postnatal brain, has shown that CaMKIV is important for late-
phase but not early-phase LTP (Kang et al., 2001). Indeed, this manipulation had the same effect
as anisomycin on late phase LTP in the CA1 hippocampus (Kang et al., 2001). Consistently, this
manipulation impairs phosphorylation of CREB and c-fos expression (Kang et al., 2001). In
contrast to others (Wei et al., 2002), this study found that freezing behaviour was reduced for
contextual but not cued-fear when studied 7 days after conditioning.
Recently, it has been shown that CaMKIV over-expressed mice demonstrate
enhancement in trace fear conditioning and enhancement in an early-stage LTP in the
hippocampus and ACC (Fukushima et al., 2008; Wu et al., 2008). Moreover, these animals do
not show any alteration in acute pain, suggesting the enhanced memory was not due to increased
pain sensitivity (Wu et al., 2008). Consistently, these animals have elevations of pCREB in the
cortex and in the hippocampus (Fukushima et al., 2008; Wu et al., 2008) and enhanced
consolidation of contextual fear memory (Fukushima et al., 2008).
CaMKIV has also been implicated in synaptic scaling (Ibata et al., 2008). Ibata et al.
(2008) used a model in which blocking calcium spikes resulted in an up-regulation of AMPA
receptors as measured by GluR2-EYFP tags. Interestingly, transfection with neurons with the
dominant negative form of CaMKIV (dnCaMKIV) blocked up-regulation of AMPA receptors.
The results suggest that if there is a drop in post-synaptic firing, there is a paradoxical increase of
CaMKIV activity. Interestingly this was found to be dependent on gene transcription. The
117
opposite may be the case for the induction of late-phase LTP which can be blocked by
dnCaMKIV or anisomycin (Kang et al., 2001). Thus under conditions of low calcium, somehow
CaMKIV is activated (Ibata et al., 2008) and under conditions of high calcium CaMKIV is
activated (Kang et al., 2001). It appears that CaMKIV may therefore be necessary to maintain a
balance, so that if a neuron is not receiving much input, it up-regulates AMPA receptors to
increase sensitivity to glutamate. By contrast, if a cortical neuron undergoes plasticity, CaMKIV
may stimulate the production of AMPA receptors. Such a mechanism may simply be related to
the amount of calcium in the cell.
Based on this summary of plasticity and CaMKIV, it appears that CaMKIV would be an
ideal candidate to examine the neurophysiological correlates of trace fear memory. In addition
since trace fear memory involves pain, and CaMKIV mice are largely normal with respect to
acute pain processing, learning processes can be potentiated without interference from different
processing streams. As the next section details, CaMKIV may also be related to sleep functions.
3.5 SLEEP and SYNAPTIC PLASTICITY
3.5.1 Possible function of sleep in learning and memory
Sleep occurs across a broad range of species, from mammals to fruit flies, yet its function
has not been firmly established and is highly debated (Benington & Frank, 2003; Sirota et al.,
To examine whether CaMKIV over-expression alters sleep, learning and memory, we
performed electroencephalogram (EEG) recordings with frontal-parietal electrodes according to
133
Figure 3.4 Experimental design Pre-conditioning sleep and wakefulness recordings were taken from mice on day one followed by trace fear conditioning. Mice were then returned to their home cage where their post-conditioning sleep was examined. Trace fear conditioning was conducted for 45 minutes in a shock chamber. This paradigm involved repeated trials of a presentation of a tone, followed by an interval (trace interval) and subsequently by a foot shock. Trace fear testing was conducted the following day in a similar fashion, only in this case the shock stimulus was omitted. Both WT and CaMKIV over-expressed mice were chronically instrumented with frontal parietal EEG recordings to examine EEG activity and freezing behavior for each part of the experimental design.
the experimental design outlined in Figure 3.4. Figure 3.5A-D shows representative traces and
grouped data of natural sleep and wakefulness states for CaMKIV over-expressed and WT mice.
A significant interaction was detected between genetics and behavioural state for delta (2-4Hz)
134
Figure 3.5 CaMKIV over-expression enhances delta sleep in the recording environment (A) Example raw traces of sleep measurements in WT and CaMKIV over-expressed mice for wakefulness, NREM and REM sleep. To the left is a drawing, showing the location of recording configurations of the electrodes. CaMKIV over-expressed mice demonstrate relatively normal EEG during wakefulness and REM sleep but had elevated slow oscillations during NREM sleep (arrow). The raw neck and the smoothed and rectified (Sm/Rt) neck signals show that EMG activity in CaMIV over-expressed mice was relatively normal across sleep and wakefulness states. (B) CaMKIV over-expressed mice have normal wakefulness EEG power compared to WT mice. (C) CaMKIV over-expressed mice have enhanced delta EEG power in the 2-4Hz range compared to WT mice (D) CaMKIV over-expressed mice have normal REM sleep EEG power compared to WT mice. (E) CaMKIV over-expressed mice have normal neck EMG across all behavioral states compared to WT mice. * indicate significant differences compared to wild type (WT) mice with P < 0.05, Values are means + SEM.
135
power (F(3,38) = 2.83; p = 0.05, two-way ANOVA). Post-hoc analysis revealed that there was a
significant increase in delta power selective to NREM sleep (t = 2.80; p = 0.008, Fig. 3.5C).
From figure 3.5D, it also appears that there was an enhancement in 7.5-13Hz power during REM
sleep, however the ANOVA yielded neither a main effect (F(1,20) = 2.49; p = 0.124, two-way
ANOVA) nor an interaction (F(3,38) = 1.46; p = 0.241, two-way ANOVA) for this frequency,
between CaMKIV and WT mice.
Neck electromyogram amplitudes were not different between CaMKIV over-expressed
mice and WT mice (F(1,20) = 2.83; p = 0.644, two-way ANOVA), confirming that the sleep
scoring was done consistently (Fig. 3.5E). It was also found that the time spent in each particular
sleep-wake state in the light (F(1,6) = 0.594, p = 0.470, two-way ANOVA) and dark (F(1,6) =
0.275, p = 0.630, two-way ANOVA) was not different between WT and CaMKIV over-
expressed mice (Table 3.1). Finally, the frequency of occurrence of each particular sleep-wake
state during the light (F(1,6) = 0.016, p = 0.904, two-way ANOVA) and dark period (F(1,6) = 0.001,
p = 0.975, two-way ANOVA) was not different between WT and CaMKIV over-expressed mice
(Table 3.1). Overall, these results show that CaMKIV over-expression augments delta-related
EEG oscillations in the sleep recording environment without impacting the time spent sleeping or
the frequency of sleep.
Table 3.1 CaMKIV over-expressed mice have normal sleep duration and bouts CaMKIV over-expressed mice demonstrate normal sleep bouts frequency and time spent sleeping. Values are means ± SEM.
136
3.7.4B CaMKIV over-expression enhances learning and memory
CaMKIV over-expressed mice exhibit learning and memory enhancements for the trace
fear paradigm (Wu et al., 2008). In the present set of trace fear experiments, animals were
trained and tested in the dark while EMG and EEG were simultaneously recorded. An EMG-
based method, previously developed in our laboratory, was used to score freezing behavior
(Steenland & Zhuo, 2009). This EMG recording method permits the measure of freezing in the
dark phase of the cycle, when nocturnal rodents are normally active. By having mice trained at
this time, mice do not need to be disturbed during the light phase for training or testing, as is
usually done in learning experiments. Secondly, this method, in conjunction with EEG can help
discriminate sleep from fear behavior (Steenland & Zhuo, 2009), and help minimize contextual
visual information which could confound our study. Brain wave and freezing behavior measures
were compared to the first 60s baseline taken in the recording chamber during trace fear
conditioning. Figures 3.6A and 3.6B show summarized and raw EMG-based freezing data for
WT and CaMKIV over-expressed mice during conditioning and testing days. There was
significant effect of time interval (F(7,140) = 24.25; p < 0.001, two-way ANOVA), with both WT
and CaMKIV over-expressed mice showing learning and memory retention. There was a
significant effect of the genetic manipulation on freezing behaviour (F(1,20) = 24.58 p < 0.001,
two-way ANOVA), with CaMKIV over-expressed mice showing more freezing during both the
conditioning phase and the testing phase. Thus, CaMKIV over-expressed mice learn and
remember better then WT controls.
137
Figure 3.6 CaMKIV over-expression enhances trace fear conditioning and subsequent memory (A) Trace fear conditioning (30 sec trace interval) learning curve for both WT and CaMKIV over-expressed mice (* is P<0.05 compared to baseline (BL)). However, CaMKIV over-expressed mice have enhanced fear behaviors during trace fear conditioning and testing above that of WT mice (# is P<0.05). (B) Example of EMG-based freezing and scoring (redlines under the trace) from WT and CaMKIV over-expressed mice. Raw EMG data show periods of quiescence during the conditioning and memory trials of the mice. During conditioning and memory testing, CaMKIV over-expressed mice demonstrate more EMG-based freezing behavior than WT mice. S/ES represents shock and expected shock, for the conditioning and test trials respectively.
3.7.4C CaMKIV over-expression and EEG responses to tone, trace interval, and shock
There are numerous reports of brain wave oscillations associated with learning, memory,
and attention in a variety of behavioural paradigms (Buzsaki, 2002; Seidenbecher et al., 2003;
Tsujimoto et al., 2006; Stark et al., 2008). We investigated the EEG responses for CaMKIV and
WT mice to tone and trace intervals, and following a shock to the foot during the trace fear
conditioning day (Fig. 3.7A-F). We also examined the EEG responses during the testing day.
However in this case, a shock was not delivered and so we examined the EEG of an expected (or
anticipated) shock.
3.7.4C-I EEG responses to Tone
It is possible that CaMKIV over-expressed mice demonstrate enhanced arousal or attention
to sound stimuli during conditioning. To investigate this, we analyzed EEG during the
138
presentation of the tone. Mice were found to have normal EEG power, for most frequencies,
during the presentation of the tone (Fig. 3.7). However, there was a significant effect of the tone
on 4-7.5Hz power (F(2,39) = 20.89; p < 0.001, two-way ANOVA) and a significant interaction
between genetics and the tone presentation (F(2,39) = 7.64; p < 0.002, two-way ANOVA). Post-
hoc analysis revealed that the presentation of tone during conditioning enhanced 4-7.5Hz power
in CaMKIV over-expressed mice (t = 6.31; p < 0.001) and WT mice (t = 2.76; p < 0.026). When
the tone was played on the testing day, post-hoc analysis revealed that 4-7.5Hz power was
increased for CaMKIV mice (t =6.73; p < 0.001) but not WT mice (t = 0.331; p = 1.0) compared
to their respective baseline. Finally, CaMKIV mice were found to have greater 4-7.5Hz power
during the tone on the testing day when compared to WT mice (t = 6.04; p < 0.001).
There was a significant alteration of 7.5-13Hz power during the tone presentation (F(2,39)
= 24.32; p < 0.001, two-way ANOVA) and a significant interaction between genetics and tone
presentation (F(2,39) = 5.79 p < 0.006, two-way ANOVA). Post-hoc analysis revealed that, for
CaMKIV over-expressed mice, presentation of a tone decreased 7.5-13Hz power during both
conditioning (t = 6.06; p < 0.001) and testing days (t = 6.92; p < 0.001) compared to baseline
(Fig. 3.7D). Additionally CaMKIV over-expressed mice had elevated 7.5-13Hz power above that
of WT mice (t = 3.04; p < 0.005) during baseline. There was also a significant alteration of 13-
20Hz power during the tone presentation (F(2,39) = 4.38; p < 0.019, two-way ANOVA). Post-hoc
analysis revealed CaMKIV over-expressed mice had reduced power during tone presentation on
trace fear testing days (t = 3.64; p < 0.002) compared to baseline (Fig. 3.7E).
139
Figure 3.7 CaMKIV over-expression and EEG responses to tone, trace interval and pain CaMKIV over-expressed mice demonstrate alterations in EEG rhythms during different phases of trace fear conditioning. EEG data was averaged across ever trial during trace fear conditioning (left side of graph) and again for every trial during memory testing (right side of graph) for WT and CaMKIV over-expressed mice. (A) None of the interventions had a significant effect on 1-2Hz frequency EEG power. (B) Trace fear training increased 2-4Hz power in both WT and CaMKIV over-expressed mice above baseline (* P <0.05) for the trace interval. (C) Trace fear conditioning significantly enhanced 4-7.5Hz (4-7.5Hz) EEG power in CaMKIV over-expressed mice for both the tone and the trace interval conditions compared to baseline. WT mice only demonstrated an enhancement in 4-7.5Hz EEG power for the tone interval (* is P<0.05). In addition, there was a significant enhancement in 4-7.5Hz EEG power of CaMKIV mice above that of WT mice for the trace interval during conditioning and the E-Shock and tone interval during memory (# is P <0.05 compared to WT). (D) The baseline 7.5-13Hz EEG power was increased in CaMKIV over-expressed mice with initial placement into the fear conditioning chamber (# is P<0.05 compared to WT) and reduced during the tone, trace and E-shock intervals (* is P<0.05 compared to baseline). (E) 13-20Hz EEG power was reduced during the tone presentation on the test day relative to baseline in CaMKIV over-expressed mice. (F) None of the interventions had a significant effect on 20-30Hz EEG power. E-shock is expected shock, BL is baseline, and values are means + SEM.
140
3.7.4C-II EEG responses to the trace interval
For a fear response to be conditioned across the trace interval, information about the tone
must be retained until the shock is perceived. Since hippocampus, in conjunction with the
prefrontal cortex, is thought to maintain contiguity of fear memory during the trace interval
(Buchel et al., 1999; Bangasser et al., 2006) we examined whether or not there would be neural
oscillations which corresponded to the trace interval. There was a significant effect of the trace
interval on 2-4Hz delta power (F(2,39) = 18.89; p < 0.001, two-way ANOVA). Post-hoc analysis
revealed that there was a significant increase in 2-4Hz power for both WT (t = 2.84; p = 0.016)
and CaMKIV mice (t = 5.89; p < 0.001) relative to baseline on the conditioning day (Fig. 3.7B).
For the testing day, CaMKIV over-expressed mice had significantly elevated 2-4Hz power (t =
3.40; p < 0.005) compared to baseline. Overall, there was no effect of genetic manipulation on 2-
4 Hz power, so this frequency is unlikely to explain the learning and memory enhancement in
CaMKIV over-expressed mice.
There was a significant effect of trace interval on 4-7.5Hz power (F(2,39) = 21.45; p < 0.001,
two-way ANOVA) and a significant interaction between genetics and the trace interval (F(2,39) =
5.99; p = 0.005, two-way ANOVA-RM). Post-hoc analysis revealed that there was a significant
increase in 4-7.5Hz power for CaMKIV mice above baseline (t = 7.42; p < 0.001) for the
conditioning day and above the baseline for the testing day (t = 3.73; p < 0.002, Fig. 3.7C). In
addition CaMKIV over-expressed mice had a significant enhancement in 4-7.5Hz power over
WT mice during the conditioning day (t = 3.10; p = 0.004). These findings are consistent with
the enhanced learning and memory of the CaMKIV mice (Fig. 3.6) and may serve as a possible
mechanism of attention processing or memory encoding.
A significant alteration in 7.5-13Hz power was detected across the conditioning interval
(F(2,39) = 23.82; p < 0.001, two-way ANOVA). In addition, a significant interaction was detected
141
between genetics and trace interval (F(2,39) = 4.46; p < 0.018, two-way ANOVA). Post-hoc
analysis revealed that 7.5-13Hz power was decreased in CaMKIV over-expressed mice
compared to baseline for the conditioning day (t = 5.78; p < 0.001) and the testing day (t = 6.58;
p <0.001, Fig. 3.7D).
3.7.4C-III EEG responses to shock and expected shock
Our previous studies have shown that CaMKIV over-expressed mice have normal
responses to acute pain (Wu et al., 2008). It is conceivable that CaMKIV mice have enhanced
cortical arousal in response to foot shock which could be detected with cortical EEG. This might
provide a possible explanation for enhanced learning and memory in these animals in the trace
fear paradigm. Thus, we analyzed whether or not EEG arousal responses to the foot shock
differed between animals (Figs. 3.7A to 3.7F). A significant alteration in 2-4Hz power was
detected for the shock/expected shock condition (F(2,39) = 15.69; p < 0.001, two-way ANOVA).
Post-hoc analysis revealed that 2-4Hz power was increased in CaMKIV over-expressed mice (t =
3.12; p <0. 0.010) during the expected shock test condition, compared to baseline (Fig. 3.7 B).
There was a significant alteration of 4-7.5Hz power following the shock/expected shock stimuli
(F(2,39) = 15.88; p < 0.001, two-way ANOVA) and a significant interaction with the genetic
manipulation (F(2,39) = 4.17; p < 0.023, two-way ANOVA). Post-hoc analysis revealed an
enhancement in 4-7.5Hz power (t = 4.16; p = 0.001) during the expected shock condition,
compared to baseline (Fig. 3.7C). In addition post-hoc analysis revealed that CaMKIV over-
expressed mice had significantly greater 4-7.5Hz power during the expected shock condition
(F(2,39) = 2.71; p < 0.009, two-way ANOVA). The increase of 4-7.5Hz power appeared to be a
result of the freezing behaviour which carried over into the expected shock interval. A significant
alteration in 7.5-13Hz power was detected during the shock/expected shock interval (F(2,39) =
142
13.17; p < 0.001, two-way ANOVA). Post-hoc analysis revealed that 7.5-13Hz power was
decreased in CaMKIV over-expressed mice compared to baseline (t = 3.92; p = 0.001) during
shock, but was unaltered in WT mice (t = 0.22; p = 1.00, Fig. 3.7D). However, when WT mice
were compared directly to CaMKIV over-expressed mice, no significant difference was detected
(t = 0.55 p = 0.583), suggesting that their actual 7.5-13Hz EEG power responses to the shocks
are not likely to account for the enhanced learning and memory of CaMKIV mice. In conclusion,
based on our previous findings (Wu et al., 2008) and the current EEG responses, we suggest that
CaMKIV over-expressed and WT mice respond similarly to pain on the conditioning day but
differently to the expectation of pain on the test day
3.7.4D Four-7.5Hz oscillation enhancements parallel learning and can be localized to the
prelimbic cortex.
Since 4-7.5Hz oscillations are enhanced in CaMKIV mice during learning and memory,
above that of baseline and WT mice (Fig. 3.7), we expected that these oscillations might
represent neural correlates of attention or memory of the expected shock during the trace
interval. We therefore expanded our 4-7.5Hz power analysis to examine its time course as the
mice learned. Figure 3.8A and B depict the 4-7.5Hz activity across intervals of trace fear
conditioning and testing. There was a significant main effect of the conditioning interval on 4-
7.5Hz power (F(7,136)= 17.28; p < 0.001, two-way ANOVA). There was neither a main effect of
genetics (F(1,20)= 2.994; p = 0.099, two-way ANOVA) nor or interaction between genetics and
trace fear interval (F(7,136)= 1.97; p = 0.064, two-way ANOVA). Since these tests were nearly
significant and an interaction between genetics and trace condition was found in Figure 3.7C, we
ran post-hoc tests to see whether theta power learning curves were different between CaMKIV
and WT animals. Four-7.5Hz power was also significantly elevated in CaMKIV over-expressed
143
mice above baseline during the testing phase (~24 hours later) (all p <0.05, as in Fig. 3.8A). The
increase in 4-7.5Hz power appeared to be related to a combination of increased voltage (Fig.
3.8B) and the number of occurrences during the trace interval. For WT mice, the 4-7.5Hz EEG
power increased during the conditioning but not the testing phase. This is the first evidence to
show that CaMKIV manipulation can enhance, in a parallel fashion, the expression of fear
behavior and 4-7.5Hz power.
To examine whether changes in 4-7.5Hz power were related to changes in freezing behavior
during the trace interval; the net change in 4-7.5Hz power from baseline, for every animal of
every trial, was quantified, plotted, and correlated against the change in freezing behavior from
baseline for the respective trace fear interval. A significant correlation was found between the
change in trace fear learning and changes in 4-7.5Hz power (r2 = 0.112, p < 0.001, Fig. 3.7C).
The results indicate that variation in freezing behavior (albeit a small amount) is related to
variation in 4-7.5Hz power.
144
Figure 3.8 CaMKIV over-expression enhances 4-7.5Hz waves which can localize to the prelimbic cortex (A) Grouped data showing that WT (open circles) and CaMKIV over-expressed mice (closed circles) have significant enhancements in 4-7.5Hz rhythm (4-7.5Hz), during the trace fear interval, throughout the conditioning protocol (* is P<0.05 compared to baseline). This effect mirrored the learning-curve of these mice, which was significantly greater in CaMKIV over-expressed mice as compared to WT for both conditioning and testing (# is P<0.05). (B) Example EEG activity corresponding to panel A, which was band-pass filtered for visual inspection of 4-7.5Hz rhythms. High voltage 4-7.5Hz activity occurred during periods of freezing behavior and appeared to be more frequent and often larger in CaMKIV over-expressed mice. (C) Trial by trial changes in freezing during the trace interval were positively correlated to changes in 4-7.5Hz power during conditioning. (D) When WT mice are trained with a 15 sec trace fear interval, prominent 4-7.5Hz rhythms can be localized to the prelimbic cortex, during conditioning (* is P<0.05 compared to baseline) but not on the testing day. Values are means +/- SEM. (E) Trace fear conditioning (15 sec trace interval) learning curve for WT mice (* is P<0.05 compared to baseline). Notice the memory enhancement when compared to Fig 3. BL is baseline, and values are means + SEM.
145
Since 4-7.5Hz frontal-parietal EEG power was increased during the trace fear interval,
we wanted to identify if this frequency band can to be localized to a particular brain structure.
Therefore, we chose to accentuate trace fear learning (analogous to the effect of CaMKIV over
expression) by reducing the duration of the trace fear interval in order to make the time between
shock and tone shorter. Since the prelimbic cortex is known to generate both theta rhythms
(Jones, 2005; Siapas et al., 2005; Paz et al., 2008) and be involved in trace fear conditioning
(Baeg et al., 2001; Gilmartin & McEchron, 2005a) we placed intracortical electrodes in this
region. Prelimbic field potential recordings were performed in WT mice under a 15 second trace
interval protocol to localize 4-7.5Hz rhythms. There was a significant effect of conditioning on
4-7.5Hz activity recorded in the prelimbic cortex (F(7,29) = 3.64; p < 0.006, one-way ANOVA).
Post-hoc analysis revealed significant enhancement of 4-7.5Hz rhythms during trace fear
conditioning (Fig. 3.8D). Consistently, there was a significant effect of conditioning on fear
Since sleep in the recording environment increased 2-4Hz EEG power, we decided to
examine the time course of these slow delta oscillations (1-4Hz) across the entire recording
session. Slow delta oscillations are thought to be related to sleep need (Borbely & Achermann,
2000) and are potentiated by learning paradigms (Huber et al., 2006). Slow delta oscillation
power was averaged for each hour starting when the lights were turned on, for pre- and post-
conditioning NREM sleep. For this analysis, separate statistical comparisons were made directly
between WT and CaMKIV over-expressed mice for each hour. There was a significant effect of
the genetic manipulation for the second (F(1,23) = 8.64; p = 0.007, two-way ANOVA), fifth (F(1,22)
= 9.52; p = 0.005, two-way ANOVA) and sixth hour (F(1,22) = 6.05; p = 0.021, two-way
ANOVA) (Figs. 3.9A and 3.9B). Further analyses revealed that CaMKIV over-expression
elevated slow delta oscillation power above WT mice during the pre-conditioning phase for the
second (t = 2.12; p = 0.034), fifth (t = 2.94; p = 0.007), sixth hour (t = 2.59; p = 0.015). This
effect may be due to the novelty to the recording environment or a selective enhancement in
brain waves that occurs during natural sleep.
We next examined the effect of trace fear conditioning on post-conditioning NREM
sleep. For this analysis, separate statistical comparisons were made directly between WT and
CaMKIV over-expressed mice for each hour. In addition we made separate comparisons between
pre and post-conditioning for each hour to examine interactions. While a similar trend for
elevated slow delta waves in CaMKIV over-expressed mice above that of WT was seen,
following trace fear conditioning, the trend did not achieve significance. However, a significant
interaction was detected for the first hour comparisons, showing that the level of slow delta
oscillation power depended, not only on whether the animal was trace fear trained but also on the
genetic manipulation (F(1,11) = 5.1; p = 0.045, two-way ANOVA).
147
Figure 3.9 CaMKIV over-expression enhances slow delta oscillation power during NREM sleep before and after trace fear training (A) Example of EEG sleep recording in the first hour, after the lights had been turned on, in the recording chamber for both CaMKIV over-expressed mice and WT mice. However if the mouse had been trace fear conditioned, slow delta oscillation (1-4Hz) enhancements were detected in the recording for CaMKIV over-expressed mice but not WT mice. (B) Group data showing that CaMKIV over-expressed mice have elevated slow delta oscillation power above that of WT mice during NREM sleep in the recording chamber across an 8 hour span. A subset of the CaMKIV over-expressed and WT mice were then recorded after trace fear training during the same time points of the next day. CaMKIV over-expressed mice were found to have elevated slow delta oscillations above that of the first hour of their pre-conditioning. However, WT mice did not show a similar enhancement. * is P < 0.05 compared to WT for the corresponding hour. # is P <0.05 between pre and post-conditioning for the corresponding hour. (C) CaMKIV over-expressed and WT mice had similar 4-7.5Hz power during REM sleep prior to and following trace fear conditioning. 4-7.5HZ power was averaged across 4 hour time periods.
148
Further analysis revealed that CaMKIV over-expressed mice have elevated slow delta oscillation
power when compared to their preceding pre-conditioning baseline for the first hour (t = 3.99; p
= 0.002). By contrast, WT mice did not show the same enhancement (t = 0.22; p = 0.827). This
result shows that genetic enhancement of learning and memory (Fig. 3.9A) also potentiates EEG
activity during subsequent sleep in CaMKIV over-expressed mice, linking genetics, sleep and
memory through the CaMKIV protein.
It was also examined whether REM sleep 4-7.5Hz power would be influenced following
trace fear conditioning. In contrast to the slow delta power analysis, in which 8 bins were used,
REM sleep was broken down into two bins (1-4 and 4-8 hours post-conditioning). This was done
because of the reliability of recording REM sleep for each individual hour was low. In contrast,
to the enhancements in slow delta oscillations during NREM sleep for CaMKIV over-expressed
mice, we found no effect of genetic manipulation on REM sleep 4-7.5Hz power before and after
training within the first four hours (F(1,20) = 1.07; p = 0.312, two-way ANOVA) and last four
It is possible that the increase in delta power in CaMKIV over-expressed mice may be
secondary to an increase in the latency of falling asleep, following fear conditioning. There was
no significant effect of genetic manipulation on the latency to NREM sleep (F(1,14) = 0.775; p =
0.393, two-way ANOVA), and REM sleep (F(1,14) = 0.331; p = 0.574, two-way ANOVA)
following trace fear conditioning (Figs. 3.10A and 3.10B). Thus, the enhancement in slow delta
oscillation power after trace fear training cannot be accounted for by alterations in sleep latency.
It is also possible that CaMKIV over-expressed mice may have increased slow delta power
because their sleep is consolidated into a few bouts of very deep sleep, following trace
conditioning. Interestingly, we found a statistically significant interaction between genetics and
149
NREM durations before and after trace fear training (F(1,13) = 6.86; p = 0.021, two-way
ANOVA). In contrast to our prediction, post-hoc analysis revealed that CaMKIV over-expressed
mice had more NREM sleep following trace fear conditioning (t = 2.23; p = 0.044). In WT mice,
NREM durations were not influenced by trace fear conditioning (t = 1.56; p = 0.142).
Additionally, REM sleep durations were not significantly different (F(1,14) = 0.244; p = 0.630,
two-way ANOVA) between WT and CaMKIV over-expressed mice (Figs. 3.10C and 3.10D).
The enhancement in delta power of CaMKIV over-expressed mice is further corroborated by the
enhancement in time spent sleeping following trace fear conditioning. Indeed the increase in
NREM sleep duration in CaMKIV over-expressed mice reinforces the concept that genetics,
sleep and memory may be linked through the CaMKIV protein. Thus, based on the findings we
expect that slow wave enhancements following trace fear training may be related to learning.
150
Figure 3.10 Impact of trace fear conditioning on sleep duration and latency (A) The latency to enter NREM sleep before (pre) and after (post) trace fear conditioning was unaltered in CaMKIV over-expressed and WT mice. (B) The latency to enter REM sleep before and after trace fear conditioning was unaltered in CaMKIV over-expressed and WT mice. (C) WT mice showed a trend for decreased NREM sleep following trace fear conditioning, while CaMKIV over-expressed mice showed a significant increase in NREM sleep duration. * is P < 0.05. (D) The duration of REM sleep before and after trace fear conditioning was unaltered in WT and CaMKIV over-expressed mice.
3.7.4F S low delta oscillation enhancements are localized to the anterior cingulate cortex
Frontal-parietal EEG recordings yield a selective enhancement in 1-4Hz slow delta power
following trace fear conditioning for CaMKIV over-expressed but not WT mice (Fig. 3.9). Thus
we needed to examine whether WT mice can also demonstrate enhanced delta power following
trace fear conditioning. We chose to accentuate trace fear learning (analogous to the effect of
CaMKIV over-expression) in WT mice by reducing the duration of the trace fear interval, in
151
expectation that this would accentuate slow delta power during subsequent NREM sleep.
Provided WT actually demonstrate enhanced delta power, we wanted to know whether it would
localize to a particular brain region. C-fos is up-regulated in the anterior cingulate cortex
following trace fear conditioning without any effect in the motor cortex (Han et al., 2003). These
results suggest that the ACC is involved in learning and memory processes associated with trace
fear conditioning. Thus we examined whether slow delta oscillations would be potentiated in the
ACC following trace-fear conditioning, indicative of memory consolidation.
Simultaneous bipolar intracortical field recordings were performed in the ACC and the
motor cortex of WT mice. The motor cortex served as a control for the ACC to examine
regionally selective slow oscillation enhancements. The behavioral paradigm for this second set
of experiments is identical to that outlined in Figure 3.4. We first standardized the location of the
recordings from the ACC to carry out this examination (Fig. 3.11). This experiment revealed that
optimal placement of the bipolar electrodes in the ACC was that straddling layers I-III, consistent
with others (Tsujimoto et al., 2006). This recording configuration was set up so as to minimize
volume conducted theta rhythms from the hippocampus and 60Hz noise from the external
environment. Data was normalized for the purposes of comparison between motor cortex and the
ACC. This was necessary as the voltage of the motor cortex was often much larger than that
recorded from the ACC. During sleep, in the recording environment, the normalized slow delta
oscillation power of the motor cortex and ACC mirrored one another across the 8 hour recording
period (Figs. 3.12A and 3.12B). Following trace fear conditioning, there was a significant
difference between the ACC and motor cortex (F(1,6) = 6.37; p = 0.045, two-way ANOVA-RM).
Post-hoc analysis revealed that there was an enhancement in slow oscillation power in the ACC
above that of the motor cortex (Figs. 3.12A and 3.12B, t = 2.52; p = 0.045, Bonferroni’s paired t
152
test). In two cases, where bipolar electrodes did not straddle layers I-III (Fig. 3.13B for electrode
positions), the ACC slow delta oscillation power did not mirror that of the motor cortex during
Figure 3.11 Optimizing electrode implantation Four electrodes were implanted into the ACC. Differential recordings were performed between adjacent electrodes. Field potential voltage is most robust when straddling layers I-III of the anterior cingulate cortex (ACC). The polarity of the recordings is opposite for the upper trace and lower trace since the electrode crosses the midline. The differences in polarity are most evident in NREM sleep. Bars indicate 200µV calibration.
pre-conditioning sleep (Fig. 3.12C), and so were not used to examine regional field potential
enhancements. Collectively, the result indicates that trace fear conditioning selectively activates
brain regions of the ACC in layers I-III during subsequent NREM sleep.
153
Figure 3.12 Trace fear conditioning increases slow delta oscillations in the ACC in layer I-III (A) Group data showing that the motor cortex and the ACC (layer I-III electrode) slow wave oscillation power paralleled one another, when their data were normalized to their average slow delta oscillation power (across 8 hours) on the first day. However, after the animals were trace fear conditioned the slow delta oscillation power significantly increased in the ACC over that of the motor cortex (* is P < 0.05 compared to motor cortex post-conditioning). (B) Example of the slow delta oscillation during NREM sleep, recorded in the motor cortex (MTC) and the ACC before and after trace fear conditioning. When the mouse goes to sleep there is an increase in slow delta oscillation power in the EEG of the ACC with only a modest increase in the motor cortex. (C) Electrodes that did not straddle layers I-III did not show and increase in slow delta oscillation power following trace fear conditioning. Indeed the motor cortex (MTC) served as a poor control for the ACC in these cases. This can be seen during the pre-conditioning session, in which the normalized slow wave oscillation power of the ACC and MTC activity did not overlap.
154
Figure 3.13. Position of electrodes Position of electrodes, in the motor cortex (MTC) and the ACC, from experiments in Figure 3.13. Data from Figure 3.13 was grouped according to whether the electrodes were located within layers I-III (A) of the ACC or outside this region (B). If electrodes were in layers I-III, the normalized baseline sleep correlated with that of the motor cortex. (Indicated as correlated or uncorrelated).
3.7.4G Learning and theta rhythms but not slow delta oscillations are correlated with trace fear
memory formation.
We first examined whether measures of fear conditioning were related to the formation of
trace fear memory. Data from CaMKIV over-expressed and WT animals were grouped together.
A statistically significant positive correlation was found between the total time spent freezing
during the trace interval and the total time spent freezing during the trace interval of the testing
day (r2 = 0.188, p<0.039, Fig. 3.14A). Interestingly, a significant positive correlation was also
found between the percent increase in 4-7.5Hz power during the trace interval and the total time
spent freezing during the trace interval of the testing day (r2 = 0.188, p<0.039, Fig. 3.14B).
155
Figure 3.14 Variations in learning and 4-7Hz oscillation power are related to memory formation (A) The average change in freezing during the trace interval was positively correlated with the average change in fear memory on the testing day. (B) The average change in 4-7.5Hz power during the trace interval was positively correlated with the average change in fear memory on the testing day. (C) The average change in slow delta oscillation power during post-conditioning sleep was not positively correlated with the average change in fear memory on the testing day.
The result indicates that some degree of variability in 4-7.5Hz power during the trace interval
accounts for a degree of variability in memory encoding. However, in spite of the enhanced slow
wave activity in CaMKIV over-expressed mice during the first hour of post-conditioning sleep,
no correlation was detected between the percent increases in slow wave activity and the fear
memory the following day (r2 = 0.0215, p<0.538, Fig. 3.14C). The result suggests that for the
trace fear paradigm, the most important factors controlling the variability of the memory is the
degree of freezing and the change in theta oscillation power during conditioning.
156
3.7.5 Discussion
Our findings provide the first direct genetic evidence that enhancing CaMKIV expression
in the forebrain enhances 4-7.5Hz oscillations during trace fear conditioning and boosts slow
delta oscillations during NREM sleep. We also show that accentuation of trace fear learning,
through reduction of the trace fear interval, produces increases in oscillation power during both
learning and subsequent sleep. This enhancement could be localized to the frontal cortex which
is also likely to be the origin of the enhancements seen in CaMKIV over-expressed mice.
Finally, correlation analyses implicate 4-7.5Hz oscillations in both the learning of the trace fear
paradigm and the memory of the conditioned response 24 hours later.
3.7.5A CaMKIV and natural sleep
CaMKIV mRNA is up-regulated following sleep (Cirelli et al., 2004) and CREB
knockout mice (Graves et al., 2003) have reduced levels of arousal, spending more time in
NREM sleep. Thus, it could be predicted that CaMKIV over-expression may also impact
behavioral states of arousal and sleep. However, unlike CREB knockout mice, CaMKIV over-
expression did not alter the amount of time spent in any behavioral state. When EEG power was
quantified in the recording environment, CaMKIV over-expressed mice showed an increase in
slow delta oscillation power values, which persisted across the first 8 hours of the light period.
There are three simple interpretations for this finding; firstly, the enhancement may be related to
processing in the novel recording environment, which then leads to and increase in slow delta
oscillation power during sleep. Indeed we did find an increase in 7.5-13Hz power during
wakefulness when animal was exposed to the novel fear conditioning chamber. Interestingly, this
is the same frequency that was decreased in CREB knockout during REM sleep (Graves et al.,
2003). Consistently, preliminary studies in our lab suggest that when a 4 day acclimatization
157
period (in which animals are hooked to the recording device in the recording environment) is
given to CaMKIV mice; the slow wave enhancements during baseline are not present. Future
studies, using sample sizes comparable to those used in the current study, will need to confirm
this possibility. Alternatively, reduced slow wave sleep is thought to correlate with both ageing
and memory impairment (Landolt et al., 1996; Carrier et al., 1997; Landolt & Borbely, 2001;
Hornung et al., 2005). Since CaMKIV over-expression has been found to rescue memory
impairment in aged mice (Fukushima et al., 2008), the enhanced slow wave delta power in the
recording environment may reflect some degree of rescue of age-related decrements in slow
wave oscillations. Finally, CaMKIV over-expressed mice may have an alteration in their sleep
homeostat. However, if this were the case it may be expected that these mice would sleep longer
or more frequently during baseline pre-conditioning sleep recordings, which they did not (Table
3.1). Indeed, our preliminary studies (not presented here) suggest that there is no obvious
difference in slow wave sleep recovery as a consequence of 4 hours of sleep deprivation, induced
with novel objects and gentle handling in CaMKIV mice.
3.7.5B CaMKIV and trace fear conditioning
We have previously shown that CaMKIV over-expressed mice have enhancement in trace
fear learning and memory (Wu et al., 2008). We replicated this finding, utilizing a new EMG-
based scoring methodology (Steenland & Zhuo, 2009) which permitted evaluation of the mouse
behaviour in low light conditions. This permitted study of these animals during their natural
sleep and wakefulness cycle. In parallel with trace fear conditioning curves, we found an
enhancement in 4-7.5Hz power, which could be localized to the prelimbic cortex. There is
evidence that these prefrontal 4-7.5Hz rhythms, which exist in mice, rats, monkeys and humans,
are involved in attention and learning processes (Tsujimoto et al., 2003; Jones & Wilson, 2005;
158
Delorme et al., 2007; Sirota et al., 2008). Indeed, trace fear conditioning requires prefrontal
cortical structures and sustained attention for learning to take place (Baeg et al., 2001; Han et al.,
2003; Knight et al., 2004; Runyan et al., 2004; Zhao et al., 2005). Thus we interpret this finding
as an increase in attention to, or recall of, the impending shock stimulus in the trace fear
paradigm. Indeed, this data is consistent with the enhanced LTP observed in the prefrontal region
of CaMKIV over-expressed mice (Wu et al., 2008).
While the hippocampus is known to be a fundamental structure exhibiting ~8Hz rhythms
(Buzsaki, 2002), we suspect that that the prefrontal 4-7.5Hz enhancement observed during the
trace interval does not necessarily correlate with the typical 8Hz CA1 rhythms. Moreover, we do
not think that the 4-7.5Hz signal recorded during the trace interval was not simply volume
conducted by the hippocampus because; the bipolar recording electrodes were placed close
together (0.3 to 0.5mm tip offset) so as to subtract common mode signals (Tsujimoto et al.,
2006). Since the hippocampal theta would be volume conducted from a distance, the signal
would be presumed to reach the bipolar electrodes. This common signal would then be
subtracted out by the inverting and non-inverting inputs of our amplifier. Thus we don’t expect a
meaningful level of volume conduction is a factor. Finally, cells demonstrating theta modulation
have been detected in the prelimbic region of the rat prefrontal cortex in response to spatial
learning tasks (Jones & Wilson, 2005; Siapas et al., 2005; Paz et al., 2008). Thus, our results
support the notions that the prefrontal 4-7.5Hz rhythms detected here originate in the prefrontal
cortex.
To examine the degree to which the 4-7.5Hz oscillations were related to learning,
correlation analyses were performed. A significant correlation was found between theta power
enhancements and learning, confirming that this brain wave frequency was indeed related to the
freezing behavior. Since, both freezing behavior and 4-7.5Hz power both demonstrate a learning
159
curve; it is reasonable to suggest that this 4-7.5 Hz frequency reflects some degree of memory
recall. Indeed, this is supported by the observation that these oscillations were localized to the
trace fear interval and not the shock period (Fig 3.7C).
3.7.5C CaMKIV over-expression enhances sleep after learning
Various studies have suggested that slow oscillations during sleep are related to memory
(Huber et al., 2004; Huber et al., 2006; Marshall et al., 2006; Vyazovskiy et al., 2008).
Moreover, brain derived neurotrophic factor (BDNF) injection in the cortex was found to
enhance local slow wave EEG activity in rats (Faraguna et al., 2008). There is also evidence that
slow delta oscillations are a consequence of GluR1 phoshorylation (Vyazovskiy et al., 2008) or
enhanced extracellular glutamate (Dash et al., 2009), possibly as a consequence basic
experiences during the day. We show that trace fear conditioning, which is enhanced by
CaMKIV over-expression, results in concomitant slow delta oscillation enhancement in the
cortex. Thus, it appears that CaMKIV, a key upstream molecule which phosphorylates CREB
(Enslen et al., 1994), can modulate the expression of slow delta oscillations as a consequence of
learning. Interestingly, slow delta oscillation enhancements could be localized to the anterior
cingulate cortex, predominating in layer I-III. Consistently, removal of the anterior cingulate
cortex impairs trace fear, but not delay fear conditioning (Han et al., 2003). In addition,
inactivation of the ACC, days after fear training, blocks fear memory retrieval (Frankland et al.,
2004). Also, it has recently been reported that neurons of the prefrontal cortex consolidate trace
eye-blink memory days after conditioning (Takehara-Nishiuchi & McNaughton, 2008). Thus,
CaMKIV mediated phosphorylation of CREB may be a major mediator of this consolidation, in
part through enhancement of local slow delta oscillations during NREM sleep.
160
It is possible that differences in post-conditioning sleep expression may be related to
differences in arousal during the training period, and not brain plasticity. However we did not
report any increase in CaMKIV mice above that of WT mice in high frequency EEG range
during the training trials (for frequencies greater than 13Hz). The most prominent change was an
increase in 4-7.5Hz EEG power and a decrease in 7.5-13Hz power (Fig. 3.7) relative to WT.
Indeed, all EEG responses to pain stimuli were identical between CaMKIV over-expressed and
WT mice. Moreover, we have previously shown that CaMKIV over-expressed mice have normal
responses to acute pain (Wu et al., 2008). Finally, we show that the slow wave enhancement
have some regional specificity to the ACC when compared to the motor cortex. This region is
thought to be involved in the gradual consolidation of memory during sleep (Takehara-Nishiuchi
& McNaughton, 2008).
3.7.5D Variations in learning and 4-7.5Hz oscillation power are related to memory formation
We found that CaMKIV over-expression enhances fear memory when studied ~24 hours
after training. While the elevations in 4-7.5Hz power during conditioning and slow delta power
(1-4Hz) during NREM sleep are seen in CaMKIV mice, suggestive of learning and memory
processes, the data remain merely suggestive. Thus, it was necessary to examine whether the
magnitude of changes in learning, 4-7.5Hz oscillations, and slow wave delta power actually
correlated with the magnitude of memory formation. To demonstrate that this was a valid test,
we first showed that the magnitude of learning and memory were correlated. We next
demonstrated that the change in 4-7.5Hz power during learning was significantly correlated with
subsequent memory. Interestingly learning and 4-7.5Hz r2 values were of identical magnitude but
were quite small. The result indicate that much of the freezing during the conditioning phase may
be related a general state of fear in that context. In future studies it may be useful to examine
161
how well trace fear conditioning predicts contextual fear memory expression. Interestingly, we
did not find any relationship between trace fear memory and changes in slow wave power. Based
on the findings we suggest that the best predictors of memory formation, at least for the trace
fear paradigm, are the degree of freezing and the change in 4-7.5Hz power during wakefulness.
This should draw attention to the importance of encoding information during wakefulness.
Indeed, a few studies have suggested that neural replay of past events even occurs in wakefulness
(Foster & Wilson, 2006; Karlsson & Frank, 2009). The current study does not discriminate
whether the observed oscillations during the trace interval reflect a point of memory
consolidation, recall or attention. However experiments examining post-training sleep
oscillations suffer from the same confound, for which there is no direct behavioral correlate. Unit
recording of neurons which demonstrate learning, combined with local pharmacological
manipulations of relevant plasticity pathways, may help resolve this issue. Taken together,
studies which examine the interactions of sleep and memory may benefit from examining brain
waves changes occurring during wakefulness in addition to brain wave changes during sleep.
One difference between the current study and other studies of learning and NREM sleep;
is that our paradigm impacts the stress levels of an animal, which may counteract the depth of
sleep. Thus we may be measuring two processes, one of stress induced sleep prevention and the
other of sleep potentiation. Indeed, this is why using the motor cortex as a control electrode,
when recording from the anterior cingulate is ideal. In such a situation, the null hypothesis would
be that both regions of cortex will respond similarly to trace fear conditioning. Irrespectively, we
found that slow wave activity in the ACC was potentiated above that of the motor cortex. Based
on the observations, the slow delta oscillation power enhancements seen in this study may be
related to either learning or memory or even a build up of extracellular glutamate left in the
162
synaptic cleft from the day’s events (Dash et al., 2009). Future experiments will have to clarify
its precise role.
163
CHAPTER 4
DISCUSSION AND FUTURE DIRECTIONS
164
4.1 Integrated discussion
Chapter 1 showed that a standard EMG recording was sufficient to score freezing
behaviour. In chapters 2 and 3 this method proved useful to examine trace fear memory. Its
utility lies in the fact that EMG-based freezing can be examined directly in parallel with brain
waves and brain cells. Moreover, this technique could be used in the dark (chapter 3) to examine
freezing behaviour during the natural wakefulness cycle of the animal. The main drawback of
using such a method is that it requires surgery and there were no automated analyses programs.
However, toward the end this thesis, I developed an automated fear scoring program, through
programming in the script language of the Spike2 software. A variety of other programs were
devised to analyze brain cell counts and EMG evoked potentials, all of which were based on
registering the recording of interest with that of the EMG.
Recording inside a shock chamber is more likely to produce 60Hz noise on neck
electrodes if the electrodes aren’t well impedance matched. The simplest way to fix this problem
was to ensure that both neck EMG electrodes were secured well to the neck muscles and that
their impedances are relatively similar before implantation. This increases the rejection of
common signals to both electrodes on the instrumentation amplifier (i.e. 60 Hz).
Another important point regarding technique is that bilateral neck recording should
theoretically be recording the same activity, and therefore produce a zero signal for the EMG.
This is one likely reason for the reduction in muscle tone during freezing behaviour. If both left
and right neck muscle contract similarly, and at the same time during freezing, the net effect of
the recording will be a flat signal. However, when the animal moves, one muscle is likely to
contract more than the other (as the head may turn left or right), in which case an increase in
muscle output is indicated. Regardless, this is the typical method to record EMG in rodents for
sleep studies and is sufficient for the experiments presented.
165
Chapter 2 shows that tone-evoked potentials in the ACC were a good predictor of future
freezing behaviour during trace fear conditioning. It is important to note that these potentials
were not likely related to change in motor movement, as neck EMG was not much changed by
the presentation of the tone. Indeed, neuron activity also appears to be related to the presentation
of the cue in the ACC. However, when an animal was found to be spontaneously freezing, we
found that neuron activity would become active just before and during the termination of
freezing. Again, neck EMG was used as a trigger point for this analysis to determine precisely
when the animal stopped freezing. If neck EMG analysis were not used, one might interpret the
ACC activity as being completely decoupled from motor output, being somehow related to the
rather nebulous concept of “affect”. Even if the concept of affect were used, motor recordings
would be helpful to provide a direction or trajectory of the animals intention.
It was found that the ACC could be activated by peripheral stimulation during anesthesia
experiments. Thus, the ACC is expected to process information from the periphery, at least
eventually. It remains to be shown which layers of the ACC receive what input and produce what
output. The experiments conducted in chapter 2 could not precisely locate the layers that were
recorded from, since the mouse brain is very small and a layer can be passed quite quickly. Thus,
future studies will need to identify layers which are important for motor and non-motor
functions. It would also be interesting to see whether graded electrical stimulation to the foot
could produce graded activation in the ACC in awake animals. For example, if 100uA current
were applied to the foot, it may not produce pain, but may cause “discomfort” in which case it
may not be sufficient to elicit motor response. By contrast, a higher level of electrical shock (e.g.
750uA) would produce pronounced behavioural activation (jumping or orienting toward the leg
that was shocked). Neurons could then be recorded across different layers of the ACC to see at
166
what intensity different networks are engaged. Moreover, in such a model, EMGs could be used
to quantify the degree of behavioural activation.
One of the major drawbacks to the experiments in chapter 2 is that this style of trace fear
paradigm does not permit repetitive training for one cell for more than 4 trials. The difficulty in
doing such experiments is that shocking the feet of the animal produces dramatic behavioural
activation, and may therefore result in loss of cell recording (for example if the animal bumps its
head). A change in design is likely needed. For example, it appears to be helpful to stimulate one
foot with a cuff electrode instead of stimulating all four feet. Reducing the shock intensity may
be helpful; however after surgery animals don’t appear to learn as well as do animals that have
never had surgery (unpublished observations) so reducing shock intensity may not be practical.
While chapter 1 and 2 highlight the periaqueductal gray as a possible site where the ACC
may exert control over the motor system, there is also evidence for the ACC connecting with the
striatum and motor cortices (chapter 2). Thus, connecting the ACC of the mouse with the
appropriate motor effectors may require careful examination using anterograde tracers in
combination with identified neuron hot-spots in the ACC.
That there was an overall decrease in pyramidal neuron activity during the trace fear
interval in the ACC comes as a surprise, especially because c-fos is elevated in the ACC during
trace fear conditioning (Han et al., 2003). However, we do show that the ACC neuronal activity
is increased at points where the animal is shocked and when the animal hears cue-related
information. It is possible that there are cell populations in the ACC which are activated during
the trace interval, but were not recorded. It is possible that these cells would be related to holding
memory online during the trace interval in expectation of shock. Based on the observation of the
electrode tracts, one may be more likely to encounter cells of this type near the midline of CG1.
167
In chapter 3 we show that genetic enhancement of CaMKIV also boosts 4-7.5Hz rhythms
during wakefulness and 1-4Hz rhythms during sleep. This finding is interesting because it is the
first to link a genetically enhanced “smart” mouse with cortical oscillations, and learning and
memory. However, the majority of the neurons recorded in the ACC showed depressed activity
during the trace fear interval coincident with increased 4-7.5Hz rhythms. Thus, we suggest that
the major structure in the preferential cortex exhibiting 4-7.5Hz oscillations is the prelimbic
cortex which tended to maintain these oscillations in parallel with fear behaviour. Interestingly
the ACC was found to have increased tone-evoked potentials which may account for the
potentiation of slow wave power in the ACC during NREM sleep. However, similar to the ACC,
the prelimbic cortex has also been shown to have enhancements in tone-evoked potentials with
fear learning (Mears et al., 2009). It remains to be shown whether the prelimbic cortex will
demonstrate enhancements in 1-4Hz potentials during sleep following trace fear conditioning.
Since the ACC is a site of remote memory storage (Frankland et al., 2004; Teixeira et al.,
2006; Ding et al., 2008), it would be of interest to examine whether slow waves in this area are
associated with remote memory formation or consolidation. However, it should be noted that the
ACC does not receive direct connections from the hippocampus (chapter 2), so a mechanism
involving the hippocampus in memory transfer to the cortex may first require the prelimbic
cortex which is known to receive a direct connection from CA1 (chapter 2).
The experiments in chapter 1 and 2 did not use delay fear as a control, nor did they use a
neutral tone as a cue control. Accordingly, the changes observed in field potentials, EEG and
brain cells are not conclusively related to processes that occur in the trace fear paradigm alone.
However, analyses of fear were often restricted to the trace interval where neither the tone nor
the shock is present. Thus, activity which occurs during this interval, whether it be field potential
or brain cells or EMG- freezing should represent that of trace fear conditioning. In any case the
168
addition of these controls to the experiments would be helpful to clarify what is specifically
related to trace fear and what could be related to simpler versions of fear conditioning.
4.2 Future Directions
Although the results of this thesis do not directly support the role of sleep and memory
processes, this field remains intriguing to me. This section will detail future directions I have
begun to take with this research.
4.2.1 Theme 1: Memory associations promoting wakefulness compete with sleep
Central question:
How can single memories keep us awake and does replay during wakefulness prevent
sleep?
Can we record individual neurons getting “sleepy”?
Rationale:
The ability to stay awake in anticipation of impending danger (e.g. pain ) or reward (e.g.
sex) has not been explored. Maintaining a wakefulness state in order to respond to danger (e.g.
trace fear memory), may be precipitated by memories which originate in the cortex.
The thalamocortical system is commonly thought to be involved in switching the cortex
to different states of wakefulness and sleep (Steriade, 2003); however, there is some suggestion
that cortiofugal projections outweigh that of corticothalamic projections (Deschenes et al., 1998).
Consistently, McCormick and von Krosigk (1992) demonstrated that stimulation of the
projection neurons from the cortex to the thalamus in-vitro, can change the thalamic bust mode
(reminiscent of sleep) to a transmission mode (reminiscent of wakefulness). In addition a recent
169
study has shown that antagonism of adenosine receptors in the prefrontal cortex can enhance the
time spent awake and reduce cortical synchronization (Van Dort et al., 2009). Thus, the top down
regulation of sleep has been demonstrated; however no model has examined whether memory
activation, related to the anticipation of a negative future event could maintain wakefulness in
spite of a competing drive to sleep. Since the trace fear memory paradigm involves making the
association of a shock and tone across a time span, it represents a unique opportunity to examine
competition between memory activation in anticipation of danger and the drive to sleep. Only
one study has shown that testing trace fear memory during NREM sleep in humans can reactivate
memory during sleep (Wamsley & Antrobus, 2009). However, a slight variation of this paradigm
might involve reactivation of the memory trace when an animal is sleepy (or falling asleep rather
than sleeping) to examine whether or not memory could compete with sleep drive. (I have done
some preliminary studies and it appears that reactivation of trace fear memory can help maintain
wakefulness when there is sleep pressure (compared to when shock and tone stimuli are
unpaired).
It has been postulated that long-range cortical-cortical connections are involved in the
maintenance of wakefulness (Massimini et al., 2005). Thus synaptic plasticity involving long
range connections may provide the strength to resist sleep or entrainment to slow oscillations. In
fear paradigms involving the pairing of a shock and tone, the auditory cortex (Quirk et al., 1997)
and prefrontal cortex will undergo plasticity (Baeg et al., 2001; Takehara-Nishiuchi &
McNaughton, 2008), as measured with brain cell recording. Indeed there is some evidence of
monosynaptic connections between these regions (Vaudano et al., 1991). If monosynaptic
cortico-cortical plasticity can oppose sleep, then auditory-frontal cortical connections in the trace
fear paradigm may be a suitable model to test this.
170
I hypothesize that cortical-cortical synaptic plasticity is likely to antagonize the drive to
sleep.
Experiments
1. Examine whether neurons involved in memory are resistant to gradual slow wave entrainment
during sleep onset as compared to other neurons.
2. Examine whether neuron pairs (prefrontal to auditory cortex or even hippocampus), which
show strong connections (cross-correlation) and are related to the memory of interest resist
entrainment to slow waves.
3. Attempt to boost memory pharmacologically or genetically and examine whether or not this
treatment impacts experiment 1 or 2.
4.2.2 Theme 2: The role of the cholinergic system in memory consolidation during sleep
Central question:
Is the cholinergic system permissive for memory consolidation during NREM sleep?
Rationale:
Neural activity is replayed in the hippocampus and prefrontal cortex during NREM sleep
(Wilson & McNaughton, 1994; Skaggs & McNaughton, 1996; Euston et al., 2007). In addition,
local manipulations, which induce synaptic potentiation in a specific cortical area, result in
higher slow wave activity in that area (Marshall et al., 2006; Hanlon et al., 2009), while
disruption of SWS impairs memory consolidation (JAeschbach et al., 2008). However, it is not
known whether local enhancements in slow waves are tied to alteration in neuron replay during
sleep. There is evidence that sharp wave ripple events in the hippocampus are related to neural
replay in the cortex (Wierzynski et al., 2009) and may be involved in the transition from the
171
down to the upstate in the neocortex (Battaglia et al., 2004). Since these findings are largely
correlative in nature, it may be of use to directly manipulate neural replay and slow waves to see
if they correspondingly change. Such manipulation may explain why the neural replay does not
persist for the whole night of sleep and is rather restricted to the first few hours.
Cholinergic neurons are crucial for synaptic plasticity, learning and sleep regulation and
their activity is high in waking and low during slow wave sleep (Jones, 1993; Vazquez &
Baghdoyan, 2001). In humans, enhancing cholinergic function during slow wave sleep via