Insights Into Amino Acid Metabolism and Incorporation by Compound- Specic Carbon Isotope Analysis of three-Spined Sticklebacks

Interpretation of isotope data is of upmost importance in ecology to build sound models for the study of animal diets, migration patterns and physiology. However, our understanding of isotope fractionation and incorporation into different consumer tissues is still limited. We therefore measured the δ 13 C values of individual amino acids from muscle and liver tissue of three-spined sticklebacks (Gasterosteus aculeatus) over time in a controlled feeding experiment. The isotope signatures of amino acids in the liver quickly responded to even small shifts in dietary isotope compositions, whereas muscle tissue remained isotopically constant over time. No signicant fractionation between diet and sh tissues was observed for half of the amino acids in both liver and muscle tissue, supporting the idea of direct nutrient routing on a protein rich diet. Small isotopic differences were observed for alanine, glycine and serine between diet and liver, indicating that metabolic processes such as glycolysis or gluconeogenesis can be tracked by the isotope signatures of their main substitutes. We demonstrate that compound-specic isotope analysis has great potential to investigate the main metabolic pathways of organisms and suggest further investigations using isotopically enriched materials to facilitate the correct interpretation of ecological eld data.


Introduction
Stable isotope analysis (SIA) of carbon is a powerful tool in ecological studies to investigate resource utilization, foraging behavior and migration patterns of animals [1][2][3] . The underlying principle is that the carbon isotope composition of diets is mostly retained by consumers, with only little fractionation occurring during incorporation. In bulk stable isotope analysis (BSIA), this fractionation increases the carbon isotope signature of consumers by 0 -1‰ 4,5 and the carbon isotope signature of an individual is therefore mostly depending on the isotope composition of primary producers. This enables tracking of nutrients from the base of food webs to higher trophic level predators. However, there can be deviations from this pattern depending on species, analyzed tissue and diet composition 6,7 . A more recent approach, compound speci c isotope analysis (CSIA), was enabled by analyzing individual compounds, rather than bulk tissues. Where BSIA of carbon typically shows little fractionation between diet and consumer, individual constituents such as amino acids (AA) can have higher fractionation, depending on which compounds they are routed from 3,8−12 .
AAs can be divided into essential amino acids (EAA) and non-essential amino acids (NEAA). EAAs cannot be synthesized by higher organisms and therefore need to be taken up directly from dietary sources, leading to no or very little isotope fractionation because these nutrients are traversing the food chain mostly unchanged from primary producers to top predators. This approach is called isotope ngerprinting and has been used for ecological studies to investigate the carbon ow and nutrient allocation of end-consumers by comparing the isotope signature of selected EAAs to those of primary producers, which should be roughly identical if the food chains are connected 3,13−15 . NEAAs on the other hand can be synthesized de novo in higher organisms and therefore originate either from dietary sources and/or other metabolic compounds during metabolism and conversion of nutrients, such as lipids and carbohydrates. The classi cation of AAs into essential and non-essential, however, is not always easy, as the requirement of some NEAAs can sometimes outmatch the organism's ability to synthesize these compounds, rendering them temporarily or conditionally essential 16 , with Glu and Pro being two examples. Another example would be tyrosine (Tyr), which can be synthesized in higher organisms but is derived from phenylalanine (Phe), which is an EAA. The carbon isotope signature of Tyr therefore typically falls into the same range as Phe 12 , although technically it should be considered non-essential.
Most studies on stable isotopes typically only consider whole body or muscle tissue 17 , as these are the easiest to access. The isotope turnover and incorporation rate can, however, be very low and as a result, turnover in muscle tissue is more affected by consumer physiology and growth phases during which protein synthesis and deposition occurs. The liver responds much more quickly to dietary isotopic changes, as its regulatory activities require continuous protein turnover and have shown to be more responsive to short-term changes in isotope signature [18][19][20][21][22] . In addition, the liver plays an important role in the metabolism of AAs and metabolic processes such as glycolysis and gluconeogenesis for energy production and storage 16,23,24 , which makes the liver an ideal tissue for studying these processes. This is especially true for CSIA, since the key compounds participating in glycolysis and gluconeogenesis are AAs, carbohydrates and lipids 25 . Regarding energy metabolism, AAs can then be further divided into glycogenic and ketogenic categories, depending on the pathways by which they are metabolized.
Glycogenic AAs are converted mainly to glucose via pyruvate, whereas ketogenic AAs are converted to ketone bodies. The differentiation is, however, not so easy, as some AAs can be metabolized via both pathways. The most important AAs for gluconeogenesis in humans are Ala and Glu 26,27 , with Ala serving as a "shuttle" to transport three-carbon skeletons from muscle to liver during proteolysis. Other important AAs participating in gluconeogenesis are Gly, Ser and Asp. Whether these compounds are directly routed from diet or synthesized de novo is further depending on dietary composition, whereas consumers fed on diets with high protein contents tend to directly route also NEAAs in order to preserve energy 11,13,28,29 .
Studying the isotope fractionation of metabolic pathways involved in the synthesis and catabolism of NEAA and other macronutrients helps researchers to interpret the ability of omnivorous and herbivorous animals to maintain energy and nutrient homeostasis on diets with different protein, lipid and carbohydrate compositions.
Since most studies using SIA are eld based, there was an initial call for more laboratory SIA studies in 1997 30 to better understand the fundamental principles of isotopic incorporation, trophic discrimination and isotope routing. Although the number of laboratory-based studies has increased since then, the call was renewed in 2009 31 , to facilitate correct interpretation of eld studies. Few controlled feeding experiments so far examined the carbon isotope fractionation of individual AAs between diet and sh consumers, with varying magnitudes and directions of carbon isotope fractionation reported [10][11][12][13]29 . signi cantly 13 C-enriched Gly carbon isotope signatures of Chinook Salmon (Oncorhynchus tshawytscha) reared on the same diet. This example demonstrates that clear differences in individual AA isotope fractionations occur among sh species.
The three-spined stickleback (Gasterosteus aculeatus) is a well-studied model sh in ecology, evolutionary biology and parasitology 32,33 , yet no carbon isotope analysis of amino acids of sticklebacks has been done to this date. There are some studies investigating carbon and nitrogen BSIA of threespined sticklebacks [34][35][36][37] , but those were also eld-based, investigating differences associated with sex, armor, phenotypes, genetics and host-parasite relationships. We conducted a laboratory-based feeding experiment, where three-spined sticklebacks were reared on a protein rich diet over the course of four months. To contribute to the advancement of our understanding of isotope incorporation, discrimination and routing, we measured the carbon isotope signature of individual AAs from muscle, liver and dietary samples by Liquid Chromatography Isotope Ratio Mass Spectrometry (LC-IRMS). Previous BSIA of dietary samples (data not published) revealed a minor carbon isotope shift over the course of the experiment, which is likely to be transferred to sh tissue to varying degrees, depending on tissue type. This is the rst time that carbon isotope signatures of AAs were measured for three-spined sticklebacks on a time series in a controlled feeding experiment. Data was analyzed to speci cally test and answer the following three main hypotheses: 1) NEAAs in stickleback tissue will be preferentially routed to diet in the same manner as EAAs, since sh are fed on high protein contents 2) Liver AA carbon isotope signatures change over the course of experiment and follow the isotope shift of dietary samples, whereas AA isotope signatures in muscle tissue remain constant or shift only slightly over time.
3) Glycogenic AAs in the liver will show different carbon isotope patterns compared to muscle and diets based on their metabolic role, but we can't predict in what way.

Isotope shift over time
One-way ANOVA for each AA with time points (days) as independent variable and δ 13 C values as dependent variable of all dietary samples revealed a signi cant isotope shift between time points for all AAs except His (DF = 3, 8, p < 0.01, Table S2, Supporting Information). Over time, average isotope signatures of AAs in dietary samples changed by -1.6 to -1.0‰ between 30 and 60 days, followed by another change of -2.6 to -0.9‰ between 60 and 90 days and a nal change of 0 to +1.5‰ between 90 and 120 days. A general trend can be seen where the isotope signature of each dietary AA decreased between 30 and 60 days as well as 60 and 90 days, followed by an increase in isotope composition between 90 and 120 days. The biggest differences between isotope signatures (~ -2 to -4‰) were therefore observed between 30 and 90 days, marking the time points where diets had the highest and lowest δ 13 C values ( Figure 1 and Figure 2). Isotope signatures of AAs in the liver were signi cantly different over time for Ala (F 3,8 = 6.8, p = 0.004), Asp (F 3,8 = 9.2, p = 0.001), Arg (F 3,8 = 13.7, p < 0.001), Lys (F 3,8 = 18.4, p < 0.001), Phe (F 3,8 = 8.9, p = 0.001) and Tyr (F 3,8 = 20.8, p < 0.001). Gly, Ser, Glu, Pro and Thr were not signi cantly different between time points in the liver, although Glu was just slightly above the signi cance level (F 3,8 = 5.1, p = 0.011). Although not all AAs in the liver revealed a signi cant isotope shift, the general trend of decreasing δ 13 C values from 30 to 60 days and from 60 to 90 days followed by an increase from 90 to 120 days was still visible, comparable to the shift in dietary samples. In contrast to liver samples, no AA in muscle tissue showed any signi cant differences between time points.

Isotope fractionation between tissues
Muscle and liver tissue of sticklebacks were separated by ASCA on the rst principal component of factor 1 (sample type, Figure 3), which contributed 22.4% to the total variance. NEAAs were the main driver of separation on the rst principal component (81.8% variance) with loadings above 0.3 for Ala, Asp, Glu and Ser, whereas Gly was the only NEAA with a negative loading of under -0.2. Ser had the overall highest loading of 0.7 on PC1. All EAAs and Pro had loadings of under ± 0.2 and therefore little to no effect on PC1 separation. Individual scorings of liver samples were negative on PC1, whereas muscle tissue had almost exclusively positive scorings. Scorings of dietary samples gathered around ~ 0. PC2 showed no clear separation between tissue types and was mostly loaded with EAAs, except for Arg and Lys. Tyr, Phe and Pro had the highest loadings of over 0.2, whereas Thr showed a noticeable negative loading of under -0.3. Scores of muscle and liver samples on PC2 showed no clear trend with a slight tendency to be more negative. Dietary samples showed positive scorings on PC2, which led to an almost distinct cluster of dietary samples with a slight overlap in the elliptical con dence ring with liver samples. Arg and Lys were the only amino acids which had almost no impact on separation between tissues for both PCs. Factor 2 (days) of the ANOVA simultaneous component analysis had an effect of 35.5% on system variance but will not be discussed because at this stage it is of no further interest for data interpretation.
Differences of individual AAs between sample types were tested by one-way ANOVA for each AA and sampling days, with tissue type as independent variable and δ 13 C in ‰ as dependent variable, followed up by Tukey post-hoc tests and a signi cance level α of 0.01 (Table S3 and S4, Supporting Information).
All NEAAs except Pro showed signi cant differences between tissue types at least once throughout the days (DF = 2, 10, p < 0.01,), although differences of Asp and Glu were only seen between liver and muscle tissue and not compared to dietary tissue. Post hoc comparisons of mean values by Tukey tests showed that diet and liver tissue were isotopically different for Ala after 60 days (p = 0.007), Gly at 90 days (p = 0.008) and Ser at 30 days (p = 0.01), whereas diet and muscle tissue only showed signi cant differences for Ala at 90 days (p = 0.001). Isotopic differences between muscle and liver tissue were more abundant, especially for Ala, Asp and Ser (Table S3, Supporting Information). The liver showed a general trend of being isotopically 13 C-depleted compared to diet and muscle tissue, except for Gly, where the liver had 13 C-enriched isotope signatures at 90 and 120 days. Differences between tissue types for EAAs were only signi cant for Phe and Tyr (DF = 2, 8; p < 0.01). These differences occurred in both cases mainly between sh tissue (either muscle or liver) compared to diets and not between liver and muscle themselves.
Muscle and liver tissue were generally 13 C-depleted compared to diets by around -1‰. Arg, Lys and Thr showed no signi cant isotopic differences between tissue types.

Isotope fractionation of Histidine
The most signi cant isotope fractionation was observed for His between diet and both liver and muscle. His in liver was isotopically enriched with mean differences of +8.2‰ compared to dietary samples (DF = 2,8; p < 0.001). Fractionation of His between sh muscle and dietary samples was lower compared to liver tissue, but still signi cant at each day with p-values <0.001 and mean differences of +5.0‰.
Differences of His isotope signatures between muscle and liver tissue were also signi cant (p < 0.001) except at 90 days and mean differences of -3.2‰. His isotope signatures in tissues were not signi cantly different over time, although dietary samples had a p-value of 0.016 and were therefore close to 0.01, with the same isotopic trend of decreasing signatures at 60 and 90 days, followed by an enrichment of isotope signatures at 120 days.

Discussion
Isotope signature over time The initial amino acid isotope signature of mosquito larvae used for feeding shows typical patterns also seen in other studies, where Gly and Ser are the most 13 C-enriched amino acids and EAAs are isotopically 13  Liver tissue, on the other hand, showed signi cant isotopic differences over time, indicating that the high nutrient turnover rate of the liver is re ected in its carbon isotope signature of AAs and even short-term changes in dietary isotope compositions are visible. EAAs showed the highest differences (one-way ANOVA, p ≤ 0.001) and might be indicative of high turnover for these compounds because they must be rooted from dietary sources. Only Ala and Asp showed signi cant differences over time from NEAAs in liver. Both AAs are important energy substrates in sh and precursors for gluconeogenesis 40 , which is the metabolic pathway in the liver to synthesize glucose from other carbon substrates. Fish therefore have a high Ala and Asp demand in liver tissue, which might outmatch the sh's ability to su ciently synthesize these compounds and lead to an increased incorporation or turnover from dietary sources. According to our results, liver carbon isotope signatures of Ala, Asp, Arg and Lys might be useful proxies to track consumer diets of small teleost sh on protein rich diets and on time scales of at least 30-day intervals.
This information could possibly be expanded to non-protein rich diets for Arg and Lys, due to the necessity of eukaryotes to directly incorporate these nutrients from dietary source, regardless of their composition 13 . Although we did not quantify the exact cellular turnover rate of individual AAs in stickleback tissues, knowing an estimated timeline for evaluating the diet of an animal is useful for ecological studies investigating e.g. migration or seasonal changes of food sources. It has been shown that choosing different tissue types in individuals enables researchers to explore temporal and spatial resource use of animals and it is critical to consider incorporation rates of isotope signatures into these tissues 41,42 . However, most studies so far on isotopic incorporation and turnover rates are still based on BSIA [43][44][45] . Enhancing our knowledge to speci c compounds in different tissues can provide powerful opportunities, e.g. to study the physiology of migrating animals, which are hard to capture multiple times. The analysis of AA isotope signatures of these animals from tissues with different incorporation rates can help to illuminate temporal variations and identify endogenous vs exogenous resources, without the need of multiple sampling points.

Fractionation between tissues
We were able to separate muscle and liver tissue by their carbon isotopic signature of AAs using ASCA ( Figure 3). ASCA provided an alternative way to PCA and LDA as classical multivariate analysis and was especially useful in our case because we could incorporate the structure of our experiment into the model with time points and sample types as two separate factors, which is not possible in PCA 46 . This removed a lot of "noise" from the dietary isotope shift when investigating isotope fractionation between sample types and to identify AAs with high or low fractionation patterns. Highest loadings were observed for Gly and Ser with opposite trends, where Gly has the most negative loading and separates liver tissue from dietary and muscle samples. Ser and Ala have the most positive loading and separate muscle from both diet and liver. It also revealed low fractionation of Arg and Lys isotope signatures between all tissue types, as those two AAs have the lowest loading overall on both PCs. Based on the results from ASCA we decided to especially focus on these AAs during ANOVA between tissue types, since the fractionation should be independent from the observed isotope shift over time induced by dietary samples.
NEAAs can be synthesized de novo in higher organisms, leading to more fractionation during metabolism and nutrient ow from primary producers to top predators in a given food web 29 . Our study however shows low isotope fractionations in general for NEAAs between diets and sh tissue and supports the idea of isotope routing for NEAAs when animals are fed protein rich diets, which is energetically favorable compared to de novo synthesis. High isotope fractionations of above 5‰ of some NEAAs between diet and consumer, as described by other studies 11,13 , did not occur in our case, but smaller differences were observed. These differences mainly occurred between muscle and liver tissue and not directly between diet and muscle/liver. It is worth mentioning here that the isotopic differences between liver and muscle tissue did not directly result from metabolic differences leading to different fractionation patterns but were a consequence of liver tissue following the isotopic shift of dietary sources over time, whereas muscle tissue remains almost constant. The only cases where NEAAs in sh tissue were signi cantly different compared to dietary samples were for Ala, Gly and Ser. The fractionation of Ser was striking, with Ser in the liver being isotopically depleted not only compared to muscle but also compared to diets after 30 days. Catabolic pathways for Ser involve deamination to pyruvate, transamination with pyruvate to form hydroxy pyruvate and alanine and formation of Gly with tetrahydrofolate 47 . Additionally, participation of Gly into gluconeogenesis requires the conversion of Gly to Ser by serine hydroxymethyltransferase and both AAs participate further in sulfur and one-carbon metabolism 16 .
These two AAs are also the only NEAAs in the liver showing abnormal isotopic behavior, with Ser being signi cantly depleted compared to muscle and diets, whereas Gly is the only AA in liver being enriched at later sampling days. The liver plays a major role in energy and amino acid metabolism by regulating and controlling both glycolysis and gluconeogenesis as well as synthesizing many of the NEAAs 24 . Synthesis of NEAA carbon backbones can occur from either glycolytic intermediates for Gly, Ser and Ala or from Krebs cycle intermediates for Asp and Glu 48 . It is, however, hard to estimate the contribution of isotopic fractionation for each of these pathways, not only because intrinsic fractionation and mass ow are unknown, but also because metabolic pathways are often intertwined and hard to separate in living organisms at natural abundance level. Enzymatic reactions and catabolism of nutrients are usually associated with discrimination against the heavier isotopes, in this case 13 C, leading to depleted signatures in the product, while the educt of the reaction becomes isotopically 13 C-enriched 49 . The small enrichment of Gly observed in liver could therefore be indicative of extensive conversion of Gly in this tissue to other compounds, with Ser being one candidate and resulting in a depletion of its isotope signature. This system could even be extended to include Thr as an EAA. Animals are able to catabolize substantial amounts of Thr to yield Gly by L-threonine 3-dehydrogenase 50 . Applying the same principle of isotope fractionation during enzymatic reactions, Thr could theoretically be enriched in consumer tissue compared to diets and although Thr in sh liver was not signi cantly different compared to sh diet, liver tissue showed tendencies to be isotopically enriched in its carbon isotope signature by 0.4 to 1.3‰. It is, however, a far stretch at this point to account the small difference in Thr and Gly to speci c metabolic or enzymatic reactions, especially since metabolism of Gly and Ser involve several different pathways. Future research and controlled feeding experiments could focus on using isotopically labeled substitutes to track individual metabolic pathways, such as the catabolism of Thr to Gly and Ser. Thr could be especially useful in that regard, since it can`t be synthesized in higher organisms and has to be taken up by diets. Pro was the only NEAA showing no difference at all over the sampling period, which points more to a behavior like an EAA being rooted by diets. As mentioned in the introduction, Pro can be synthesized in higher organisms, but the classi cation as a NEAA is sometimes misleading as metabolic requirements might heavily outmatch de novo synthesis. Our results indicate that sh fed on high protein content incorporate Pro mostly from dietary sources leading to no or very little fractionation in both liver and muscle tissue.
Phe and Tyr were the only EAAs showing different isotope signatures between diet and sh tissue, which is surprising for Phe since it also needs to be directly taken up and rooted from dietary sources. Tyr however is synthesized from Phe and the depleted isotopic signature of Tyr compared to diets could therefore also be explained by enzymatic fractionation discriminating against the heavy carbon isotopes. Arg and Lys were the EAAs with lowest fractionation overall, strongly following the trend of dietary isotope signatures even in muscle tissue over short time periods. Arg is highly abundant in sh protein and tissue uid 16 , serving as a precursor for the synthesis of proteins, nitric oxide, urea, Pro, Glu and creatine 51 . Arg is of special interest here because it is not often reported in the literature, since most studies apply GC-IRMS for measurement of carbon isotope signatures of AAs, which requires derivatization and results in the loss of Arg 52,53 . Because of the high abundance, demand and turnover of dietary Arg in sh leading to very little fractionation even in muscle tissue over a short-term dietary isotope shift, it could serve as another proxy in addition to Phe, Leu, Ile and Lys 10-13 to track carbon sources from primary producers in muscle or liver tissue. Identifying EAAs with very low fractionation patterns is pivotal for ecological studies using isotope ngerprinting to better constrain the source of endmember signatures and study the carbon ow in terrestrial and oceanic environments 3,8,9,[13][14][15]29 . Although EAAs cannot be synthesized by eukaryotes, the fundamental assumption that these constituents must be solely sourced from dietary protein resulting in low isotope fractionation of ~0‰ has not always been met. There are studies suggesting that gut microbes can contribute to the homeostasis of EAAs in animals by de novo synthesis, which of course complicates their use as stable and robust proxies for isotope ngerprinting 29,54−57 . The contribution of microorganisms in the gut of sticklebacks to the EAAs homeostasis could explain the observed difference of Phe and Tyr isotope signatures in sh muscle and liver tissue compared to diets and highlights the importance to experimentally explore the variation of EAA isotope fractionation by gut microbiome. According to our results, Arg and Lys might therefore be better suited in sticklebacks and possibly other teleost species to estimate their carbon ow and allocate resource consumption, if the contribution of gut symbionts is of no interest.

Isotope fractionation of Histidine
His showed overall the highest fractionation among the analyzed AA. This represents an unusual outcome, as EAAs are taken up directly from diet and incorporated into consumer tissue with very low trophic discrimination. The previously mentioned de novo synthesis and contribution of EAAs from gut microbes could of course be one explanation for the highly 13 C-enriched isotope signatures of His in in stickleback liver and muscle tissue. Other reasons could include the enzyme histidine decarboxylase, which is produced by bacteria and causes histamine sh poisoning 58, 59 . It has been speculated before that histidine decarboxylase might still be released from autolyzing of bacterial cells during and after freeze drying and convert histidine to histamine 60 . Enzymatic reactions are known to cause isotopic fractionation, which usually discriminates against the heavier isotope 49 and could result in enriched isotope signature of the remaining histidine in sh tissue. Interestingly, an early study investigating histamine concentrations during storage of esh and liver tissue of mackerel under different conditions showed a higher increase of histamine concentrations in the liver 61 , which would t to the higher isotope values of histidine in stickleback liver compared to muscle tissue, if it was caused by enzymatic reactions. Testing this, however, would require histamine isotope measurements in dietary and sh samples to compare isotope signatures and was out of scope of this study. Another explanation would be that sticklebacks fed on mosquito larvae lack dietary histidine to match their metabolic requirements. We did not directly measure the protein and AA content of dietary mosquito larvae, but His has been mentioned in the literature as a more frequent limiting AA when animals are fed with insect meals 62 . His is an important amino acid for growth, tissue formation and hemoglobin synthesis in sh 63 and a lack of dietary His may lead to increased fractionation in consumer tissue due to catabolism during starvation. There is however no consensus so far on the isotopic effect of starvation or nutritional stress on consumer tissue 64 especially for single compounds, and such effects could be subject for further studies.

Conclusion
Studying the isotope signature of AAs between diet and consumer gained much attraction during the last years, yet our knowledge of the fundamental principles behind isotope incorporation and fractionation of individual substitutes is still limited. Our study shows that direct isotope routing even of NEAAs might still be the preferred way for nutrient assimilation when sh are fed with high protein diets, and it seems that there are not only differences in isotopic turnover rates between muscle and liver, but also that individual AAs in these tissues show different isotopic behaviors based on their metabolic role. However, investigating isotope fractionation of AAs on natural abundance levels might not be the best approach to study fundamental incorporation and turnover patterns, since a lot of information might be lost to natural variations and measurement uncertainty. This calls for controlled feeding experiments with heavily 13 Cenriched diets or single constituents to accurately track metabolic pathways. One good example to investigate could be the mentioned Thr, Gly and Ser catabolism in animals, and has the potential to greatly improve our understanding of nutrient assimilation and conversion, which is needed and long overdue for the correct analysis and interpretation of eld data. On the other hand, it is promising to see that even small differences and uctuations of isotope values, which are more realistically encountered in nature, can be investigated using CSIA of different sample materials.

Feeding experiment
Three-spined Sticklebacks were laboratory-raised offspring and reared in twelve 14 L tanks (VewaTech, Germany) as part of a parasitic infection experiment. Water was recirculated and held at 18°C with a 15 h light and 9 h dark cycle. Stickleback offspring were produced by in vitro fertilization from individuals collected from a brook in North-West Germany (52°17'33.11"N, 7°36'46.48"E), about eight months old at the beginning of the experiment and fed daily with washed red mosquito larvae (Chironomidae) over the course of four months. Each lab tank contained twelve individuals, which were divided into two groups.
The rst group consisted of 3 individuals per tank, which were exposed to an uninfected copepod as a sham-exposed control group, while the other 9 individuals per tank were exposed to an infected copepod as a parasite-exposed group. Three complete lab tanks with 36 individuals were sampled on each sampling date (30, 60, 90 and 120 days post exposure) and ve out of the fteen sham-exposed individuals per sampling date were randomly selected for CSIA. Sub samples of mosquito larvae were taken once a week, stored at -20°C and pooled between 1-30, 31-60, 61-90 and 91-120 days for CSIA. Insect larvae are considered to be a protein rich diet for sh, with a crude protein content of up to 60% and an adequate amino acid pro le, although His, Lys or Try can be lacking depending on the used insect  for each of the four sampling days taken at 30, 60, 90 and 120 days were hydrolyzed, and the small sample amount of obtained stickleback tissue did not allow for within-individual replicate analysis. Each hydrolyzed sample was injected in triplicate into the LC-IRMS system and outliers were determined by Grubbs Test on a con dence level of 0.95 and excluded from further analysis. We calculated mean values and SD from triplicate injections to estimate instrumental precision before referencing our data. with its corresponding standard deviation (SD). Data was tested for normality with Kolmonogorov -Smirnov-Test because of the small sample size on a con dence level of 0.95, which was not violated for any given AA and sample. One-way analysis of variance (ANOVA) was used to test for differences of AAs over time points (days) and between tissue type. Using time points as an independent variable in ANOVA can be problematic because it is not strictly categorial, but each of our sample represents an independent sh individual which couldn't be sampled multiple times and is therefore not a repeated measure. The signi cance level α was set to 0.01 to compensate for the low number of biological replicates (n = 5 for sh tissue) that were analyzed, and each ANOVA analysis was accompanied by Brown-Forsythe tests (α = 0.05) to check for equality of group variances and followed up with Tukey post hoc tests to identify signi cant differences between group values. Reducing the signi cance level avoids false-positive results for small sample sizes, but it consequently increases false-negative results, and we are therefore only discussing the most signi cant differences of our data, while smaller differences might be lost. We further conducted ANOVA simultaneous component analysis (ASCA) as a multivariate analysis in a design of experiment approach, with days and tissue as xed factors and δ 13 C of each AA as multivariable. ASCA combines the principles of ANOVA and PCA 46 and allows to investigate differences in the isotope signature between tissues without the in uence of the expected isotope shift over time dominating our analysis. Isotope data were used without any transformation.

Data availability
The datasets generated during and/or analysed during the current study are available in the Figshare repository: https://doi.org/10.6084/m9. gshare.17014220.v1