2012 Colinet et al Funct Ecol .pdf

File information

Original filename: 2012 Colinet et al Funct Ecol.pdf
Title: Exploring the plastic response to cold acclimation through metabolomics

This PDF 1.3 document has been generated by Arbortext Advanced Print Publisher 9.0.114/W / PDFlib PLOP 2.0.0p6 (SunOS)/Acrobat Distiller 7.0 (Windows), and has been sent on pdf-archive.com on 07/07/2015 at 20:33, from IP address 93.182.x.x. The current document download page has been viewed 851 times.
File size: 1.1 MB (12 pages).
Privacy: public file

Download original PDF file

2012 Colinet et al Funct Ecol.pdf (PDF, 1.1 MB)

Share on social networks

Link to this file download page

Document preview

Functional Ecology 2012, 26, 711–722

doi: 10.1111/j.1365-2435.2012.01985.x

Exploring the plastic response to cold acclimation
through metabolomics
Herve´ Colinet*,1,2, Vanessa Larvor2, Mathieu Laparie2 and David Renault2

Earth and Life Institute ELI, Biodiversity Research Centre BDIV, Catholic University of Louvain, Croix du Sud 4-5,
B-1348 Louvain-la-Neuve, Belgium; and 2Universite´ de Rennes 1, UMR CNRS 6553 Ecobio, 263 Avenue du Ge´ne´ral
Leclerc CS 74205, 35042 Rennes Cedex, France

1. Adaptive responses to thermal stress typically involve a range of plastic acclimatory responses
in ectothermic animals. The mechanisms underlying phenotypic plasticity in inducible cold tolerance are complex and not fully understood.
2. Here we investigated how thermoperiodic cold acclimation affected the cold tolerance and
the metabolome of adult Drosophila melanogaster. We have used targeted GC ⁄ MS metabolomic
profiling to address whether cold acclimation induced specific metabolic changes and affected
the dynamics of the homeostatic response following different types of cold stress (acute and
3. Developmental combined with gradual adult acclimation strongly promoted cold tolerance.
This phenotypic variation was associated with significant metabolic changes, among which some
sugars, polyamines and metabolic intermediates are fingerprints of these changes. Cold acclimation allowed individuals to maintain metabolic homeostasis, whereas non-acclimated counterparts suffered from deep and persistent homeostatic perturbations.
4. This study gives a fertile ground for future research in disentangling the role of several metabolites putatively involved in cold acclimation and cold stress response. It also provides insight
into the mechanisms by which cold acclimation is achieved in D. melanogaster and gives a basis
for elucidating the evolution of plastic responses to thermal variations.
Key-words: acclimation, cold stress, Drosophila melanogaster, metabolomics, recovery,
phenotypic plasticity

Temperature affects all aspects of the biological organization of organisms and this is particularly true for coldblooded (ectothermic or poikilothermic) animals. These
organisms possess diverse responses for dealing with thermal stress, they can adapt genetically (i.e. long evolutionary
process) and ⁄ or they can acclimate (i.e. phenotypic adjustments) (Fischer & Karl 2010). Seasonal acquisition of cold
tolerance is a long-term process (weeks or months) that generally involves thermal and non-thermal cues, such as
humidity or photoperiod. Overwintering process has been
widely studied in arthropods (e.g. Leather, Walters & Bale
1993; Chown & Terblanche 2006). This phenomenon
involves slow changes in physiology and biochemistry
before a long-term cold protection is realized (Korsloot,
Van Gestel & Van Straalen 2004). Seasonal changes in cold
tolerance involve quantitative and qualitative variations in
*Correspondence author. E-mail: herve.colinet@uclouvain.be

compatible solutes (i.e. polyols, sugars or amino acids) (Lee
1991, 2010; Leather, Walters & Bale 1993), synthesis of
diverse proteins such as heat shock proteins (HSPs) or antifreeze proteins (AFPs) (Korsloot, Van Gestel & Van Straalen 2004; Lee 2010) and membrane remodelling (reviewed
by Kosˇ ta´l 2010). In spite of this knowledge, a good picture
of how Drosophila melanogaster (Diptera: Drosophilidae)
builds up cold tolerance and survives winter low temperatures has not yet been established (Korsloot, Van Gestel &
Van Straalen 2004; Doucet, Walker & Qin 2009).
Like many species, D. melanogaster has the capacity to
enhance thermotolerance in response to a pre-exposure to
sub-lethal temperature (Hoffmann, Sørensen & Loeschcke
2003), a phenomenon referred to as thermal acclimation
(Angilletta 2009). Acclimatization refers to the same process
but when occurring under natural conditions (Chown &
Nicolson 2004). Laboratory studies have shown that D. melanogaster possesses a range of plastic responses for dealing
with cold stress. These include developmental, gradual and
rapid acclimation (refer to Angilletta 2009 for definitions)

2012 The Authors. Functional Ecology 2012 British Ecological Society

712 H. Colinet et al.
that differ according to the timing and the length of the preexposure (Hoffmann, Sørensen & Loeschcke 2003; Rako &
Hoffmann 2006; Colinet & Hoffmann 2012). Different
acclimatory treatments affect different proxies of D. melanogaster cold tolerance, and when applied in combination, these
treatments promote cold tolerance considerably (Rako &
Hoffmann 2006; Colinet & Hoffmann 2012). The underlying
mechanisms of rapid acclimation (also called rapid cold hardening, RCH) have been rather well characterized (reviewed
by Lee & Denlinger 2010). In contrast, much less is known
about those related to gradual and developmental cold acclimation (Chown & Terblanche 2006). In this study, we focused
on these two acclimation responses by exposing D. melanogaster to different fluctuating thermal conditions during immature development and adult life.
Over the past years, the so-called ‘omics’ techniques have
emerged as powerful tools for studying organism–environment interactions. Changes in metabolite levels can provide
biochemical fingerprints of the integrated response of an
organism to environmental stressors (Bundy, Davey & Viant
2009). It has long been known that biochemical changes occur
during cold acclimation (Lee 1991, 2010; Leather, Walters &
Bale 1993; Chown & Nicolson 2004); however, the degree to
which a cold stress alters the metabolome and how cold acclimation affects the metabolic disruption remain a matter of
speculation. So far only a few metabolomics studies have been
interested in cold stress in terrestrial invertebrates (reviewed
by Bundy, Davey & Viant 2009). These have mainly focused
on freezing (Bundy, Ramlov & Holmstrup 2003; Hawes et al.
2008; Michaud et al. 2008; Bundy, Davey & Viant 2009) or
RCH (Michaud & Denlinger 2007; Overgaard et al. 2007),
while the response to longer acclimation treatments has not
been studied to the same extent.
System-wide coordination of metabolic fluxes is crucial to
maintain metabolic homeostasis in multicellular organisms.
Environmental stresses, such as temperature, can transiently
disturb cellular homeostasis, and in response to this, cells activate a so-called cellular homeostatic response which aims to
restore homeostasis progressively (Ku¨ltz 2005). The mechanisms behind recovery from cold stress are complex, and it
seems that more genes ⁄ proteins are activated during the
recovery than during the actual period of stress (Clark &
Worland 2008). It is thus essential to differentiate between the
cold exposure and the subsequent recovery phase (Colinet,
Lee & Hoffmann 2010a). In metabolomics, it is of paramount
importance to measure temporal patterns because metabolic
variations can occur very rapidly (Nicholson et al. 2002).
Malmendal et al. (2006) have underlined that the ‘metabolic
trajectories’, defined as the temporal disruption in the multivariate metabolic space following a stress, can serve as a sensitive tool for monitoring homeostasis status.
In the present study, we assumed that combined
acclimatory treatments (i.e. developmental and gradual)
will strongly promote cold tolerance, and hypothesized
that this plastic change may be associated with detectable
metabolic changes. More precisely, we made four explicit

1. Accumulation of cryoprotective sugars or free amino acids
(FAAs). Increased cold tolerance, including that afforded
by RCH, is correlated with the production of polyols in
some species (Lee 1991; Storey & Storey 2005), but there is
no evidence of the role of polyols in promoting cold tolerance of D. melanogaster (Kelty & Lee 2001; Overgaard
et al. 2007). Other compatible solutes such as sugars and
FAAs could contribute to cold tolerance (Lalouette et al.
2007; Overgaard et al. 2007; Kosˇ ta´l, Zahradnı´ cˇkova´ &
Sˇimek 2011). We thus tested whether these putative
metabolites (sugars and FAAs) are involved in the plastic
changes resulting from developmental and gradual acclimation in D. melanogaster.
2. Modulation of energy production. Regulation of metabolism by changes in mitochondrial number and capacities
often take place during cold acclimation (e.g. Lefebvre &
Fourche 1985; Joanisse & Storey 1995; McMullen & Storey 2008). We thus expected that cold acclimation would
result in altered levels of intermediary metabolites
involved in energy production.
3. Differential homeostatic response. Heat-hardening protects Drosophila flies from a subsequent heat stress (Hoffmann, Sørensen & Loeschcke 2003) and it was previously
shown that heat-hardened flies were able to restore the
general homeostasis much faster than control flies following exposure to a severe heat stress (Malmendal et al.
2006). Likewise, we hypothesized that cold acclimation
will allow fast homeostatic regeneration or favour metabolic maintenance. We thus expected post-cold stress metabolic trajectories to deviate little versus markedly from
the control state in cold- versus non-acclimated adults,
4. Stress-specific metabolic disorder. The nature of chilling
injuries resulting from acute and chronic cold stress is supposed to be different (e.g. Chown & Nicolson 2004). Acute
cold stress likely induces severe damages to proteins and
membranes, while chronic cold stress causes progressive
metabolic disorders (e.g. loss of ion homeostasis) (see
Chown & Nicolson 2004; Kosˇ ta´l, Vambera & Bastl 2004;
Lee & Denlinger 2010). Based on these differential mechanisms, we hypothesized that metabolic disorders will vary
according to the type of cold stress inflicted, with an acute
cold stress presumably producing more severe perturbation than chronic cold stress.

Materials and methods

We conducted our experiments on a mass-bred D. melanogaster line
derived from the mix of two wild populations collected in October
2010 at Plancoe¨t and Rennes (Brittany, France). The flies were maintained in laboratory by rearing several hundred individuals in
uncrowded conditions (by restricting the oviposition period and density of adults) in 200 mL bottles at 25 ± 1 C (16L : 8D) on standard
fly medium consisting of sugar, brewer yeast and agar. Flies have been
in the laboratory for five generations at the time of experiments.

2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

Metabolomics of cold acclimation response 713

Flies were acclimated using two combined treatments: developmental
followed by gradual acclimation (DA + GA) (see Colinet & Hoffmann 2012). To generate flies for the experiments, groups of forty
6-day-old females were allowed to lay eggs in 200 mL bottles containing medium during a restricted period of 12 h. This semi-controlled
procedure allowed the flies to develop under uncrowded conditions.
Females were then removed and bottles with eggs were randomly
placed under either cold or moderately warm thermoperiod to continue development until adult emergence. We used thermoperiod
because it enhances cold hardiness and cryoprotectant synthesis more
than constant temperatures (Kosˇ ta´l, Sˇlachta & Sˇimek 2001). The temperature fluctuated from 10 to 20 C or from 18 to 28 C (using
12 : 12 h cycles) for the cold or warm thermoperiod, respectively. The
abbreviations ‘cold-acclimated’ (CA) and ‘warm-acclimated’ (WA)
are used to distinguish the two experimental conditions, acknowledging that temperatures between 18 and 28 C should be considered
more as benign than really warm for this species (Hoffmann 2010).
A 12L : 12D photoperiod was used with the scotophase occurring
during the cold period. Similar acclimatory conditions had previously
been used and successfully promoted cold tolerance (Colinet & Hoffmann 2012). Flies took 11 and 28 days to emerge under warm and
cold conditions, respectively. After development, newly hatched
adults were kept under the same rearing conditions for 5 days (i.e.
additional adult gradual acclimation). Adult flies were tested for cold
tolerance and for metabolic profiling at the end of a thermoperiod
cycle (i.e. when temperature was at a minimum in a treatment). All
assays were performed using females that had been sexed visually
without CO2 anaesthesia using an aspirator. Programmed thermoregulated incubators (Model MIR-153; SANYO Electric Co. Ltd,
Munich, Germany) were used and temperature was checked using
automatic recorders (Hobo data logger, model U12-012, accuracy
±0Æ35 C; Onset Computer Corporation, Bourne, MA, USA).


Cold tolerance was compared between CA and WA flies. Tolerance
to both acute and chronic cold stress was assessed using the method

previously described by Colinet, Lee & Hoffmann (2010b). Briefly,
tolerance to acute cold stress was scored after an exposure at )5 C
for 2 h. For each acclimation treatment (WA versus CA), eight sets of
25 flies were randomly constituted from eight independent rearing
bottles. These sets were separately placed in 42 mL glass vials and
then immersed for 2 h in a glycol solution cooled to )5 C for 2 h.
After the stress, the flies were returned to 25 ± 1 C (16L : 8D) to
recover with food for 24 h. Flies were then categorized as: dead (no
movement), knockout (alive but not able to stand on their legs) or
active (walking with coordinated movement). For each type of cold
stress (acute and chronic), proportions among the three categories
were compared between WA and CA flies using Chi-square contingency tests. To assess chronic cold tolerance, 50 females were exposed
to 0 C for 18 h by placing a vial in a cold incubator (Model MIR153; SANYO Electric Co. Ltd.). Flies were then allowed to recover at
25 ± 1 C (16L : 8D) and recovery times were individually recorded.
Flies were considered recovered when they stood up. Data were used
to generate temporal recovery curves which were compared with
Mantel–Cox analysis using PRISM V 5.01 (GraphPad software, Inc.
2007) (see Colinet, Lee & Hoffmann 2010b). Chronic cold tolerance
was also assessed using survival assays. Eight independent sets of 25
flies were exposed to chronic cold stress (0 C for 18 h) and then kept
on food at 25 ± 1 C (16L : 8D) for 24 h before being scored as dead
or knockout or active, as described above.


To assess the effects of acclimation on metabolic profiles, we sampled flies at the end of each thermoperiodic acclimation period, just
before the stress exposure (control basal condition, see Fig. 1). In
addition, metabolic profiles were compared during the stress by
sampling flies at the end of the acute ()5 C ⁄ 2 h) and the chronic
(0 C ⁄ 18 h) cold stress (stress conditions, Fig. 1). Finally, to analyse whether the dynamics of homeostatic response varied according to the acclimation (CA versus WA) and the type of cold stress
(chronic versus acute), samples were taken after 1, 2 and 4 h of
recovery at 25 ± 1 C (16L : 8D) (recovery conditions, Fig. 1).





Fig. 1. Schematic diagram of the experimental design. Moderately warm (red lines, panels a & b) or cold thermoperiods (blue lines, panels c & d)
were used to acclimate Drosophila flies during development and as adults. After the acclimation, the flies were exposed to either a chronic (0 C
for 18 h, panels a & c) or an acute ()5 C for 2 h, panels b & d) cold stress. Arrows indicate sampling times for metabolomic analysis: (i) at the
end of acclimation period (basal control condition), (ii) at the end of the stress and (iii) after 1, 2 and 4 h of recovery.
2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

714 H. Colinet et al.
For each sampling point, flies were removed from experimental
conditions, directly snap-frozen in liquid nitrogen and stored at
)80 C until extractions. Flies were checked under stereomicroscope during the recovery period to ensure that they were not dead
when sampled for metabolomics. For each condition, eight biological replicates, each consisting of a pool of 15 females, were used.
Each sample was weighed (fresh mass) using a Mettler micro-balance (Mettler Toledo, Model UMX2, Greifensee, Switzerland;
accurate to 0Æ001 mg) before the extractions.


The samples were homogenized in 600 lL of cold ()20 C) methanolchloroform solution (2 : 1) using a tungsten-bead-beating apparatus
(Retsch MM301; Retsch GmbH, Haan, Germany) at 25 agitations
per second for 1Æ5 min. Then, 400 lL of ice-cold ultrapure water was
added to each sample and vortexed. After centrifugation at 4000 g for
5 min at 4 C, two aliquots of the upper aqueous phase (which contained polar metabolites) were transferred to new chromatographic
glass vials: one containing 300 lL of extract and another with 30 lL.
The 300 lL aliquot (i.e. 10 · concentrated extract) was used to quantify the majority of metabolites, whereas the 30 lL aliquot was used
to quantify the few very abundant compounds. Therefore, for each
individual sample, two distinct runs of GC-MS were performed. The
vials containing the aliquots were vacuum-dried using a Speed Vac
Concentrator (MiVac; Genevac Ltd., Ipswitch, England). Samples
were then resuspended in 15 lL of 20 mg mL)1 methoxyaminehydrochloride (Sigma-Aldrich, St. Louis, MO, USA) in pyridine before
incubation under automatic orbital shaking at 40 C for 90 min.
Then, 15 lL of N-methyl-N-(trimethylsilyl) trifluoroacetamide
(MSTFA; Sigma-Aldrich, Saint Quentin, France, #394866) was
added to make a total volume of 30 lL and the derivatization was
conducted at 40 C for 30 min under agitation. All the derivatization
process was automatized using CTC CombiPal autosampler (GERSTEL GmbH & Co.KG, Mu¨lheiman der Ruhr, Germany), ensuring
identical derivatization time and process for all samples.


The GC-MS system consisted of a Trace GC Ultra chromatograph
and a Trace DSQII quadrupole mass spectrometer (Thermo Fischer
Scientific Inc., Waltham, MA, USA). The injector temperature was
held at 250 C. The oven temperature ranged from 70 to 147 C at
9 C min)1, from 147 to 158 C at 0Æ5 C min)1, from 158 to
310 C at 5 C min)1, and then the oven remained 4 min at
310 C. We used a 30 m fused silica column (TR5 MS, I.D.
25 mm, 95% dimethyl siloxane, 5% Phenyl Polysilphenylene-siloxane) with helium as the carrier gas at a rate of 1 mL min)1. One
microlitre of each sample was injected using the splitless mode
(25 : 1). We completely randomized the injection order of the samples to avoid generating data skewed by a possible instrument drift.
The temperature of the ion source was set at 250 C and the MS
transfer line at 300 C. Detection was achieved using MS detection
in electron impact. GC-MS full scan mode is generally used for the
screening of unknown compounds. In the present work, we used
the selective ion monitoring mode (SIM) (electron energy: )70 eV),
ensuring a precise annotation of the detected peaks. In the SIM
mode, the MS instrument is set to only look for specific masses of
interest (two or three ions) rather than looking for all masses over
a wide range. SIM analysis provides more sensitivity than full scan
analysis but it only provides information regarding targeted metab-

olites (Waller et al. 2007). We thus only searched for the metabolites that were included in our spectral data base, which currently
includes 60 pure reference compounds. The peaks were accurately
annotated using both their mass spectra (two specific ions) and
their retention index. One quality control containing the 60 pure
compounds at 200 lM was run every 15 samples to check for
instrument performance and control for instrument drifts (Fiehn
et al. 2008). Calibration curves were set using samples consisting of
60 pure reference compounds at levels of 10, 20, 50, 100, 200, 500,
700 and 1000 lM. Chromatograms were deconvoluted using XCalibur v2.0.7 software (Thermo Fischer Scientific Inc., Waltham, MA,
USA). Metabolite levels were quantified according to their calibration curves. Arabinose was used as internal standard to account for
potential loss during sample preparation and injection (Fiehn et al.
2008). Calculated concentrations were adjusted according to their
internal standard. Finally, the concentrations were reported according to the fresh mass of each sample.


For the metabolic profiles of the control basal conditions in CA versus WA treatments (see Fig. 1), all metabolite levels were first compared using t-tests (a = 0Æ05) to determine those that varied
according to acclimation. All metabolites were ranked in a volcano
plot according to their statistical P-value (t-test) and their difference
of abundance (i.e. fold change). A PCA was then performed on all the
significant metabolites to detect those contributing the most to the
structure. For the comparison of temporal metabolic profiles among
the different conditions, the main patterns and significance were
searched by performing a between-class PCA analysis. This PCA
focuses on the differences among the classes defined as qualitative
instrumental variable (Dole´dec & Chessel 1987). A Monte Carlo test
was then performed to examine the significance of the difference
between the classes (Romesburg 1985). A dendrogram was also created using Hierarchical Ascendant Classification (HAC) with the centroid projections on the first principal component (PC1) as data, and
using Ward’s aggregation method. All data were scaled and meancentred prior to the PCAs. All analyses were performed using the
‘ade4’ library in the statistical software ‘R 2.13.0’ (R Development
Core Team 2008). Finally, three-way ANOVAs were also reported for
individual metabolites with acclimation (WA versus CA), stress
(acute versus chronic) and time (0–4 h) as factors (see Supporting


Thermoperiodic cold acclimation significantly promoted
both chronic and acute cold tolerance. CA flies recovered
much faster than WA flies (Mantel–Cox Test: v2 = 118Æ1;
d.f. = 1; P < 0Æ0001; Fig. 2a). Mortality assessed 24 h after
chronic cold stress was also significantly reduced by cold
acclimation (v2 = 157Æ51; d.f. = 2; P < 0Æ0001; Fig. 2b); a
high proportion of flies had fully recovered (active) when cold
acclimation was applied, while many flies were either dead or
knocked out without cold acclimation. Finally, mortality
assessed 24 h after acute cold stress was also significantly
reduced by cold acclimation (v2 = 76Æ45; d.f. = 2;
P < 0Æ0001; Fig. 2c), with a high proportion of active flies

2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

Metabolomics of cold acclimation response 715



Fig. 2. Cold tolerance assessments. Panel (a) shows the temporal recovery curves of cold-acclimated (CA) (blue) versus warm-acclimated (WA)
(red) adults exposed to chronic cold stress. Each dot represents the mean proportion of recovering flies in relation to time after cold stress (±SE,
n = 50). Cold tolerance of CA (blue) versus WA (red) flies assessed 24 h after chronic (panel b) or acute cold stress (panel c). Individuals were
categorized as dead, knockout or active. Bars represent mean percentages (±SE) derived from eight replicated groups of 25 flies (n = 8).

when cold acclimation was applied against a high mortality
rate without acclimation.


The separation of polar metabolites from female D. melanogaster yielded well-separated peaks and chromatograms were
generally consistent across samples. Among the 60 metabolites included in our library, some of them had to be discarded
for different reasons: (i) signal ⁄ noise ratio inferior to 10, or
(ii) concentration under the quantification limit, or (iii) inability to obtain an exploitable calibration curve (see Fiehn et al.
2008 for details on quality requirements). Finally, 35 identified metabolites were conserved for quantitative metabolic
profiling. Among these, we found 11 FAAs, six sugars, seven
polyols, seven metabolic intermediates and four amines or
diverse metabolites (see Table S1 for list of compounds and


Among the 35 metabolites detected, 16 varied significantly
between CA and WA treatments (P < 0Æ05) (Fig. 3). A number of sugars (Suc, Fru, Tre), polyols (erythritol, sorbitol),
FAAs (Val, Leu, Gly) and polyamines (Cad, Put) were more
abundant in CA flies, while some metabolic intermediates,
such as fumarate, malate, citrate and glycerate, were less
abundant in CA flies (Fig. 3). All the metabolites were represented in a volcano plot (enclosed within Fig. 3). On this synthetic representation, metabolites located on the left side were
on average less abundant in CA flies, while those located on
the right side were more abundant in CA flies. Malate, fumarate, Suc, Fru and sorbitol appeared clearly off-centred,
meaning high magnitude variations between CA and WA flies
(Fig. 3). A PCA was performed using these 16 significant
metabolites and resulted in a clear-cut separation between the
two treatments along the PC1 (Fig. 4a). PC1 accounted for
51Æ1% of total inertia. The metabolites that contributed the
most to this separation were Suc, Fru, sorbitol, malate, citrate, fumarate, glycerate and Cad (see Fig. 4b). PC2 and PC3
accounted for 12Æ2% and 9Æ8% of total inertia, respectively
(Fig. 4c), and mainly represented within-treatment inertia.


System-wide temporal metabolic responses were characterized using between-classes PCA on the full data set, with
respect to an instrumental variable corresponding to the
combination of acclimation treatment (WA or CA), type
of cold stress (chronic or acute) and time (0, 1, 2 or 4 h).
The ordination of classes within the first plane (PC1,
PC2) is represented in Fig. 5a, where the PC1 illustrates
the time-course of the metabolic response (mainly for
WA treatments) and PC2 shows a clear opposition
between the two acclimation treatments (WA versus CA).
PC1 and PC2 accounted for 56Æ1% and 17Æ2% of the
total inertia, respectively. The subsequent PCs were not
considered as they accounted for only small proportion of
the inertia: 6Æ3%, 5Æ5% and 4Æ3% for PC3, PC4 and PC5,
respectively. The structure suggested that the CA treatments varied little along PC1, as the centroids remained
relatively clustered. On the contrary, WA treatments displayed a continuous sequence along PC1, indicative of
increasing temporal metabolic changes. Monte Carlo randomizations (based on 1000 simulations) confirmed the
significance of differences among classes (P < 0Æ001),
which accounted for 47Æ3% of the total inertia. The
metabolites contributing the most to the temporal metabolic changes (PC1), observed mainly in WA treatments,
were erythritol, glucuronate, ribitol, xylitol, ETA, sorbitol,
inositol, Leu and Val (Fig. 5b). The metabolites contributing the most to the separation between acclimation treatments (PC2) were Fru, fumarate, citrate, malate and Cad
(Fig. 5b). The PC1 scores of each treatment were plotted
against the time in Fig. 5c. From this figure, it became
clear that metabolic profiles departed progressively from
their initial status (i.e. control) in WA treatments, suggesting a progressive and irreversible loss of homeostasis.
On the contrary, PC1 scores in CA treatments showed little temporal variation, suggesting a capacity to maintain
(or quickly restore) metabolic homeostasis. The three-way
ANOVA performed on PC1 scores revealed that they varied
significantly with the acclimation treatment and the time,
but not with the type of cold stress (see Table S2). The
significant ‘acclimation · time’ interaction indicated that
the time-course of PC1 scores varied according to

2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

716 H. Colinet et al.

Fig. 3. Comparison of metabolite levels assessed by the end of acclimation period (i.e. basal controls, refers to Fig. 1). Quotients of mean content
of cold-acclimated (CA) over warm-acclimated (WA) are plotted (i.e. fold change). Blue bars or red bars for metabolites that were, respectively,
more or less abundant in CA flies. Stars indicate a significant difference between CA and WA treatments (t-test, P < 0Æ05). A volcano plot is
enclosed within this figure; metabolites were ranked according to their statistical P-value (y-axis) and fold change (log2-based) (x-axis). Offcentred metabolites are those that varied the most between CA and WA conditions.




Fig. 4. PCA on the metabolites that varied between cold-acclimated (CA) and warm-acclimated (WA) flies, when sampled at the end of the acclimation period (i.e. control conditions in Fig. 1). Red and blue is used for WA and CA treatments, respectively. A plot of the first two principal
components (PCs) is shown in panel a. Lines link individuals to their respective centroids (n = 8). Projection of the 16 variables on the correlation circle is shown in panel b. The barplot in panel c shows the amount of inertia accounted by successive PCs. Refer to Table S1 for metabolites

acclimation. Also, the significant ‘stress · time’ interaction
translated the fact that the temporal variations of PC1
scores depended on the type of stress. Indeed, it appeared
from the Fig. 5c that the acute cold stress induced a
slightly more severe temporal metabolic disruption than

the chronic stress did. The dendrogram resulting from the
HAC further showed that (i) all CA treatments were scattered throughout a separated cluster, (ii) the temporal
sequence of metabolic disruption of the WA treatments
was manifested within a clearly dissimilar cluster and

2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

Metabolomics of cold acclimation response 717




Fig. 5. PCA of all treatment’s metabolites. Projection of the significant variables (correlation circle: panel b) and of the 18 different treatment
groups (panel a) onto the first factorial plane of the between-class PCA analysis. Lines link individuals to the centroid of their respective treatment
(n = 8). In all panels, red and blue is used for warm-acclimated (WA) and cold-acclimated (CA) treatments, respectively. Panel c illustrates individual scores on PC1 according to the recovery time. Control condition before the stress ()1 h), during the stress (0 h) and during the time-course
of recovery (1, 2 and 4 h) (refer to Fig. 1 for experimental design). Values are mean scores ±SE. Panel c shows the dendrogram derived from
CHA based on PC1 scores, showing dissimilarities between CA and WA treatments. Refer to Table S1 for metabolites abbreviation.

(iii) the WA control and stress had a position closer to
CA treatments (Fig. 5d).


Changes in individual metabolite levels (Fig. 6) were compared with three-way ANOVAs with acclimation, stress and
time as factors (plus interactions; Table S2). We also reported
the mean changes in total polyols, sugars, FAAs, metabolic
intermediates and the total metabolite pool (Fig. S1). Almost
all metabolites showed significant changes owing to either
acclimation or stress or time or interactions among these factors (Table S2). Metabolites that varied according to acclimation included compounds from all families (20 compounds of
35) (see Table S2). Even though this translated an overall difference, combining all sampling times and stress types, we
found the same significant compounds as in the previous
approach (see Figs 3 and 4). Two interesting aspects of these
analyses were the ‘acclimation · time’ and ‘stress · time’
interactions, which indicated whether the time-course of
metabolite variation was differentially affected by acclimation and the type of stress, respectively. Not surprisingly, the
interaction ‘acclimation · time’ was significant in the majority of metabolites (27 compounds of 35) (Fig. 6 and
Table S2), which corroborated our observation that acclimation strongly impacted the temporal homeostatic response.
The temporal patterns of the total levels of polyols and sugars
were also differentially affected by acclimation (Fig. S1 and
Table S2). Progressive and lasting concentration increases

were noted in the WA treatments, while there were only moderate and temporary increases in the CA treatments (Fig. S1).
Finally, we found only a few metabolites that had temporal
patterns differentially affected by the type of stress (eight
compounds of 35) (Table S2). The concerned metabolites
included mainly polyols, FAA and succinate.


Similar to earlier studies on D. melanogaster, we found that
cold acclimation strongly promoted cold tolerance (Hoffmann, Sørensen & Loeschcke 2003; Rako & Hoffmann 2006;
Kristensen et al. 2008; Colinet & Hoffmann 2012). We
hypothesized that this phenotypic variation may be associated with significant metabolic changes. More specifically, we
hypothesized that compatible solutes with assumed cryoprotective functions, such as sugars or FAAs, may contribute to
the increased cold tolerance. Adults were gradually acclimated for the same duration (i.e. 5 days). All flies from both
groups were sexually mature when tested; however, it is possible that their physiological age was not perfectly synchronized, WA flies ageing slightly faster than CA flies. Thus, we
cannot totally exclude the possibility that phenotypic difference in cold tolerance and associated metabolic profiles were
completely independent from this effect. Even so, a small distortion of age has minor effect on cold tolerance compared to
acclimation (Kristensen et al. 2008). In spite of this, we found

2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

718 H. Colinet et al.

Fig. 6. Temporal variations of individual metabolites for all treatment combinations. Control condition before the stress ()1 h), during the stress
(0 h) and during the time-course of recovery (1, 2 and 4 h). Symbols (s) and ()) for acute and chronic cold stress, respectively, in warm-acclimated (WA) flies (red lines). Symbols (4) and (r) for acute and chronic cold stress, respectively, in cold-acclimated (CA) flies (blue lines). Values
are means ±SE (n = 8). Refer to Fig. 1 for experimental design.

that thermoperiodic cold acclimation (DA + GA) was
mainly associated with an accumulation of sugars (Suc, Fru
and Tre) and polyamines (Cad and Put), and a reduction of
metabolic intermediates (citrate, fumarate, malate and glycerate). A previous study also found that the content of sugars
(Tre, Suc and Glc) increased with decreasing culturing temperatures in other drosophilids, D. lutescens and D. takahashii (Kimura 1982). Likewise, a GC ⁄ MS-based study found
accumulation of Suc and Tre in CA D. melanogaster larvae
(Kosˇ ta´l et al. 2011). Increase in sugars thus appears to be a
shared response to cold acclimation among various stages of
drosophilids. Changes in sugar concentrations, even moderate, seem to correlate with cold tolerance in D. melanogaster

(Overgaard, Sørensen & Loeschcke 2010). In this study, Suc
was particularly accumulated in CA flies. This carbohydrate
cryoprotectant is also known to accumulate in various overwintering insects (e.g. Baust & Edwards 1979; Ditrich & Kosˇ ta´l 2011). While our data, together with earlier studies,
support a growing consensus regarding the implication of
sugars in the cold tolerance of D. melanogaster, it is also conceivable that differential rate of ingestion ⁄ digestion during
acclimation might affect sugar levels. Further targeted
research is required to decipher the precise functional role(s)
of sugars in cold tolerance of drosophilids. This could be
tested in simple systems such as cell lines or in vitro

2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

Metabolomics of cold acclimation response 719
In addition to their accumulation during cold acclimation,
we also found that sugars, such as Suc and Fru, showed treatment-dependent temporal variations following cold stress.
The level of Suc dramatically fell after the stress exposure,
while this of Fru increased. This opposite pattern was only
detected in the cold-tolerant CA flies (see Fig. 6). Suc is a precursor for Fru synthesis which might explain the opposite
variations of these compounds. A metabolomic study
(1H-NMR) also found large temporal accumulation of sugars
(Tre, Glc and maltose) following cold stress in D. melanogaster adults (Overgaard et al. 2007). It is not known whether the
accumulation of sugars during cold acclimation and also following cold stress is a direct protective response or indirect
consequence. This metabolic pattern was found across various
stages, acclimation types and metabolomic methods suggesting that sugars have a role to play in acclimation and stress
responses in D. melanogaster. The exact causative roles of sugars still need to be elucidated in future work, but it is becoming
clear that compatible solutes, such as sugars, can protect cells
in various ways other than osmotically (Kosˇ ta´l, Sˇlachta &
Sˇimek 2001), for instance by stabilizing membranes and macromolecules, even at low physiological concentrations (e.g.
Gekko 1981; Leslie et al. 1995; Cacela & Hincha 2006).
Polyols, such as glycerol, are by far the most common cryoprotectants found in insects (Storey & Storey 2005). Except
for sorbitol, we found no indication of polyol accumulation
in CA control flies. Although sorbitol was initially more
abundant in CA flies, it is unlikely that such a small difference
was responsible for increased cold tolerance. Previous targeted and holistic studies also found no evidence of the role of
polyols in promoting cold tolerance in drosophilids (Kimura
1982; Kelty & Lee 2001; Overgaard et al. 2007; Kosˇ ta´l et al.
Another putative metabolic fingerprint of cold acclimation
response was the relatively higher level of polyamines (Put
and Cad) in CA flies. Put accumulated in response to heat
shock in several invertebrate species (Hamana, Hamana &
Shinozawa 1995; Michaud et al. 2008) and it also accumulated in CA Drosophila larvae (Kosˇ ta´l et al. 2011). Polyamines play important roles in plant stress tolerance (e.g.
freezing tolerance) and cold acclimation response (e.g. Gill &
Tuteja 2010). Orn is an immediate precursor for Put synthesis. Orn was the unique FAA that was significantly less abundant in CA flies (Fig. 3), which seems consistent with this
mechanism for Put metabolism. Conversion of Orn to polyamine is essential for the induction of HSP response in intestinal epithelial cells (Iwashita et al. 2011), and Put regulates
HSP synthesis in plants (Ko¨nigshofer & Lechner 2002). There
could also be a connection between polyamines and HSPs in
insects exposed to thermal stress (Michaud et al. 2008),
although this link is still hypothetical. As there is an acclimation-related plasticity of stress genes expression (Hsps) in
Drosophila species (Goto & Kimura 1998; Colinet &
Hoffmann 2012), this putative function might be of interest.
Similarly, Cad acts as a signal for heat shock induction in
plants (Shevyakova et al. 2001); however, such information is
lacking in insects.

We expected that cold acclimation would result in altered
levels of intermediary metabolites involved in mitochondrial
energy production. We found indeed that thermoperiodic
cold acclimation was associated with a reduction of glycolytic
(glycerate) and TCA cycle intermediates (citrate, fumarate
and malate). A reduction of energy metabolism after cold
acclimation is considered as an adaptive energy-saving strategy (Lee 1980; Evans 1981). In Dendrolimus tabulaeformis,
cold hardiness is promoted via an accumulation of sugars
concomitant with a metabolic depression after cold acclimation (Zeng et al. 2008). We found that TCA cycle intermediates were less abundant at the end of cold acclimation period
and this pattern remained during the recovery period. Likewise, the concentration of organic acids, including succinate
and maleate (i.e. the trans-isomer of fumarate), decreased in
CA larvae of D. melanogaster (Kosˇ ta´l et al. 2011), which supports our observations and suggests a common response
across different life stages. The reduction of these intermediary metabolites could translate a reduction of TCA cycle (i.e.
metabolic depression), but it could also result from a more
rapid utilization of these compounds (i.e. faster turnover) in
the case of metabolic compensation. It is clear that, whatever
the causal mechanism, modulation of TCA cycle represents
an interesting fingerprint of the cold acclimation response in
D. melanogaster. Energy metabolism is plastic, not only
owing to phenotypic adjustments, but also owing to genetic
cold adaptation in D. melanogaster species (Laayouni et al.


We assumed that cold acclimation will allow fast homeostatic
regeneration or favour metabolic maintenance, and thus we
expected post-cold stress metabolic trajectories to deviate little from the control state in CA flies. CA flies recovered
quickly from cold stress and tended to reach their initial metabolic status within a short time (Fig. 5). In contrast, WA flies
were unable to recover within the sampling time, suffered
from lethal chilling injuries and were thus likely pushed
beyond their homeostatic limits. Exposure to environmental
stress perturbs the general cellular homeostasis, and as
response, cells first activate a so-called ‘stress response’ which
is a first defence reaction to the strain imposed. Example of
this includes rapid expression of an arsenal of Hsp genes (e.g.
Bettencourt et al. 2008; Colinet, Lee & Hoffmann 2010a). In
addition to this stress response, cells activate a second set of
adaptations which aims to restore cellular homeostasis progressively and which is exemplified by the activation of proteins catalysing the accumulation of compatible organic
osmolytes (Ku¨ltz 2005). The cold stress itself may induce a
rapid acclimation response that is an attempt for the organism to maintain whole-organism homeostasis. As WA flies
were less protected from chilling injuries than CA flies, this
rapid response could have been galvanized in these flies. The
temporal metabolic variations observed during the recovery
phase could thus represent a homeostatic response and also

2012 The Authors. Functional Ecology 2012 British Ecological Society, Functional Ecology, 26, 711–722

Related documents

2012 colinet et al funct ecol
2013 colinet et al metabolomics
2018 colinet renault exp gerontol
2012 colinet hoffmann funct ecol
2018 enriquez et al front physiol
2014 colinet renault cbp a

Link to this page

Permanent link

Use the permanent link to the download page to share your document on Facebook, Twitter, LinkedIn, or directly with a contact by e-Mail, Messenger, Whatsapp, Line..

Short link

Use the short link to share your document on Twitter or by text message (SMS)


Copy the following HTML code to share your document on a Website or Blog

QR Code

QR Code link to PDF file 2012 Colinet et al Funct Ecol.pdf