Metabolic fingerprinting of gilthead seabream (Sparus aurata) liver to track interactions between dietary factors and seasonal temperature variations

Introduction

Metabolomics is usually defined as the holistic study of metabolites in living systems, which includes most molecules present apart from polynucleotides (mostly studied by genomics and transcriptomics) and proteins (mostly studied by proteomics). In a way, this field can be seen as an extension of analytical chemistry and chemometrics, given that the simultaneous study of the wide range of molecules present in cells and biological fluids only became possible with the continuous instrumentation and methodological developments in these areas. It is not surprising, then, that most metabolomic studies apply popular analytical chemistry techniques (from mass spectrometry and nuclear magnetic resonance, to optical spectroscopy, atomic spectroscopy and gravimetry, among others) optionally coupled to separation techniques (usually chromatography or electrophoresis-based) (Dunn & Ellis, 2005; Goodacre et al., 2004).

Metabolomic studies are often of a comparative nature and usually take either a metabolic profiling and/or a metabolic fingerprinting approach. In the first case, the focus is on obtaining a quantitative estimate of all (or a specific subset of the) metabolites present in a certain biological sample. Given the high complexity of biological samples, metabolic profiling usually requires high-resolution separation methods prior to analysis, which explains the popularity of “hyphenated mass spectrometry” methods (such as GC-MS, HPLC-MS and CE-MS) for this purpose (Dettmer, Aronov & Hammock, 2007; Scalbert et al., 2009). The second approach, metabolic fingerprinting, avoids altogether the deconvolution of the mixture into its different components and simply attempts to capture a multivariate fingerprint of a biological sample in some arbitrary feature space in a way that the similarity of samples in the “real” metabolomic feature space can be deduced/estimated from their similarity in the fingerprinting space. For this purpose, techniques such as ¹H-NMR, ¹³C-NMR, DIMS (Direct Injection Mass Spectrometry), vibrational (FT-IR or Raman) spectroscopy and pyrolysis mass spectrometry have been commonly used, which, although not nearly as sensitive and specific as hyphenated methods, are often less expensive, quicker and/or require less sample preparation (Dettmer, Aronov & Hammock, 2007; Dunn & Ellis, 2005; Ellis et al., 2007).

This manuscript describes the results of a trial in which we explored the use of transmissive FT-IR spectroscopy for metabolic fingerprinting of liver tissue from gilthead seabream (Sparus aurata), to better understand how seasonal temperature variations and dietary factors affect the hepatic metabolic content. FT-IR spectroscopy (as other types of IR spectroscopy) is based on the differential absorption of IR radiation with specific wavelengths by different molecules. The characteristic wavelengths absorbed by a specific molecule depend on the energy differences between vibrational states of the molecule, being mostly defined by the presence of particular functional groups, as each functional group tends to display specific vibration modes, thereby conditioning the energy of allowed vibrational transitions (Ellis & Goodacre, 2006). Though the use of FT-IR spectroscopy as a metabolic fingerprinting technology in fish biology and aquaculture research is still incipient, several studies have already pointed out its usefulness in such diverse contexts as the differentiation between wild vs. farmed gilthead seabream (Ceylan, Tanrikul & Özgener, 2014), the study of the effects of estrogens and other pollutants in rainbow trout (Çakmak, Togan & Severcan, 2006; Çakmak et al., 2003), analysis of the effects of zinc and arsenic in indian carp (Palaniappan, Nishanth & Renju, 2010; Palaniappan & Vijayasundaram, 2008; Palaniappan & Vijayasundaram, 2009) and analysis of lipids in frozen hake fillets (Sánchez-Alonso, Carmona & Careche, 2012), underlying the flexibility of this technique in the study of distinct fish tissues within the context of several different biological problems.

As mentioned in the previous paragraph, the target of this study was the gilthead seabream, a species of high commercial value reared in the Mediterranean coast. The reason we are interested in studying the effect of seasonal variations in temperature on gilthead seabream metabolism is the occasional occurrence of a syndrome (called the “winter disease” or the “winter syndrome”), which induces chronic mortalities during the low temperature months, followed by acute mortality when the temperature progressively rises, with significant economic impact for fish farmers. It is characterized by both behavioral (e.g., erratic swimming, voluntary fasting, hyposensitivity to stimuli) and physiological (e.g., impaired growth, pale and fatty liver, tissue necrosis, infections) changes that can ultimately lead to systemic dysfunction and death (Contessi et al., 2006; Doimi, 1996; Gallardo et al., 2003; Ibarz et al., 2010b; Ibarz et al., 2010c; Kyprianou et al., 2010; Tort et al., 1998; Tort et al., 2004). In etiological terms, though many factors seem to be relevant to the onset of this syndrome (namely, the occurrence of metabolic distress, nutritional imbalances and/or deficits, immunosuppression and the presence of opportunistic pathogens (Contessi et al., 2006)), the common underlying factor seems to be exposure to low environmental temperatures, which induce slower metabolic rates and disrupts feeding behaviour (with fish displaying little to no feed consumption when temperatures drop below 13 °C).

Although the exposure of reared gilthead seabream to low temperatures might be unavoidable during cold months, there are indications that the administration of specifically-formulated diets prior, during and/or after low temperature periods might mitigate the ultimate impact of thermal challenge on fish nutritional status and health, helping to prevent the occurrence of full-blown “winter syndrome” (Bavčević et al., 2006; Ibarz et al., 2010a).

Within this context, a nutrient-enriched diet (“Winter Feed” or WF) was designed to serve as reference for a high-quality feed, appropriate to maintain improved nutritional and metabolic status in gilthead seabream during the cold months. Although hardly economically feasible, this diet was formulated to constitute a high-quality reference (against which other possible diets can be compared), containing a higher proportion of marine-derived ingredients (i.e., fish meal and krill protein hydrolysate) and supplemented with phagostimulants (i.e., betaine), marine phospholipids, soy lecithin, antioxidant vitamins and taurine, compared to a challenging diet with low levels of fishmeal and a concomitant partial replacement of fish oil by rapeseed oil (“CTRL” diet), representing a low-cost commercial formulation. The used formulations were thus chosen to induce (as far as possible) two extreme nutritional states in gilthead seabream so that we can confirm, on one hand, the possibility of modulating the challenging effect of seasonal temperature variations through a nutritional approach and, on the other hand, the feasibility of using FT-IR spectroscopy for the purposes of metabolic fingerprinting.

Materials & Methods

Fish rearing and sampling

The experiment was conducted at the Experimental Research Station of CCMAR (37°00′ N, 07°58′ W, Faro, Portugal) and took place between November and June of the following year. Four homogenous groups of 25 gilthead seabream (Sparus aurata) each, with a mean initial body weight of 87 ± 5 g, were stocked in 1000 L outdoor circular plastic tanks supplied with flow-through seawater (rearing density of about 2.18 kg m⁻³). Throughout the trial, fish were subjected to a natural temperature regime, which was logged every hour (Fig. 1), with mean daily temperatures ranging from 7.6 °C to 25.0 °C. Similarly, other physicochemical parameters varied within the natural ranges (natural photoperiod, salinity: 33 ± 2%, dissolved oxygen: above 5 mg L⁻¹). Each dietary treatment was tested in duplicate tanks over 213 days. Fish were fed to apparent satiety, by hand, either once a day (at 10.00 h, during the winter period), or twice a day (at 10.00 and 16.00 h, during the spring period) and feed intake was recorded. Prior to harvesting for sampling, fish were starved for 48 h.

Over its course, two samplings were performed: in March (end of winter sampling), during the lowest temperature regime, and in June (end of spring sampling), during the temperature rise period (Fig. 1). Fish were weighted, measured and liver samples (from 10 fish/tank) taken and immediately frozen in liquid nitrogen for further characterization. At sampling, macroscopic observation of liver allowed identification of abnormal characteristics (discoloration, firmness and exudation). Analysis of whole-body composition was performed on the fish carcasses after liver excision.

The experiment described was conducted in accordance with the Guidelines of the European Union Council (Directive 2010/63/EU) and the Portuguese legislation for the use of laboratory animals, and under a “Group-1” licence (permit number 0420/000/000-n.99-09/11/2009) from the Veterinary Medicine Directorate, the competent Portuguese authority for the protection of animals, Ministry of Agriculture, Rural Development and Fisheries, Portugal.

Results and Discussion

The results of the fish trial appear to support positive effects of the WF diet on fish performance, with fish fed this diet generally displaying higher relative growth rates (RGR), higher thermal-unit growth coefficients (TGC), higher hepatosomatic indices (HSI) and lower feed conversion ratios (FCR), compared to CTRL-fed fish (see Table 3, Figs. 2, 3 and S3). Though CTRL-fed fish displayed significantly higher feed consumption during the spring period (when expressed as a fraction of body weight), this did not result in any appreciable compensatory growth. It is interesting to note that the TGCs estimated for the winter period are consistently lower than those estimated for the spring period, which suggests that observed differences in terms of growth performance between the two periods should be attributed not only to the temperature differences, but also to the voluntary fasting effect due to low temperatures. No mortalities were reported for any of the tanks, which is a sign that the fish coped with the seasonal challenge to some degree, displaying none of the behavioural symptoms of “winter disease”. Nevertheless, some of the fish (particularly in the CTRL tanks) displayed at least one of the phenotypic traits of “winter disease”: pale and friable liver. No significant differences between tanks were observed in terms of whole body composition (results not shown).

Multivariate analysis of the FT-IR dataset suggests that the biggest observed differences in terms of hepatic metabolic fingerprint are between samples from the June vs. March sampling (Fig. 4A). Unsupervised embedding of the samples from the two samplings in separate also shows a clear effect of the WF diet vs. CTRL diet (Figs. 4B and 4C), particularly for the June sampling. Comparatively, no clear tank effect (i.e., when comparing Tank 1 vs. Tank 2 and Tank 3 vs. Tank 4) on the hepatic metabolic fingerprints can be observed.

Looking at the actual spectra and the results of univariate analysis (Fig. 5), it becomes clear that the biggest season and diet effects can be explained in terms of peaks 11–16 (i.e., 930–1155 cm⁻¹), which correspond to IR absorptions attributable to either carbohydrates and/or nucleic acids. If we assume that these IR absorptions correspond to nucleic acids, we would expect to see the same behaviour for peak 10 (highly characteristic of nucleic acids), which does not occur. As such, these observed differences are more likely to represent changes in hepatic carbohydrate content (e.g., glycogen) than changes in nucleic acid content. The results indicate that season had a strong negative impact on the hepatic glycogen reserves of gilthead seabream and that the dietary treatment generally had the opposite effect: not only are carbohydrate stores higher in WF-fed fish compared to CTRL-fed, at the end of winter, but this increase is still visible at the end of spring, despite observed depletion of carbohydrate stores between the two sampling points, for both diets. This type of hepatic glycogen depletion in gilthead seabream due to thermal shifts and fasting has already been documented (Ibarz et al., 2007b).

Another interesting observation regards peaks mostly associated with lipids (both saturated and unsaturated fatty acids, triglycerides, cholesterol esters and phospholipids), such as peaks 1–2, 4, 6–9 and 19, which display an increasing trend over time (i.e., when comparing the March against the June sampling) for the CTRL diet and the inverse trend (decrease over time) for the WF diet, though not significantly so, according to our criterion (FDR > 0.01). Looking at the results in more detail, we can see that this lack of significance is mostly due to high inter-individual variability, with only some of the CTRL fish displaying this trend of increased hepatic lipid content at the end of spring. This observation was confirmed by looking at the 2840–3030 cm⁻¹ spectral range (not shown), where lipid-related IR absorptions related to CH stretching are expected to be present, which also displayed the same general trend. This suggests that some CTRL-fed fish appear to be mobilizing lipid stores (probably from perivisceral adipose tissue) at the end of spring, to face increasing energy demands, while none of the WF-fed fish display this effect (suggesting an improved metabolic status). It should be noted that one of the specific symptoms of the “winter disease” is precisely the progressive hepatic accumulation of lipids resulting in a steatotic-like liver (Ibarz et al., 2010c). Though this could be a sign that WF does indeed provide some protective or mitigating effect of the conditions that lead to the onset of “winter disease” in gilthead seabream, no differences were observed in terms of viscerosomatic index to suggest extensive lipid mobilization by CTRL-fed fish, confirming the observation that most fish (regardless of diet) successfully coped with the seasonal temperature challenge, in this particular trial.

Finally, we also applied sPLS to try to predict the experimental factors and zootechnical parameters (Y-matrix) using the FT-IR dataset (X-matrix). What we observed was that the HSI was the only variable predicted by the FT-IR data with high accuracy. Plotting the correlation of the different FT-IR features against the HSI (Fig. 6A), it becomes clear that there is a strong relation between HSI and the IR absorptions assigned to carbohydrates. This relation can be confirmed by comparing the area under the carbohydrate peaks against the HSI (Fig. 6B), which shows a direct relation between the two (apart from some level of “saturation” for fish with an HSI below 1.25%). This suggests that observed variations in HSI (between diets and between sampling points) can be mostly attributed to changes in hepatic carbohydrate content, again reinforcing the notion that the positive effects observed for WF-fed fish in this trial should be mostly explained in terms of a modulating effect on the hepatic carbohydrate stores.

Interestingly, previous studies on the physiological impact of low temperatures (with fasting) on gilthead seabream (Ibarz et al., 2007a; Ibarz et al., 2005; Ibarz et al., 2007b) suggest that they generally induce an increase in the HSI, which is explained by higher mobilization of lipids (due to higher energy demands caused by thermal shifts). Taken together with the present study, this suggests that increased HSI, by itself, might not be a sufficient predictor of the metabolic state of gilthead seabream, underlining the need to concurrently assess which biomolecules are associated with observed changes in HSI.

Conclusions

Over the course of this trial, we observed seasonal changes in zootechnical parameters, and in terms of hepatic metabolic fingerprint, which suggest fish were challenged, displaying a progressive loss of hepatic carbohydrate stores. Nevertheless, under experimental conditions, fish still displayed somatic growth over this period and none of the fish displayed all of the symptoms of “winter disease”. Still, some of these, particularly in the CTRL group, displayed possible early signs of the metabolic dysfunction associated to the “winter disease”, namely hepatic accumulation of lipids.

Regarding the WF diet, all obtained information points towards a generally beneficial effect in terms of performance and nutritional/metabolic status, with WF fish consistently displaying higher liver weight and HSI, at both sampling points, mostly due to a difference in the hepatic abundance of carbohydrates. Although the performance parameters (for all tanks) were far from optimal, the WF diet seems to have a mitigating effect regarding the seasonal challenge, not only in terms of impaired growth and carbohydrate depletion, but also in terms of the observed hepatic accumulation of lipids in the later sampling. This effect can be attributed to, among other things, the higher content in fish meal and/or higher protein digestibility, for the WF diet. This suggests that, indeed, the strategy of using a nutritional approach to mitigate the effects of seasonal thermal variations on gilthead seabream metabolism seems feasible and that, in this sense, diet WF is an adequate candidate as a “positive control” diet against which to compare alternate (and possibly more cost-effective) formulations.

Also, we feel confident that the present work reinforces the notion that an assessment of covariates (such as HSI, in this case) on the same set of experimental units can be important and increase the interpretability of observations.

Finally, we have confirmed that the use of FT-IR as a metabolic fingerprinting technology can be useful in such a context to obtain untargeted information on the nutritional and metabolic status of gilthead seabream and other fish species.

Supplemental Information