How metabolomics can contribute to bio p
! " ! # $
% &
!
" #$ ## % #&# # !
1 23
Your article is protected by copyright and all
rights are held exclusively by Springer Science
+Business Media New York. This e-offprint is
for personal use only and shall not be selfarchived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.
1 23
Author's personal copy
Metabolomics
DOI 10.1007/s11306-013-0532-y
ORIGINAL ARTICLE
How metabolomics can contribute to bio-processes: a proof
of concept study for biomarkers discovery in the context
of nitrogen-starved microalgae grown in photobioreactors
Fre´de´rique Courant • Arnaud Martzolff • Graziella Rabin • Jean-Philippe Antignac
Bruno Le Bizec • Patrick Giraudeau • Illa Tea • Serge Akoka • Aure´lie Couzinet •
Guillaume Cogne • Dominique Grizeau • Olivier Gonc¸alves
•
Received: 20 December 2012 / Accepted: 8 April 2013
! Springer Science+Business Media New York 2013
Abstract Microalgae appear to be one of the most
promising sustainable resources as alternative crops for the
production of renewable transport fuel. The exploitation of
this bioresource requires, however, a fine monitoring of the
culture conditions, for example by using more relevant
control variables than usual macroscopic indicators (biomass or pigment estimation). In this proof of concept study,
we propose to search potential biomarkers of progressive
nitrogen regime culture conditions using an untargeted
metabolomic approach based on LC-HRMS combined to a
non-invasive analysis based on FTIR spectroscopy. One
microalgae model was investigated i.e. Chlamydomonas
reinhardtii to characterize the effect of progressive
Electronic supplementary material The online version of this
article (doi:10.1007/s11306-013-0532-y) contains supplementary
material, which is available to authorized users.
F. Courant ! J.-P. Antignac ! B. Le Bizec
Laboratoire d’E´tude des Re´sidus et Contaminants dans les
Aliments LABERCA, LUNAM Universite´, Oniris,
44307 Nantes, France
A. Martzolff ! G. Rabin ! G. Cogne ! D. Grizeau !
O. Gonc¸alves (&)
LUNAM Universite´, Universite´ de Nantes, CNRS, GEPEA,
UMR 6144, Baˆt. CRTT, 37 bd de l’Universite´, BP 406,
44602 Saint-Nazaire Cedex, France
e-mail: olivier.goncalves@univ-nantes.fr
P. Giraudeau ! I. Tea ! S. Akoka
LUNAM Universite´, Universite´ de Nantes, CNRS, Chimie et
Interdisciplinarite´ : Synthe`se, Analyse, Mode´lisation CEISAM,
UMR 6230, 2 rue de la Houssinie`re, BP 92208, 44322 Nantes
Cedex 03, France
A. Couzinet
LUNAM Universite´, Universite´ de Nantes MMS, 2 rue
de la Houssinie`re, 44322 Nantes Cedex 03, France
nitrogen regime in batch culture conditions on its metabolome. FTIR allowed assessing the intracellular macrometabolic perturbations, highlighting the over-accumulation
of carbohydrates. LC-HRMS complemented the macromolecular information by revealing the dependence of
microalgae metabotypes on nitrogen regime conditions
tested for cells culture. Patterns of significantly modulated
metabolites were also detected during those slight contrasted nitrogen regimes and interesting features were
structurally elucidated. This included metabolites belonging to the pantothenate, branched chain and aromatic
amino acids pathways. In the last step of this proof of
concept study, amino acid targets proposed by metabolomic investigations were assessed on nitrogen-limited continuous culture on photobioreactors. This was performed to
test the validity of proposed targets in real small-scale
industrial production conditions. Results were very
encouraging and suggested the possibility of using potentially relevant metabolites as intracellular biomarkers only
(tryptophan) or as both intra and extracellular biomarkers
(e.g. 2-methylbutyric acid and ketoleucine).
Keywords LC-HRMS ! FTIR spectroscopy !
Nitrogen progressive regime ! Chlamydomonas reinhardtii !
Biofuel ! Photobioreactors
Abbreviations
ATR
Attenuated total reflectance
BCAAs
Branched-chain amino acids
CE
Capillary electrophoresis
FTIR
Fourier transform infrared spectroscopy
GC
Gas chromatography
HC
Hierarchical clustering
HPLC
High performance liquid chromatography
HTS-XT
High throughput screening eXTension
123
Author's personal copy
F. Courant et al.
LC-HRMS
MS
NMR
PLS-DA
RSD
TAP
TIC
Liquid chromatography—high resolution mass
spectrometry
Mass spectrometry
Nuclear magnetic resonance
Partial least square-discriminant analysis
Relative standard deviation
Tris acetate phosphate
Total ionic current
1 Introduction
Microalgae are often cited as ‘‘green gold mines’’ for
generating renewable energy since they literally present
almost all the characteristics of the perfect sustainable
alternative crop (Larkum et al. 2012; Ratha and Prasanna
2012). However mastering such organisms for the purpose
of industrial biofuel production still represents a tremendous challenge with both scientific and industrial bottlenecks (Lam and Lee 2012). When dealing with biomass
production in closed photobioreactors, the fine monitoring
of culture conditions appears to be one of the major
problems to be solved, since the accumulation of energetic
molecules depends on the operational growth conditions of
microalgae and especially on nutrient supplementation
(Gonzalez-Fernandez and Ballesteros 2012; Wang et al.
2012). Indeed mineral starvation—especially nitrogen
regime—is often exploited for the accumulation of lipids or
carbohydrates, with culture monitoring solely performed
with macroscopic indicators such as biomass evolution or
pigment quantification (Breuer et al. 2012).
However those macroscopic indicators cannot reflect all
the complexity of the phenotype expressed by algae cultivated in nitrogen regime conditions, since it results from
multilevel interactions including genetic, transcriptomic,
proteomic or metabolic factors (Jamers et al. 2009). Understanding this complex biological process should ideally be
performed by integrating information associated to all those
‘‘omics’’. Dealing with such an approach is however not
straightforward since large biases exist when attempts are
made to acquire or interpret experimental high-throughput
data (Martinez-Gomez et al. 2012). One strategy could
consist in focusing on the modulation of intracellular small
molecules, thus assessing the end-result of this complex
biological process. This approach should highlight pertinent
biomarkers or patterns of biomarkers characteristic of microalgae cultured under nitrogen regime.
With regards to the metabolome, the monitored signals
correspond to chemical substances (metabolites) accessible
to the analysis, which are final products formed after
the complex transcription, translation, and regulation
123
mechanisms. Metabolomics enables the differential
assessment of the levels of a broad range of endogenous
and exogenous metabolites and has been shown to have a
great impact on the investigation of the physiological status, discovering biomarkers and identifying related pathways (Wang et al. 2010). From an analytical point of view,
metabolomics deals with the comprehensive analysis of
metabolites present in a biological sample by the combined
use of a fingerprint technology and multivariate statistical
analyses. The most widely used technique for metabolomic
purposes remains nuclear magnetic resonance (NMR)
(Nicholson et al. 2002; Nicholson et al. 1999). Non-invasive approaches such as vibrational spectroscopy are also
exploited, not for metabolome description purpose, but
more as rapid and complementary explanatory techniques
allowing the biochemical profiling of cultures at a cellular
level (Winder et al. 2011; Wu et al. 2011). Fourier transform infra-red (FTIR) spectroscopy demonstrated its ability
to highlight reproducible biochemical differences when
comparing the effect of various treatments, and even to
estimate the in vivo content of modulated molecules
without cellular preparation or metabolite extraction
(Alvarez-Ordonez et al. 2011; Goff et al. 2009). However,
despite its advantages, FTIR lacks ability in characterizing
hydrosoluble compounds. Mass spectrometry (MS)-based
methods, on the other hand, have recently proved to be
valuable for metabolomic studies, especially thanks to
recent technological advances. Furthermore, they present
some incomparable advantages over NMR and IR in terms
of sensitivity (Antignac et al. 2011; Theodoridis et al.
2008).
In the present study, both FTIR and MS were exploited.
The first one was used with the objective to obtain a fast
and simple method for evaluating and monitoring culture
conditions by providing information on the macromolecular content of cells. The second one, which is more sensitive and has powerful resolution, was exploited to provide
detailed information on metabolites. The objective of our
proof-of-concept study is to evaluate the potential of the
metabolomic approach to assess modulated metabolite
signatures characteristic of microalgae cells cultivated
under limited nitrogen culture conditions. Candidate biomarkers or patterns of biomarkers should be in the near
future exploited as control indicators for the fine monitoring of photobioreactors utilized to grow microalgae under
nitrogen regime. The choice of culture conditions progressively limited in nitrogen and not fully depleted was
motivated by the final applicative objective of this study.
Indeed, for industrial applications continuous cultures are
often chosen for high productivity reasons, therefore
imposing to perform nutrient limitation in order to work
with living cells (Van Vooren et al. 2012). Performing our
proof of concept with fully nitrogen-depleted cultures
Author's personal copy
Assessing biomarkers of nitrogen stressed cultured microalgae using metabolomics
would have stopped growing biomass and provided data
too far from industrial operational conditions. Results and
perspectives emerging from this proof-of-concept study are
described and discussed in this manuscript.
2 Materials and methods
2.1 Chemicals and reagents
All chemicals and reagents were obtained from SigmaAldrich (Taufkirchen, Germany) in the highest quality
available (analytical grade).
2.2 Microalgae culture conditions
2.2.1 Batch cultures
Wild-type Chlamydomonas reinhardtii 137AH cells from
the culture collection of the French Alternative Energies
and Atomic Energy Commission (CEA Cadarache, France)
were conserved on solid (agar 1,75 %) standard tris–acetate-phosphate (TAP) medium (Gorman and Levine 1965)
at 16 "C and replicated every month. From the latter, liquid
precultures were performed in flasks containing TAP
medium and nursed at 25 "C under constant agitation
of 135 rpm and illumination of approximately
100 lmol m-2 s-1 in an incubator (New Brunswick
Innova# 44). Precultures were replicated every 4 days to
keep cells in mid-log phase. After five replications and
having reached a cell density of about 4–7.106 cells mL-1
(corresponding to a late logarithmic growth state), cells
were harvested by centrifugation (4,5009g, 5 min, 20 "C
in a Sorvall# RC-6 Plus centrifuge) and the cell pellet was
resuspended in fresh TAP medium. Resuspended cells
were then used to inoculate batch cultures fed with the
same fresh medium for reference culture conditions R or
with medium progressively depleted with nitrogen i.e.
NH4Cl 0.2 g L-1 for culture conditions N- and NH4Cl
0.1 g L-1 for culture conditions N–. Standard culture
conditions were then applied until a homogenous cell
density of 4–6.106 cells mL-1 and a typical observable
phenotype (i.e. pigment discoloration) were observed for
limitation experiments (Wegener et al. 2010).
2.2.2 Continuous cultures in photobioreactor
Microalgae cells coming from preculture flask (2.2.1) were
harvested by centrifugation (4,5009g, 5 min, 20 "C in a
Sorvall# RC-6 Plus centrifuge) and the cell pellet was
resuspended in fresh medium (Hutner et al. 1950) where
nitrogen is only the growth limiting substrate solely.
Resuspended cells were then inoculated in a torus-shaped
photobioreactor operating in chemostat mode. The culture
volume was kept constant by feeding the culture with the
same fresh medium using a dosing pump (Stepdos# pump
03/RC, KNF Neuberger) at constant flow rate, and by
harvesting the culture using a peristaltic pump (Masterflex# LS Easy-Load II) at the same flow rate. Steady-state
conditions were achieved by maintaining constant dilution
rate (0.0077, 0.0079 or 0.0104 h-1) and light intensity (200
or 400 lmol m-2 s-1). pH was kept constant at 7.5 ± 0.1
by CO2 injection using an on–off controlled mass flow
controller (EL-FLOW F-201CV-020-RAD-33-Z, Bronkhorst#) at a constant flow rate of 3 mL min-1. Temperature was kept constant at 25 ± 1 "C. Mixing was
performed by a marine impeller under a constant stirring
speed (300 rpm). In order to prevent any metabolic shift
toward photorespiratory pathways, nitrogen (N2) as inert
gas allowed to strip oxygen from the culture broth at a
constant flow rate of 10 mL min-1 using a mass flow
controller (EL-FLOW F-201C-RAD-33-V, Bronkhorst#).
A septum on the top of the photobioreactor enabled daily
manual sampling for analysis. On-line data acquisition
used a data processing system (DAQ 6023E-National
Instruments#) that enabled automated data acquisition and
process control. The software was written with the LabView Virtual Instruments programming techniques
(National Instruments#).
2.3 Microalgae cells sampling for analytical data
acquisition
2.3.1 Cell sampling for FTIR analysis
For the acquisition of FTIR spectra, a 1 mL sample of fresh
culture was taken from each replicate flask or automatically
sampled via the PBR sampling device. Cells were centrifuged (4,5009g, 5 min, 20 "C in a Mikro R22 centrifuge)
and washed with 2 mL physiological water. The washing
centrifugation protocol was repeated three times. Depending on the number of cells per milliliter, pellets were
resuspended in about 100 lL of physiological water. Volumes of 1 lL (approximately 105–106 cells) were deposited on a 384 well silicon microplate and oven-dried for
5 min at 40 "C.
Solid microalgal pellet residues or liquid organic
extracts coming from solvent extraction test experiments
were directly dried for approximately 15 min on the diamond crystal of the ATR accessory of the FTIR spectrometer using an infrared lamp (220 V, 50 Hz, 100 W).
2.3.2 Cell sampling for LC-HRMS analysis
For the acquisition of LC-HRMS metabolic profiles, a
rapid sampling protocol was used on cell batch cultures
123
Author's personal copy
F. Courant et al.
identical to those analyzed with FTIR spectroscopy. It
consisted in a fast filtration step coupled to a manual rapid
quenching step in liquid nitrogen. Homogeneous microalgae samples of 1 mL of fresh culture (106–107 cells mL-1)
were deposited on polyamide filters (Sartorius#,
25007-47N, 0.2 lm) placed on a vacuum sintered glass
filtration unit. In order to remove extracellular salts, filters
were subsequently rinsed with 3 mL of rinsing solution
(NH4Cl 1 g L-1; CaCl2, 2H2O 0.05 g L-1; KH2PO4
0.2 g L-1) (Martzolff et al. 2012). Filters were then
quickly quenched in liquid nitrogen (-196 "C) and stored
at -80 "C for further analysis (Bolten et al. 2007).
2.4 Metabolite extraction of sampled microalgae cells
Hydrosoluble compounds extraction was performed using
(hydro)-alcoholic solvent mixtures on sampled microalgae
cells. For this proof of concept, two combinations were
used solely, 100 % methanol and hot ethanol–water mix
(75–25 v/v) (El Rammouz et al. 2010).
Filters resulting from rapid sampling and quenching
methods (Sect. 2.3.2) were plunged into methanol. Samples
were vortexed for 3 min, centrifuged (4,5009g, 5 min,
20 "C in a Sorvall# RC-6 Plus centrifuge) and the organic
extract was finally removed and filtered on glasswool
(glasswool was systematically heated at 460 "C during 6 h
to avoid further contamination). The solvent extractioncentrifugation-filtration operation was repeated until the
maximum impoverishment of microalgal cell pellets was
monitored with ATR-FTIR analysis. The organic phases
were pooled and evaporated under dry analytical nitrogen
stream (Nitrogen evaporator—Organomation Associates
Inc., USA) before analysis.
For the hot ethanol–water extraction, the filters resulting
from rapid sampling and quenching methods (Sect. 2.3.2)
were plunged into 5 mL of boiling mixture (90–95 "C).
The tubes were strongly shaked for 3 s and then extracted
for 30 s, 1 min, 1 min 30 s and 2 min in a water bath
maintaining the extraction temperature at 90–95 "C. The
extraction process was then quenched by plunging tubes
for 3 min in ethanol–water (75–25 v/v) maintained at
-80 "C. The filter was removed, the tube was centrifuged
(4,5009g, 5 min, 20 "C in a Sorvall# RC-6 Plus centrifuge) and the organic extract was filtered on glasswool
(glasswool was systematically heated at 460 "C during 6 h
to avoid further contamination). The organic phase was
evaporated under dry analytical nitrogen stream (Nitrogen
evaporator—Organomation Associates Inc., USA) before
analysis.
Twenty microliters of a mixture of internal standards
(triamcinolone and ponasterone A at 1 ng/lL) were added
to each sample to allow the monitoring of the instrument
analytical variability during the fingerprinting process.
123
2.5 Analytics
2.5.1 HTSXT-FTIR spectra
Infrared spectra were recorded in transmission mode
directly on a silicon 384 well microplate loaded with dried
microalgae samples. The Bruker tensor 27 FTIR spectrometer equipped with the HTS-XT plate reader module,
with a deuterated triglycine sulphate detector RT-DLaTGS
and the OpusLab v 7.0.122 software (Bruker Optics, Germany) was set up with the following parameters. The
spectral resolution was fixed to 1 cm-1, the number of
scans to 32, the selected spectral range between 4,000 and
400 cm-1, and the microplate reader pinhole aperture was
set up to 3 mm. The diameter of the IR beam was therefore
sufficient for the detection of all dried cells within the spot
diameter of around 3 mm. Background spectra were collected using the same instrument settings as those
employed for the samples. Spectra were recorded for five
replicates per sample.
2.5.2 ATR-FTIR spectra
Infrared spectra were recorded in reflection mode directly
on the diamond crystal of the ATR accessory loaded with
dried microalgae or extract samples. The Bruker tensor 27
FTIR spectrometer equipped with the ATR platinum
module, with a deuterated triglycine sulphate detector RTDLaTGS and the OpusLab v 7.0.122 software (Bruker
Optics, Germany) was set up with the following parameters. The spectral resolution was fixed to 1 cm-1, the
number of scans to 32, the selected spectral range between
4,000 and 400 cm-1. Background spectra were collected
using the same instrument settings as those employed for
the samples and was performed against air. Spectra were
recorded for 3 replicates per sample.
2.5.3 LC-HRMS metabolic fingerprints
LC-HRMS experiments were performed on an Agilent
1200 HPLC system including an autosampler and a binary
pump coupled to a Finnigan LTQ-OrbitrapTM hybrid mass
spectrometer (Thermo Fisher Scientific, Bremen, Germany) fitted with an electrospray source operated in the
negative ion mode. 10 lL of each sample were injected on
a Hypersil-Gold column (100 mm 9 2.1 mm 9 1.9 lm
particle size, Thermo Fisher Scientific). The mobile phase
consisted in water containing 0.1 % acetic acid (A) and
acetonitrile containing 0.1 % acetic acid (B). The elution
gradient (A:B, v/v) was as follow: 95:5 from 0 to 2.4 min;
75:25 at 4.5 min; 30:70 at 11 min; and 0:100 at 14 min.
The flow rate was 0.4 mL/min. The HPLC column was
connected without splitting to the electrospray interface
Author's personal copy
Assessing biomarkers of nitrogen stressed cultured microalgae using metabolomics
operating in negative mode. The electrospray voltage was
set to -4 kV, and the capillary voltage and tube lens offset
were set to -20 and -90 V, respectively. The sheath and
auxiliary gas flows (both nitrogen) were set to 55 and 6
arbitrary units (a.u.), respectively, and the drying gas
temperature was set to 325 "C. Mass spectra were recorded
from 65 Th up to 1,000 Th at a resolution of 30,000
(FWHM at m/z 400). Mass spectra were acquired in the
centroid mode.
2.6 Data analysis
2.6.1 FTIR spectra preprocessing
The spectra were recorded to the background spectra and
baseline corrected—CO2, H2O, ATR (when needed)—and
smoothed. A second derivative calculation was also performed on the corrected spectra. All these treatments were
achieved using integrated functions of OpusLab v 7.0.122
software (Bruker Optics, Germany).
2.6.2 Multivariate analysis of FTIR spectra
Preprocessed spectra or preprocessed second derivative
spectra were exported as text files for file format modification on Microsoft Excel v14.0.0 or statistical treatment
on R 2.15.2. Second derivative spectra were systematically
used to improve the infrared band resolution and thus
enhance the discrimination of vibrators contributing to the
shape of raw FTIR spectra. Since one vibrator is associated
to one molecule family, maximum chemical information
should be extracted from second derivative FTIR spectra
(Mecozzi et al. 2011). Spectral data were centered and
reduced to minimize the unit range influence before statistical analysis.
2.6.3 Metabolomic LC-HRMS fingerprint preprocessing
Following their acquisition, metabolomic fingerprints
were deconvoluted to allow the conversion of the threedimensional raw data (m/z, retention time, ion current) to
time- and mass-aligned chromatographic peaks with
associated peak areas. Xcalibur# software (Thermo Fisher
Scientific) was used to convert the original Xcalibur data
files (*.raw) to a more exchangeable format (*.cdf). Data
processing was then performed using the open-source
XCMS software (Smith et al. 2006). XCMS parameters
for the R language were implemented in an automated
script. The interval of m/z value for peak picking was set
to 0.1, the signal to noise ratio threshold was set to 6, the
group bandwidth was set to 15 and the minimum fraction
was set to 0.75.
2.6.4 Statistical analysis of LC-HRMS data
A non-parametric analysis of variance, the Kruskal–Wallis
test, was performed on all signals [(m/z, rt) features)]
generated by the XCMS treatment of the LC-HRMS data to
determine if the nitrogen regime was a parameter significantly affecting the metabolome of microalgae cells. A
non-parametric Mann–Whitney U test was then performed
on the structurally identified metabolites which presented a
significant Kruskal–Wallis p value (p \ 0.05) to determine
more finely the effect of the progressive nitrogen regime.
The Kruskal–Wallis test and the Mann–Whitney U test
were carried out using Statistica Data Miner v 7.1 (StatSoft, Maisons-Alfort, France).
In order to visualize the patterns of (m/z, rt) features
presenting a Kruskal–Wallis p value lower than 0.05 across
the three nitrogen regimes, a hierarchical cluster analysis
(HC) was performed using the web server analysis pipeline
Metaboanalyst 2.0, dedicated to metabolomic data exploitation (Xia et al. 2012). A blue/red colored heat map was
then plotted to summarize the data.
A partial least square—discriminant analysis was performed on all the signals stemming from the XCMS
analysis of the LC-HRMS data. PLS-DA was conducted on
SIMCA-P?# software (version 12, Umetrics Inc., Umea˚,
Sweden). Prior to analysis, data were log10 [1 ? x]transformed and Pareto scaled.
2.7 Metabolite identification
Compounds identification was performed on the basis of an
internal data bank created for LC-ESI-LTQ-Orbitrap
acquisitions and using ACD/Labs software (Courant et al.
2012).
2.8 Targeted analytical biochemistry
2.8.1 Total carbohydrate and starch concentration
determination
The total carbohydrates content was determined by the
phenol–sulphuric acid method (DuBois et al. 1956).
2–10 mL of culture were sampled and centrifuged
(12,0009g, 5 min, 20 "C). The cell pellet was resuspended
in de-ionized water so that carbohydrates concentration
was comprised between 0.02 and 0.1 g L-1. 0.5 mL of
phenol (50 g L-1) and 2.5 mL of fuming sulphuric acid
(98 %) were added to 0.5 mL of resuspended culture. After
10 min of incubation, tubes were mixed (vortex, 10 s) and
then incubated in water bath (35 "C, 30 min). Sample
absorbances (483 nm, Perkin–Elmer, Lambda 2S) were
then read and compared to a standard curve for
quantification.
123
Author's personal copy
F. Courant et al.
Starch determination was performed using a method
slightly modified from that of Klein and Betz (1978). 2 mL
aliquots of cell suspension were sampled and centrifuged
(12,0009g, 5 min, 20 "C). The cell pellets were suspended
in 1.5 mL of methanol, incubated (45 min, 44 "C) and then
centrifuged again. The pellets were rinsed with 1 mL of
sodium acetate buffer (100 mM, pH 4.5), resuspended in
500 mL of sodium acetate buffer, and autoclaved (60 min,
135 "C) for starch solubilization. Starch assays were performed with a commercial kit (starch assay kit SA-20,
Sigma-Aldrich) based on an enzymatic method following
the supplier’s recommendations.
2.8.2 Protein concentration and amino acids profiling
Total proteins content was quantified with the Lowry and
colleagues method (Lowry et al. 1951) using the Folin
phenol reagent.
Amino acids profiling was performed as follows. 1 L of
culture was centrifuged (Sorvall centrifuge RC-6 Plus,
5,0009g, 10 min, 4 "C). The cell pellet was washed with
de-ionized water and centrifuged again. Then, the cell
pellet was freeze-dried (RP2 V Se´rail freeze dryer, 24 h).
Amino acids, except tryptophan were quantified using an
amino acid analyser. Cysteine and methionine were oxidized at 0 "C with a mixture of performic acid and phenol.
Amino acids, except tryptophan, were hydrolyzed for 23 h
with hypochlorous acid, pH was adjusted to 2.20. Amino
acids were separated using ionic chromatography and
quantified after ninhydrin reaction by spectrophotometry
(570 nm, except for proline, 440 nm). Tryptophan Tryptophan was hydrolyzed with saturated barium hydroxide
solution at 110 "C for 20 h. After hydrolysis, the internal
standard was added and samples were quantified by HPLC
equipped with a fluorimetric detector.
3 Results and discussion
3.1 First objective of this proof of concept study:
estimating the quality of the metabolomic protocol
3.1.1 Qualitatively, the number of detected features
increased with the tested solvent polarity
and appeared to be heat shock dependent for the hot
water–ethanol mix
The enrichment of hydro-alcoholic extract features was
assayed with both FTIR and HRMS techniques. The second derivative FTIR signatures of organic extracts were
qualitatively interpreted hypothesizing that one wave
number was characteristic of one chemical family feature
(Mecozzi et al. 2011). LC-HRMS profiles were interpreted
123
assuming that one feature corresponded to one massretention time ([m/z; rt]) couple extracted from raw TIC by
XCMS software. It appeared that very few detected features were shared between the tested extracts whatever the
technique considered (methanol or hot water–ethanol mix).
This suggested the complementarity of the tested extraction
approaches on C. reinhardtii cells (data not shown). The
methanol and hot water–ethanol mix approaches exhibited
an increasing number of LC-HRMS detected features with
the increasing polarity of the tested solvent. The hot water–
ethanol mix presented the highest number of detected
features (846 for methanol compared to 1,681 for hydroalcoholic for a time of contact of 1 min), except when
exceeding 1 min of contact time, where a clear decrease
was monitored (data not shown).
3.1.2 The repeatability of the extraction protocol
was assayed quantitatively for the approach
exhibiting the highest number of detected features
i.e. the hot water–ethanol mix
The repeatability of the extraction protocol was determined
in order to ensure that the monitored metabolite level
change was associated to the tested stimulus and not to
technical sources. Five replicates of one sample were
extracted following the chosen protocol (i.e. hot water–
ethanol mix for four contact times) and injected on the LCHRMS system. As these replicates were identical, the
different signals (metabolites) constituting the fingerprints
and extracted through the solvent protocol should lead to
the same intensity across the replicates, with low variability (i.e. RSD \ 40 %). 65 % of the ions constituting
the global fingerprints were found to present relative
standard deviations (RSD) lower than 40 % for two contact
times, i.e. 30 s and 1 min, and superior to this threshold
value starting from 1 min 30 s. Such results allowed considering the hot water–ethanol mix data obtained for 30 s
and 1 min contact times as satisfactory since the calculated
RSD took into account the variability of the extraction
protocol and of the fingerprinting process.
3.1.3 The LC-HRMS fingerprint acquisition appeared
to be of good quality and therefore interpretable
In this proof of principle study, the assessment of LCHRMS metabolite fingerprints was performed using the
extraction protocol providing the apparent highest feature
diversity (i.e. hot ethanol–water mix with 1 min of contact
time) and the most repeatable extraction protocol. Prior to
analyzing information coming from the LC-HRMS metabolic signatures, the repeatability of the fingerprinting
process was evaluated by means of the added internal
standards. The RSD calculated on both internal standards
Author's personal copy
Assessing biomarkers of nitrogen stressed cultured microalgae using metabolomics
3.2 Second objective of this proof of concept study:
evaluating the effect of the progressive nitrogen
regime on Chlamydomonas reinhardtii observable
metabolome
3.2.1 The progressive nitrogen regime provoked
a carbohydrate accumulation
Raw FTIR spectra of fresh cells indicated an immediate
effect of such a limitation on the relative abundance of
microalgae carbon storage molecules (Fig. 1). A significant increase of the relative amount of carbohydrate as a
function of the nitrogen limitation was first observed. This
finding was then accompanied by an increase of relative
lipid content, but in a less marked manner. Such macromolecular observations were consistent with the data initially obtained by Ball et al. (1990) and described for
nitrogen limitation (Dean et al. 2010) or deprivation
protocol (James et al. 2011). FTIR observations were
moreover reinforced by Longowrth’s recent results
(Longworth et al. 2012). Indeed, during their nitrogen
starvation protocol, kinetic results also indicated that the
accumulation of carbohydrate was systematically followed
by an increase of neutral lipids. In order to complement
the FTIR macromolecular observations, LC-HRMS metabolic profiles assessed on similar cell samples were
analyzed.
3.2.3 The pattern of regulated features were clearly
detected according to the nitrogen regime
Among the 1,681 [m/z, rt] detected features on the LCHRMS metabolic fingerprints, 560 presented a Kruskal–
Wallis p value lower than 0.05 across the three nitrogen
regimes, suggesting significant effect of such regime on our
observable fraction of the metabolome. Double hierarchical
clustering was performed on these data and the results were
summarized on a heatmap in order to visualize if metabolite patterns could be detected according to the nitrogen
regime (Fig. 3). Sample clustering results in rows highlighted the effect of the nitrogen regime on metabolic
signatures since three distinct groups could be established
(one group per nitrogen regime). Feature clustering results
in columns highlighted several ion patterns presenting
different behaviors. Some features were whether up or
down regulated when increasing the nitrogen regime
B
0.6
0.4
Ratio
0.2
1800
1600
1400
1200
1000
800
0.0
Absorbance units
The PLS-DA analysis of LC-HRMS metabolic signatures
of nitrogen-deprived microalgae cells suggested a major
effect of the nitrogen regime (Fig. 2). Indeed, the score plot
on the first component (accounting for 39.5 % of the initial
information) indicated a clear separation of metabolic
signatures arising from reference nitrogen cultured cells
and those coming from the two tested nitrogen regimes. On
the second component (accounting for 16.9 % of initial
information), a clear discrimination was performed
between metabolic signatures obtained with the two kinds
of nitrogen regime.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
A
3.2.2 Metabolic fingerprints were nitrogen regime
dependent
0.8
detected across the different samples (n = 15) were found
acceptable (RSD = 29 % for Triamcinolone and
RSD = 31 % for Ponasterone A) and suggested a good
quality of the whole fingerprint acquisition.
C/P
L/P
–1
wave number (cm )
Fig. 1 FTIR spectra of Chlamydomonas reinhardtii entire cells
cultured under progressive nitrogen limited conditions. Raw infrared
spectra (left panel) illustrated global changes in the macromolecular
composition of microalgae when comparing the reference culture
conditions R (green) to limitations N- (yellow) and N– (red). The
evolution (right panel) of lipid (1,740 cm-1): amide I (1,655 cm-1)
‘‘L/P’’ and carbohydrate (1,150 cm-1): amide I ratio ‘‘C/P’’ is
represented as a function of the nitrogen regime R (green), N(yellow) and N– (red). These ratios are representative of the relative
evolution of lipid and carbohydrate contents (Dean et al. 2010; James
et al. 2011) (Color figure online)
123
Author's personal copy
F. Courant et al.
nit1
nit2
ref
EtOH_1min_T-L1_into(1) - metaboanalyst2 (FC).M1 (PLS-DA)
t[Comp. 1]/t[Comp. 2]
Colored according to Obs ID (label)
30
25
20
L23
15
L25
L22
L21
10
L24
5
t[2]
Fig. 2 PLS-DA score plot of
LC-HRMS metabolic signatures
of Chlamydomonas reinhardtii
cells grown on nitrogen regime
R (blue triangle: ref), N- (red
triangle: nit1) and N– (green
triangle: nit2). Replicates are
indicated with suffix numbers 1,
2, 3, 4 and 5. The model
presented the following
characteristics R2(Y) = 0,989
and Q2 = 0,869 and was
validated using 200
permutations (R2(Y)int = 0,555
and Q2int = -0,343) (Color
figure online)
0
R2
R1
-5
L13
L11
L14
R4
R5
-10
L12
R3
L15
-15
-20
-25
-30
-40
-30
-20
-10
0
10
20
30
40
t[1]
SIMCA-P+ 12 - 2012-10-04 09:13:18 (UTC+1)
stringency. Others patterns presented intermediate regulation schemes where sudden up regulation was followed by
down regulation and inversely.
3.2.4 A biological interpretation was attempted
on the identified LC-HRMS features, suggesting
interesting potential biomarkers
In order to complete the observed tendencies, features
selected using the Kruskal–Wallis test were investigated
for annotation and subsequent biological interpretation
attempts. Interesting biomarkers were confirmed experimentally (Table 1).
(i) Over-accumulation of carbohydrates and lipids could
be assessed through a regulation of the branched-chain
amino acids (BCAAs) and the pantothenate pathways.
Increasing relative amounts of carbohydrate and lipids
observed with FTIR were also suggested with LC-HRMS
annotated data, highlighting the regulation of metabolites
involved in the leucine biodegradation and the pantothenate biosynthesis pathways (Table 1). Indeed, leucine and
pantothenic acid were found to be significantly down regulated whatever the nitrogen regime. The regulation of
detected metabolites involved in the catabolism of leucine
was however not clearly established for both tested
123
nitrogen regimes in this proof of concept study. This was
the case of ketoleucine, found to be down regulated significantly in the N- regime and 2-methylbutyric acid, significantly up regulated in the N– regime. For those two
metabolites, absence of significant regulation observed for
N– or N- regime respectively precludes further interpretations about the distinction of both nitrogen regimes. Nevertheless, our data could suggest an original pathway for
microalgae CoenzymeA biosynthesis, inferring a relationship between branched-chain amino acids and the pantothenate pathway as depicted by Chassagnole et al. (2002)
(Fig. 4a). The monitored consumption of leucine could be
here partially explained by the detection of its direct degradation intermediate (ketoleucine) and the presence of an
alternative end product of its catabolism i.e. the 2-methylbutyric acid (Ganesan et al. 2006). Those observations
suggested an original alternative pathway for microalgae
leucine biodegradation, but were not informative enough to
elucidate the connectivity of the leucine pathway to the
pantothenate pathway. Indeed, the carbon flux arising from
leucine degradation could be directed toward the ketoisovalerate node, or toward another metabolite common to
both the pantothenate and the leucine pathway. Concerning
the down regulation of pantothenate vitamin, it could be
straightforwardly explained by an increasing mobilization
Author's personal copy
Assessing biomarkers of nitrogen stressed cultured microalgae using metabolomics
Fig. 3 Heatmap displaying the 560 LC-HRMS features selected
according to the Kruskall Wallis p value results. Pearson distance
measurements and Ward clustering algorithm were used for calculations. Samples with nitrogen regimes R (blue: ref), N- (red: nit1)
and N– (green: nit2) are disposed in rows and replicates are indicated
with suffix numbers 1, 2, 3, 4 and 5. LC-HRMS features are disposed
in columns and each coloured cell corresponded to the normalized
relative abundance values obtained after data processing from lower
to highest relative abundance value (from blue to red) (Color figure
online)
of this precursor for coenzyme A synthesis, the latter
mobilized itself to biosynthesize carbohydrates and fats
(Smith et al. 2007).
(ii) The hydrophobic amino acid metabolism was also
impacted. LC-HRMS suggested complementary information that was not highlighted with the FTIR approach.
Particular attention was paid to the regulation of several
aromatic heterocyclic compounds that were significantly
down regulated (tryptophan) and up regulated (indole
acetic acid) for both nitrogen regimes, and significantly up
regulated (phenyl pyruvic acid) for nitrogen regime N–
(Table 1). Tryptophan regulation was not surprising, since
it was already observed under salt acclimation for Oryza
sativa (Sanchez et al. 2008) or low nitrogen growth regime
for Arabidopsis thaliana (Kusano et al. 2011). In 2005,
Bolling and Fiehn (2005) also performed a nitrogen starvation experiment on C. reinhardtii cells but rather
observed a positive modulation of this amino acid. Difference of experimental protocol could mainly explain this
opposition with our results. Recent proteomic results
experiments (Longworth et al. 2012) indicated that the
limitation in nitrogen affected the metabolism of
123
123
Table 1 List of confirmed annotated features
Compound
RT
(min)a
L-Tryptophan
3.13
Phenyl pyruvic
acid/phenyl
pyruvate
Delta
RT
(min)
Delta
m/z
(ppm)b
Ratio of
median
N- vs R
p value
N- vs
Rc
Ratio of
median
N– vs R
p value
N– vs
Rc
Taxonomy
Function—comment
0.10
203.0826
-0.22
0.01
0.008
0.02
0.008
Essential amino
acid
Essential amino acid metabolism
5.26
0.14
163.0400
-0.29
0.54
0.310
5.06
0.032
Phenylpyruvic acid
derivatives
Intermediate or catabolic byproduct of phenylalanine metabolism
Indoleacetic acid/
indoleacetate
8.52
0.07
174.0564
1.74
2.13
0.008
9.87
0.008
Aromatic
heteropolycyclic
compounds
Tryptophan Metabolism (growth hormone plant)
L-Leucine
0.97
0.03
130.0874
0.17
0.06
0.008
0.06
0.008
Essential amino
acid
Essential amino acid metabolism
N-acetyl-valine
6.61
0.03
158.0824
0.88
0.22
0.008
0.24
0.008
Amino acids and
derivatives
Valine is involved in carbohydrate metabolism.
Ketoleucine
(KLeu)
3.01
0.12
129.0557
0.24
0.36
0.032
1.90
0.151
Branched fatty
acids
Component of Leucine catabolic pathway
2-Methylbutyric
acid (2-MBA)
7.48
0.03
101.0609
0.85
1.44
0.690
3.59
0.008
Branched fatty
acids
Component of leucine catabolic pathway
Pantothenic acid/
pantothenate
2.26
0.47
218.1033
-0.32
0.40
0.016
0.47
0.032
Essential vitamins
Pantothenic acid is needed to form coenzyme-A (CoA), and is thus critical
in the metabolism and synthesis of carbohydrates, proteins, and fats
Uridine
0.91
-0.04
243.0621
-0.85
0.12
0.008
0.03
0.008
Pyrimidine
nucleosides
Pyrimidine metabolism
Guanosine
1.10
0.28
282.0842
-0.52
0.50
0.016
0.31
0.008
Purine nucleosides
Ribonucleosides metabolism
Guanosine-5monophosphate
1.05
0.37
362.0506
-0.35
0.34
0.008
0.50
0.056
Purine nucleosides
Purine metabolism
Adenosine
1.27
0.06
266.0903
3.11
0.26
0.008
0.27
0.008
Purine nucleosides
Important biological roles in addition to being components of DNA and
RNA. For instance, adenosine plays an important role in energy
transfer—as adenosine triphosphate (ATP) and adenosine diphosphate
(ADP)
5-Deoxy-5methyladenosine
5.95
-0.02
296.0822
-0.36
0.16
0.008
0.19
0.008
Purine nucleosides
Purine metabolism
Author's personal copy
Observed
mass
Nitrogen regime most discriminant features selected a through Kruskal–Wallis or Mann–Whitney test performed between the three culture conditions were annotated using commercial standards. First four
columns correspond to experimental annotation parameters, following ones to statistical indicators and last ones to annotation, chemical taxonomy, and biological function of identified features
R normal condition culture
N nitrogen depleted culture
Bold values indicate an up-regulation for the nitrogen depleted regime
Italicized values indicate a down-regulation for the nitrogen depleted regime
RT retention time
b
d (Observed mass - theoretical mass)/theoretical mass 9 106
c
p value of Mann–Whitney test
F. Courant et al.
a
Author's personal copy
Assessing biomarkers of nitrogen stressed cultured microalgae using metabolomics
hydrophobic amino acids. Indeed, the shikimate pathway,
the biosynthetic pathway by which the aromatic amino
acids, phenylalanine, tyrosine and tryptophan were
assembled, was found to be impacted in their proteomic
experiments, thus supporting the regulation observed in our
experiment for tryptophan and a catabolic byproduct of
phenylalanine (Phenyl pyruvic acid). The fact that coherent
pathways were identified with two independent omics
experiments suggests the biological relevance of such a
reversible trigger.
(iii) The auxin signaling pathway could be assessed in
our proof of principle study. Surprisingly, indole acetic
acid (also called auxin) was detected among the most
positively impacted LC-HRMS features during this progressive nitrogen regime. This phytohormone, described to
be mainly implicated in superior plant growth during its
whole development, has also been detected in microbial
organisms, fungi, animals and algae lineage (Lau et al.
2009). The auxin function remains however unclear in
microalgae even if known precursors were identified in
other organisms [5-hydroxy-tryptamine for O. sativa for
example (Sanchez et al. 2008)]. In our case, no tryptamine
derivatives were detected among the most impacted features. Nevertheless, the opposite progressive regulation of
both auxin (accumulation) and its distant precursor tryptophan (reduction) remained the strongest evidence suggesting the existence of an auxin-like signaling pathway in
microalgae as already proposed by De Smet et al. (2011),
Lau et al. (2011). According to the observed regulation
pattern, it was reasonable to hypothesize that the progressive nitrogen regime induced a stress response regulatory
loop where the reduction of tryptophan levels could be due
to its constant mobilization as a precursor to biosynthesize
auxin phytohormone. It could also be the consequence of a
metabolic switch involving an unknown regulatory pathway that would be tryptophan independent and able to
activate the biosynthesis of auxin phytohormone from an
indole derivative for example as suggested by Soeno et al.
(2010) (Fig. 4b).
(iv) The nucleoside metabolisms were found to be
impacted. Nucleoside compounds were found to be down
regulated during our nitrogen limitation protocol. Adenosine, adenosine derivative, guanosine and guanosine
derivative were triggered in response to a low level of
available nitrogen (Table 1). This finding was quite surprising, since those metabolites are not known to be
directly involved in nitrogen sensing in microalgae. Nevertheless, the fact that those metabolites—and moreover
their regulation—were highlighted by independent experiments with different techniques—i.e. ours and Lee do and
colleagues’s investigations (Lee do et al. 2012)—, reinforce the novelty of such metabolites as being implicated in
nitrogen induced signaling pathway and the pertinence of
the discovered biomarkers in this proof of concept.
3.3 Last objective of this proof of concept study:
evaluating the biomarkers suggested
by the untargeted metabolomic investigation
for nitrogen limited microalgae growing
in photobioreactor
3.3.1 Chlamydomonas reinhardtii cells were grown
in photobioreactors under two continuous regimes
providing reference and nitrogen limited culture
conditions
The non-targeted metabolomic investigation performed on
C. reinhardtii cells grown in nitrogen-limited batch cultures condition proposed several biomarkers involved in
carbon storage molecules, amino-acids, nucleosides
metabolism or putative algae hormone signaling pathway.
The last objective of this proof of concept study consisted
in assessing the preliminary verification of the metabolomic proposed targets in real small-scale industrial production conditions. For that purpose, a continuous culture
was performed to grow C. reinhardtii cells in nutrient
optimized and highly controlled environment (photobioreactor) providing a reference culture in chemostat mode. By
modifying the dilution rate of the growing medium, it was
possible to obtain a nitrogen-modified chemostat mode
thus providing a nitrogen-limited continuous culture assay
(Table S1). Biochemical analyses were performed on both
C. reinhardtii cultures, targeting carbon storage molecule
and proteins preferentially. Total carbohydrates, starch and
total protein analyses were performed to confirm the
accumulation of energetic macromolecules and the direct
effect of nitrogen starvation on protein content. Aminoacids profiling was moreover accomplished to assess the
pertinence of amino-acids biomarkers suggested by untargeted-metabolomics.
3.3.2 Targeted biochemical analysis of macromolecules
confirmed FTIR analysis
The biochemical results confirmed the accumulation of
carbohydrate monitored by FTIR analysis on batch cultures. The amount of total carbohydrates rose from 17 % of
biomass dry weight (%DW) in the non-limited culture up
to 60–70 %DW in the nitrogen-limited culture. Moreover
starch represented more than 50 % of total carbohydrates
in our nitrogen-limited photobioreactor culture conditions.
As expected, the total protein content was also impacted by
the nitrogen-limitation protocol with content decreasing
from 50 %DW to 10 %DW in limited condition.
123
Author's personal copy
F. Courant et al.
A
Glucose
B
Pyruvate
Alanine
Acetolactate
Diacetyl
?
Valine
Ketoisovalerate ?
2-MBA
?
Kleu
Tryptamine
D-Tryptophan
N-hydroxyltryptamine
Indole-3-pyruvic acid
Leucine
Indole-3-acetaldoxime
Indole-3-glucosinolates
Pantoate
Indole-3-acetaldehyde
Indole
?
?
N-(5-phospho-β-D
-ribosyl) anthranilic acid
Phenyl pyruvic acid
Tyrosine
Phenylalanine
Anthranilic acid
Indole-3-acetonitrile
?
Ketopantoate
1-(OCarboxyphenylamino)-1’
-deoxy-D-ribose-5’
-phosphate
Indole-3-glycerol
-phosphate
Indole
L-Tryptophan
Aspartate
Indole-3
-acetamide
Indole-3-acetic acid
β-alanine
Camalexin
Indole-3-acetic
acid conjugate
Chorismic acid
5-Enolpyruvylshikimate-3
-phosphate
Pantothenate
Coenzyme A
Fig. 4 Maps of putative impacted pathways suggested by metabolomics investigations. MS detected metabolites are highlighted in bold
on the pathways. Relative level of biomarkers are indicated with
arrows, first one accounting for N- regime and second one for N–
regime. Green arrow corresponds to down regulation (regime
compared to standard) and red arrow to up regulation (regime
compared to standard). White question mark indicates that observed
regulation presented a p value above 0.05 threshold. Indirect
connections between metabolites are indicated with dashed arrows.
a pantothenate pathway as depicted by Chassagnole et al. (2002). 2MBA stands for 2-methylbutyric acid, Kleu for ketoleucine. Putative
connections between the leucine pathway and the pantothenate
pathway are indicated with red dashed arrows with red question mark
on it. b Auxin biosynthesis pathway as depicted by