7/17/2019 Barcoding T Cell Calcium Response Diversity With Methods for Automated and Accurate Analysis of Cell Signals http://slidepdf.com/reader/full/barcoding-t-cell-calcium-response-diversity-with-methods-for-automated-and 1/23 Barcoding T Cell Calcium Response Diversity with Methods for Automated and Accurate Analysis of Cell Signals (MAAACS) Audrey Salles,#1,2,3,¤a Cyrille Billaudeau,#1,2,3,¤b Arnauld Sergé ,1,2,3,*¤c Anne-Marie Bernard,1,2,3 Marie-Claire Phélip!,1,2,3 "iclas Ber!au,$,% Ma!hieu &alle!,1,2,3 Pierre 'ren!,1,2,3 (idier Margue!,1,2,3 )ai-a )e,1,2,3 and +annic )an1,2,3,* 'régire Al!an-Bnne!, .di!r Au!hr in/ra!in 0 Ar!icle n!es 0 Cpyrigh! and icense in/ra!in 0 his ar!icle has been ci!ed by !her ar!icles in PMC Astract We introduce a series of experimental procedures enabling sensitive calcium monitoring in T cell populations by confocal video-microscopy. Tracking and post-acquisition analysis was performed using Methods for Automated and Accurate Analysis of ell !ignals "MAAA!#$ a fully customi%ed program that associates a high throughput tracking algorithm$ an intuitive reconnection routine and a statistical platform to provide$ at a glance$ the calcium barcode of a population of individual T-cells. ombined with a sensitive calcium probe$ this method allowed us to unravel the heterogeneity in shape and intensity of the calcium response in T cell populations and especially in naive T cells$ which display intracellular calcium oscillations upon stimulation by antigen presenting cells. Author Summary The adaptive immune response to pathogen invasion requires the stimulation of lymphocytes by antigen-presenting cells. We hypothesi%ed that investigating the dynamics of the T lymphocyte activation by monitoring intracellular calcium fluctuations might help explain the high specificity and selectivity of this phenomenon. &owever$ the quantitative and exhaustive analysis of calcium fluctuations by video microscopy in the context of cell-to-cell contact is a tough challenge. To tackle this$ we developed a complete solution named MAAA! "Methods for Automated and Accurate Analysis of ell !ignals#$ in order to automate the detection$ cell tracking$ raw data ordering and analysis of calcium signals. 'ur algorithm revealed that$ when in contact with antigen- presenting cells$ T lymphocytes generate oscillating calcium signals and not a massive and sustained calcium response as was originally thought. We anticipate our approach providing many more new insights into the molecular mechanisms triggering adaptive immunity. !ntroduction alcium ion acts as a universal second messenger in response to most cellular stimuli ()*. The pattern of the calcium response is biphasic$ and primarily results from the production of inositol-+ phosphate ",+# which triggers the release of calcium from the endoplasmic reticulum "/ store release# into the cytoplasm. This decrease is sensed by the stromal interaction molecules "!T,M# that secondarily trigger the capacitative entry of extracellular calcium via the calcium release activated channels "/A# of the '/A, family (0*1(2*. Measuring the intracellular concentration of calcium is therefore of primary interest when analy%ing transduction processes in living cells. urrently$ this is achieved by methods which combine flow cytometry with intracellular diffusive fluorescent calcium-sensitive dyes in immunological relevant cells such as macrophages$ 34 cells$ T or 5 cells. As an example$ the calcium response is routinely monitored in T cells (6*1()6* as a functional read-out of the outside-in signal transduction subsequent to T-cell receptor "T/# engagement by ma7or histocompatibility complex "M&# molecules with agonist peptide. &owever$ when naive T cells encounter antigen-
23
Embed
Barcoding T Cell Calcium Response Diversity With Methods for Automated and Accurate Analysis of Cell Signals
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
7/17/2019 Barcoding T Cell Calcium Response Diversity With Methods for Automated and Accurate Analysis of Cell Signals
Ma!hieu &alle!,1,2,3 Pierre 'ren!,1,2,3 (idier Margue!,1,2,3 )ai-a )e,1,2,3and +annic )an1,2,3,*'régire Al!an-Bnne!, .di!r Au!hr in/ra!in 0 Ar!icle n!es 0 Cpyrigh! and icense in/ra!in 0his ar!icle has been ci!ed by !her ar!icles in PMC
AstractWe introduce a series of experimental procedures enabling sensitive calcium monitoring
in T cell populations by confocal video-microscopy. Tracking and post-acquisition
analysis was performed using Methods for Automated and Accurate Analysis of ell
!ignals "MAAA!#$ a fully customi%ed program that associates a high throughput
tracking algorithm$ an intuitive reconnection routine and a statistical platform to provide$
at a glance$ the calcium barcode of a population of individual T-cells. ombined with a
sensitive calcium probe$ this method allowed us to unravel the heterogeneity in shape and
intensity of the calcium response in T cell populations and especially in naive T cells$
which display intracellular calcium oscillations upon stimulation by antigen presenting
cells.
Author SummaryThe adaptive immune response to pathogen invasion requires the stimulation of
lymphocytes by antigen-presenting cells. We hypothesi%ed that investigating the
dynamics of the T lymphocyte activation by monitoring intracellular calcium fluctuations
might help explain the high specificity and selectivity of this phenomenon. &owever$ the
quantitative and exhaustive analysis of calcium fluctuations by video microscopy in thecontext of cell-to-cell contact is a tough challenge. To tackle this$ we developed a
complete solution named MAAA! "Methods for Automated and Accurate Analysis of
ell !ignals#$ in order to automate the detection$ cell tracking$ raw data ordering and
analysis of calcium signals. 'ur algorithm revealed that$ when in contact with antigen-
presenting cells$ T lymphocytes generate oscillating calcium signals and not a massive
and sustained calcium response as was originally thought. We anticipate our approach
providing many more new insights into the molecular mechanisms triggering adaptive
immunity.
!ntroductionalcium ion acts as a universal second messenger in response to most cellular stimuli ()*.
The pattern of the calcium response is biphasic$ and primarily results from the productionof inositol-+ phosphate ",+# which triggers the release of calcium from the endoplasmic
reticulum "/ store release# into the cytoplasm. This decrease is sensed by the stromal
interaction molecules "!T,M# that secondarily trigger the capacitative entry of
extracellular calcium via the calcium release activated channels "/A# of the '/A,
family (0* 1 (2*. Measuring the intracellular concentration of calcium is therefore of
primary interest when analy%ing transduction processes in living cells. urrently$ this is
achieved by methods which combine flow cytometry with intracellular diffusive
fluorescent calcium-sensitive dyes in immunological relevant cells such as macrophages$
34 cells$ T or 5 cells. As an example$ the calcium response is routinely monitored in T
cells (6* 1 ()6* as a functional read-out of the outside-in signal transduction subsequent to
T-cell receptor "T/# engagement by ma7or histocompatibility complex "M&#
molecules with agonist peptide. &owever$ when naive T cells encounter antigen-
the physiological conditions of stimulation. ,n addition$ recent works have demonstrated
that T/ triggering by the M& molecules follows unusual physico-kinetic parameters
of serial engagement-disengagement ()8*$ ()9*$ which could be the molecular basis for
the broad selectivity$ high specificity$ extreme sensitivity ():* and the capacity to induce a
rapid intracellular response that characteri%e T/ triggering ();*. While soluble anti T/
or anti <+ antibodies (0=*$ antibody coated beads (0)*$ (00*$ and phorbol myristate
acetate>ionomycin (0+* can all induce a productive calcium signal in T cells that
ultimately leads to their activation$ proliferation and cytokine production$ the calcium
elevation triggered by these strong irreversible stimuli is usually sustained. ,t may not
therefore be representative of the response to physiological stimulations$ which is more
likely to consist in calcium spikes and oscillations (;*$ (02* 1 (08*. ,n order to capture the
true calcium responses triggered during cell-cell contacts such as those occurring during
T-cell and A stimulation$ video-imaging is compulsory in that it provides informative
parameters on individual cell behavior "i.e. displacements$ shape and intensity
fluctuation# (09*.
'btaining such imaging data requires a complex custom-built experimental set-up usually
dedicated to the detection of ?@-excitable calcium probes and to the maintenance of
physiological parameters for long-term recordings (;*$ (06*$ (0:*. ,n any case cell
tracking is mandatory and is often performed by manual approaches (0:* 1 (+=* however$
in addition to being time-consuming$ manual analysis is prone to systematic errors due to
sub7ective choice. !uch pre-selection is an unavoidable step in any manual analysis.
Automating the simultaneous tracking of hundreds of cells over hundreds of time frames
would overcome these issues. 3evertheless$ simultaneously tracking moving cells at high
density represents a considerable challenge$ particularly considering the need to correctly
resolve interlaced tracks of stretching cells while providing valuable statistical confidence
and robustness. While many software packages do incorporate a cell tracking module or plugins$ the normali%ation of the calcium signal for each cell as well as the classification
of calcium responses and any quantification generally have to be performed manually
involving tedious excel datasheets (+)*.
&ere we have developed a complete approach named Methods for Automated and
Accurate Analysis of ell !ignals "MAAA!# which enables the simultaneous tracking
of a population of individual moving cells "multiple target tracking$ MTT# (+0* and the
automatic extraction of robust statistics on pertinent observables. The MAAA! program
has been conditioned to synchroni%e$ normali%e and assemble the recorded cell traces to
provide an at a glance calcium barcoding of a heterogeneous cell population and facilitate
the a posteriori data mining and interpretation. We used MAAA! to examine the
calcium responses induced in T cells upon interaction with As and with it were able toreveal the oscillatory calcium responses in mouse naive <2B T cells upon antigen
recognition.
' !ResultsDevelopment of a protocol enaling the sensitive detection of intracellularcalcium fluctuations in TCR mediated T"cell stimulation via its naturalligand
Aiming to establish an easy$ robust$ sensitive and reliable way of evaluating calcium
fluctuations in T cells$ we assessed many visible calcium indicators such as Cluo-2 AM$
Cluo-+ AM$ and Cluo-:. All displayed short term leakiness of the loading without
7/17/2019 Barcoding T Cell Calcium Response Diversity With Methods for Automated and Accurate Analysis of Cell Signals
gradients could be visuali%ed under such experimental conditions since few discrete
hotspots were detectable among the homogenous fluorescence. 5< 5D and mitotracker
red loaded cells were imaged to decipher whether mitochondria would accumulate
calcium indicator. ,n +A; T cells the two signals were not mutually exclusive "Figure
S1.C,D# unlike in Eurkat cells or primary human T cells (+6*$ (+;*. art of the signals was
correlated under stimulation$ most presumably caused by C/T between the two dyes
"Figure S1.C, D#. We detected no significant sequestration of the 5< 5D calcium dye$unlike Cura-0 in Eurkat cells although this phenomenon had a limited impact on whole
cell calcium measurement (+6*.
Cigure )
BD PBX, a highly sensitive calcium probe allows ratiometric analysis an
antiboy labeling o! primary cells. "A# Comparison o! e$citation anemission spectra o! Fluo%&%A' an the calcium inicator inclue in the
PBX calcium assay (it, BD PBX.
These inconsistencies with previous reports (+2* motivated us to consider 5< 5D as a
close relative of Cluo-2 AM but harboring subtle differences$ that required the dedicated
loading buffer to avoid the dye leakage displayed by Cluo-2 AM "Figure S1A#. <ue to
the lack of information about the 5< 5D from the manufacturer$ we thus determined the
in vitro 4d of the 5< 5D as described previously (2=* "Figure 1B#$ which gave a
consistent value for 4d of +)0 nMF++$ comparable to the 4d for Cluo-2 AM of +09F2;
nM and that in the literature "4dG+26 nM# (2)* "see Materials and Methods#. The great
similarity between 5< 5D and Cluo-2 AM led us to assume the in situ 4d of 5< 5D
to equal the reported in situ 4d value ") HM# of Cluo-2 AM (+2*$ deemed acceptable
when accurate determination by electrophysiology is either not feasible or not available
(0:*. 5ased on this 4d value$ we estimated that the intracellular calcium concentration in
resting +A; T cell hybridomas would be around ;= nM and consistent with previously
published values for these hybridomas (20* as well as leukemic cell lines Eurkat (+;*$
mouse thymocytes (2+* and human peripheral blood lymphocytes (+6*.
We used flow cytometry to compare the detection sensitivity of 5< 5D with the
ratiometric ,ndo-) AM "routinely used in calcium assays$ ?@ excitable#. ,n terms of
response pattern$ T cell hybridomas loaded with 5< 5D do not strictly speaking respond
the same as those loaded with ,ndo-) AM$ as the non-ratiometric calcium indicator 5<
5D does not allow the evaluation of intracellular free calcium concentration "Figure
1C#. We therefore sought to estimate the intracellular calcium concentrations in 5< 5D
loaded T cell hybridomas. ?pon various concentrations of ionomycin$ we compared
intracellular calcium elevation in 5< 5D versus ,ndo-) AM loaded +A; T cell
hybridomas. As previously mentioned$ the kinetics were not fully stackable and the
calculated intracellular calcium concentration differed between the two methods due to
4d discrepancy and loss of linearity in the relationship linking calcium concentrations
and high fluorescence amplitudes (2=*. ,ndeed at such high fluorescence values under
strong ionomycin concentrations$ the calcium concentration was overestimated and is the
reason for us reporting fluorescence amplitude instead of erroneous calcium
concentrations throughout this manuscript. 3evertheless to our surprise$ lower
concentrations of ionomycin rapidly abrogated fluorescence elevations of indo-) AM
whereas 5< 5D fluorescence elevation remained detectable even at subnanomolar
ionomycin concentrations. This indicated that 5< 5D is sensitive to low intracellular
calcium elevation. Thapsigargin "Figure 1D# or the cross-linking of the T/><+
complex by anti-<+I "0)) biotin>streptavidin# induced in cells loaded with 5< 5D
responses similar to those induced by ionomycin. This method was extendable to naive
primary <2B T cells labeled or not with an anti <2 monoclonal antibody "Figure 1)#.
Automatic trac#ing of high density moving cells y MAAACSWhile surface receptor crosslinking with antibodies is a convenient way of stimulating
calcium responses in T cells$ it cannot physiologically reproduce the dynamics of
T/>M&-antigen interactions in the context of T-cell>A contacts. 'ur goal was to
decipher calcium signals arising from cellular contacts and requiring imaging approaches.
We chose to perform all recordings on a conventional argon laser equipped-confocal
scanning microscope$ widely found in laboratories. onsidering the lack of available
methods (22* able to combine automatic tracking of cells and calcium signal processing$
we set up our own procedure to automatically track moving cells at high densities with a
minimum of input parameters. ,t is a customi%ed version of our previously developed
MTT algorithm (+0*$ originally dedicated to tracking single fluorophores coupled to
plasma membrane molecules at high density. We built a plugin that converts cellularimages into cell position images that are comparable with the single molecule images
supported by MTT "Figure *A + Figure S*#. /aw fluorescence images were first passed
through a median filter to eliminate the electronic noise emanating from detection and an
appropriate mask$ consisting of a disk of adequate radius "see Materials and Methods for
details on mask si%e#$ was applied to identify cells as single ob7ects. ach cell was thus
defined by the xy coordinates of its centroid which then served to reconstruct the cellJs
tra7ectory over the stack of images "Figure *A + ieos S1, S*, S-, S&, S, S/, S0#.
3oteworthy$ more complex cellular shapes could be handled if using other appropriate
detection schemes. 3ext$ we generated synthetic images containing Kaussian peaks at the
corresponding positions$ with a radius optimi%ed for the tracking performed by MTT and
an intensity equal to the integrated cell signal$ itself proportional to the intracellular
calcium concentration. The resulting sequence of single molecule like images could then be treated by MTT "ieos S, S/#. /eplacing each cell by a Kaussian peak of smaller
si%e prevented the occurrence of two targets crossing over each other$ initially a ma7or
concern for MTT$ thus rendering the reconnection of traces during the MTT procedure far
more efficient. 'verlapping was then handled at the detection stage$ where crossing cells
presenting as a Lpeanut shape were detected as two targets. &owever$ strongly
overlapping cells resulting in more of a spherical shape were detected as a single target
and thus required %-stack acquisitions and appropriate analysis to recogni%e the
MAAA! generates traces which are defined by the position of the detected cells overtime and intrinsically calculates their instantaneous velocity and their fluorescence
intensity. Kiven that the basal level varies from cell to cell due to differences in
intracellular calcium concentrations$ or due to heterogeneous efficiency in calcium
indicator loading$ we needed to accurately set a baseline of fluorescence that we defined
as the median of fluorescence calculated until the maximum fluorescence value had been
reached for each cell "Figure *B#. To establish the best mode of normali%ation$ we
analy%ed for each cell in non-stimulating conditions how the mean signal amplitude was
correlated with signal fluctuations. We found that this relationship was proportional$ thus
implying that the normali%ation can be performed by division. alcium responses are
highly diverse$ both in terms of magnitude and oscillation "Figure S-#. The amplitude of
calcium mobili%ation varies according to the type of stimulus and the addition of
inhibitors$ but also within a cell population for a given stimulus>treatment. Moreover$ the
shape of these signals$ their maintenance and their oscillations are also varied. We
therefore defined analytical parameters to describe and characteri%e these response
diversities "Figure *C#. Cor each cell$ the response magnitude is described as the
fluorescence amplitude (FA) of calcium mobili%ation$ corresponding to the time-average
of normali%ed intensities on the whole trace. The temporal fluctuations are deciphered by
analy%ing the persistence and the oscillations of the calcium signals. We defined as the
responsefraction (RF) the ratio of two phasesN the time when the normali%ed intensity is
greater than the threshold over the total time during which the intensity is detected. We
also calculated the number of bursts/min "BP'# defining the number of peaks detected
above the threshold divided by the duration of the detected trace "able S1#. ,n order to
provide a global$ comprehensive view of the calcium response in a substantial number ofcells for any given condition$ we color coded the normali%ed calcium intensities with a
gradient of blue and orange for values below or above the threshold "see below#$
respectively. The resulting values$ for each cell at each time-point$ could then be pooled
to generate a heat map$ the dimensions of which hence corresponded to the cell number
and time "Figure *C le!t panel#. 3on-relevant pixels$ either before or after detection of a
given cell in the time-lapse movie$ were left in black. This representation simultaneously
depicts the global tendency$ together with the intra-population variability of response
(26*. ollectively$ all calcium signal parameters are summari%ed on scatter dot plots
where responding cells are represented by a single dot "Figure *C right panel#. 5y
integrating these different parameters$ the heterogeneous behavior of activated cells
"maintained$ oscillatory and unique# can be determined and inactive cells identified$ the
proportions of which are then represented in a pie chart diagram "see Materials and
MAAACS trac#ing performance versus e$haustive manual trac#ing
An endemic problem in automatic tracking approaches is reaching a level of
completeness that manual tracking only can guarantee. We analy%ed several videos in parallel by the automatic and by manual methods and determined the percentage of cells
tracked by MAAA! compared to that by manual tracking in the observation field
"detection percentage#. +A; T cell hybridomas or naive <2B T-cells loaded with 5<
5D were seeded onto a monolayer of '!-9 cells stably expressing the molecules of the
ma7or histocompatibility complex (28*. We chose experimental situations with high cell
densities on a rough and irregular surface. We only considered tracked cells detected by
MAAA! over more than 6 images. The superimposition of the tra7ectories obtained
manually or with MAAA! "Figure -A# illustrated the efficiency of our algorithm$ the
detection percentage of which was greater than ;8O "3G)06#. !urprisingly$ more traces
were generated by MAAA! than obtained manually. onsequently$ the MAAA! cell
traces were fragmented into several parts as shown in Figure -B illustrated by the
number of fragments needed to reconstitute the full length tra7ectories ").+ for primary T
cell and ).; for hybridomas$ Figure -B#. The lower efficiency for the latter can be
explained by significant cellular shape changes over time preventing their detection by
the circular mask and thus their reconnection. To fully document the cell response over
time and to improve the quantitative analysis of cell signaling$ we implemented a
program allowing the reconnection of fragments of tra7ectories "Figure -C#. Cirst$ the
method selected among the set of tra7ectories terminating before the end of the
acquisition "plotted in gray in Figure -C# to be reconnected to traces starting after the
breaking off "plotted in red and blue in Figure -C#. Then the algorithm classified the
candidates for reconnection by minimi%ing the interval between the stop and start times
"Pt#$ the distance between the final and initial positions "Pr# and the difference of the
mean fluorescence amplitude of the fragments "P,#. The user is free to decide whether thetra7ectories should be reconnected by consulting the original video. ,n this way$ the
tracked time percentage was clearly improved ";6O for primary T cells and :+O for
hybridomas$ Figure -D#.
'AAACS trac(ing per!ormance versus e$haustive manual trac(ing. "A# 2verlay o!
cell tra3ectories analy4e manually or with 'AAACS.Threshold of specificactivation
Within the course of this study$ we noticed that without specific stimulation$ cytoplasmic
calcium concentrations displayed spontaneous oscillations$ the amplitude of which was
almost negligible for T cell hybridomas but not for primary naive T cells ()+*. To account
for this$ we performed a median filtering with a sliding window of 9 frame-si%e on eachfluorescence amplitude to remove aspecific oscillations in the absence of any stimulus. ,t
was then important to carefully define the threshold of peptide-specific activation. As
mentioned by many authors (08*$ (20*$ (29* and in our observations "Figure -#$ calcium
signaling exhibits high diversity even within the same cell line and depends on the
applied perturbations "stimulation$ drug treatment or mutation#. ,t should be noted that
defining any biological threshold for activation could be misleading since it is highly
dependent upon the experimental conditions. Moreover$ the definition of a criterion for
specific activation should respect the response heterogeneity without favoring a subset of
responding cells. Accordingly$ we set up a detector which compares the fluorescence
amplitude to a threshold of activation in order to identify activated cells in our
experiments "Figure S&#. ,f the fluorescence amplitude is greater than the threshold$ then
the cell is declared as activated. To determine this threshold$ we investigated the
statistical properties of the fluorescence amplitude of cells in the absence of any
stimulation compared to that in cells that have been activated as a result of stimulation.
Cor a given probability of false alarm "CA#$ an activation threshold could be deduced
from the cell responses in the absence of any stimulation. The probability of detection
"<# could then be calculated as the percentage of activated cells revealed by the detector.
We then aimed to identify threshold values that minimi%ed false detections "low CA#without decreasing the identification of activating cell "high <#. A robust method to
ob7ectively determine the activation threshold is to perform a receiver operating
characteristic analysis "/' curves in Figure S&A, B# to explore the values of activation
threshold that maximi%e the overall score < x ")-CA#. All activation thresholds are
reported in Figure S&C. !urprisingly$ the activation threshold was quite stable whatever
the cell type "hybridoma or naive T cells# or stimulation process "antibody or A#. Cor
T-cell hybridomas$ this method exhibited very high < "Q=.;;# and low CA "R=.=0#. <
was also high in primary T cells "Q=.;:# and the CA was reasonable "R=.=:# though
slightly higher due to a higher diversity in the cell signaling.
%valuation of the calcium response of a heterogeneous population of
individual T cell hyridomas
To test our MAAA! algorithm$ we analy%ed a population of T hybridomas +A; loaded
with 5< 5D and seeded at the bottom of a well of a Sab-Tek before their stimulation
with thapsigargin after ) minute. We generated a sequence with a frequency of ) confocal
image every 9 seconds for += minutes. Cluorescence intensity was not affected by
repetitive illuminations as previously mentioned "Figure S1B#. 3o a priori assumption
was made and raw recordings were sub7ected to MAAA!. To compare the calcium
signals$ we normali%ed the fluorescence intensities as a fold of the basal fluorescence. A
mean curve of variations in fluorescence over time was obtained "Figure &C# and
compared to flow cytometry measurements "Figure &B#. Cluorescence rose immediately
after thapsigargin addition$ reaching a plateau + minutes after induction (2:* before
slowly decreasing "Figure &C#. This response was fully stackable with the kineticsmonitored by flow cytometry "Figure &B#. We noted that the response>baseline ratio was
higher according to imaging recordings$ presumably due to a better sensitivity of
detection on confocal photomultipliers. ,n this case the benefit of MAAA! is limited
since all individual cells responded homogenously by a strong$ sustained non-oscillating
sustained or oscillatory behavior. This implies that within a cell population and for any
given stimulus$ the observed differences in response are not limited to intensity dispersion
but also to the mode of response. We sought to document this point using different
experimental stimuli each producing their own calcium response in terms of shape$
intensity and heterogeneity "Figure # (6=*. ,ndeed$ when the cells were seeded onto an
activating surface "anti T/ antibody coated pits#$ while most responses were sustained
"CAG2.2F=.2+$ /CG=.80F=.=+$ 5MG=.=;F=.=)#$ heterogeneous responses were also
observed. These heterogeneities were even more obvious when the +A; cells were
stimulated by ,-A4-&S expressing '!-9 As. The asynchronous landing of the T
cells and the heterogeneous M& ,, agonist peptide expression levels are parameters that
affect the calcium response in addition to irregular crawling and scanning activities of the
T cells on an A monolayer. ,ndeed$ during the first += minutes$ 6=O of the cells
displayed a specific calcium rise ()6*. Most of these exhibited a maintained fluorescence
amplification but which was weaker in term of intensity and response fraction as
compared to that in response to the stimulating antibody "Figure C#$ supporting the
notion that abundant$ immobile$ highly affine ligands are not strictly recapitulating the
stimulation by the natural membrane ligand of the T/. !upporting this view$ weanaly%ed the cell motility by MAAA!$ as an integral parameter of T cell activation (0;*$
(20*$ (6)*. MAAA! analysis of cell velocity showed that inducers of strong and
sustained calcium responses "thapsigargin or anti T/ coated slides and to a lesser extent
anti <26 unstimulating surfaces# negatively impacted the motility of cells ()0*$ since
instant speed measurements did not exceed 0 Hm>min in the few minutes after landing on
the slide$ indicating that the cells were almost immobile. ,n contrast$ T-cells migrating on
)valuation o! calcium response o! a heterogeneous population o! iniviual cell
hybriomas upon i!!erent e$perimental stimuli.Although experimental$ '!-9 ,-A4-
&S have been shown to efficiently simulate physiological situations of T cell activation
by inducing productive calcium signals$ in a context of immunological kinapses (68*$ (69*rather than stable immunological synapses leading to specific cytokine secretion such as
,nterleukin-0 "data not shown#. Additionally$ as previously demonstrated$ a clear
correlation exists between activation and motility since unactivated T cells appear to
move faster than activated ones within the same population (6:* and T cell hybridoma
mobility appears to decrease rapidly after calcium rise and rounding of the cells ()=*$ ()0*
"Figure S/A#. MAAA! was conditioned to automatically analy%e the velocity and
shape of the cells in addition to fluorescence signals$ although no link between these
parameters was found that was as tight as previous reports in cell systems expressing co-
stimulatory or specific adhesion molecules.
,n the presence of 0-A5$ the fluorescence amplitudes upon either anti T/ "CAG
+.08F=.02# or '!-9 ,-A4-&S stimulation "CAG).68F=.=8# were strongly reducedcompared to /A active control conditions. &owever$ although the response fraction
and the number of bursts per minute were left unaffected "/CG=.68F=.=+$ 5MG
=.=;F=.=0# "Figure C# upon anti-T/ T cell activation$ short and low calcium
oscillations (08* were predominant in most T-cells "/CG=.08F=.=)$ 5MG=.)2F=.=)#
"Figure C an Figure S-B# seeded onto '!-9 ,-A4-&S.
,n this case$ the lack of co-stimulatory or specific adhesion molecules that usually sustain
T cell>A interactions and signaling (6;* suggested that following T/ engagement by
M&-peptide$ signaling events would occur through waves such as displayed by the
calcium oscillations (8=*. We therefore titrated the T/-dependent calcium signaling in
the presence of 0-A5 in T-cells as a function of the amount of peptide loaded onto '!-
9 ,-A4. As shown in Figure /$ the peptide-specific$ 0-A5 sensitive calcium response
was dependent upon the amount of peptide presented by '!-9 ,-A4. ,n the absence of
peptide$ about 6O of the cells displayed a weak but significant calcium response above
threshold. The percentage of responding cells increased proportionally to the peptide
concentration to reach a plateau at a &S 2818) peptide concentration of 6= nM "Figure
/A#$ while being constantly oscillatory "Figure /)#. The fluorescence amplitude of the
responding cells was significantly higher than that observed in the absence of antigenic
peptide$ even at low antigen doses "down to =.6 nM#. "Figure /B# !urprisingly$ the
fluorescence amplitude$ response fraction$ and burst frequency were independent of the
antigen concentration "except at higher antigen concentrations# "Figure /B8D#. These
data show that$ at least in +A; T cell hybridomas$ the T/-mediated antigen-dependent
/-store-operated calcium response is digitally triggered irrespective of the antigenic peptide concentration. This is consistent with results from an earlier study (20* focusing
however on the global calcium responses in +A; T cells.
)!!ect o! peptie concentration on the )5 release o! calcium.&aive CD' T cellsdisplay heterogeneous intracellular calcium flu$es
The antigen-dependent calcium response of mouse primary T cells has previously been
investigated in vivo by microscopy on explants or sections of lymphoid organs (;*$ (0:*$(6+* or ex vivo on artificial activating surfaces (66*. &owever$ most of these studies were
performed on lymphoblasts obtained by continuous activation in the presence of ,S-0 for
several hours and which therefore differ from naive T cells ():*$ (29*$ (8)*. <ata from
studies that have examined the calcium response of naive T cells suggest that the calcium
homeostasis of naive <2B T cells ex vivo is complex and at least in part antigen-
independent ()+*$ (80*. We evaluated these calcium responses of T-cells with MAAA!
"Figure 0#. 3aive +A; transgenic <2B T cells (8+*$ (82* seeded onto a surface coated
with anti T/ antibodies showed a strong increase in fluorescence "CAG2.8)F=.06# that
was maintained over time "/CG=.9)F=.=+$ 5MG=.)+F=.=)# similar to that observed
with +A; hybridomas "Figure 0A, B,
$C#.
C#. &owever$ we also observed that when seeded onto non-stimulating ,-A4 expressing
'!-9 cells$ around 0=O of naive T cells responded spontaneously with short weak
calcium pulses "CAG).0+F=.=+$ /CG=.)+F=.=) 5MG=.)2F=.=)# reminiscent of those
previously reported ()+* and (;*. eptide specific calcium signals in the presence of '!-
9 ,-A4-&S were mostly oscillatory in more than 8=O of the cells "CAG0.=0F=.=;$ /CG
=.+0F=.=0$ 5MG=.00F=.=)#$ in contrast to those observed in hybridomas " Figure 0B#
which were mainly sustained. Another fundamental difference with hybridomas is that we
found no clear correlation between calcium fluxes and cell velocity "Figure S/B#$ which
nevertheless was expected considering in vivo reports (86*. We then wondered whether
these calcium responses in <2B naive T cells were sensitive to 0-A5. /A channel
activity in T cells is characteri%ed upon stimulation by thapsigargin or soluble anti-<+antibody "0))#$ generating sustained calcium responses that are absent in patients
suffering from an inherited form of severe combined immune deficiency "!,<#
syndrome or upon 0-A5 treatment (+* ,ndeed$ upon 0-A5 treatment$ the thapsigargin-
induced calcium response was drastically reduced. quivalent kinetics "evaluated by flow
cytometry# was obtained with 0)) stimulation in the presence of 0-A5 or <TA
"Figure S0A8B#. 3o additive or competitive effect was detected under these
influx (88* 1 (8:*. !imilarly$ we analy%ed the calcium response to '!-9 ,-A4-&S in the
presence of 0-A5. There was a moderate but significant decrease in the amplitude of the
calcium response "CAG).88F=.=9$ /CG=.0;F=.=)$ 5MG=.)8F=.=)# compared to
conditions in absence of 0-A5$ together with a slight decrease of the oscillation
frequency "Figure 0A8C#. ,n addition 0-A5 on naive T-cells seeded onto '!-9 ,-A4
did not show any significant impact on calcium fluxes "CAG).0:F=.=6$ /CG=.)+F=.=0$
5MG=.)6F=.=0#$ remaining lower to the calcium response in presence of antigenic
peptides. Altogether upon blockade of the /A channel activity by 0-A5$ we evidence
that the !' dependent calcium entry plays a limited role in mouse naive T-cells upon
T/ triggering by-M&-peptide.
9aive CD&: cells isplay mainly intracellular calcium oscillations upon antigenic
challenge. "A# Barcoing o! stimulation.' !DiscussionThe first aim of this study was to significantly increase the sensitivity$ accuracy$
completeness and statistical reliability of video-microscopy approaches to record calcium
fluxes$ by combining a strong calcium probe with a robust algorithm for high density cell
tracking coupled to an automated interface for rigorous post-acquisition analysis. The
second ob7ective was to use this method to describe the characteristic parameters ofintracellular calcium fluxes within a population of T cells in response to different stimuli.
We highlighted the heterogeneous nature and dynamics of these fluxes after T/
engagement by its natural ligand in a cell>cell context$ which cannot be documented by
flow cytometry. The T/-M&-peptide is a paradigm for unconventional intercellular
receptor-ligand interaction (8;* based upon successive cycles of engagement>release ()8*$
()9*. &owever$ more functional data supporting this current view are needed$ taking care
to account for the free motility of cells prospecting for cognate antigens supported by
M& molecules in a 0< cell membrane environment. 'ur goal was to develop
experimental tools to contribute to the understanding of these mechanisms. The first
challenge was finding a bright and stable fluorescent calcium probe in the visible range of
the spectrum and that was easy to monitor both by cytometry and on conventionalconfocal microscopes.
T cells dedicated to calcium imaging are usually loaded with calcium indicators the
emitted fluorescence of which is ?@-shifted upon calcium elevation thus allowing
ratiometric measurements "such as Cura-0#. The sensitivity of these probes can however
be impacted by their intracellular compartmentali%ation "adsorption by proteins$
interaction with membranes or sequestration by organelles e.g. mitochondria#$ or their
extrusion by organic cation transporters. To overcome these technical issues$ loading can
be performed with diluents "pluronic acid# or transporter blockers "probenecid#$ although
these compounds can be noxious to T-cells and thereby affect their response (+:*. <espite
these potential drawbacks$ Cura-0 loaded T cells are routinely used for long experimental
procedure followed by transcriptomics without limitations by reduced cell viability (06*.
,n our study$ cells loaded with 5< 5D were used over extended periods of time without
any evidence of mortality$ compartmentali%ation$ or photobleaching which have been
reported to affect Cluo-2 AM use (+2*. Although they could be considered as anecdotic or
trivial$ such properties enable more reproducibility and the use of 5< 5D in a greater
number of experiments compared to other fluorescent visible probes. 'ur here proposed
MAAA! method incorporates our previously reported MTT algorithm dedicated to
single particle tracking (+0* and nanoscopy (9=*$ (9)* set up to enable the detection$
monitoring and reconnection of tra7ectories of moving T cells acquired by conventional
confocal microscopy. The ability to simultaneously track a great number of targets is in
itself a challenge but in particular encounters difficulties when tracks are interlaced or
crossing over. The performance of MTT was found to be slightly superior compared to
existing algorithms however the implementation of a program of assistance proposing
candidate traces to be reconnected to aborted traces was a ma7or breakthrough in terms of
improving the accuracy and completeness. ,n addition$ during the analysis process$
MAAA! enabled the re7ection if necessary of dead or dividing cells. onsequently$
while MAAA! is not yet a fully unsupervised method$ we speculate that +< time lapse
video-acquisition methods "on a spinning disk confocal microscope$ for example# wouldgreatly reduce the number of aborted traces due to focus loss that occurs on a 0<Btime
acquisition scheme such as that in this study "in particular for primary T cells#.
ompleteness is a ma7or issue in this kind of study$ since the baseline calculation could
be under-estimated or incorrect when the first time points after cell landing are missing.
&ere$ the automated normali%ation of calcium signals facilitates their comparison among
a population of cells. The MAAA! analysis makes simple the analysis and$ more
importantly$ the quantification of signaling. MAAA! deciphers a video sequence in
about )= minutes$ where manual tracking and analysis would take at least 0 hours
"depending on the duration of the time lapse and the number of cells#. @ideo microscopy
records the behavior of individual cells over time and not 7ust part of a population of
anonymous cells. This allowed us to demonstrate that calcium oscillations are highly
diverse among cells ()6* both in terms of intensity and frequency they are mostlytransient oscillations in primary T cells in contact with antigen loaded As. This
diversity in cell responses supports the notion that T-cell triggering is stochastically
linked to heterogeneity in the T cell population (6=*. This is conceivable for T-cell clones
due to genomic drift$ but may seem more surprising for primary <2B T cells. Siterature
reports that oscillatory calcium fluctuations are associated to effector function of T cells
and proliferation (0)* in contrast to memory T cells which display unique increases in
calcium (90*. ,n addition$ sustained calcium responses are observed mainly in apoptotic T
cells (9+*. Altogether$ our approach would be able to reveal in a seemingly homogenous
population$ T cell diversity in terms of function or fate$ based upon antigen dependent
calcium response mode. Another interesting finding is that /A channel dependent
activity does not support a sustained calcium response in naive T cells encountering
As$ and that the predominant calcium response modes in T cells are oscillations$ inagreement with literature (6*$ (;*$ (06*$ (92*$ at least in part sensitive to 0-A5 blockade.
This result supports recent works showing an intriguing role of voltage dependent a0B
channels "av).2# (88* 1 (8:* in the calcium influx into naive T cells. 'ur results also
suggest that membrane calcium channel openings are tightly correlated to /-calcium
waves upon T/ triggering (96*$ (98*$ and that sustained calcium fluxes such as those
triggered by stimulating antibodies and revealed by flow cytometry or video imaging are
not strictly physiologic$ at least not in naive T-cells. ,t could be valuable to consider our
results in the light of recent evidence suggesting a role for cytoplasmic calcium sustained
elevation in the orientation of the cytoplasmic domains of the <+ chains of the
T/><+ complex upon activation (99*.
As a ma7or conclusion$ the introduction of MAAA! emphasi%es the urgent need to
7/17/2019 Barcoding T Cell Calcium Response Diversity With Methods for Automated and Accurate Analysis of Cell Signals
record the effects of cell-to-cell stimuli using real-time videos. We believe that MAAA!
holds huge scope that could be easily adapted to study various kinds of targets "such as
dots$ vesicles$ cells$ animals# based on various types of emitted signal$ however one
immediate application would be to compare our in vitro results to 0-photon imaging of
calcium indicator-loaded T cells migrating in lymph nodes (6+*$ (86*.
' !Materials and MethodsReagents and antiodies
0-Aminoethoxydiphenylborate "0-A5# ")= HM final concentration used for hybridomas$
0= HM for naive T cells# and thapsigargin ") HM final concentration# were purchased
from albiochem$ and ,onomycin "=.) Hg>mS$ final# from !igma. The 5D calcium
assay kit$ the antibody against <+I "clone )26-0))# "8.6 Hg>ml$ final#$ 0)) biot ")=
Hg>mS final# and the C0+.) anti-T/ @U : )-+ antibody ")= Hg>mS$ final# were supplied
by 5ecton <ickinson. The mitochondrion label$ Mitotracker red MD-/os$ and the
calcium indicators ,ndo-) AM$ Cluo-2 AM and Cura /ed AM were supplied by Sife
technologies "Molecular probes#. !treptavidin "6 Hg>mS# was supplied by Eackson
,mmunoresarch.2&+ "anti ,-Ak-&S#$ K4).6 "anti <2#$ and &);+.)8.+ "anti <26#
")= Hg>mS$ final# monoclonal antibodies were produced and purified in the lab from
hybridoma supernatants according to standard protocols.
Cell culture
+A; hybridoma T <2B cells are specific for hen egg lyso%yme peptide "&S# bound to
M& ,, ,-Ak molecules (9:*. These cells were cultured in /M, medium supplemented
with 6O C!$ ) mM sodium pyruvate and )= mM &epes.
'!-9 cells were cultured in <MM medium with 6O C!$ ) mM sodium pyruvate and
)= mM &epes. xperimental antigen-presenting cells "As# were generated by stablytransfecting '!-9 cells "Amaxa$ @ solution$ A=02# with plasmid c<3As coding for the
V chain of M& ,, and the U chain of M& ,, ,-A4 alone or covalently fused to a peptide
derived from &S "provided by <.A. @ignali# (28*. ells were sorted according to their
positivity to surface labeling by 2&+ antibodies "Cacsvantage$ 5ecton <ickinson#. A
monolayers were generated by seeding 6.6 )=2 cells into poly-S-Sysine-coated :-well
Sab-Tek chamber "3unc#.
!pleens and lymph nodes were recovered from 5A>E x +& non-transgenic mice and
+A; T/ transgenic mice (8+*$ (82*. After the extraction of cells onto nylon membrane in
<MM C)0 medium "Son%a#$ splenic erythrocytes were removed via 3&2l lysis. <2B
T cells were isolated by depleting the <2 negative cells according to manufacturer
BD B*6 )=2 T cells per well were plated in ;8 well plates in )== HS of complete medium. ells
were loaded with 5< 5D diluted in )D dye loading solution "according to manufacturer
instructions# at )>)===e "∼) HM# for +A; cells ")>)888 for primary T cells i.e. ∼=.8
HM#. )== HS of this solution were dispatched to each well before incubation at +9
during ) hour "+= min for primary T cells# in the dark. ells were then washed twice in
&ankJs balanced salt solution "&5!!# &epes buffer containing ) mM calcium and
resuspended in the same medium. Cive wells were pooled "0.6 )=6 cells# and analy%ed
either by flow cytometry or microscopy. When mentioned$ cells were resuspended in
&5!! without a0B>Mg0B$ &epes ) mM. 0-A5 was added 7ust before recordings.
Additional experiments were performed by adding Cura /ed to the 5< 5D dye loading
buffer to a final concentration of 6.6 HM. Alternatively$ Cluo-2 AM was added to the
loading medium in the presence or in the absence of dye loading solution to a final
concentration of )= HM.
!ndo"+ AM0.6 )=8 cells were resuspended in 6== HS of complete &5!! buffer "&5!! without
calcium$ without Mgl0$ supplemented with ) mM al0$ ) mM Mgl0$ =.)O 5!A and
) mM &epes# with 6 HS ,ndo-) AM ")= HM# and incubated for += minutes at +9. ells
were washed twice in warm complete &5!! "+9# and ad7usted to the concentration of
6 )=6 cells>mS in complete &5!! before acquisition.
,low cytometry
ells were analy%ed on a S!/ , flow cytometer "5ecton <ickinson# with ell uest
software or S!/ ,, "for Cura /ed>5< 5D and ,ndo-) AM acquisitions# using the
CA!<iva software. 5D calcium indicator was observed over time on the CS) channel
with an excitation by an Argon laser 2:: nm and a 6+=>+= nm emission filter at +9$
maintained using a water bath. <ata analysis was performed with ClowEo software and
the median intensity of fluorescence was plotted vs. time after exclusion of dead cells and
cell debris.
Confocal microscopy
Movies were made on a Xeiss S!M 6)= Meta confocal microscope equipped with a +=
mW argon laser "06O output$ )O A'TC#. ictures were taken with a -Apochromat
2=Y>).0 water immersion ob7ective$ using the 2:: nm line of the argon laser$ &CT
?@>2:: dichroic mirror and a 6=6 nm long pass filter at +9$ maintained using a hot
plate. Time-lapse movies were composed of +== images "6)0Y6)0 pixels : bit 006
HmY006 Hm pinhole set to + airy units# taken every 9 seconds. Additional observationswere performed on an ?ltraview @oD erkin lmer spinning disk confocal microscope.
MAAACS analysis
All scripts$ including multiple target tracking "MTT# (+0*$ were developed under Matlab
"The Mathworks#. The source code of MTT$ deposited at the Agence pour la Protection
des Programmes$ n ,<<3.C/.==).09==0).=== !..0==:.===.+)0+=$ is freely available
for research purposes at httpN>>www.ciml.univ-mrs.fr>lab>he-marguet.htm. ell tracking
and automated analysis of cell signals with MAAA! can be done either in command
line "directly in Matlab# or using a graphical user interface "K?,#. While the K?, is more
intuitive it is limited to the analysis of a single acquisition whereas the command line
solution permits the sequential analysis of several video-acquisitions.
seeded onto '!-9 experimental antigen presenting cells with peptides "see materials and
methods# "nG:#. "
B# !uperimposed tra7ectories of tracked +A; T cell hybridomasnormali%ed to their starting coordinates "each cell trace has been plotted in a randomly
chosen color#. ?pon stimulation with A$ the traces were plotted separately to
distinguish the behavior of unactivated from activated cells. "C# Analytical parameters
!or velocity uner i!!erent stimuliN The mean value "B>Z !M# for each analytical
parameter "velocity$ mobile fraction## is shown in red for each activated cell depicted as a
dot on each scatter plot.
"!#
lick here for additional data file."2.:M$ eps#
,igure S1
Diversity o! correlation between cell velocities an calcium in!lu$ in cell
populations. The velocity of a set of individual T cell was concomitantly color coded
with fluorescence intensity and displayed as a function of time. "A# anel of non-
activated or activated +A; T cell hybridomas. "B# anel of non-activated or activated
primary naive <2B +A; T cells.
"!#
lick here for additional data file.").8M$ eps#
,igure S2
C5AC channel activity in naive cells. "A# )!!ect o! )DA on the calcium responsewithin naive CD&: cells. 3aive <2B T cells were stimulated by anti <+ antibody
and calcium mobili%ation was measured by flow cytometry in the presence of increasing
concentrations of <TA. "B# ;lobal an intracellular mobili4ation o! calcium in naive
CD&: cells. The effect of 0-A5 on calcium mobili%ation was measured by flow
cytometry in naive <2B T cells after stimulation by thapsigargin. "C# Comparison o!
the e!!ects o! )DA an *%APB on intracellular calcium mobili4ation. The effects of
0-A5 and <TA were compared by flow cytometry in naive <2B T cells after
T/-peptide-M& interactions in situ show accelerated kinetics and increased affinity.
3ature 28+N ;8+1;89 (M free article* (ubMed*
):. ,rvine <E$ urbhoo MA$ 4rogsgaard M$ <avis MM "0==0# <irect observation of
ligand recognition by T cells. 3ature 2);N :261:2; (ubMed*
);. &use M$ 4lein S'$ Kirvin AT$ Cara7 EM$ Si E$ et al. "0==9# !patial and temporaldynamics of T cell receptor signaling with a photoactivatable agonist. ,mmunity 09N 981
:: (ubMed*
0=. ,mboden E5$ !tobo E< ");:6# Transmembrane signalling by the T cell antigen
receptor. erturbation of the T+-antigen receptor complex generates inositol phosphates
and releases calcium ions from intracellular stores. E xp Med )8)N 2281268 (M free
synapses formed between T cells and dendritic cells. 3at ,mmunol 0N ;061;+) (ubMed*
8+. &o W]$ ooke M$ Koodnow $ <avis MM ");;2# /esting and anergic 5 cells are
defective in <0:-dependent costimulation of naive <2B T cells. E xp Med )9;N
)6+;1)62; (M free article* (ubMed*
82. Tourne !$ 4ouskoff @$ &o W$ <avis M$ 5enoist $ et al. ");;;# ,nhibition of
thymocyte positive selection by natural M&N peptide ligands. ur E ,mmunol 0;N +;21
2=0 (ubMed*
86. Criedman /!$ 5eemiller $ !orensen M$ Eacobelli E$ 4rummel MC "0=)=# /eal-timeanalysis of T cell receptors in naive cells in vitro and in vivo reveals flexibility in synapse
and signaling dynamics. E xp Med 0=9N 09++1092; (M free article* (ubMed*
88. 'milusik 4<$ 3ohara SS$ !tanwood !$ Eefferies WA "0=)+# Weft$ Warp$ and WeaveN
The ,ntricate Tapestry of alcium hannels /egulating T Symphocyte Cunction. Cront
,mmunol 2N )82. (M free article* (ubMed*
89. 'milusik 4$ riatel EE$ hen D$ Wang ]T$ Du &$ et al. "0=))# The a"v#).2 calcium
channel is a critical regulator of T cell receptor signaling and naive T cell homeostasis.
,mmunity +6N +2;1+8= (ubMed*
8:. 3iemeyer 5A$ &oth M "0=))# xcitable T cellsN a"v#).2 channel contributions and
controversies. ,mmunity +6N +)61+)9 (ubMed*
8;. @alitutti !$ oombs <$ <upre S "0=)=# The space and time frames of T cell activation
at the immunological synapse. C5! Sett 6:2N 2:6)12:69 (ubMed*