How to Process Samples for Metabolomics in Ecology: A Comparative Study of Preanalytical Storage Methods

ABSTRACT

The technical progress of metabolomics—the analysis of metabolite composition of cells, tissues, and biofluids—and its ability to unravel phenotypical responses of organisms to their environment was accompanied by the democratization of their use in a wide range of scientific fields. While the reliability of analytical procedures was thoroughly assessed, the effects of preanalytical processing on sample metabolic profiles (i.e., conditioning, storage, transportation, etc.) remain quite unclear. This uncertainty is especially notable for samples collected in the frame of ecological studies, which often involve field sampling and remote sample production, leading to extended transportation durations and suboptimal storage conditions. In this study, we evaluated the impact of storage duration and temperature, along with the effects of freeze‐drying (or lyophilization; the process of stabilization through dehydration by sublimation), on the metabolic profiles of samples relevant to ecological studies. Specifically, we focused on the lesser mealworm (Alphitobius diaperinus), the fruit fly (Drosophila melanogaster), and the perennial ryegrass (Lolium perenne). The levels of 60 different metabolites were quantitatively analyzed using targeted metabolomics through gas chromatography–mass spectrometry (GC–MS). We report significant metabolic shifts associated with freeze‐drying, resulting in both increases and decreases in the contents of more than half of the quantified metabolites across all assessed chemical families. Several amino acids exhibited more than a fourfold increase in all investigated matrices. Furthermore, while samples stored at −80°C exhibited profiles most similar to those of samples analyzed right after collection, the metabolic profiles of these samples gradually changed over the 6 months of storage. Interestingly, metabolic shifts related to sample preanalytical processing and storage were relatively consistent across the biological matrices studied, particularly between the two insect species. Based on these observations, we propose several recommendations for reliable preanalytical sample processing in ecological studies, considering logistical and economic constraints.

1. Introduction

Metabolomics, also known as metabolic profiling, is the holistic approach to determine the metabolite composition of cells, tissues, and biofluids. The comprehensive analysis of metabolites allows to undercover the influence of an environment on the physiology, transcriptome, proteome, and overall phenotype of a given organism [1, 2, 3]. Over the past 40 years, metabolomics has been increasingly used in a wide range of research fields, including medicine [1, 2], microbiology [4], food [5, 6, 7], and ecology [8, 9, 10, 11], offering insights into the physiology of organisms, either under standard conditions or in response to treatments, diseases or stressful conditions. Targeted and untargeted analysis can be carried out to study metabolomic pathways or to provide a comprehensive view of the metabolome, enabling the discovery of new biomarkers signaling, for instance, diseases, effects of drugs, or environmental stress.

To ensure the reproducibility of generated data, metabolomics follows a general workflow that can be adapted depending on the type of sample and the field of study. The metabolomics workflow involves the collection and preprocessing of samples (e.g., plasma separation from blood), quenching, aliquoting, transporting, storing, acquiring data using specific technologies, and analyzing the data using tools and metabolite databases.

Numerous studies have focused on optimizing various steps of the workflow, such as the extraction method, sample‐processing stages, and data collection. This has sometimes resulted in the production of standard operating procedures [1, 2, 11, 12, 13, 14, 15]. Research, mostly focusing on biological human samples (e.g., urine and blood), has been able to list the important factors that may affect metabolomics results on these matrices, non‐exhaustively including the sample collection time, as well as the time and method employed for sample pre‐processing, transport, and storage [6, 15, 16, 17]. However, there is a lack of studies on how storage conditions affect the measurement of primary metabolites on nonhuman biological samples. Ecological studies can be particularly affected by this issue, since biological samples are frequently collected far from analytical facilities. In these cases, samples must be stabilized until they can be analyzed to prevent alteration in the metabolic profile and the production of artifacts. One widely used method involves quenching samples in liquid nitrogen and storing at −80°C before sending them to the laboratory under dry ice [3, 18]. However, this method involves equipment that is not usually available at collection sites. Another limitation of this method is the high cost of transport under dry ice, coupled to the risk of sample degradation in the event of a transportation issue. For this reason, samples in metabolomics are often stabilized by freeze‐drying. It involves exposing the sample to low temperature and low pressure, thus dehydrating it by sublimation during 24–72 h. This process stops enzymatic activity and limits metabolite oxidation due to the absence of free water from the sample [19]. It is commonly used in metabolomics due to its convenience and is considered to make biological samples stable enough to allow storage and transportation at room temperature (RT) [17]. Moreover, freeze‐drying generally facilitates sample grinding, especially in hard tissues such as plants and insects, and thus the efficiency of extractions [20]. Despite procuring several advantages, freeze‐drying was reported to modify metabolic profiles in various biological samples [21, 22, 23, 24]. For instance, after freeze‐drying wheat roots, 7% of metabolites became undetectable and 43% of metabolites presented both statistically increased and decreased abundances [21]. While similar alterations were also found in several other photosynthetic organisms [22, 23, 24], the effects of freeze‐drying on invertebrate samples remain relatively widely unexplored, especially in the context of transport and storage.

In addition to freeze‐drying, the duration and temperature of storage may further influence metabolite concentrations in ecological samples. In a recent study, Lang et al. (2024) defined a metabolite extraction method adequate to field conditions on maize leaves, in a context where freeze‐drying and non‐stop cooling could not be carried out [20]. They tested three different storage temperatures (30°C, 4°C, and −20°C) for a duration ranging from 1 to 75 days, and compared extraction protocols performed in the field versus in the laboratory. Both duration and temperature of storage influenced metabolic profiles [20]. While this study provided valuable insights, the combined effects of storage duration and temperature along with freeze drying were not explored.

Since freeze‐drying, storage duration, and temperature can influence metabolic profiles, we questioned their combined impact on metabolite levels in biological samples relevant in the scope of ecological studies. Here, we addressed this matter by examining the metabolic profiles of three biological models: two insect species the fruit fly Drosophila melanogaster and the lesser mealworm Alphitobius diaperinus, and one gramineous plant Lolium perenne (perennial ryegrass), using a quantitative, targeted GC–MS approach. To that aim, these samples were subjected to three storage durations (0 days, 15 days, and 6 months) and three storage temperature conditions (RT, −20°C, and −80°C), with and without freeze‐drying.

We expected strong metabolic shifts in freeze‐dried samples, with potential decreases leading to undetectability due to metabolite degradation, but also possible increases linked to macromolecule degradation/increased extractability [25]. However, we expected freeze‐dried samples to be more stable through time due to reduced enzymatic activities associated with dehydration [19, 26]. Lastly, while −80°C without freeze‐drying was expected to be the optimal storage condition, characterized by marginal shifts in metabolite contents, extended storage durations may lead to increased alterations in the concentration of quantified metabolites due to possible degradation processes.

2. Material and Methods

2.1. Experimental Design

Samples from three biological models—the fruit fly D. melanogaster (Diptera: Drosophilidae), the lesser mealworm A. diaperinus (Coleoptera: Tenebrionidae), and the perennial ryegrass L. perenne (Angiosperm: Poaceae)—were subjected to six different storage treatments before extraction, plus a fresh control. All samples were immediately snap‐frozen in liquid nitrogen to kill the individuals and stop metabolic activity. The fresh mass (FM) of each sample was then quickly measured using a high‐precision microbalance (Mettler Toledo UMX2, accuracy 0.001 mg). Each condition was tested with five independent biological replicates. The storage treatments, listed in Table 1, were:

• Control (fresh, Day 0)—extracted immediately without conservation (Code C: gray).

• Frozen at −80°C for 15 days (Code 15d_80: light blue) and 6 months (Code 6m_80: dark blue).

• Lyophilized (freeze‐dried) and stored at RT for 15 days (Code 15d_L_RT: bright red) and 6 months (Code 6m_L_RT: dark red).

• Lyophilized and stored at −20°C for 15 days (Code 15d_L_20: light green) and 6 months (Code 6m_L_20: dark green).

TABLE 1.

Storage conditions, durations, and corresponding treatment codes and colors used in the figures. Each condition was applied to Drosophila melanogaster, Alphitobius diaperinus, and Lolium perenne samples with five replicates per treatment.

Storage duration	Control (fresh)	−80°C	Lyophilized/RT	Lyophilized/−20°C
0 days	C (gray)	—	—	—
15 days	—	15d_80 (light blue)	15d_L_RT (bright red)	15d_L_20 (light green)
6 months	—	6m_80 (dark blue)	6m_L_RT (dark red)	6m_L_20 (dark green)

These seven conditions were applied independently to each of the three models to assess the effect of conservation method and duration on sample metabolic integrity.

2.2. Live Material

Experiments were conducted on an outbred laboratory population of D. melanogaster derived from wild individuals collected in September 2015 in Brittany (France). Fly stocks were maintained in incubators (MIR‐154‐PE, Panasonic) set at 25°C, 60% RH and under a 12:12 L photoperiod, on standard fly medium comprising 80 g.L⁻¹ of inactive brewer yeast (MP Biochemicals 0290331205), 50 g.L⁻¹ of sucrose, 10 g.L⁻¹ of agar (Sigma‐Aldrich A1296), supplemented with 8 mL.L⁻¹ of 10% methyl 4‐hydroxybenzoate (Sigma‐Aldrich H5501). Flies were cultured in 100 mL glass bottles filled with 20 mL of artificial food. A minimum of 20 bottles was kept every generation.

The stock of beetles, A. diaperinus, was field‐collected from a poultry house in northern Brittany (France) (Miniac sur Bécherel: X: 48.29, Y: −1.95) in 2018 [27]. After collection, hundreds of individuals were continuously maintained in large plastic boxes at 25 ± 1°C and in continuous dark. The insects were supplied with dry dog food ad libitum, pieces of carrots, and Styrofoam to stimulate pupation as described in Gouesbet et al. (2025) [37]. Water was supplied with moist cotton wool.

L. perenne (diploid Brio variety) seeds were obtained from the Phytosem seed company. Seeds were sterilized as in Serra et al. (2015) [38]. After 2 days in the dark at 4°C to synchronize germination, seeds were sown in vermiculite moistened with sterile Hoagland solution medium (No.2HOP01, Caisson Labs, 1.63 g.L⁻¹, pH 6, 2 mL.g⁻¹ of vermiculite). Devices were maintained in controlled conditions at 20 ± 1°C, 50 ± 10%RH, D = 16 h/N = 8 h, 130 µmol.s⁻¹.m⁻² for 12 days for germination and growth. Standardized seedlings of 5–8 cm at the two‐leaf stage were used for the experiments.

2.3. Sample Extractions

Plant shoots (∼150 mg FM), pools of 15 fruit fly females (∼20 mg FM), and pools of three final‐instar lesser mealworm larvae (∼50 mg FM) were extracted as described in Serra et al. (2015) for plants and in Li et al. (2025) for insects [38, 39].

In all cases, five biological replicates were processed. Samples were promptly homogenized in an ice‐cold methanol‐chloroform solution (2:1, v/v) using two tungsten beads and a bead‐beating device (Retsch MM301, Retsch GbmH, Haan, Germany) at 25 beats per second for 90 s. The volume of methanol‐chloroform used was 750 µL for D. melanogaster, 1000 µL for A. diaperinus, and 600 µL for L. perenne. After homogenization, samples were incubated at −20°C for 3 h. Then, ice‐cold ultrapure water was added: 500 µL for D. melanogaster samples and 666 µL for A. diaperinus, and 400 µL for L. perenne. Samples were vortexed thoroughly. Finally, the samples were centrifuged at 4000 g for 10 min at 4°C. An aliquot of the upper aqueous phase, which contains polar metabolites, was transferred to new Eppendorf tubes: 750 µL for D. melanogaster and 1000 µL for A. diaperinus, and 600 µL for L. perenne.

A specific volume of this upper aqueous phase, which contained polar metabolites, was transferred to GC–MS glass vials: 50, 50, 30 µL for D. melanogaster, A. diaperinus, and L. perenne, respectively. Two deuterated internal standards (glucose‐d7 Merck CAS: 23403‐54‐5, glycine‐d5 Merck CAS: 23403‐54‐5) were added in each vial to normalize the GC–MS responses. These aliquots were then dried under vacuum at 33°C using MiVac concentrator (Genevac Ltd., Ipswitch, England).

2.4. Online Derivatization and GC–MS Data Acquisition

Targeted GC–MS analyses were performed as described in Prud'Homme et al. (2018) [40]. Briefly, prior to GC–MS analysis, polar metabolites from extracts were derivatized online using a CTC PAL auto sampling/injection robot (CTC Analytics AG, Zwingen, Switzerland). First, the dry residue was resuspended in 30 µL of 25 mg.mL⁻¹ methoxyamine hydrochloride (CAS: 593‐56‐6) in pyridine (CAS: 110‐86‐1). This methoxymation step was performed at 40°C for 60 min under automatic orbital shaking. Then, the derivatization through silylation is processed with the addition of 30 µL of BSTFA (N,O‐bis(trimethylsilyl)trifluoroacetamide; CAS: 25561‐30‐2) at 40°C for 60 min under agitation. Polar metabolites were separated, identified, and quantified using a GC–MS platform consisting of a 7890 chromatograph and a 5977B quadrupole mass spectrometer (Agilent Technologies, CA, USA). Each derivatized sample (1 µL) was introduced into the injector at 250°C in split mode with a ratio of 25:1. Chromatographic separation was performed using a HP‐5MS Ultra Inert capillary column (Agilent Technology 30 m × 25 mm × 0.25 µm) with helium as carrier gas at a constant flow rate of 1 mL.min⁻¹. The oven temperature ranged from 60°C to 325°C at 10°C.min⁻¹, and then remained at 325°C for 10 min. The column effluent was ionized by electron ionization at 70 eV at an ion source temperature set to 230°C and the MS transfer line at 280°C. Detection was achieved under the full‐scan mode at a scan speed of 1.562 from 60 to 600 amu after a 4.9 min solvent delay. We screened for the 61 pure reference compounds included in our custom spectral database (Table S1). The calibration mixture was freshly prepared for each run from pure reference compounds stored individually at −80°C. Quadratic model calibration curves were constructed with the ratio of compound area to internal standard area for seven concentration levels: 5, 10, 20, 50, 100, 200, 500 µM. Sample order was randomly distributed in the injection sequence to avoid any bias and to reduce the influence of potential confounding factors. Chromatograms were deconvoluted using MassHunter Quantitative Analysis B.09.00. Target compounds should display a signal‐to‐noise of 10:1 for quantification and 3:1 for qualitative data. Peaks considered valid needed to be correctly integrated, and present one quantitative ion and two qualitative ions with the correct qualitative to quantitative ion ratios. If the response of a compound validated all the previous criteria but fell outside the calibration range, only the qualitative response was used (glucose for all samples, proline for A. diaperinus, and malic acid, fructose, sucrose, citric acid for L. perenne). Two types of data were therefore processed in this study, quantitative data (µmol.mg⁻¹ of fresh sample) or qualitative data (compound/ISTD area.mg⁻¹ of fresh sample).

2.5. Statistical Analyses

Missing values (i.e., NA) were treated in three different ways depending on the number of missing values (nNA) within each experimental condition, which included five independent true biological replicates. When nNA = 5, the missing values were considered a nonrandom absence of the molecule (putatively due to concentrations falling below the limits of quantification [LOQ]; [28]). In this case, they were replaced with zero. When nNA = 1, it was assumed that the metabolite was most likely present but that a random analytical error had occurred in one replicate (i.e., random error or stochastic fluctuation during the data acquisition process; [28]), so the missing value was replaced with the mean of the other four replicates. However, when nNA was between 2 and 4, the molecule was considered unsuitable for multivariate analysis and was therefore removed from the dataset.

All statistical analyses and plots were performed in RStudio [29]. The effects of the storage treatment on each variable were assessed using ANOVA, followed by post hoc comparisons with the “emmeans” package. Metabolic profiles were compared using between‐class PCA with the “ade4” package [30], a semi‐supervised ordination method that highlights between‐group structure similar to PLS‐DA. Monte Carlo tests were then performed to assess the significance of the differences among the classes (based on 1000 simulations). To identify the variables (i.e., metabolites) contributing most to the between‐class PCA structure, the correlations to the principal components (PCs) were extracted and visualized/ranked in heatmaps.

3. Results and Discussion

All biological matrices considered, sample storage—whatever its type, temperature, and duration—significantly affected metabolite contents (Figures 1, 2, 3, Figure S1). Indeed, in samples stored for at least 15 days after sample preparation, metabolic patterns were differentiated in the between‐PCA from those of samples processed immediately (Figures 1, 2, 3). These variations consisted in both significant decreases and increases in several metabolite contents, occurring in all metabolic families (e.g., amino acids, soluble sugars, Krebs cycle intermediates; Figures S2–S4). In addition, metabolites were often present in control samples but absent in other conservation treatments (e.g., quinic acid and sucrose in A. diaperinus; Figure S3) or absent in control samples but present in other treatments (e.g., lactic acid and ethanolamine in D. melanogaster or glycerol in L. perenne; Figures S2 and S4). These patterns are associated with metabolites that display concentrations near to the Limit of Quantification (LOQ). Consequently, they are considered absent when their levels fall below this LOQ.

FIGURE 1.

Comparison of metabolic profiles from Drosophila melanogaster depending on the conservation method (lyophilized or not), duration (15 days or 6 months), and temperature (room temperature, −20°C, or −80°C). (a) Between‐PCA of metabolic profiles depending on the conservation method (C = unprocessed control, 15d_80/6m_80 = no freeze‐drying then 15 days/6 months at −80°C, 15d_L_20/6m_L_20 = freeze‐drying then 15 days/6 months at −20°C, 15d_L_RT/6m_L_20 = freeze‐drying then 15 days/6 months at room temperature) and (b) associated heatmaps of each metabolite contribution to the first two axes (metabolite abbreviations are provided in Table S1). Monte Carlo randomizations were conducted to confirm the significance of the differences among classes in the between‐PCA (p < 0.001; Figure S1a). (c) Boxplots of the top metabolites contributing to each axis, one for the positive and one for the negative side: ethanolamine (+) and malic acid (−) for Axis 1, tyrosine (+) and lactic acid (−) for Axis 2. Associated statistical tests are provided in Table S2, and responses of all metabolites and associated statistical differences are available in Figure S2.

FIGURE 2.

Comparison of metabolic profiles from Alphitobius diaperinus depending on the conservation method (lyophilized or not), duration (15 days or 6 months) and temperature (room temperature, −20°C, or −80°C). (a) Between‐PCA of metabolic profiles depending on the conservation method (C = unprocessed control, 15d_80/6m_80 = no freeze‐drying then 15 days/6 months at −80°C, 15d_L_20/6m_L_20 = freeze‐drying then 15 days/6 months at −20°C, 15d_L_RT/6m_L_20 = freeze‐drying then 15 days/6 months at room temperature) and (b) associated heatmaps of each metabolite contribution to the first two axes (metabolite abbreviations are provided in Table S1). Monte Carlo randomizations were conducted to confirm the significance of the differences among classes in the between‐PCA (p < 0.001; Figure S1b). (c) Boxplots of the top metabolites contributing to each axis, one for the positive and one for the negative side: adonitol (+) and glucose‐6‐P (−) for Axis 1, lactic acid (+) and tyrosine (−) for Axis 2. Associated statistical tests are provided in Table S2, and responses of all metabolites and associated statistical differences are available in Figure S3.

FIGURE 3.

Comparison of metabolic profiles from Lolium perenne depending on the conservation method (lyophilized or not), duration (15 days or 6 months) and temperature (room temperature, −20°C, or −80°C). (a) Between‐PCA of metabolic profiles depending on the conservation method (C = unprocessed control, 15d_80/6m_80 = no freeze‐drying then 15 days/6 months at −80°C, 15d_L_20/6m_L_20 = freeze‐drying then 15 days/6 months at −20°C, 15d_L_RT/6m_L_20 = freeze‐drying then 15 days/6 months at room temperature) and (b) associated heatmaps of each metabolite contribution to the first two axes (metabolite abbreviations are provided in Table S1). To maximize the difference between treatments and present patterns similar to the other biological matrices, Axis 3 is presented here rather than Axis 2. Monte Carlo randomizations were conducted to confirm the significance of the differences among classes in the between‐PCA (p < 0.001; Figure S1c). (c) Boxplots of the top metabolites contributing to each axis, one for the positive and one for the negative side: putrescine (+) and glycerol acid (−) for Axis 1, tyrosine (+) and malic acid (−) for Axis 3. Associated statistical tests are provided in Table S2, and responses of all metabolites and associated statistical differences are available in Figure S4.

Interestingly, the different variations observed following storage were relatively consistent between biological models, especially in A. diaperinus and D. melanogaster. For example, sugars such as glucose‐6‐phosphate or trehalose and maltose were systematically less present in stored samples, while amino acids such as leucine or valine were more present. This recurrence suggests a general pattern that could occur in all biological matrices, and thus that sample storage should be avoided as much as possible to produce reliable metabolic profiling results. Such effects of sample storage on the quality of metabolomic analyses were previously underlined in the scope of medical biology [17, 31, 32]. In the scope of ecology in field conditions, Lang et al. (2024) recently demonstrated the influence of storage duration (from 1 to 75 days), storage temperature (from −20°C to 30°C), and air‐drying on maize leaf tissues metabolite levels [20]. This study provides further evidence of the impact of storage on metabolic contents concerning plant and animal samples [20].

Concerning insect samples, D. melanogaster and A. diaperinus, the between‐PCA of metabolic profiles strongly differentiated storage conditions along Axis 1 (52.08% and 72.46% of the variance, respectively, Figure 1a and Figure 2a), with control samples and samples kept at −80°C being located on the negative side of Axis 1 while freeze‐dried samples were located on the positive side of this axis (L_RT and L_20; Figure 1a and Figure 2a). These differences were mostly related to soluble sugar contents and organic acids from the Krebs cycle being overrepresented in control and −80°C samples, and to amino acids being overrepresented in freeze‐dried samples (Figure 1b,c and Figure 2b,c, Figures S2 and S3). As a striking example, glucose‐6‐phosphate and citric acid were overrepresented in control and −80°C samples from both D. melanogaster and A. diaperinus, while alanine and ethanolamine were overrepresented in freeze dried samples (Figure 1b,c and Figure 2b,c).

Concerning plants, storage conditions were less differentiated in the between‐PCA (42.96% of the variance explained along Axis 1; Figure 3). However, control samples and samples stored at −80°C were again differentiated from freeze‐dried samples, this time along Axis 3 (Ctrl and −80°C on the positive side, freeze‐dried on the negative side; 16.03% of the variance; Figure 3). Freeze‐dried samples were characterized again by low contents in glucose‐6‐phosphate and citric acid and high amounts of amino acids such as leucine, isoleucine, and valine (Figure 3, Figure S4). However, in contrast to the patterns observed in insect samples, plant freeze‐dried samples exhibited higher fructose levels and lower serine levels compared to the control samples.

Freeze‐drying led to strong shifts in metabolic profiles of all investigated biological matrices. These shifts were mostly related to significant increases in metabolite contents. In the striking case of A. diaperinus, 24 metabolites out of 34 displayed increased contents in freeze‐dried samples compared to control ones, while only 3 displayed significantly lower concentrations (Figure S3). These 24 metabolites include all investigated chemical families, but amino acids seem particularly concerned. Such increases in freeze‐dried samples could be related to increased extractability already observed in marine polychaetes [33]. However, different outcomes of freeze‐drying were observed in other plant samples, such as decreased levels of several metabolites in Arabidopsis thaliana and pear fruits [23], and a tendency toward a decrease in wheat roots metabolic contents (e.g., 64 increased and 128 decreased abundances; [21]). In addition, the degradation of macromolecules into simpler compounds may be responsible for metabolic profile shifts (e.g., proteolysis; [17]). Such degradation can occur when samples are not completely dried due to marginal enzymatic activity, and freeze‐drying was often found to produce samples with residual humidity [23] and thus subjected to potential biodegradation of macromolecules. However, it remains unclear whether the observed shifts in metabolic profiles are linked to the freeze‐drying process itself (e.g., increased extractability; [33]) or to subsequent storage and its duration. To better tackle this matter, analyzing the metabolic profiles of samples immediately after freeze‐drying would provide valuable information. Nevertheless, our results clearly indicate that freeze‐drying should be avoided whenever possible when analyzing metabolic profiles, and that storage at −80°C—while not without limitations—is preferable over freeze‐drying.

Aside from the effects of freeze‐drying, Axis 2 also allowed to discriminate between the different D. melanogaster samples (21.77% of the variance, Figure 1a) and, to a lesser extent A. diaperinus samples (13.41% of the total variance, Figure 2a). While the contribution of this axis to the total variance was much lower than Axis 1, samples were differentiated based on the duration of the conservation (15 days vs. 6 months). Indeed, D. melanogaster samples stored for 15 days were located on the same side of Axis 2 than control samples (positive side), while those stored for 6 months were located on the other side and were characterized by lower levels in amino acids such as tyrosine and methionine and higher levels in organic acids such as aspartic and lactic acid (negative side, Figure 1a–c). Concerning A. diaperinus samples, an opposite pattern was observed, with samples stored for 6 months being closer to control samples than those stored for 15 days (Figure 2a). In plants, these duration‐related patterns were observed in all samples, whether they were stored at −80°C or freeze‐dried and stored at RT/−20°C (Figure 3a). Metabolic profiles of plant samples stored for 15 days were thus closer to those of control samples than those of samples stored for 6 months, displaying, for example, intermediate levels of glucose‐6‐phosphate and malic acid (Figure 3c, Figure S4).

These results overall underlined variability in metabolic profiles through time, including in samples stored at −80°C. Due to the acknowledged reliability of sample storage at −80°C [34], samples maintained at this temperature were expected to remain stable compared to freeze‐dried samples, which were stored either at RT or −20°C. However, significant variations were still observed, with both increases and decreases in metabolite contents occurring at this temperature. Although such variations are generally attributed to the defrosting process [17], we observed strong differences between the metabolic profiles of samples stored at −80°C for 15 days compared to 6 months. For example, several metabolites were significantly reduced in samples stored for 15 days compared to control samples, and showed an even greater decrease after 6 months of storage (e.g., tyrosine in D. melanogaster). This suggests that samples are unstable at this temperature, which could be linked to enzymatic activities often occurring at temperature as low as −73°C [35]. Therefore, storage at −80°C seems to introduce bias in metabolomic analyses, but remains quite close to control samples, especially for short durations (15 days here).

The differences in metabolic profiles observed between samples stored for 15 days and those for 6 months may also be partly attributable to changes in analytical conditions over time. Factors such as gradual column contamination or minor adjustments during routine maintenance can indeed affect analytical performance. However, to minimize the impact of such potential variations, several precautions were implemented: internal standard was added to samples during analysis, quality control (QC) samples were regularly injected throughout the analytical runs; scheduled preventive maintenance was performed on the instrument; mass spectrometer was tuned before each run, calibration curves were freshly prepared for each analytical session, and LOQs were adjusted accordingly. However, based on our results, conducting metabolomic analyses at precise and similar time‐points between samples seems crucial to provide increased reliability. In addition, a more precise assessment of metabolite temporal dynamics during low temperature storage would provide more detailed information to assess optimal storage duration and provide key information for sample treatment procedures. Moreover, as suggested by Lang et al. (2024), storing metabolites in solvent extract might increase stability over time, yet this would require further experiments since artifacts linked to conservation in solvents have often been observed [20, 36].

4. Conclusion

Throughout this study, we underlined the effects of various processes linked to sample conservation on quantitative targeted metabolomics output. These effects were assessed in biological matrices on which information is scarce when it comes to metabolic profiling, and for which guidelines are rarely given concerning preanalytical sample processing. Based on our results, several recommendations can be made. First, storage associated to freeze‐drying is strongly discouraged due to the significant metabolic shifts it induces; instead, samples should be snap‐frozen and stored at −80°C whenever possible. Second, storage duration should be kept as short as possible, as progressive changes in metabolite contents were observed over time, even in samples stored at −80°C. Finally, to ensure reliable comparisons between treatments, all samples should be analyzed after the same storage duration; samples from different treatments should never be compared if they have undergone different conservation durations. However, due to logistical and experimental constraints associated with ecological field studies, some of these recommendations may not be feasible. In such cases, we refer to the recommendation of Fiehn et al. (2016) who suggested creating a pooled sample composed of all individual samples, and repeatedly analyzing its metabolic profile at each run. This pooled QC can provide a robust basis for data normalization, helping to prevent misinterpretations related to sample processing and storage conditions.

Author Contributions

Nathalie L. Bris: conceptualization, data curation, formal analysis, investigation, methodology, project administration, writing – original draft, writing – review and editing. Axel Beringue: formal analysis, investigation, resources, visualization, writing – original draft, writing – review and editing. Johanna Rivas: investigation, resources, writing – original draft, writing – review and editing. David Renault: conceptualization, resources, writing – review and editing. Marion Chorin: conceptualization, data curation, formal analysis, investigation, writing – review and editing. Hervé Colinet: conceptualization, formal analysis, investigation, methodology, project administration, resources, visualization, writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: jssc70333‐sup‐0001‐SuppMat.pdf.

Supporting File 2: jssc70333‐sup‐0002‐TableS1.xlsx.

Supporting File 3: jssc70333‐sup‐0003‐TableS2.xlsx.

Le Bris N., Beringue A., Rivas J., Renault D., Chorin M., and Colinet H., “How to Process Samples for Metabolomics in Ecology: A Comparative Study of Preanalytical Storage Methods.” Journal of Separation Science 48, no. 12 (2025): e70333. 10.1002/jssc.70333

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.