Trophic discrimination factor of nitrogen isotopes within amino acids in the dobsonfly Protohermes grandis (Megaloptera: Corydalidae) larvae in a controlled feeding experiment

Abstract The trophic discrimination factor (TDF) of nitrogen isotopes (15N/14N) within amino acids, between a stream‐dwelling dobsonfly larva (Protohermes grandis: Megaloptera; Corydalidae) and its diet (chironomid larvae), was determined in controlled feeding experiments. Last‐instar larvae of P. grandis were collected from the Yozawa‐gawa River, central Japan, and reared in the laboratory. After fed to satiation for 1 month, one group of larvae was each fed one living chironomid larva per day for 4 weeks, while a second group was starved for 8 weeks. The larvae were harvested at intervals and the nitrogen isotopic composition of glutamic acid (δ 15NGlu) and phenylalanine (δ 15NPhe) were determined to calculate TDF. The mean TDF of satiated and starved larvae were 7.1‰ ± 0.5‰ (n = 3) and 7.3‰ ± 0.5‰ (n = 5), respectively. Thus, the TDF for P. grandis larvae in this study was similar to that reported for other arthropods (approximately 7‰) and was independent of satiation or starvation. A previous study of wild P. grandis larvae, based on the δ 15NGlu and δ 15NPhe values, estimated its trophic position (TP) as approximately 2.0 ± 0.1 (n = 5), a low value close to that of algivores, although they are generally characterized as carnivores (usually accepted as TP ≥ 3). The TDF for P. grandis larvae suggests that their low TPs in nature were caused by incorporation of vascular plant‐derived amino acids (with a different δ 15N profile from that of algae) and not by an unusually low TDF or by the effects of the satiation/starvation on amino acid metabolism.

one group of larvae was each fed one living chironomid larva per day for 4 weeks, while a second group was starved for 8 weeks. The larvae were harvested at intervals and the nitrogen isotopic composition of glutamic acid (δ 15 N Glu ) and phenylalanine (δ 15 N Phe ) were determined to calculate TDF. The mean TDF of satiated and starved larvae were 7.1‰ ± 0.5‰ (n = 3) and 7.3‰ ± 0.5‰ (n = 5), respectively. Thus, the TDF for P. grandis larvae in this study was similar to that reported for other arthropods (approximately 7‰) and was independent of satiation or starvation. A previous study of wild P. grandis larvae, based on the δ 15 N Glu and δ 15 N Phe values, estimated its trophic position (TP) as approximately 2.0 ± 0.1 (n = 5), a low value close to that of algivores, although they are generally characterized as carnivores (usually accepted as TP ≥ 3).
The TDF for P. grandis larvae suggests that their low TPs in nature were caused by incorporation of vascular plant-derived amino acids (with a different δ 15 N profile from that of algae) and not by an unusually low TDF or by the effects of the satiation/starvation on amino acid metabolism.

K E Y W O R D S
compound-specific isotope analysis of amino acids, deamination, starvation, top predator, trophic position

| INTRODUCTION
The trophic position (TP) of an animal is an important dimension of its ecological niche in a given food web (Chase & Leibold, 2003).
Knowledge of the TP of top-predator animals in a food web is useful to assess the food chain length and is a key factor for understanding ecosystem function (e.g., Vander Zanden & Fetzer, 2007). To accurately and precisely estimate TPs of animals in various ecosystems, the compound-specific isotope analysis of amino acids (CSIA-AA) has recently been used (e.g., Bowes & Thorp, 2015;Chikaraishi, Kashiyama, Ogawa, Kitazato, & Ohkouchi, 2007;Chikaraishi et al., 2009;McClelland & Montoya, 2002;Steffan et al., 2013). In amino acid metabolism, glutamic acid is deaminated, leading to enrichment of 15 N at each trophic level (TL) (e.g., Δδ 15 N consumer−diet = 8. 0‰, McClelland & Montoya, 2002;Chikaraishi et al., 2009). In contrast, phenylalanine retains its amino group because of its low deamination flux during the metabolism, resulting in a negligible enrichment in 15 N per TL (e.g., McClelland & Montoya, 2002;Chikaraishi et al., 2009). Based on this principle, the δ 15 N values of glutamic acid and phenylalanine within a single animal can provide information concerning its TP, using the following equation with an analytical precision (1 σ) of 0.12 units (Chikaraishi et al., 2009): where δ 15 N Glu and δ 15 N Phe are the δ 15 N values of glutamic acid (as a representative "trophic" amino acid) and phenylalanine (as a representative "source" amino acid), respectively. The constant, β, represents the difference between δ 15 N Phe and δ 15 N Glu in a primary producer at the base of the food chain (i.e., −3.4‰ for algae and cyanobacteria; +8.4‰ for vascular plants; Chikaraishi et al., 2009;Chikaraishi, Ogawa, Doi, & Ohkouchi, 2011). TDF is the trophic discrimination factor (i.e., Δδ 15 N Glu − Δδ 15 N Phe ) at each shift of TL, which has been shown to be universal across various taxonomic groups (Steffan et al., 2013. Compared with traditional approaches (e.g., δ 15 N analysis of bulk animal tissues), the CSIA-AA method has significant advantages. In particular, the TP can be estimated from the combined δ 15 N Glu and δ 15 N Phe values within a single organism of interest, and the δ 15 N values of primary producers at the base of food webs are approximately indicated by the δ 15 N Phe values of organisms at any TP in the food webs (e.g., Chikaraishi et al. 2014;Choy, Popp, Hannides, & Drazen, 2015;McCarthy, Benner, Lee, & Fogel, 2007;Popp et al., 2007;Steffan et al., 2015).
CSIA-AA has been applied to the stream food web analysis in a previous study . In stream ecosystems, dobsonfly larvae (Megaloptera: Corydalidae) are one of the carnivorous insects that are widely distributed throughout the world. Larvae of Protohermes grandis (Thunberg, 1781) are considered to be nearly the top predators in stream macroinvertebrate food webs in Japan, based on their large mandibles and foregut contents (Hayashi, 1988a;Takemon, 2005). However, CSIA-AA estimated the TP of P. grandis larvae as 2.0 ± 0.1 (n = 5, based on β = −3.4‰), considerably lower than the value of TP ≥ 3 expected from its biological and functional position as a carnivorous animal . This could be explained by (1) species-specific unusually low TDF, (2) reduction in the TDF associated with prolonged starvation inherent in their lifestyle, or (3) increased β value caused by incorporation of vascular plant-derived amino acids into the basal resources of the stream food web.
With respect to the first hypothesis, if the TDF of P. grandis larva is lower than that reported in the literature for other organisms (e.g., 7.6‰, Chikaraishi et al., 2009), the TP of P. grandis larvae would be lower than expected. In relation to the second, the resource availability in stream ecosystems is typically low compared with coastal marine and terrestrial ecosystems and the foreguts of wild P. grandis larvae are often empty (Hayashi, 1988a). This implies that P. grandis larvae may, at times, be at various stages of starvation. Thus, their metabolism might be specifically adapted to such starvation. Third, incorporation of amino acids derived from vascular plants into the diet animals of P. grandis could increase the β value in Equation 1 from −3.4‰ (used for algal-based food webs) toward +8.4‰ (used for vascular plantbased food webs), resulting in an apparent underestimation of TP for P. grandis. In this case, the mixing proportion of amino acids derived from aquatic autotrophs and vascular plants to the basal diet resources of P. grandis must be calculated to find the TP accurately. To examine the above hypotheses (1) and (2), in this study we reared wild-caught P. grandis larvae in controlled feeding experiments in the laboratory, under satiated and starved conditions, and determined the δ 15 N Glu and δ 15 N Phe values to calculate TDF.

| Sample collection and laboratory experiments
We collected last-instar larvae of P. grandis from the Yozawa-gawa River (35°44′N, 139°11′E), a tributary of the Tama-gawa River, central Japan, on 01 October 2014. We used last-instar larvae in this study because the larval period is 2-3 years in the field and it is impossible to rear the larva from the first instar to the last instar in the laboratory (Hayashi, 1988b). The last-instar larvae stay in a stream for about 8 months (from autumn to the following spring) and increase greatly in body mass before pupating. Therefore, we chose this stage of larvae that just molted to the last instar in early autumn. Great care was taken to prevent injury caused by the larvae biting each other.
Larvae were placed into individual plastic vessels for transfer to the laboratory. As a wild reference, one individual was immediately frozen until the isotope measurement. Other larvae were placed individually in glass vessels (62 mm diameter × 82 mm) with several stones as a refuge ( Figure 1). Well-aerated tap water was supplied, not exceeding 5 mm in depth, and replaced daily. The rearing vessels were kept in an incubator at constant temperature of 20 ± 1°C, and 12:12-h L:D light cycle.
Larvae of P. grandis were fed a single living chironomid larva, Propsilocerus akamusi (Tokunaga, 1938), per day for a month, to establish uniform δ 15 N values among individuals. Chironomid larvae are one of the main prey items for wild P. grandis larvae (Hayashi, 1988a).
Two clumps of the last-instar larvae of P. akamushi were purchased from a fishing-tackle shop and were kept alive at a low basal metabolic rate at 4°C in a refrigerator during the experimental period.
This chironomid has a univoltine life cycle, and the last-instar larvae enter a long summer diapause from May to September (Yamagishi & Fukuhara, 1971). Although the δ 15 N value of the chironomid larvae was predicted to change during the experimental period, this variation should not theoretically propagate to their TP estimates as long as all the cohorts had been fed the same diet. This is because the δ 15 N value of trophic amino acid (i.e., glutamic acid) is subtracted by that of source amino acid (i.e., phenylalanine) in Equation 1 (Chikaraishi et al., 2009), with the assumption that isotopic fluctuations occur in parallel among amino acids. To assess the variability of both the δ 15 N values and TP estimates in the diet of P. grandis during the experimental period, three chironomid larvae were randomly chosen for the isotope measurement.
After the 1-month period of satiation, one individual of P. grandis larva was chosen as an initial representative and was frozen. The remaining larvae were divided into two groups: "Satiated" larvae were continuously reared with one living chironomid larva per day (n = 3 including the initial individual), and "starved" larvae were left unfed (n = 5). One larva was harvested and frozen from the satiated group at 2 and 4 weeks after the experiment started and from the starved group after 1, 2, 4, 6, and 8 weeks. One pupa of P. grandis that had been reared with one living chironomid larva per day for 14 weeks of its larval period was also used for the isotope measurement. Isotopic variability among P. grandis individuals should reflect that among the chironomid larvae, and the isotopic composition of amino acids in the three representative chironomid larvae was determined. Therefore, we analyzed one individual of the P. grandis larva at each time point. USA, and SI Science Co., Sugito-machi, Japan) were analyzed every five or six samples to confirm the reproducibility of the isotope measurements. Analytical errors (1 σ) of the standards were better than 0.5‰, with a minimum sample quantity of 0.5 nmol N.

| Isotope measurement
TPs of the P. grandis and chironomid larvae were estimated using Equation 1. β value was set at −3.4‰ (used for algal-based food webs).
The TDF was calculated using the following equation: where δ 15 N GluPro and δ 15 N PhePro are δ 15 N Glu and δ 15 N Phe values, respectively, of P. grandis larvae, and δ 15 N GluChi and δ 15 N PheChi are mean δ 15 N Glu and δ 15 N Phe values, respectively, of chironomid larvae (n = 3).
Two-way ANOVA was used to test the effects of treatment (i.e., satiated vs. starved groups) and time (i.e., experimental period) on the TP of P. grandis larva.

| RESULTS
The estimated TP of the wild P. grandis larva (the reference individual) from the Yozawa-gawa River, based on the δ 15 N values of amino acids, was 2.2 (Table 1), although the natural variability in this site was unknown. The estimate was consistent with the TP observed in wild P. grandis larvae from different locations, reported in a previous study (2.0 ± 0.1, n = 5, Ishikawa et al., 2014). However, after 1-month satiation (pre-experiment), the estimated TP of P. grandis larvae was elevated to 2.7, a value approximately one-unit higher than that of chironomid larvae (TP = 1.8 ± 0.1, n = 3), the only diet available for P. grandis in the experiment (Table 1). Although isotope variability among individuals was not available due to the lack of replications, the TP of the P. grandis larvae satiated for 1 month was quite consistent with that of the pupa (TP = 2.8, satiated for 14 weeks) (Table 1).
These values are similar to those reported for other insects in feeding experiments on terrestrial food chains (7.6‰ ± 0.1‰, Steffan et al., 2013), for white shrimps fed mixed pellets of microalgae and squid (6.5‰, Downs, Popp, & Holl, 2014 and references therein), and for an aquatic vertebrate (tadpoles) fed dried (dead) bloodworms (7.5‰ ± 0.4‰, Chikaraishi, Steffan, Takano, & Ohkouchi, 2015) ( Table 2). The TDF of starved P. grandis larvae was surprisingly similar to that of satiated larvae. This clearly indicates that P. grandis did not exhibit an unusually low TDF during the feeding experiments and that the magnitude of the TDF was independent of the state of satiation or starvation for at least 8 weeks.

| DISCUSSION
Ecologically unrealistically low TP values estimated for wild P. grandis larvae in a previous study  could be explained by one of the following three hypotheses: (1) species-specific unusually low TDF; (2) an effect of starvation on TDF; or (3) high β value caused by incorporation of vascular plant-derived amino acids into the stream food web. However, the hypotheses (1) and (2) are rejected by the feeding experiments in this study, although the quality and quantity of the diet differed from that in the natural settings. In addition to the chironimid larvae, various kinds of macroinvertebrates have been found in the foreguts of the wild P. grandis larvae, but in many cases, their foreguts are empty (Hayashi, 1988a). The third hypothesis is a possible explanation, supported by both the observed TDF in this study and biological observations in the literature. Larvae of P. grandis are generalist predators that feed on a range of aquatic animals, including algivorous and detritivorous insects, and other T A B L E 1 Values of δ 15 N (‰) of amino acids for the larvae and pupa of Protohermes grandis and their experimental diet (living chironomid larvae) The wild group was larva collected from the field. Satiated and starved groups refer to larvae that were fed chironomid larvae (1 day −1 ) and unfed, respectively. After the 1-month preconditioning (one chironomid per day), the values of δ 15 N (‰) were analyzed for the larvae at 0, 2, and 4 weeks in the satiated group and at 1, 2, 4, 6, and 8 weeks in the starved group. Ala, alanine; Gly, glycine; Val, valine; Leu, leucine; Ile, isoleucine; Glu, glutamic acid; Phe, phenylalanine; TP, trophic position (see Equation 1). ND, not determined.
T A B L E 2 Trophic discrimination factor (TDF) between satiated (n = 3) and starved (n = 5) Protohermes grandis larvae and their chironomid diet, calculated using Equation 2 (see text), and comparison with three other systems observed in feeding experiments: a terrestrial insect food chain (apple leaves, apple aphids, hover flies, parasitoids, and hyperparasitoids); white shrimps (Litopenaeus vannamei) fed a mixture of microalgae and squid; and tadpoles of Bufo japonica fed dried (dead) bloodworms predators (Hayashi, 1988a). Detritivores may feed on the remains of vascular plants (e.g., leaf litter) in stream environments, transferring amino acids from vascular plants to P. grandis larva. If a β value of +8.4‰ (used for vascular plant-based food webs) was applied in Equation 1, the TP of the wild P. grandis larva would be estimated as 3.8. Therefore, in case that the "real" TP of the wild P. grandis larva is ≥3, which is between 2.2 (i.e., β = −3.4‰) and 3.8 (i.e., β = +8.4‰), a hypothetical mixing model would estimate that ≥49% of amino acids in P. grandis larva are derived from vascular plants.
The trophic discrimination of nitrogen isotopes is attributable to the isotopic fractionation associated with deamination of amino acids in animals (Chikaraishi et al., 2007;Ohkouchi, Ogawa, Chikaraishi, Tanaka, & Wada, 2015). Deamination is a major initial step in the breakdown of trophic amino acids to produce metabolic energy for growth (Bender, 2012), and the balance of input (assimilation) and output (deamination) is theoretically an essential factor to determine the magnitude of the TDF McMahon, Thorrold, Elsdon, & McCarthy, 2015). This isotopic discrimination was empirically supported by an enzymatic experiment that showed that the δ 15 N value of the remaining amino acids was gradually elevated as the deamination flux increased (Macko, Estep, Engel, & Hare, 1986;Miura & Goto, 2012). The fairly constant TDF values observed in previous feeding experiments (Downs et al., 2014;Steffan et al., 2013) suggest that the proportions of deaminated amino acids are almost identical within arthropods that are fully satiated. In other words, an unusual isotopic discrimination is expected if an animal is forced into a state of severe famine, in which the budget of the available amino acids is considerably reduced McMahon et al., 2015). This is because consumers may have to continuously facilitate deamination to obtain metabolic energy even without additional input of amino acids derived from their diets.
However, in this study, we observed no substantial difference in the TDF between satiated and starved P. grandis larvae. Furthermore, we found that P. grandis larvae tolerate starvation for at least 8 weeks.
These observations suggest that the animal overcomes starvation without a substantial increase in output (i.e., metabolic breakdown) of amino acids, although they would need some metabolic energy for homeostasis. This phenomenon leads to several critical questions as follows: Is it possible that, in the wild, animals do not produce metabolic energy from amino acids? If yes, then from where is the metabolic energy derived? With respect to the P. grandis larvae, the respiratory quotient (i.e., carbon dioxide emission divided by oxygen consumption in a given time) is typically approximately 0.7, suggesting that they obtain energy in part by combusting storage lipids (Hayashi & Yoshida, 1987). Therefore, we suggest that starved dobsonfly larvae produce metabolic energy from storage lipids or carbohydrates rather than from proteins. These alternative sources serve as an "energetic buffer" against a starvation, resulting in unchanging δ 15 N and TDF values for amino acids because of disuse of amino acids as metabolic fuel. This adaptation to starvation means that ecologists might observe relatively unvarying (or almost constant) TP estimated by using CSIA-AA in wild animals, even when exposed to changing conditions of satiation and starvation. Thus, starvation-independent TDF will be key to evaluating the broad applicability of CSIA-AA methodology to a number of heterogenetic ecosystems.