General isotopic implications
The observed isotope tissue-diet shifts were both large in comparison with those previously found (McCutchan et al. 2003) and surprisingly these tissue-diet shifts were a function of the food source consumed. In this study, Ophryotrocha labronica fed all the domains of life had a Δ13C range of −3.6 to +3.6. This range is larger than the −2.7‰ to +3.4‰ found in a recent review covering many animal phyla (McCutchan et al. 2003). Previous investigators have found large ranges of Δ13C for polychaetes: −0.2 to 2.1 for Capitella capitata [sic] (Haines & Montegue 1979); +1.3 to +2.1 for Pseudopolydora kempi japonica (Hentchell 1998); and −2.2 to +3.9 for Hediste (Nereis) diversicolor (Jackson & Harkness 1987). Both C. capitata [sic] and He. diversicolor also had distinct shifts depending on their food source. Yet, these studies either lacked replication (Haines & Montegue 1979) or started with adults and relied upon tissue turnover during a 4-month period to evaluate Δ13C (Jackson & Harkness 1987). Data from the present study, which raised the model species from birth and took advantage of the increased throughput and decreased cost of isotopic analysis since 1979, support many of their findings. Yet, this phenomenon is not limited to annelids. Two recent studies found Δ13C values with a larger range than that observed in O. labronica, including a mammal whose Δ13C ranged from −8.8 to 0.6 (Caut et al. 2008) and an amphipod whose Δ13C ranged from −10 to −2 (Crawley et al. 2007), both of which were a function of the food source provided. In addition, similar findings that Δ13C is a function of a food source's δ13C have been found in a variety of cross-phyla studies (Hilderbrand et al. 1996; Felicciti et al. 2003; McCutchan et al. 2003; Caut et al. 2008, 2009).
Whereas a universal tissue-diet shift has been commonly applied to food-web studies, food quality and food amount in relation to metabolic needs impact the magnitude of Δ13C and Δ15N. Consumed food can be directly incorporated into a consumer's tissue or catabolized and respired/excreted or resynthesized prior to incorporation into its tissues (Auerswald et al. 2010). Within a food source, molecules are not homogeneous in their isotopic composition (DeNiro & Epstien 1977). This means that a consumer's tissue reflects the sum of many different tissue-diet shifts whose values are a function of differential incorporation and starting isotopic composition. Differential incorporation of food source molecules is manifested by proteins being largely directly incorporated into a consumer's tissue whereas carbohydrates and many lipids are catabolized before being rebuilt from components – as such proteins would have very low Δ13C and carbohydrates and lipids would have a larger Δ13C (Ambrose & Norr 1993; Fantle et al. 1999). This non-uniform reprocessing of compounds, called isotopic routing, can be especially important in mixed diets where one food source is higher in protein than the other food source. The high protein item will appear to be the dominant food source based on isotopic metrics even though it may be a minor component of the consumer's diet (Schwarcz 1991). Food sources with divergent C : N ratios cause greater tissue-diet shifts. A larger C : N ratio causes a greater Δ13C because of increased respiration and lower Δ15N as a result of nitrogen limitation (Webb et al. 1998; Adams & Sterner 2000; Phillips & Koch 2002). Greater amounts of food also allow rapid deposition of body mass, and consumer tissues will take less time to reflect the isotopic signature of the food source (Carleton & Martinez del Rio 2010). In addition, lipid synthesis causes a large isotopic fractionation for FAs (DeNiro & Epstien 1977), meaning that the more de novo FA synthesis performed by an organism the greater the depletion of 13C will be in that consumer's tissue. Neither archaeal food source provided any FAs to O. labronica, and both had negative Δ13C values as expected if de novo lipid synthesis was driving a portion of Δ13C (Table 1). However, these Δ13C were not significant, and the only significant and negative Δ13C value was from Photobacterium profundum, which appeared to provide at least some FAs assimilated by O. labronica, including 16:0, a FA that forms the basis for FA synthesis (Kattner & Hagen 1995). Therefore the balance between de novo lipid synthesis and assimilation did not appear to be a key factor in Δ13C. In times of famine, Δ15N is extremely high because the consumer's tissues are degraded during starvation and so 14N is excreted, creating an essentially infinite tissue-diet shift (Hobson et al. 1993). However, catabolic enzymes in marine organisms may not discriminate against 15N, leading to zero fractionation during starvation (Frazer et al. 1997; Mayor et al. 2011). In addition, Δ15N may be quite different for dissolved nitrogen sources. In any case, Δ15N values in this study were not indicative of starvation.
Acidification treatments can reduce the δ15N of a sample, which would appear as a reduced Δ15N (changing a predicted Δ15N from +2.4‰ to +1.1‰; McCutchan et al. 2003); however, this is not always the case and varies among groups analysed (e.g. Bosley & Wainright 1999; Carabel et al. 2006). In this study, the positive Δ15N for O. labronica fed Spinacia oleracea in the absence of antibiotics provides evidence that the acidification did not cause the negative or nonsignificant Δ15N that I observed for O. labronica fed most of the food sources. Many researchers use separate samples for δ13C and δ15N analyses to avoid this potential problem; however, in small organisms, especially those from deep-sea and chemosynthetic environments, where only a few individuals are available for analysis and single individuals may not provide sufficient biomass for separate analysis, coupled analysis is still necessary and common.
In this study I attempted to minimize many of the aforementioned artifacts associated with tissue-diet fractionation and yet still observed a food source-dependent tissue-diet shift. By raising O. labronica from hatching on a prescribed diet I avoided problems with incomplete isotopic turnover from previous food sources. Isotopic routing impacts on these estimates were minimized as all food sources were fed in monoculture and the whole organism was sampled, eliminating divergent assimilation among tissue types analysed. The C : N ratio did impact Δ13C but to a lesser extent than the starting food source δ13C.
The negative Δ15N values present in the S. oleracea treatment that included antibiotics potentially identified a methodological bias within this experiment. A variety of plausible explanations may explain this feature. Invertebrates have been shown to use dissolved organic food sources to augment particulate food (Rau & Anderson 1981; Manahan 1990). In this study, the antibiotics provided a potential organic nitrogen source that was on the same order of magnitude in concentration as the nitrogen provided by the food sources and the antibiotic's δ15N could explain the negative Δ15N (Fig. 1). If the antibiotics were consumed it would be expected that the carbon from the antibiotics would also be consumed and the food sources that provided the least nitrogen would have the greatest assimilation of antibiotic nitrogen. Neither of these phenomena occurred, although differential incorporation of nitrogen and carbon can result, especially if there are discrepancies among the nutritional values of the food sources (Podlesak & McWilliams 2006). Δ15N for Oryza sp. was not significant and was less than Δ15N for Halobacterium salinarium and S. oleracea even those these latter two provided five times the nitrogen of Oryza sp. (Table 1). An alternate mechanism by which the antibiotics may have impacted Δ15N is through altering any gut symbiont−host relationships. This is potentially the most likely explanation for O. labronica fed S. oleracea, as this appeared to be the only feeding assay that was influenced by the antibiotics. Heterotrophic annelids have been known to have a variety of bacterial symbionts, including denitrifying bacteria (Karsten & Drake 1997), a group that can cause δ15N changes much greater than 10‰ (Mariotti et al. 1981). Additionally, it cannot be ruled out that the nitrogen was taken up through absorption rather than dietary assimilation (Montagna & Bauer 1988).
It is not conclusive that the negative Δ15N values were caused by the antibiotics as (i) negative, often unexplainable δ15N values are found in a variety of field studies, (ii) the range of values observed within the individuals fed in the absence of antibiotics encompassed that of the food source and (iii) the negative Δ15N of individuals fed Halob. salinarium was not impacted by the presence of antibiotics. δ15N values reduced below 0‰ have been seen in a variety of heterotrophic fauna from deep-sea methane seep settings (e.g. Levin & Michener 2002; Levin & Mendoza 2007; Thurber et al. 2010), hydrothermal vents (Colaço et al. 2002), and lab-based studies (McCutchan et al. 2003). In addition, if the deviation from literature values of Δ15N was caused by the presence of antibiotics, when the antibiotics were not added there should have been a significant difference between food source and tissue and a Δ15N >2.3‰, neither of which occurred, even for individuals fed S. oleracea. While it is possible that the antibiotics may have augmented O. labronica's diet, they were not the main source of food. The gut of O. labronica was obviously full of the food source it was provided, and anecdotally, when this species was not fed sufficiently it did not survive on antibiotics alone at the concentration provided. In the instance where the same source of S. oleracea was fed to O. labronica with and without antibiotics, the range did overlap and thus there may be a potential influence on the nutritional source of S. oleracea used. Further study is clearly necessary to identify the role of antibiotics, host–gut symbiont interactions, dissolved organic nitrogen and food source in measured values of Δ15N. Regardless, this study shows that negative Δ15N occur and that food-web studies that apply a uniform and positive Δ15N may misidentify the trophic level of a consumer.
Essential fatty acids in annelids
A surprise from this laboratory study was the ubiquity of the essential FAs in individuals that were fed diets that did not contain essential FAs, suggesting the presence of desaturases not known to occur within most animal phyla. Essential FAs are those that cannot be synthesized using desaturases, the enzymes that add double bonds to FAs, known to occur in most Metazoa (Berge & Barnathan 2005); however, this paradigm is rapidly changing with further directed study (Monroig et al. 2013). Ophryotrocha labronica appears to be able to synthesize the polyunsaturated fatty acids (PUFAs), 20:4(n − 6) and 20:5(n − 3) from diets that do not include any FAs, i.e. the two archaeal food sources, and diets that do not provide the precursors for their synthesis, the two bacterial food sources. Although Spinacia oleracea provided both 18:3 and 18:2 FAs, the precursors that are used in terrestrial systems to form PUFAs, O. labronica fed S. oleracea had lower PUFAs than those fed other diets. There is some debate as to whether or not these starting points are functional for marine systems and thus may not act as precursors for this species (Pond et al. 2002). Without showing the specific enzymatic pathway or the presence of those necessary desaturases, this study can only suggest that O. labronica is able to synthesize essential fatty acids. In addition to this study, the most compelling evidence that supports the ability of Polychaeta to synthesize PUFAs comes from reducing habitats: 20:5(n − 3) and 20:4(n − 6) are both found in the symbiont-bearing, mouthless and gutless hydrothermal vent worm Ridgea piscesae (Pond et al. 2002), a methane seep frenulate (Lösekann et al. 2008) and a methane seep archivorous dorvilleid (Thurber et al. 2012). The latter example was found within authigenic carbonates and had 20:5(n − 3) and 20:4(n − 6) FAs with δ13C isotopic signatures of −103‰ and −109‰, respectively, clearly indicating nonphotosynthetic origin and likely synthesis; these PUFAs were not found in the habitat that the worms were collected from. More evidence of de novo synthesis by an annelid comes from a heterotrophic, hydrothermal vent species, Paralvinella palmiformis, that subsists on bacteria but has PUFA concentrations similar to shallow water species (Taghon 1988). Nematodes (Spychalla et al. 1997), harpactacoid but not calanoid copepods, mollusks (Monroig et al. 2013), heterotrophic ciliates and flagellates (Klein Breteler et al. 1999; Zhukova & Kharlamenko 1999) are now known to be able to synthesize these FAs and further research is warranted to determine if annelids may as well.
Fatty acid analyses in food webs
While effort was made to minimize microbial reprocessing of the food sources throughout this experiment, it is unlikely that all microbial growth was completely eliminated. Microbial fungal growth and input to the feeding trials is a possibility; however, if fungus were a dominant food source a uniform isotopic signature among individuals fed each of the food sources should have been present, which was not the case (Fig. 3). In addition, it is unlikely that fungi would have impacted the results similarly at three different laboratories and when both Pacific and Atlantic source water was used. 15:0, 17:0 and branched FAs were present in Ophryotrocha labronica fed all of the food sources. These FAs are thought be derived exclusively from microbial biomass. The least affected of all was O. labronica fed the Archaea, Haloferax volcanii, and this food source still resulted in 9.6% of the total FAs within O. labronica appearing to be bacterial (15:0, 17:0 and branched FAs). Unlike the aforementioned siboglinid polychaete, Ridgea piscesae, the dorvilleids are not known to have chemoautotrophic symbionts. Annelids however, are known to have microbial flora that aid in digestion and could impact biomarker uptake and modification, including essential FA formation (Sampedro et al. 2006). As O. labronica was able to grow to reproduction on all food sources, with no change in growth rate (Thurber et al. 2012), in no instance did the antibiotics used eliminate necessary processing of the food sources by potential gut symbionts. However, increased gut flora processing would have further differentiated the FA profile of O. labronica from that of its food sources. Microbial processing may be an integral part of how a food source is reflected in the tissue of a consumer and an inherent aspect of using biomarkers to identify a species diet.
Fatty acid analysis has been shown to be a powerful tool to identify qualitative food web links and can provide quantitative application in certain systems, yet as shown here, great care must be used when interpreting field results without careful laboratory study. In a landmark paper, Iverson et al. (2004) was able to use laboratory-based analysis of a marine mammal to derive a model sufficient to provide reliable tissue-diet shifts for each of the FAs present within the consumer's tissue. They were then able to accurately apply this in the field. Within the Annelida, Capitella sp. has been shown to have a FA composition reflecting its diet (Marsh et al. 1990), yet the diets tested in that study provided 18 different FAs in contrast to this study in which each of the food sources provided between zero and seven FAs. This resulted in O. labronica having to synthesize most of its FAs and likely led to FA profiles distinct from those of this species’ diet (Fig. 5). In a study similar to this one, ciliates and flagellates were fed food sources that lack FA diversity, and this resulted in a similarly large diversity of FAs in these consumers’ tissues (Zhukova & Kharlamenko 1999). This finding suggests that when a food source is limited in FA diversity, the FA signature of the consumer reflects its diet less. In addition, many short chain (i.e. 16:0 and shorter) FAs are catabolized rather than incorporated, contributing to why O. labronica does not reflect its diet – 16:0 was the dominant FA of three of the food sources present, with other short chain FAs being present along with them. As bacteria commonly have a profile dominated by short chain FAs, it is likely that bacterial FAs are routinely under-represented in bactivorous fauna. As no unique archaeal biomarkers were found within the FAs of the consumer, even when analyses aimed at targeting archaeal biomarkers were used, Archaea are not resolvable within a FA-derived food web. Instead, archaeal lipids are digested and the C from them likely used to synthesize FAs (Thurber et al. 2012).
For complex invertebrate food webs, FA analysis has been applied successfully to determine divergent food sources, and is therefore a valid approach for identifying interactions (e.g. Kharlamenko et al. 2001; Colaco et al. 2007; Drazen et al. 2008; Thurber et al. 2013). Yet as shown here, not all biomarkers are reflected in a species’ diet. When temperature, metabolism and species were held constant, only Bacillus subtilis resulted in enrichment of O. labronica's tissues in diet-provided FAs, and even then O. labronica's lipid pattern was discernible from only two of the other food sources. The absence of unique dietary FAs in the tissue of a consumer does not suggest that it avoids a particular food source. These results show that cellular machinery can be more important than diet in the FA profile of certain invertebrates. This may be because, unlike ‘higher’ organisms, annelids are adapted to FA-poor diets and thus have a greater abundance of desaturases, or gut symbionts, than is currently recognized. It remains to be shown if this species’ FA signature will more closely represent its diet if it is fed a high quality food source.
Food web implications for deep and chemosynthetic habitats
Fatty acid and stable isotopic approaches have been routinely employed since shortly after vents and seeps were discovered (e.g. Childress et al. 1986; Fiala-Médioni et al. 1986; Taghon 1988), largely because of the remoteness of these habitats and the inherent challenges that confront those who study them. Deep-sea vent and seep production is largely performed by Bacteria and Archaea (Fisher 1990; Levin 2005), and, just like the surrounding deep-sea fauna, whose food is stripped of essential FAs as the food sinks through the water column (Wakeham et al. 1997) provide limited access to those FAs thought to be needed for growth and reproduction. Many, but not all Bacteria (Nichols 2003) do not form essential FAs and therefore deep-sea animals that normally subsist on Bacteria either require periodic photosynthetic deposition, such as seasonal phytodetritus pulses (Billett et al. 1983), or the ability to synthesize FAs de novo to gain PUFAs. Through the results of the present study, in addition to the enigmatic presence of PUFAs in many annelids that do not appear to rely on photosynthetic production, it is clear that any conclusions based on the presence of these compounds alone could be incorrect as to the relative role of photosynthetic input into deep-sea habitats. Thurber et al. (2013) found PUFAs in fauna that was clearly (based on isotopic results) gaining the vast majority its energy through chemosynthetic production, and only through clustering analysis was it clear that there was a critical threshold of PUFAs that indicated surface derived versus in situ production. This finding supports the idea that FAs are still pertinent biomarkers to study food webs in chemosynthetic habitats yet understanding their caveats is critical to their implementation. In addition, the application of compound specific fatty acid analysis or amino acid analysis can further resolve any conflicting isotopic and FA data (McClelland & Montoya 2002; Thurber et al. 2012).