MSc dissertation

Peripheral changes in respiratory chain genes from the mitochondrial

genome in Alzheimer's disease as a potential biomarker

David Robinson.

Supervisors: Dr. Angela Hodges and Dr. Aoife Keohane.

Department of Neuroscience, Institute of Psychiatry, King's College London, University of

London.

Project report in partial fulfilment for the degree of MSc in Neuroscience August 2012.

Statement of work

The following describes which individuals were responsible for the various aspects of the research.

Research design: Dr. Angela Hodges.

Diagnoses and sample collection: (see Lunnon et al., 2012).

RNA extraction: Dr. Katie Lunnon

cDNA synthesis: Dr. Katie Lunnon

Primer design: Dr. Angela Hodges, Phillip Mcguire, and David Robinson.

Production of standard solutions, including: PCR, gel electrophoresis, gel extraction, purification,

Nanodrop analyses of concentration and purity, and serial dilutions: Dr. Aoife Keohane, Dr. Katie

Lunnon, and David Robinson.

Production of standard curves: David Robinson.

qRT-PCR: David Robinson.

Statistical analyses: David Robinson.

2

Abstract

Alzheimer’s disease (AD) is a neurodegenerative disease that is predicted to affect as many as 90 million

people world-wide by 2040 (Ferri et al., 2005). The symptoms of AD include memory loss, psychosis,

depression, language defects, and general cognitive decline. Pathologically, it is characterised by

accumulations of extracellular amyloid plaques and intracellullar neurofibrilliary tangles of

hyperphosphorylated tau. Another prominent feature is mitochondrial dysfunction. Mitochondria contain

complexes that are responsible for oxidative phosphorylation (OXPHOS), the process via which

adenosine diphosphate (ADP) and inorganic phosphate (Pi) are used to produce adenosine triphosphate

(ATP) – considered to be the body’s universal energy currency. Mitochondria also contain their own

genome, which encodes 13 of the 88 subunits that constitute the OXPHOS complexes. The abundance of

OXPHOS transcripts encoded by mitochondrial DNA (mtDNA) is altered in AD brains (Chandrasekaran

et al., 1997; Aksenov et al., 1999; Manczak et al., 2004), as are OXPHOS transcripts encoded by nuclear

DNA in MCI and AD blood (Lunnon et al., 2012), indicating that peripheral responses are evident in

early AD. The current study used quantitative real-time PCR (qRT-PCR) to investigate whether a

selection of mtDNA-encoded OXPHOS transcripts (ND3, ND4, ND4L, ND5, and CYB) were

dysregulated in MCI and AD blood. These transcripts were found to be significantly more abundant in

MCI and AD relative to controls (p<.05). These results suggest that data regarding the abundance of

OXPHOS transcripts encoded by the mitochondrial genome might be used to develop biomarkers of early

AD. Biomarkers could be used to monitor responses to current or novel therapies, and to facilitate early

diagnosis, thereby capitalising on the effectiveness of treatments.

3

Contents

1. Introduction.

1.1. Alzheimer’s disease: symptoms, pathology, and genetics.

1.2. The need for early diagnosis – biomarkers.

1.3. Mitochondria and reactive oxygen species in Alzheimer’s disease.

1.4. The mitochondrial genome.

1.5. Mitochondrial transcription.

1.6. Post-transcription.

1.7. Aims.

2. Methods and materials.

2.1. Subjects and samples.

2.2. RNA extraction and cDNA synthesis.

2.3. Primer design.

2.4. Principles of qRT-PCR and production of standard solutions.

2.5. Testing primers.

2.6. Producing standard solutions.

2.7. Standard curves.

2.8. Assaying samples.

3. Results.

3.1. The effect of disease status on mRNA abundance.

3.2. Relative abundance of mRNA species.

3.3. Correlations in mRNA abundance.

4. Discussion.

4.1. mRNA abundance in blood and brain and the implications for biomarkers.

4.2. Factors conferring relative abundance of mRNA species.

4.3. mRNA correlations and division of the precursor transcript.

4.4. Conclusions.

5. Acknowledgements.

6. References.

7. Appendices.

4

1. Introduction

1.1. Alzheimer’s disease: symptoms, pathology, and genetics.

Alzheimer’s Disease (AD) is the most common form of dementia, estimated to affect as many as 30

million people world-wide, and predicted to affect 90 million by 2040 (Ferri et al., 2005). In addition to

memory defects and general cognitive decline, symptoms can include language deficits, depression, and

psychosis. Ultimately, it leads to death. Given the debilitating nature of AD, patients inevitably become

dependent on the assistance of others. It has been suggested that direct health and social care costs in the

UK are approximately £6 billion, whilst informal care might amount to an additional £8 billion. Costs are

predicted to rise with aging populations and the increases in prevelance that they entail (Lowin, Knapp,

and McCrone, 2001).

The hallmark pathological characteristics of AD are accumulations of extracellular amyloid plaques and

intracellullar neurofibrilliary tangles of hyperphosphorylated tau. Additional features include alterations

to synaptic plasticity, an increased abundance of reactive nitrogen and oxygen species,

neuroinflammation, and mitochondiral dysfunction, culminating in neuronal loss.

Early-onset AD is caused by mutations and duplications in the amyloid precursor protein (APP) gene

(Goate et al., 1991; Rovelet-Lecrux et al., 2006), and mutations in the presenilin 1 (PS1) and presenilin 2

(PS2) genes (Sherrington et al., 1995; Levy-Lehad et al., 1995), which are subunits of the gamma

secretase enzyme complex, which is involved in the cleavage of APP into beta amyloid (A) - found in

plaques. In sporadic, late-onset AD, which accounts for around 95% of cases (Kern and Behl, 2009), the

4 allelle of the apolipoprotein E (APOE) gene represents the largest risk factor (Corder et al., 1993).

Despite familial AD representing a small proportion of total AD cases, much weight has been given to the

postulation that A accumulations represent the start of sporadic pathogenesis, referred to as the ‘amyloid

cascade hypothesis’. At present, however, no comprehensive account of the aetiology is widely accepted.

5

1.2. The need of early diagnosis – biomarkers.

Whilst currently no treatments are proven to prevent or reverse the pathophysiological features of AD,

there exist a number of pharmacological, psychosocial, and behavioural interventions that target

symptoms. Generally, the effectiveness of these is increased by early administration (Osborn and

Saunders, 2010). For example, with drugs that reduce rates of decline, such as acetylcholinesterase

inhibitors, their eventual impact positively correlates with earlier use. However, it is thought that

neurodegeneration related to AD might occur several decades prior to the appearance of clinical features

(Morris et al., 1996), and, therefore, symptom-based diagnositc methods are unlikely to capitalise on the

potential efficacy of treatments. Likewise, an inability to recognise the presence of nascent AD is

prohibitive of early intervention trials, which, when using large samples, provide the best hope for

establishing the effectiveness of new treatments (Lovestone et al., 2009).

A suitable surrogate biomarker that can distinguish between AD cases and controls prior to clinical

manifestations, or between controls and those that are at increased risk of developing AD, would be of

value. People with mild pre-clinical symptoms or Mild Cognitive Impairment (MCI) have an increased

likelihood of developing AD (Mitchell, 2009), and represent a population in which biomarkers could be

sought. As many as 70% of those with MCI might have prodromal AD. Whilst a substantial amount of

research has investigated whether diagnoses of AD might be posssible by assessing levels of A and tau

in cerebrospinal fluid (CSF) (e.g., Maruyama et al., 2001), a biomarker should ideally be detectable

within a medium for which the logistics of acquisition and processing are efficient and well-established.

Lunnon and colleagues (2012), using microarrays, compared levels of blood RNA in controls to those

with MCI and AD. Approximately a quarter of probes suggested significant differential gene expression

between controls and MCI/AD patients, indicative that peripheral changes are detectable at early stages of

pathogenesis. A number of nuclear-encoded transcripts required for the mitochondrial complexes

responsible for oxidative phosphorylation (OXPHOS), in addition to transcripts for subunits of the core

mitochondrial ribosomal complex, were down-regulated. Earlier research has demonstrated that

expression of mitochondrially-encoded OXPHOS transcripts is also dysregulated in AD brains

6

(Chandrasekaran et al., 1997; Aksenov et al., 1999; Manczak et al., 2004). Assessing aberrant expression

of OXPHOS genes might provide a means of detecting pre-symptomatic AD or of monitoring responses

to treatments that target mitochondrial dysfunction in AD.

1.3. Mitochondria and reactive oxygen species.

Human cells contain multiple mitochondria, which are implicated in apoptotic signalling (Hengartner,

2000), calcium homeostasis (Rizutto, Bernadi, and Pozzan, 2000), and are the main cellular source of

adenosine triphosphate (ATP) – the body’s universal energy currency. ATP is generated by oxidative

phosphorylation (OXPHOS), which is orchestrated by five complexes embedded within the inner

mitochondrial membrane: NADH dehydrogenase (complex I), succinate dehydrogenase (complex II),

ubiquinol cytochrome c oxireductase (complex III), cytochrome c oxidase (complex IV), and ATP

synthase (complex V). In OXPHOS, electrons are transferred unidirectionally from one complex to

another, forming an electron transport chain (ETC). Electrons are donated from nicotinamide adenine

dinucleotide (NADH) and flavin adenine dinucleotide (FADH), which are produced in the mitochondrial

matrix by the Kreb’s cycle. As electrons are accepted, protons are transferred from the matrix to the

intermembrane space, producing a proton gradient, which is utilized by complex V to power the synthesis

of ATP from adenosine diphosphate (ADP) and inorganic phosphate (Pi).

Mitochondria have long been implicated in neurodegenerative diseases (e.g., AD, Parkinson’s disease,

Amyotrophic lateral sclerosis, Huntington’s disease). Dysregulation of OXPHOS produces reactive

oxygen species (ROS), which are capable of causing damage to proteins, lipids, and DNA when present at

high levels for prolonged periods, resulting in cell death. Neurons are especially susceptible to free radical

damage because they contain low levels of the antioxidant, glutathione, conferring upon them a decreased

capacity to remove oxygen free radicals (Christen, 2000).

1.4. The mitochondrial genome.

Due to its close proximity, a probable target of ROS generated by the ETC is mitochondrial DNA

(mtDNA). Indeed, its location, in addition to its lack of protective histones, is thought to contribute to the

high mutation rate in mtDNA relative to nuclear DNA (Maruszak and Żekanowski, 2011). MtDNA is

7

organised in a single, circular, double-stranded 16.5kb molecule, containing 13 protein-coding genes

(Figure 1), encoding subunits of the OXPHOS complexes (the remaining 75 OXPHOS subunits are

encoded by nuclear DNA), 2 ribosomal RNAs (rRNA), 12s RNA and 16s RNA, which constitute

components of the small and large mitoribosomal subunits, respectively (Sharma et al., 2003), and 22

transfer RNAs (tRNA), required for mitochondrial genome gene translation.

Damage to the mitochondrial genome is likely to entail the production of aberrant OXPHOS subunits or

dysregulation of translational processes. It is conceivable that there exists a ‘vicious cycle’, whereby

mtDNA mutations, via their impact upon the functionality of the OXPHOS system, increase ROS

abundance, increasing the likelihood of new mutations (Linnane et al., 1989). It is possible that the altered

levels of RNA related to OXPHOS observed in AD patients in several studies (Chandrasekaran et al.,

1997; Aksenov et al., 1999; Manczak et al., 2004; Lunnon et al., 2012) reflect alterations to

transcriptional, post-transcriptional, or translational processes that occur in response to this putative cycle.

With regard to its initiation in AD, it has been reported that A interacts with a number of mitochondrial

proteins, including A binding alcohol dehydrogenase (ABAD) and cyclophilin D (CypD), which

promotes ROS generation (Lustbader et al., 2004; Du and Yan, 2010).

8

Figure 1. The mitochondrial genome (modified from Kyriajouli et al., 2008). The mitochondrial genome encodes subunits

for 4 of the 5 OXPHOS complexes: NADH dehydrogenase (green), ubiquinol cytochrome c oxireductase (dark blue),

cytochrome c oxidase (orange), and ATP synthase (light blue). It also encodes 22 tRNAs (black) and 2 rRNAs (purple). The

location of the 3 promoter regions, HSP1, HSP2, and LSP1, are indicated towards the top-left (discussed below).

1.5. Mitochondrial transcription.

Reflecting its bacterial origins, transcription of the mitochondrial genome is largely polycistronic. It

consists of 3 promoter regions (Figure 1): 2 on the heavy strand – heavy strand promoter 1 (HSP1) and

heavy strand promoter 2 (HSP2) – and 1 on the light strand – light strand promoter 1 (LSP1). HSP1

transcription produces transcripts containing the 12s and 16s rRNAs, whilst HSP2 transcription produces

a transcript of the entire heavy strand, including 12 protein-coding genes (subunits of complex I: ND1,

ND2, ND3, ND4L, ND4, and ND5; a subunit of complex III: CYB; subunits of complex IV: CO1, CO2,

and CO3; and subunits of complex V: ATP6 and ATP8), both rRNAs, and 14 tRNAs. The LSP1

transcript incorporates the entire light strand, which includes the 1 remaining protein-coding gene (a

subunit of complex I: ND6) and 8 tRNAs (reviewed in Smits, Smeitink, and van den Huevel, 2010).

9

Transcription of the human mitochondrial genome is maintained by 3 nuclear-encoded components: an

RNA polymerase (POLRMT) and 2 mitochondrial transcription factors, TFAM and TFB2M (Litonin et

al., 2010). Transcripts are divided into their constituent genes by the endonucleolytic excision of tRNAs

by mitochondrial RNase P and TRNase Z at the 5’ and 3’ ends, respectively (Ojala, Montoya, and Attardi,

1981). Some genes are not separated by tRNAs and, therefore, are translated from a single transcript: (1)

ATP8/ATP6 (overlapping genes) and CO3, (2) ND4L/ND4 (overlapping genes), and (3) ND5 and CYB.

1.6. Post transcription.

Human mitochondrial mRNA lack 5’ untranslated regions (Montoya, Ojala, and Attardi, 1981), so there

is limited potential for regulation of translation by gene-specific initiation factors. Translation is likely to

be mediated by steady-state levels of mRNA, which are determined by RNA stability and turnover

(Piechota et al., 2006). A small number of factors have, however, been identified, including a translational

activator of CO1 (TACO1) (Weraapachai et al., 2009).

Microdeletions affecting the stop codon of ND3 in a patient presenting with mtDNA disease resulted in

decreased levels of the transcript, suggesting that surveillance mechanisms coupled to translation promote

degradation of transcripts containing aberrant stop codons (Temperley et al., 2003). Turnover of ND3,

ND2, and CYB is faster than for other transcripts. Each of these genes only contain the first nucleotide of

their stop codons – the other 2 being added during adenylation. Conceivably, their stop codons are more

likely to be affected by the imperfect excision of tRNAs from the precursor. Increased stability of more

abundant transcripts (i.e., ATP6/8, CO2) might be conferred by secondary structures or by interactions

with stabilising proteins that are mediated by levels of OXPHOS (Leary et al., 1998; Piechota et al.,

2006).

1.7. Aims.

The current study employed quantitative real-time PCR (qRT-PCR) to investigate whether levels of

mitochondrial mRNA in blood differs between controls, those with MCI, and those with AD. As previous

research has observed altered levels of nuclear-encoded OXPHOS-related transcripts in blood in those

with MCI and AD (Lunnon et al., 2012), it was anticipated that abundance of transcripts encoded by the

10

mitochondrial genome would also be altered. Given the high levels of conversion from MCI to AD,

identification of abnormalities in the former might be of use in the development of an early AD diagnostic

biomarker. Also, assuming that altered levels reflect disease processes, assessing abundance of transcripts

might provide a means of evaluating the efficacy of drugs that target pathological features of AD.

Insights into the extent to which transcripts encoded by mtDNA correlate were also sought, as well as

information regarding the relative abundance of mtDNA-derived transcripts. Knowledge of these would

provide a basis for assessing current ideas regarding the nature of mitochondrial transcription and

turnover of the resultant mRNA. Understanding of these processes is of value for interpreting why

transcripts are more or less abundant in disease, including AD.

2. Methods and Materials

11

2.1. Subjects and samples.

Blood samples (n = 437) were taken from individuals participating in the AddNeuroMed study

(Lovestone et al., 2009) (Table 1). Participants were from Kuopio, Lodz, London, Perugia, Thessaloniki,

and Toulouse. Diagnoses of probable AD were made using the National Institute of Neurological and

Communicative Disorders and Stroke and Alzheimer’s Disease and Related Disorders Association

Alzheimer’s criteria (McKhann et al., 1984) and the Diagnostic and Statistical Manual of Mental

Disorders (American Psychiatric Association, 2000). Diagnoses for MCI were made in accordance to the

Petersen’s criteria of amnestic MCI and scores on the Total Clinical Dementia Rating Scale (CDR).

Participants were excluded from the study if they were under 65 years of age, or if they suffered from any

other psychiatric or significant non-psychiatric illness, or depression. Further details of the diagnostic and

recruitment procedures are described by Lunnon and colleagues (2012). Standardized operating

procedures for sample collection and subject assessment were followed at all locations. Consent was

obtained in accordance to the declaration of Helsinki (1991) and ethical approval was obtained at all

locations.

Table 1. Age, gender, location, and number of APOE 4 alleles by disease status.

Control (n = 162) MCI (n = 134) AD (n = 141)

Mean age (years) 72.99 74.73 76.56

Number of each gender

Males: 66

Females: 96

Males: 59

Females: 75

Males: 46

Females: 95

Number from each location

Kuopio: 41

Lodz: 14

London: 47

Perugia: 41

Thessaloniki: 15

Toulouse: 4

Kuopio: 28

Lodz: 18

London: 19

Perugia: 43

Thessaloniki: 16

Toulouse: 10

Kuopio: 37

Lodz: 28

London: 14

Perugia: 36

Thessaloniki: 21

Toulouse: 5

Number of APOE 4 alleles

0: 114

1: 44

2: 4

0: 81

1: 47

2: 6

0: 60

1: 62

2: 19

2.2. RNA extraction and cDNA synthesis.

12

Prior to the current study, 2.5ml blood samples were collected in PAXgene blood RNA vacutainer tubes

(BD Diagnostics) and RNA was extracted using the PAXgene bloodRNAkit (Qiagen), according to the

manufacturer’s protocol. A 2100 Bioanalyser (Agilent Technologies) was used to assess RNA integrity.

cDNA was synthesized using 250ng of total RNA using the QuantiTect Reverse Transcription Kit

(Qiagen) according to the manufacturer’s protocol from samples with RNA integrity numbers >7.

2.3. Primer design.

Primers were designed using Primer 3 (Table 2). The National Center for Biotechnology Information’s

(NCBI) Basic Local Alignment Search Tool (BLAST) was used to check for homology between target

and non-target sequences. Primers were redesigned if they exhibited complementarity for a given non-

target RNA, contained any single nucleotide polymorphisms (SNP), particuarly at the 5’ end, produced

more than 1 product, or an incorrect product (verified by sequencing), or could not be optimised to

produce a standard curve of the correct efficiency (see below). Primers were purchased from Sigma

Aldrich.

Table 2. Primers used in qRT-PCR to quantify levels of mitochondrial transcripts.

Target

produc

t

Forward primer

5’ 3’

Reverse primer

5’

3’

Produc

t size

ND3

ND4L

ND4

ND5

CYB*

TTACGAGTGCGGCTTCGAC

C

TAGTATATCGCTCACACCTC

CTAGGCTCACTAAACATTC

TA

TCGAATAATTCTTCTCACCC

TATCCGCCATCCCATACATT

CCTAAGTCTGGCCTATGAGT

CACATATGGCCTAGACTAC

CGCAGTACTCTTAAAACTA

GG

CGCAGGATTTCTCATTACTA

ACAACCCCCTAGGAATCAC

C

209bp

209bp

206bp

137bp

188bp

*Designed during the current project.

13

2.4. Principles of qRT-PCR and production of standard solutions.

qRT-PCR utilises the exponential amplification of products by the PCR to quantify the original

abundance of a product (e.g., cDNA). It entails the use of dyes that emit a fluorescent signal, though only

when in the presence of double-stranded DNA (dsDNA). Therefore, as the level of a dsDNA product

doubles with every cycle of the reaction, the strength of the signal does too. There is, then, a negative

correlation between the number of cycles that are required for the signal to reach a given threshold –

referred to as the ‘cycle threshold’ (CT) value - and initial levels of the product. Before assaying samples,

it is necessary to establish CT values for solutions that contain a known number of target molecules (i.e.,

standard solutions). Subsequently, it is possible to estimate the quantity of target molecules in a sample

based on the CT value.

2.5. Testing primers.

Standard PCR, followed by gel electrophoresis, was used to test primers. For each set of primers, a 10μl

solution, containing 2μl 5x HOT FIREPol® EvaGreen® qPCR Mix Plus (ROX) (Solis Biodyne), 2μl

diluted cDNA, 1μl forward primer (10μM), 1μl reverse primer (10μM), and 4μl nuclease-free water, was

made up in a single well of a 384-well plate. No-template controls (NTC), in which cDNA was replaced

by nuclease-free water, were also produced for each set of primers to make ongoing checks for

contamination. All reagents (except the nuclease-free water), having been kept at -20C, were thawed on

ice, then vortexed (~5s) and centrifuged (~5s), before being aliquoted. An optical adhesive film

(Microamp) was used to cover the plate after reagents had been aliqoted. Plates were centrifuged for 1min

at 2000RPM before reactions were run.

Reactions were run in a 7900HT fast RT-PCR machine (Applied Biosystems) in conjunction with SDS

2.3 software (Applied Biosystems). In order to activate the DNA polymerase, plates were firstly heated to

95C for 15min. Next, 40 cycles of the following steps were run: 95C for 15s (denaturation), 60C for 20s

(annealing), 72C for 20s (elongation). Finally, to check for homegeneity of PCR products, a dissociation

14

stage was run in which the temperature was raised to 95C for 15s, then lowered to 60C for 15s. A

manual Ct threshold of 0.2 was used.

PCRs were then run in an agarose gel to confirm whether the primers had worked. A 1% gel was created

by melting 1g of agarose powder (Invitrogen) with 100ml of 1X TAE buffer in a microwave (~2min).

Before it had set, 5μl of GelRed (Biotium) was added to the gel, which was poured into a casting tray

containing a comb with 8 wells. 1μl of 10X BlueJuice™ Gel Loading Buffer (Invitrogen) was added to

both the 10μl PCR and the 10μl NTC, before aliquoting them into separate wells in the gel. 10μl of 100bp

DNA ladder (Invitrogen) was aliquoted into a well adjacent to the PCR to allow for sizing of the PCR

product. Gels were run at 100V for ~1.5h.

Gels were inspected to see whether they contained a band in the column containing the PCR and to check

that bands were absent in the column containing the NTC. Bands in the former were compared to the

100bp ladder to confirm that the PCR product was of the expected size.

2.6. Producing standard solutions.

Having established that a set of primers worked, standard PCR was used to amplify cDNA for use in

standard solutions. A solution of 144μl, containing 36μl 5x HOT FIREPol® EvaGreen® qPCR Mix Plus

(ROX), 36μl diluted cDNA, 18μl forward primer (10μM), 18μl reverse primer (10μM), and 72μl

nuclease-free water, was aliquoted into a 384-well plate, divided across 8 wells, each containing 18μl of

the solution. The rest of the procedure for setting up and running the PCR was as described above. Gels

were also made, run, and inspected as described above, with the exception that larger gels were used,

containing 2g of agarose powder and 200ml 1X TAE buffer.

PCRs were purifed using a MinElute Gel Extraction Kit (Qiagen), followed by further purification and

concentrating using a MinElute PCR Purification Kit (Qiagen) according to the manufacturer’s protocols.

DNA concentration was assessed using a Nanodrop 1000 (Theremo Scientific) and was used to calculate

the volume needed for making a stock containing 1010 copies/1l (referred to as ‘standard 10’). All other

standard solutions (9-1) were made by a 1:10 serial dilution (i.e., standard solution 9, theoretically,

contains 109 copies/1l).

15

2.7. Standard curves.

Standard curves were tested for all genes prior to assaying samples. 4l of a solution containing 0.5μl

forward primer (10μM), 0.5μl reverse primer (10μM), 2μl nuclease-free water, and 1μl HOT FIREPol®

EvaGreen® qPCR Mix Plus (ROX), was aliquoted using an electronic pipette into 22 wells of a 384-well

plate. Next, 1l of each standard solution was aliquoted using a multi-pipette into wells containing the

solution described above. Duplicates of each standard were assayed and a NTC (also duplicated) was

included in each plate. The rest of the procedure for setting up and running the PCR was as described

above.

Data were subsequently analysed in Microsoft Excel. For each gene, the mean CT values of duplicates for

each standard solution were calculated and used to generate a curve and a corresponding linear regression

equation. The coefficient from this equation was entered into an online efficiency calculator (Agilent

Technologies). Efficiencies between 90% and 110% were considered acceptable as they suggest that the

PCR product approximately doubled with each cycle of the reaction. Standards that produced values

outside of this range required futher optimisation and so were re-made, or assayed again having re-

designed primers. The differences between CT scores for duplicates were also examined; standard curves

were re-tested in instances where duplicates were in excess of one CT score from each other.

2.8. Assaying samples.

Samples were assayed in random order and the identities of the individuals from whom they came were

not known to the researcher. Standard solutions were assayed alongside samples in each plate as

described above. Samples, having been stored at -80C, were thawed on ice, vortexed (~5s), and

centrifuged (~5s), before aliquoting. cDNA was made prior to the current study and diluted 1:5. 1l of

this was aliquoted using a multi-pipette into wells containing 4l of the solution described above (i.e.,

primers, dye mix, and nuclease-free water). As with standard solutions, each sample was duplicated; if

CT values for duplicates were in excess of one cycle apart, the samples were assayed again. Reactions

were set up and run as described above.

16

3. Results.

3.1. The effect of disease status on mRNA abundance.

Data were normalized to two house keeping genes, measured in the same samples, selected following

analysis of a panel of geNorm reference genes (Primer Design Limited), to account for technical

variations (e.g., pipetting errors, plate variation, etc.): ATP5B and SF3A1 (previously found to be the

most stable in these samples). To achieve normal distributions, data were log10 transformed and samples

with estimates in excess of 2 standard deviations from the overall mean for a given gene were removed

from the analysis. Samples were also removed if CT scores for duplicates were consistently over 1 apart

from each other, or if demographic information was missing. Normality was achieved for the majority of

data when categorised by gene and status (Appendix 1). All analyses were completed in SPSS (IBM).

Analyses of covariance (ANCOVA) were conducted to establish whether disease status (AD, MCI, or

control) could account for variations in RNA abundance for each gene, whilst controlling for age, gender,

number of APOE 4 alleles, and the location of the centre at which samples were taken. There were

significant effects of status on estimates for ND4, F(2, 430) = 9.354, p<.001, ND4L, F(2, 430) = 14.037,

p<.001, ND5, F(2, 430) = 19.382, p<.001, and CYB, F(2,430) = 33.651, p<.001. There was no such effect

for ND3, F(2, 430) = .643, p>.05. For all genes, effects of covariates were non-significant, except for

gender on ND4 (p<.05) and centre location on ND4L (p<.05).

Levene’s Test of Equality of Error Variances indicated equal variance between status groups in all genes

except ND3. Also, whilst kurtosis tests suggest that ND3 estimates resembled other genes, the ND3 data

appeared to be positively skewed to a greater extent than estimates for other genes (Table 3). A Kruskal-

Wallis non-parametric analysis of variance (ANOVA), however, also indicated that differences between

status groups for ND3 estimates were non-significant (p>.05), as did an ANCOVA, with the above

covariates, performed on square root transformed data (p>.05).

Table 3. Skewness and Kurtosis

ND3 ND4 ND4L ND5 CYBSkewness .431 -.195 .341 .071 -.150Kurtosis .123 .058 -.291 -.806 -.283

17

Bonferroni tests were conducted to establish which status groups differed within individual genes. There

were no significant differences between status groups for ND3 (p>.05). For all other genes, estimates for

the MCI group were significantly higher than for the control group (p<.05 for ND4L; p<.001 for all

others), as were estimates for the AD group (p<.001). There were no significant differences between

estimates for MCI and AD groups for any genes (p>.05). On average, there was a 1.46 fold increase in

estimates of mRNA abundance between control and MCI groups, and a 1.72 fold increase between

control and AD groups (Table 4).

Table 4. Changes in abundance of mRNA from genes of the mitochondrial genome in MCI and AD groups relative to

controls.

Mean MCI Mean AD

ND3 1.39 1.6

ND4 1.21 1.03

ND4L 1.6 2.7

ND5 1.92 2.2

CYB 1.19 1.06

3.2. Relative abundance of mRNA species.

A one-way repeated ANOVA revealed a significant effect of gene on mRNA estimates, F(3.39, 1478.8) =

16786.23, p<.001 (Figure 2). Ratios of estimates for all genes compared to all others were also calculated,

based on pre-log10-transformed data (Table 5). Estimates for ND4 mRNA abundance were highest,

followed, in descending order, by ND4L, ND3, ND5, and CYB.

18

Figure 2. Estimated marginal means for ND3 (1), ND4 (2), ND4L (3), ND5 (4), and CYB (5), based on log10

transformed data.

Table 5. Ratios of estimates of mRNA abundance for all genes (genes on the left are listed first).

ND3 ND4 ND4L ND5 CYB

ND3 1:241.1854 1:1.3773 1:0.4087 1:0.0983

ND4 1:0.0041 1:0.0057 1:0.0017 1:0.0004

ND4L 1:0.7261 1:175.1169 1:0.2968 1:0.0714

ND5 1:2.4465 1:590.0643 1:3.3695 1:0.2405

CYB 1:10.1723 1:2453.3983

1:14.0101 1:4.1578

A potential source of variation between estimates for different genes is variation in PCR product

concentrations between standards for different genes. These variations are probably attributable to

pipetting errors made whilst making the standard 10 solutions. To investigate this, mean CT scores,

calculated from CT scores from individual plates, were produced for standard 4 for each gene (standard 4

is present in most standard curves and typically produces CT scores that resemble sample CT scores). A

Pearson product-moment coefficient was computed to assess the correlation between standard 4 means

19

and mean mRNA abundance estimates derived from log10 transformed data. There was a significant

positive correlation between the two variables, r = .851, n = 5, p<.05 (Figure 3), with standard 4 means

calculated to explain ~72% of the variance.

Figure 3. Correlation between standard 4 CT means and estimates of mRNA abundance.

From left to right: CYB, ND4L, ND3, ND5, ND4.

3.3. Correlations in mRNA abundance.

Pearson product-moment coefficients were computed to assess the correlation between estimates for all

genes. There were significant positive correlations between all genes (p<.001). Pearson product-moment

coefficients were also produced for all genes having split the data according to status. Within the control

group, there were significant positive correlations between the majority of estimates. ND3 estimates,

however, were not significantly correlated with any other gene (p>.05), except ND4L (p<.001) (Appendix

2). Within the MCI group, there were significant positive correlations between all estimates (Appendix 3).

Within the AD group, there were significant positive correlations between the majority of estimates. ND4

20

estimates, however, were not significantly correlated with ND4L or ND5 (p>.05) (Appendix 4). Since

ND3 estimates did not exhibit homogeneity of variance (HOV), Spearman’s rank correlation coefficients

were also produced. The results resembled those described above, with the exception that, within the

control group, ND3 estimates were significantly, positively correlated with all other genes (Appendix 4).

4. Discussion

21

4.1. Abundance of mRNA by status groups.

The ANCOVA analyses indicate that, controlling for covariates, levels of transcripts for ND4, ND4L,

ND5, and CYB are significantly increased in blood in MCI and AD populations relative to controls.

Differences between MCI and AD groups, however, were not statistically significant. No differences

between any groups for ND3 estimates were significant. The fold changes for ND3 transcripts for MCI

and AD groups relative to controls do not appear to be markedly different to those for the other genes.

Failing to reach significance might be attributable to the distribution of ND3 data, which does not exhibit

HOV, as evidenced by the Levene’s test. A non-parametric ANOVA, however, also failed to produce a

significant result, as did an ANCOVA performed on square root transformed data. It is possible that the

removal of samples in excess of 2 standard deviations from the mean for each status group (i.e., not just

relative to the overall gene means) would produce HOV, though this would entail reductions in power,

which might affect whether other genes achieve significance.

Manczak and colleagues (2004) investigated differences in 11 transcripts encoded by the mitochondrial

genome in frontal cortex brain specimens from controls, and in those with early and definite AD. In

contrast to the current findings, they found that complex I (NADH dehydrogenase) subunit transcripts

were generally reduced in early and definite AD – an occurrence that has also been observed in the

hippocampus and inferior parietal lobule (Aksenov et al., 1999). All other transcripts, however, tended to

be more abundant. They suggest that the decreases in complex I transcripts might be attributable to an

increased susceptibility of complex I genes to mutation, which might mean that they are targeted for

degradation. It is possible that the brain, as an environment, entails a higher likelihood of mutations to

mtDNA than blood, which might explain the incongruence between findings. However, they also reported

an increase in abundance of ND6 (also a complex I subunit) in both AD groups, suggesting that

explanations need to be gene-specific. Moreover, whilst degradative mechanisms that target aberrant

mRNA might account, to some extent, for differences, the increased abundance of ND6, which is

transcribed separately from the other genes, implies that transcriptional processes are relevant.

A number of nuclear-encoded OXPHOS genes and components of the core mitochondrial ribosome

22

complex are down-regulated in blood in MCI and AD (Lunnon et al., 2012). A shortage of translational

machinery might explain the increases in transcripts in the current study, since translation-coupled

degradation would be reduced. It is also conceivable that the increases reflect a compensatory response,

whereby transcription is up-regulated in response to decreases in OXPHOS resulting from the down-

regulation of the nuclear-encoded genes. However, this seems unlikely given that the ETC depends on the

correct functioning of all complexes, presumably rendering the up-regulation of only a small set of

subunits ineffective.

An alternative explanation is that the respective down- and up-regulations of nuclear and mtDNA genes

both occur in response to complex I mutations, whereby nuclear genes produce fewer subunits to match

the reduced number of functional complex I subunits encoded by mtDNA, and transcription of mtDNA is

up-regulated to compensate for the dysfunctional subunits. There are, however, are at least 2 potential

problems with this suggestion. Firstly, if mutations to complex I genes cause an up-regulation of mtDNA

transcription, this would not necessarily be reflected in steady-state levels, since aberrant transcripts are

more likely to be targeted for degradation (Temperley et al., 2003; Piechota et al., 2006). Secondly, since

it has been suggested that mutations to complex I genes in mtDNA are more likely to occur in the brain

than in blood, it would be expected that the increases in complex I mRNA abundance observed in blood

would be exaggerated in brain, though, in actuality, their presence is decreased (Aksenov et al., 1999;

Manczak et al., 2004). Regarding the first of these issues, since nuclear-encoded mitochondrial

translational components are down-regulated, degradation would also be reduced, allowing for mRNA

accumulation. Regarding the second point, it is conceivable that there exists a threshold in the level of

mutations beyond which transcription cannot occur, which would explain the up- and down-regulations

observed in blood and brain, respectively.

The initiation of these processes in AD might be related to interactions, which promote ROS generation in

mitochondria, between A and certain mitochondrial proteins, including ABAD and CypD (Lustbader et

al., 2004; Du and Yan, 2010). Factors that confer upon the brain a decreased capacity to remove ROS,

such as the lack of glutathione in neurons (Christen, 2000), might result in the majority of mtDNA

23

molecules, in brain regions affected in AD, harbouring a mutation level that exceeds the putative

threshold, whilst in blood such levels are not reached.

According to this interpretation of the seemingly incongruent observations, made between blood and

brain, of mtDNA-derived mRNA abundance, both reflect the same disease response. Conceivably, then,

assessing abundance of these mRNA species in blood could provide a means of judging AD progression

in brain. On one hand, this may be of value in evaluating the efficacy of novel drugs that target

pathological features of AD, in which case a reduction in these mRNA would be anticipated. On the other

hand, assessing mtDNA-derived mRNA might provide a means for early diagnosis of AD. Detecting AD

before symptoms emerge would allow for the application of existing interventions sooner than is possible

with symptom-based diagnosistic methods, thereby capitalising on their potential. Early diagnosis would

also allow for early-intervention trials.

Future reseach should assess the abundance of other mRNA species encoded by mtDNA, in blood in MCI

and AD, to compare how abundance of these transcripts relates to those assayed here and to their

equivalents in brain (Chandrasekaran et al., 1997; Aksenov., 1999; Manczak et al., 2004). Patterns of

abundance of mtDNA-encoded transcripts in AD might resemble those of other neurodegenerative

diseases that exhibit mitochondrial dysfunction, and, therefore, information about a range of mRNA

species is likely to be needed to distinguish early AD from other diseases. Once data is available for all

mtDNA genes, classifiers should be tested to assess the degree of accuracy that can be achieved with this

data. It would also be interesting to compare their performance to other, more invasive, diagnostic

methods, such as the analysis of CSF for A or tau, which depends on lumbar punctures.

4.2. Abundance of mRNA between genes.

There are large differences between estimates for some transcripts and others. ND4 mRNA, for example,

is estimated to be ~2500 times more abundant than that of CYB . Other estimates are characterised by

similarity; ND4L transcripts are estimated to be only ~1.3 times more abundant than ND3 transcripts.

Variations in the concentrations of PCR products in standards between genes are likely to account for a

substantial amount of the variation in estimates. It is probable that the concentration of the ND4 PCR

24

product, for instance, is reduced in the standards, leading to higher CT scores for standards, and higher

estimates for samples.

Whilst variation in standard concentrations make it difficult to produce accurate estimates, it is possible to

attempt to rank estimates for different genes. For instance, since the mean estimate for ND3 exceeds that

of ND5, despite ND3 having a lower mean for standard 4 CT scores, it is likely that ND3 transcripts are

more abundant than those of ND5. The same logic can be applied to infer that ND4L transcripts are

probably more abundant than those of ND3, and, therefore, of ND5. More generally, it is possible to get

an idea of their ranking, whilst controlling for standard concentrations, by looking at the graph and

observing the extent to which a dot falls below or above the fit line. For example, whilst ND3 and ND5

are both below the line, the distance is greater for ND5, suggesting that ND3 transcripts are, in actuality,

in greater abundance. According to this approach, the genes ranked in ascending order of transcript

abundance are: ND5, ND3, ND4L, CYB, and ND4. Though adopting an approach such as this might, to

an extent, help to control for standard PCR product concentration between genes, it would be preferable

to make ongoing checks of standard CT score variations to address the issue prior to analysis.

Assuming the ranking described immediately above is accurate, some of the relative levels are

incongruent with other accounts of mitochondrial mRNA relative abundance. For example, the model

proposed by Temperley and colleagues (2003) would suggest that ND3 and CYB transcripts would be

least abundant because of their susceptibility to containing aberrant stop codons, which make them targets

for translation-coupled degradation. Previous research has since demonstrated that these transcripts, in

addition to ND2, are less stable than other mtDNA-derived transcripts in HeLa cells (Piechota et al.,

2006). Conceivably, translation-coupled degradation of aberrant transcripts has less influence in MCI and

AD because of a down-regulation of mitochondrial translational components, though the control group do

not appear to exhibit patterns of relative abundance that diverge from the overall pattern. The fact that

levels of ND3 and CYB transcripts are not the lowest imply that factors other than the stop codon position

also influence gene-specific turnover rates.

Prior research has shown that ATP6/8 and CO2 transcripts are more abundant than other mtDNA-derived

25

transcripts in HeLa cells, which might be related to their secondary structures or the presence of

stabilising proteins. Half-lives of these transcripts are reduced in the presence of ethidium bromide, which

is thought to interfere with secondary structures and mRNA-protein interactions (Piechota et al., 2006).

The binding of such proteins might be increased by reductions in OXPHOS as a way of preventing ATP

deficiency, as evidenced by the stabilisation of the CO2 transcript when complex IV (cytochrome c

oxidase) is inhibited by sodium azide (Leary et al., 1998). Differential abundance might also be conferred

by the presence of gene-specific translation initiation factors, whereby an increase in a specific factor

would promote turnover of its associated transcript. However, whilst initiation factors, such as TACO1

(Weraarpachai et al., 2009), have been identified in humans, there is limited potential for their

involvement in translation-coupled degradation since human mtDNA-derived transcripts lack 5’

untranslated regions (Montoya, Ojala, and Attardi, 1981).

It is apparent that a number of factors are likely to determine relative abundance of mRNA species

encoded by the mitochondrial genome, and that there is probably a complex set of interactions between

these factors. Moreover, the influence of these factors conceivably varies between tissue types, which

might partly explain the incongruence between the current findings and those of Piechota and colleagues

(2006). Differential abundance of OXPHOS transcripts is an interesting occurrence as it would appear

that the correct functioning of the complexes would depend on the presence of all subunits. One

explanation, of course, is that this is not true, and that some subunits are more essential than others, and

so measures are taken to enhance their stability. The relative importance of subunits could be investigated

by monitoring mitochondrial metabolic rates whilst silencing expression of individual OXPHOS genes

with, for example, small interfering RNA (siRNA). Understanding the relative importance of subunits, as

well as how their levels are regulated, will enrich interpretations of mitochondrial dysregulation in

neurodegenerative disease.

4.3. Gene correlations.

When not separated by status, all gene estimates exhibited significant positive correlation with all others.

This is expected given that all of the genes assayed are transcribed into 1 polycistronic transcript. Within

26

the control group, ND3 was not found to be significantly correlated with any other genes, except ND4L,

when using a parametric test. Using a non-parametric test, however, returned significant results. Whilst

all estimates were significantly, positively correlated within the MCI group, ND4 was not significantly

correlated with ND4L and ND5 in the AD group, showing a weak, negative correlation with the former.

Since ND4 and ND4L overlap, this occurrence is unexpected. An inspection of the scatter plot (Appendix

6), however, reveals that a large number of the samples occupy a cluster that is indicative of a positive

correlation between the 2 genes, and that the coefficient will have been influenced by a smaller number of

samples that deviate extensively from this group. Nonetheless, the lack of an overall linear relationship is

surprising. A possible explanation is that degradation of cDNA occurred in the samples between the times

at which each of the genes were assayed, though this would have had to have affected samples

differentially for linearity to be compromised.

Across status groups, the strongest positive correlations were between estimates for ND5 and CYB,

which is expected given that they are not separated by a tRNA-encoding gene, and, therefore, occupy a

bicistronic transcript (Ojala, Montoya, and Attardi, 1981). Thus, it is unlikely that they would be

differentially affected by transcription or mRNA degradation. It would be interesting to assay CO3 and

ATP6/8 transcripts, which are also thought to be contained within a bicistronic molecule. Presumably, the

strength of the correlation between these 2 genes would closely resemble that of the correlation between

ND5 and CYB, though any differences might indicate the presence of degradative or stabilising

mechanisms that exhibit specificity for individual transcripts contained within a bicistronic molecule.

4.4. Conclusions.

It has been demonstrated that a number of transcripts encoded by the mitochondrial genome, namely

those derived from ND4, ND4L, ND5, and CYB, are significantly more abundant in blood in MCI and

AD. Though this finding contrasts to observations made in the brain (Aksenov., 1999; Manczak et al.,

2004), a model has been proposed that reconciles these differences, and which considers them to be part

of the same disease response. Consequently, it has been suggested that abundance of these mRNA species

might be of value for early diagnosis of AD, or for monitoring responses to novel treatments. Future

27

research should obtain data regarding abundance of the other transcripts of the mitochondrial genome in

blood in MCI and AD, and investigate whether classifiers can be trained to distiguish AD from controls,

and from other neurodegenerative diseases characterised by mitochondrial dysfunction. Their

performance should be assessed in relation to classifiers that have been trained with data derived from

more invasive procedures (e.g., lumbar puncture).

Additionally, the relative abundance of gene transcripts has been investigated to assess existing accounts

of the processes that confer relative mRNA abundance. Some authors have argued that the location of a

gene’s stop codon increases its susceptibility to produce aberrant transcripts, thereby increasing their rate

of turnover by surveillance mechanisms that are coupled to translation (Temperley et al., 2003). However,

2 of the gene transcripts, ND3 and CYB, thought to be especially susceptible to aberrant expression, were

not the least abundant transcripts in the current study, emphasising the need to produce models of mRNA

turnover that take into account a range of factors, including secondary structures, stabilising proteins,

levels of OXPHOS, translation initiation factors, and tissue type. Moreover, such models will have to

consider the interactions between these factors. In doing so, researchers will be better placed to interpret

mitochondrial dysfunction in neurodegenerative disease.

The gene correlation data support the idea that transcription of the mitochodrial genome is a generally

uniform process, whereby initiation from the HSP2 produces a polycistronic transcript containing 12 of

the 13 protein-coding genes. The especially strong correlation between 2 of the genes, CYB and ND5,

also lends support to the tRNA punctuation model of RNA processing in human mitochondria (Ojala,

Montoya, and Attardi, 1981). In addition to the processes of stabilisation and degradation described

above, the nature of mitochondrial transcription is, of course, also a key determinant of expression, the

understanding of which is crucial for a comprehensive account of mtDNA-derived mRNA abundance,

whether in AD or not.

5. Acknowledgements.

28

I would like to thank Dr. Angela Hodges and Dr. Aoife Keohane for their guidance throughout the

completion of this project.

6. References.

29

Aksenov MY, Tucker HM, Nair P, Aksenova MV, Butterfield DA, Estus S, Markesbery WR (1999) The

expression of several mitochondrial and nuclear genes encoding the subunits of electron transport chain

enzyme complexes, cytochrome c oxidase, and NADH dehydrogenase in different brain regions in

Alzheimer’s disease. Neurochem Res 24:767-774.

American Psychiatric Association (2000) Diagnostic and statistical manual of mental disorders (IV-TR).

Washington, DC: American Psychiatric Press.

Chandrasekaran K, Hatanpää K, Rapoport SI, Brady DR (1997) Decreased expression of nuclear and

mitochondrial DNA-encoded genes of oxidative phosphorylation in association neocortex in Alzheimer’s

disease. Mol Brain Res 44:99-104.

Christen Y (2000) Oxidative stress and Alzheimer disease. Am J Clin Nutr 71:621-629.

Corder EH, Saunders AM, Strittmatter WJ, Schemechel DE, Gaskell PC, Small GW, Roses AD, Haines

JL, Pericak-Vance MA (1993) Gene dose of apolipoprotein E type 4 allele and the risk on Alzheimer’s

disease in late onset families. Sci 261:921-923.

Du H, Yan SS (2010) Mitochondrial permeability transition pore in Alzheimer’s disease: cyclophilin D

and amyloid beta. Biochim Biophys Acta 1802:198-204.

Ferri C, Prince M, Brayne C, Brodaty H, Fratiglioni L, Ganguli M, Hall K, Hasegawa K, Hendrie H,

Huang Y, Jorm An, Mathers C, Menezes PR, Rimmer E, Scazufca M (2005) Global prevalence of

dementia: a Delphi consensus study. Lancet 366:2112-2117.

Goate A, Charter-Harlin M, Mullan M, Brown J, Crawford F, Fidani L, Giuffra L, Haynes A, Irving N,

James L, Mani R, Newton P, Rooke K, Roques P, Talbot C, Pericak-Vance M, Roses A, Williamson R,

Rossor M, Owen M, et al. (1991) Segregation of a missense mutation in the amyloid precursor protein

gene with familial Alzheimer’s disease. Nature 349:704-706.

Hengartner MO (2000) The biochemistry of apoptosis. Nature 407:770-776.

Kern A, Behl C (2009) The unsolved relationship of brain aging and late-onset Alzheimer disease.

Biochim Biophys Acta 1790:1124-1132.

30

Kyriakouli DS, Boesch P, Taylor RW, Lightowlers RN (2008) Progress and prospects: gene therapy for

mitochondrial DNA disease. Gene Ther 15:1017-1023.

Leary S, Battersby B, Hansford R, Moyes CD (1998) Interactions between bioenergetics and

mitochondrial biogenesis. Biochim Biophys Acta 1365:522-530.

Levy-Lahad E, Wasco W, Poorkaj P, Romano DM, Oshima J, Pettingell WH, Yu C, Jondro PD, Schmidt

SD, Wang K, Crowley AC, Fu Y, Guenette SY, Galas D, Nemens E, Wijsman EM, Bird TD,

Schellenberg GD, Tanzi RE (1995) Candidate Gene for the chromosome 1 familial Alzheimer’s disease

locus. Sci 269:973-977.

Linnane AW, Marzuki S, Ozawa T, Tanake M (1989) Mitochondrial DNA mutations as an important

contributor to ageing and degenerative diseases. Lancet 1:642-645.

Litonin D, Sologub M, Shi Y, Savkine M, Anikin M, Falkenberg M, Gustafsson CM, Temiakov D (2010)

Human mitochondrial transcription revisited: only TFAM and TFB2m are required for the transcription

of the mitochondrial genes in vitro. J Biol Chem 285:18129-18133.

Lovestone S, Francis P, Kloszekska I, Mecocci P, Simmons A, Soininen H, Spenger C, Tsolaki M, Vellas

B, L Wahlund, Ward M (2009) AddNeuroMed – the European collaboration for the discovery of novel

biomarkers for Alzheimer’s disease. Ann NY Acad Sci 1180:36-46.

Lowin A, Knapp M, McCrone P (2001) Alzheimer's disease in the UK: comparative evidence on cost of

illness and volume of health services research funding. Int J Geriatr Psychiatry 16:1143-1148.

Lunnon K, Ibrahim Z, Proitsi P, Lourdusamy A, Newhouse S, Sattlecker M, Furney S, Saleem M,

Soininen H, Kloszewska I, Mecocci P, Tsolaki M, Vellas B, Coppola G, Geschwind D, Simmons A,

Lovestone S, Dobson R, Hodges H (2012) J Alzheimer’s Dis 29:1-26.

Lustbader JW, Cirilli M, Lin C, Xu HW, Takuma K, Wang N, Caspersen C, Chen X, Pollak S, Chaney

M, Trinchese F, Liu S, Gunn-Moore F, Lue LF, Walker DG, Kappusamy P, Zeweir ZL, Arancio O, Stern

D, Yan SS, et al. (2004) ABAD directly links Abeta to mitochondrial toxicity in Alzheimer’s disease.

Science 304:448-452.

31

Manczak M, Park BS, Jung Y, Reddy PH (2004) Differential expression of oxidative phosphorylation

genes in patients with Alzheimer’s disease. Neuromol Med 5:147-162.

Maruszak A, Żekanowski C (2011) Mitochondrial dysfunction and Alzheimer’s disease. Prog neuro

psychopharmacol biol psychiatr 35:320-330.

Maruyama M, Arai H, Sugita M, Tanji H, Higuchi M, Okamura N, Matsui T, Higuchi S, Matsushita S,

Yoshida H, Sasaki H (2001) Cerebrospinal fluid amyloid beta (1-42) levels in the mild cognitive

impairment stage of Alzheimer’s disease. Exp Neurol 172:433-436.

Mckhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM (1984) Clinical-diagnosis of

Alzheimer’s disease – report of the nincds-adrda work group under the auspices of department-of-health-

and-human-services task-force on Alzheimer’s disease. Neurol 34:939-944.

Mitchell AJ (2009) The prognosis of mild cognitive impairment – is it better than expected. Acta

Psychiatr Scand 9:9-10.

Montoya J, Ojala D, Attardi G (1981) Distinctive features of the 5’-terminal sequences of the human

mitochondrial mRNAs. Nature 290:465-470.

Morris JC, Storandt M, McKeel DW, Rubin EH, Price JL, Grant EA, Berg L (1996) Cerebral amyloid

deposition and diffuse plaques in presymptomatic and very mild Alzheimer’s disease. Am Acad Neurol

46:707-719.

Ojala D, Montoya J, Attardi G (1981) A punctuation model of RNA processing in human mitochondria.

Nature 290:470-474.

Osborn G, Saunders A (2010) Current treatments for patients with Alzheimer’s disease. J Am Osteopath

Assoc 110:516-526.

Piechota J, Tomecki R, Gewartowski K, Szczesny R, Dmochowska A, Kudla M, Dybczynska L, Stepien

PP, Bartnik E (2006) Differential stability of mitochondrial mRNA in HeLa cells. Acta Biochim Polonica

53:157-167.

Rizzutto R, Bernardi P, Pozzan T (2000) Mitochondria as all-round players in the calcium game. J

Physiol 529:27-47.

32

Rovelet-Lecrux A, Hannequin D, Raux G, Le Meur N, Laquerrié A, Vital A, Dumanchin C, Feuillette S,

Brice A, Vercelletto M, Dubas F, Frebourg T, Campion D (2006) APP locus duplication causes dominant

early-onset Alzheimer disease with cerebral amyloid angiopathy. Nature Genet 38:24-26.

Sharma MR, Koc EC, Datta PP, Booth TM, Spremulli LL, Agrawal RK (2003) Structure of the

mammalian mitochondrial ribosome reveals an expanded functional role for its component proteins. Cell

115:97-108.

Sherrington R, Rogaev EI, Liang Y, Rogaeva EA, Levesque G, Ikeda M, Chi H, Li G, Holman K, Tsuda

T, Mar L, Foncin J, Brunl AC, Montesi MP, Sorbi S, Rainero I, Pinessl L, Nee L, Chumakov I, Pollen D,

et al. (1995) Nature 375:754-760.

Smits P, Smeitink J, van den Heuven L (2012) Mitochondrial translation and beyond: processes

implicated in combined oxidative phosphorylation deficiencies. J Biomed Biotechnol 2010:1-24.

Temperley RJ, Seneca SH, Tonska K, Bartnik E, Bindoff LA, Lightowlers RN, Chrzanowska-Lightowers

ZMA (2003) Investigation of a pathogenic mtDNA microdeletion reveals a translation-dependent

deadenylation decay pathway in human mitochondria. Hum Mol Genet 12:2341-2348.

Weraarpachai W, Antonicka H, Sasarman F, Seeger J, Schrank B, Kolesar JE, Lochmüller H, Chevrette

M, kaufman BA, Horvath R, Shoubridge EA (2009) Mutation in TACO1, encoding a translational

activator of COX 1, results in cytochrome c oxidase deficiency and late-onset Leigh syndrome. Nature

Genet 41:833-837.

7. Appendices.

Appendix 1. Tests of normality for genes organized by status.

Tests of Normality

Status Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

ND3_LOG 1.00 .100 171 .000 .951 171 .000

33

2.00 .042 141 .200* .992 141 .625

3.00 .066 144 .200* .986 144 .150

ND4_LOG

1.00 .085 171 .004 .986 171 .086

2.00 .060 141 .200* .990 141 .418

3.00 .045 144 .200* .995 144 .902

ND4L_LOG

1.00 .105 171 .000 .962 171 .000

2.00 .068 141 .200* .987 141 .187

3.00 .067 144 .200* .984 144 .090

ND5_LOG

1.00 .061 171 .200* .982 171 .024

2.00 .066 141 .200* .975 141 .011

3.00 .057 144 .200* .985 144 .113

CYB_LOG

1.00 .063 171 .095 .984 171 .053

2.00 .059 141 .200* .993 141 .731

3.00 .083 144 .017 .975 144 .009

* This is a lower bound of the true significance.

a. Lilliefors Significance Correction

The Kolmogorov-Smirnov test is typically applied when n>50.

Appendix 2. Correlations between mRNA abundance between all genes in the control group.

Controls

ND3 ND4 ND4L ND5 CYB

ND3

Pearson Correlation 1 .083 .510** .117 .104

Sig. (2-tailed) .294 .000 .139 .188

N 162 162 162 162 162

34

ND4

Pearson Correlation .083 1 .275** .481** .484**

Sig. (2-tailed) .294 .000 .000 .000

N 162 162 162 162 162

ND4L

Pearson Correlation .510** .275** 1 .367** .198*

Sig. (2-tailed) .000 .000 .000 .011

N 162 162 162 162 162

ND5

Pearson Correlation .117 .481** .367** 1 .545**

Sig. (2-tailed) .139 .000 .000 .000

N 162 162 162 162 162

CYB

Pearson Correlation .104 .484** .198* .545** 1

Sig. (2-tailed) .188 .000 .011 .000

N 162 162 162 162 162

** Correlation is significant at the 0.001 level (2-tailed).

* Correlation is significant at the 0.05 level (2-tailed).

Appendix 3. Correlations between mRNA abundance between all genes in the MCI group.

MCI

ND3 ND4 ND4L ND5 CYB

ND3 Pearson Correlation 1 .292** .449** .418** .556**

Sig. (2-tailed) .001 .000 .000 .000

35

N 134 134 134 134 134

ND4

Pearson Correlation .292** 1 .262** .228** .406**

Sig. (2-tailed) .001 .002 .008 .000

N 134 134 134 134 134

ND4L

Pearson Correlation .449** .262** 1 .579** .514**

Sig. (2-tailed) .000 .002 .000 .000

N 134 134 134 134 134

ND5

Pearson Correlation .418** .228** .579** 1 .524**

Sig. (2-tailed) .000 .008 .000 .000

N 134 134 134 134 134

CYB

Pearson Correlation .556** .406** .514** .524** 1

Sig. (2-tailed) .000 .000 .000 .000

N 134 134 134 134 134

** Correlation is significant at the 0.01 level (2-tailed).

Appendix 4. Correlations between mRNA abundance between all genes in the AD group.

AD

ND3 ND4 ND4L ND5 CYB

ND3 Pearson Correlation 1 .271** .448** .192* .266**

36

Sig. (2-tailed) .001 .000 .022 .001

N 141 141 141 141 141

ND4

Pearson Correlation .271** 1 -.062 .103 .326**

Sig. (2-tailed) .001 .463 .225 .000

N 141 141 141 141 141

ND4L

Pearson Correlation .448** -.062 1 .502** .318**

Sig. (2-tailed) .000 .463 .000 .000

N 141 141 141 141 141

ND5

Pearson Correlation .192* .103 .502** 1 .515**

Sig. (2-tailed) .022 .225 .000 .000

N 141 141 141 141 141

CYB

Pearson Correlation .266** .326** .318** .515** 1

Sig. (2-tailed) .001 .000 .000 .000

N 141 141 141 141 141

** Correlation is significant at the 0.01 level (2-tailed).

* Correlation is significant at the 0.05 level (2-tailed).

Appendix 5. Correlations between mRNA abundance between all genes in the control group using Spearman’s rank

coefficients.

Controls

37

ND3 ND4 ND4L ND5 CYB

ND3

Correlation Coefficient 1.000 .194* .457** .210** .246**

Sig. (2-tailed) . .013 .000 .007 .002

N 162 162 162 162 162

ND4

Correlation Coefficient .194* 1.000 .429** .509** .479**

Sig. (2-tailed) .013 . .000 .000 .000

N 162 162 162 162 162

ND4L

Correlation Coefficient .457** .429** 1.000 .445** .271**

Sig. (2-tailed) .000 .000 . .000 .000

N 162 162 162 162 162

ND5

Correlation Coefficient .210** .509** .445** 1.000 .547**

Sig. (2-tailed) .007 .000 .000 . .000

N 162 162 162 162 162

CYB

Correlation Coefficient .246** .479** .271** .547** 1.000

Sig. (2-tailed) .002 .000 .000 .000 .

N 162 162 162 162 162

** Correlation is significant at the 0.01 level (2-tailed).

* Correlation is significant at the 0.05 level (2-tailed).

Appendix 6. Correlation between ND4 and ND4L.

38

39