Incremental analysis of vertebral centra can reconstruct the stable isotope chronology of teleost fishes

Isotope analysis has high potential for understanding fish ecology and food‐web structure in aquatic ecosystems. The utility of isotope analysis will be greatly improved if we can reconstruct the chronology of several isotopes at multiple growth stages of individual fish. However, no practical methods exist for reconstructing the chronology of light‐element isotopes (e.g. δ13C, δ15N, δ34S, and Δ14C) in teleost fishes. Here, we present and test a new analytical approach for reconstructing the isotopic ratios of light isotopes at multiple life‐stages in teleost fishes. We sampled an anadromous salmon species, masu salmon Oncorhynchus masou (n = 3), along with water from its natal stream and from the ocean. We subdivided the vertebral centra of the salmon equally into 10 sections and extracted bone collagen from each sample. We then measured the stable sulphur isotope ratios of each vertebral section and compared them with δ34S values of the river water and sea water. We also measured the 87Sr/86Sr ratios of otoliths as a reference indicator of salmon migration between fresh water and the ocean. In all samples, the bone section closest to the centre of the centrum had the lowest δ34S values, which were similar to those of fresh water. The δ34S values gradually increased from the centre to marginal sections, finally reaching constant values similar to those of seawater. The 87Sr/86Sr ratios of sagittal otolith sections had significant inter‐individual differences and were consistent with the patterns of variation of the δ34S values of the vertebral sections. Our results show that the vertebral centra of teleost fishes record isotopic information from juvenile to adult life stages. We suggest that our method can provide reproducible isotopic chronology, even in teleost fishes smaller than 50 cm. This method can be used in isoscape studies and in studies of the ecology of marine teleost fishes.


| INTRODUCTION
Isotope analysis is a powerful tool in ecological studies of fishes for reconstructing their dietary sources, trophic positions, and movements.
Furthermore, the utility of isotope analysis would be greatly improved if we could reconstruct the history of these isotopes at multiple growth stages of individual fish.
One common method for reconstructing the isotopic chronology of teleost fishes is otolith analysis (Kerr & Campana, 2014). The otoliths of teleost fishes are composed mostly of calcium carbonate (Campana, 1999) and can provide accurate chronological isotopic records because they are not subject to metabolic activity and have high temporal stability (Campana & Thorrold, 2001). However, chronological isotopic analysis of otoliths is generally applicable to the isotopes of only a few elements (δ 13 C, δ 18 O and 87 Sr/ 86 Sr), and there are no practical methods for providing chronological information on the isotope ratios of other elements, such as Δ 14 C, δ 15 N and δ 34 S, in teleost fishes. For elasmobranch species, on the other hand, isotopic analysis of collagen in vertebrae is commonly used to reconstruct annual isotopic records (e.g. Estrada, Rice, Natanson, & Skomal, 2006;Kerr, Andrews, Cailliet, Brown, & Coale, 2006). Although, unlike otoliths, the vertebrae may not have perfect chronological isotope information, Δ 14 C analysis of elasmobranch vertebrae suggests that the vertebral centra of elasmobranch fishes record a certain degree of isotopic chronology (Ardizzone et al., 2006;Campana, Natanson, & Myklevoll, 2002). However, this method has been applied exclusively to elasmobranch species because separating these bones into multiple sections and measuring stable isotope ratios in the sections are technically difficult because of the small size of most teleost fishes. Nevertheless, we expect that incremental isotope analysis of the vertebral centrum can be applied to teleost fishes if enough bone collagen can be collected from their vertebral sections to measure stable isotope values.
Here, we present a new experimental protocol for preparing and extracting collagen from the vertebral sections of teleost fishes, and we test whether bone collagen in the vertebral sections provides a chronological isotopic record of teleost fish. We use the stable sulphur isotope (δ 34 S) values from an anadromous salmon species, masu salmon, Oncorhynchus masou Brevoort, 1856. The δ 34 S values are good for identifying terrestrial and marine signals because sea water has a markedly higher δ 34 S value than freshwater (Thode, 1991). The δ 34 S values of fresh water are determined by the type of bedrock in the region and show no seasonal variation (Rubenstein & Hobson, 2004). Marine sulphates have uniform δ 34 S values (21.0‰; Rees, 1978), which are spatially and temporally homogeneous (Rubenstein & Hobson, 2004). Furthermore, δ 34 S values in bone collagen are minimally affected by trophic enrichment because methionine, which is the dominant sulphur-bearing amino acid in fish bone collagen (Eastoe, 1957), is one of the essential amino acids for which there is little preypredator isotopic discrimination (Chikaraishi et al., 2009). Therefore, there should be negligible trophic enrichment of stable sulphur isotopes in bone collagen (Barnes & Jennings, 2007;Peterson, Howarth, & Garritt, 1985), further increasing their usefulness as an accurate reflection of marine versus freshwater habitat use by fishes.
We also used the strontium isotope ratios ( 87 Sr/ 86 Sr) in otoliths of the salmon as a reference record of these fishes' migrations between fresh water and the ocean. This ratio in otoliths has been used as a record of the migration of anadromous fishes between fresh water and the ocean (e.g. Kennedy et al., 2002;Padilla, Brown, & Wooller, 2015. This is because different water bodies have distinct 87 Sr/ 86 Sr values that are reflected in the 87 Sr/ 86 Sr in the otoliths: there is little to no biological fractionation from uptake to incorporation into the otoliths (Pouilly, Point, Sondag, Henry, & Santos, 2014;but see de Souza, Reynolds, Kiczka, & Bourdon, 2010;Halicz, Segal, Fruchter, Stein, & Lazar, 2008).
Masu salmon are distributed in the Western Pacific Ocean off East Asia. They hatch in fresh water and generally stay there for about 1.5 years, after which some individuals migrate to the ocean where they spend another approximately 1.5 years before returning to their natal streams to spawn. Other individuals may stay in the river mouths for several months without completing their migration to the ocean, and then return to freshwater (Machidori & Kato, 1984). Masu salmon are about 11-14 cm fork length when they migrate to the ocean, and they reach about 35-70 cm at the spawning stage (Machidori & Kato, 1984). Therefore, if chronological records of the δ 34 S values reflecting their freshwater and seawater habitat use remain in the vertebral centra, it should be possible to detect both freshwater and marine signals from the central and marginal sections, respectively, of the vertebrae, as well as from the 87 Sr/ 86 Sr values in otoliths.

| Study site and field sampling
Field sampling was conducted on the Churui River in the town of Shibetsu, in eastern Hokkaido, Japan ( Figure 1). The mean annual temperature in Shibetsu is 6.2°C and the mean annual precipitation is 1204.0 mm (Japan Meteorological Agency, 2016).
Anadromous salmonids in the Churui River include masu salmon, pink salmon Oncorhynchus gorbuscha Walbaum, 1792, and chum salmon Oncorhynchus keta Walbaum, 1792. Masu salmon return to their natal streams from June to July, pink salmon from August to October, and chum salmon from September to November. Pink and chum salmon are artificially hatched and stocked, but masu salmon are not. Masu salmon in Hokkaido are known for their strong homing ability, even to tributary level (Miyakoshi et al., 2012). Masu salmon are, therefore, a suitable species for the investigation of natural isotopic signatures from fresh water to the ocean.
We sampled masu salmon, river water, and sea water on 12 and 13 July 2016. We collected three masu salmon (identified as OM-01, OM-02 and OM-03) from upstream in the Churui River ( Figure 1); river-water samples were collected at the same point (upstream) and from the middle section of the Churui River ( Figure 1). Sea water was sampled about 10 km south of the mouth of the Churui River ( Figure 1) so that the sampling would not be influenced by river-water discharge.
River-water samples were initially filtered through a 0.7μm glass-fibre filter (GF/F, Whatman, Buckinghamshire, UK); we then added 1.0 M (mol L −1 ) HCL at 3 ml/L of filtered sample. We next added 10% (wt:vol) BaCl 2 solution at 15 ml/L and allowed the mixture to react for about 24 hr at ambient temperature. The precipitated BaSO 4 was then collected on a 0.7μm membrane filter (C300A025A, Advantec, Tokyo, Japan) and dried for approximately 24 hr at 60°C in a drying oven.
Because of the homogeneity of sulphur isotopes in sea water, we did not use the δ 34 S value from our seawater samples; instead we used a representative δ 34 S value for sea water of 21.0‰ (Rees, 1978).

| Otolith preparation and stable strontium isotope analysis
Sagittal otoliths extracted from skulls were sonicated in Milli-Q water (Millipore, Bedford, Massachusetts, USA) for 30 s and then freeze-dried. The otoliths were embedded in epoxy resin (SpeciFix Resin, Struers, Ballerup, Denmark) and polished to the core on polyester sheets coated with aluminium oxide powder in decreasing particle sizes (e.g. 500, 800 and 1,000 grit). Each sample was fixed to a frosted-glass slide with epoxy-based adhesive (Quick 30, Konishi, Osaka, Japan) and ground parallel to the slide surface with a grinding machine (Discoplan-TS, Struers, Ballerup, Denmark). The slides were sonicated in pure water for 30 s and freeze-dried. We then took photomicrographs of each thin section and counted the annuli on the surface of the otoliths ( Figure S1). Finally, multiple samples were taken from the otoliths using a micro-drill (Micro Mill, Electro Scientific Industries, Portland, Oregon, USA), staring at the core and ending at the rim.
Sample scrapings were digested in 1 ml 3 M HNO 3 on a hotplate at 80°C for 6 hr. After sample digestion, 0.5 ml of the sample solution was diluted with 5 ml 1% HNO 3 for analysis of elemental concentrations of calcium and strontium by quadrupole inductively coupled plasma mass spectrometry (7500cx, Agilent Technologies, Waldbronn, Germany). Indium was added inline as an internal standard for drift correction. External standard curves were generated by using a properly diluted multi-elemental standard solution (XSTC-622, SPEX SertiPrep, Metuchen, New Jersey, USA). Concentrations of calcium and strontium were measured with a typical precision of ± 3% to 5%.
All reagents used were of high purity, and the blank contribution was <1% of the analyte.
Strontium was separated from sample solutions by using columns filled with strontium resin (Eichrom Technologies Inc., Lisle, Illinois, USA), as follows. Samples re-dissolved in 0.3 ml 3.5 M HNO 3 were loaded onto the columns. The strontium fractions were then collected with 1.8 ml 0.05 M HNO 3 after elution of other elements using 0.5 ml 3.5 M HNO 3 and 0.5 ml 7 M HNO 3 , in that order. Finally, the F I G U R E 1 Location of Hokkaido Island, Japan, and sampling locations for masu salmon Oncorhynchus masou, river water and sea water. This figure was made by using GIS software (

| Bone segregation and collagen extraction
For stable sulphur isotope analysis, all vertebrae were first extracted from the fish body; any remaining muscle, spines and cartilage were removed from the vertebral centra by using scissors and a microgrinder ( Figure 2). Relatively small vertebrae from around the tail and head were excluded from further analysis. The thickness of each vertebral centrum was measured to the nearest 1 μm with a micrometer, and then each centrum was frozen at -20°C in Milli-Q water (MC-802A Electro Freeze, Yamato Kohki Industrial Co. Ltd., Saitama, Japan). Finally, we subdivided each centrum equally into 10 sections from the centre to the margin using a sliding microtome (REM-710 Retratome, Yamato Kohki). We combined corresponding vertebral sections from all vertebral centra to obtain enough bone collagen to determine the δ 34 S values.
Bone collagen was extracted from each set of combined vertebral sections by using the general gelatinization method (Longin, 1971;Yoneda et al., 2004), with slight modifications. We first immersed the samples in a methanol:chloroform mixture (1:1, vol:vol) for approximately 6 hr. The samples were rinsed twice with 99.5% methanol and then the remaining solvent was allowed to completely evaporate at ambient temperature. Next, the samples were immersed for about 12 hr in 0.1 M NaOH and then washed twice with Milli-Q water. The samples were subsequently treated with 1.0 M HCl for about 12 hr, and then rinsed twice with Milli-Q water. To obtain a high collagen yield, we did not crush the vertebral sections into powder (Sealy, Johnson, Richards, & Nehlich, 2014

| Stable sulphur isotope analysis
Approximately 3 mg of bone collagen or 0.5 mg of BaSO 4 were put into tin capsules. We then measured the stable sulphur isotope ratios with a Delta V Plus mass spectrometer (Thermo Fisher Scientific)

| DISCUSSION
Sulphur in bone collagen and strontium in otoliths follow different physiological pathways, and therefore these elements have different origins and biological turnovers. Stable sulphur isotope ratios of fish tissues, including bone collagen, are influenced by diet rather than ambient water (Hesslein, Hallard, & Ramlal, 1993;Nehlich, Barrett, & Richards, 2013) because methionine-the major sulphur-containing amino acid in bone collagen-is one of the essential amino acids, and most vertebrates cannot synthesize methionine. On the other hand, the 87 Sr/ 86 Sr ratio in otoliths of marine fishes is directly determined by that of ambient water and is not influenced by diet (Berg, 1968;Farrell & Campana, 1996;Walther & Thorrold, 2006). The importance of diet and ambient water to 87 Sr/ 86 Sr ratios in otoliths of freshwater fishes has not yet been determined.
The 87 Sr/ 86 Sr ratio in otoliths, which is known to record the migration history of fishes (e.g. Kennedy et al., 2002;Muhlfeld, Thorrold, McMahon, & Marotz, 2012;Padilla et al., 2015), shows a trend concordant with the δ 34 S of vertebrae. The lowest 87 Sr/ 86 Sr values, observed 400-500 μm from the core of otoliths from salmon OM-02 and OM-03, were comparable to the value for the ambient river water, strongly suggesting that the natal river of these fish was the Churui River. The otoliths of all three fish sampled shared 87 Sr/ 86 Sr values of about 0.707 at their cores. This is possibly due to the considerable contribution of maternal strontium to the development of otoliths in juvenile fish (Volk, Blakley, Schroder, & Kuehner, 2000). The 87 Sr/ 86 Sr ratios of the otolith sections from OM-01 did not decrease to the value of the river water; we interpret this to indicate that fish OM-01 migrated to the sea before all maternal strontium had been eliminated from its body. The unexpected decrease of 87 Sr/ 86 Sr at 558 μm from the core of the OM-01 otolith suggests that this fish returned to the river once and then went back out to sea. Such multiple returns to fresh water by masu salmon have not been reported previously but are possible given the flexible migratory pattern of this species (Machidori & Kato, 1984). water is the eggs of anadromous salmon such as chum, pink and masu salmon, given that juvenile salmon sometimes consume the eggs of spawning salmon (Reed, 1967). In either case, it is not clear why only OM-01 showed a different isotopic pattern, and more field surveys are necessary to reveal the specific reason.
The  Figure 3). It is likely that this mismatch results from differences between growth mechanisms in vertebrae and otoliths. The growth of vertebral bones is generally considered to parallel growth in total body size (Campana & Thorrold, 2001). In contrast, otoliths often grow even when somatic growth is very slow or completely stopped (e.g. Campana & Thorrold, 2001;Maillet & Checkley, 1990;Reznick, Lindbeck, & Bryga, 1989).
Presumably, all salmon in our study rapidly increased in body size during their early life stage; specimen OM-01 then immediately migrated to the ocean, whereas the other two specimens stayed in the river for a substantial amount of time. At higher latitudes, growth rates of anadromous salmon are typically higher in the ocean than in freshwater (Gross, 1987;Vøllestad, Peterson, & Quinn, 2004).
Therefore, despite the comparatively short residence time of OM-01 in freshwater, the body and vertebral growth patterns of all three fish specimens might have been almost the same, regardless of the time they spent in the river. However, their otoliths would have kept growing even during the period of freshwater residence with slower somatic growth, thereby resulting in different patterns of variation between the 87 Sr/ 86 Sr ratios of the otoliths and the δ 34 S values of the vertebral sections. We therefore suggest that both the vertebral centrum and the otolith contain chronological isotopic information, but that the time resolutions obtained from these tissues differ: during the juvenile stage, analysis of the vertebrae yields a lower time resolution than otolith analysis.
The three fish sampled in this study had vertebrae of similar size but different ages and growth rates, and these differences likely influenced the isotopic patterns in the vertebral sections and otoliths.
For example, OM-01 was aged 2 years, whereas the other fish were 3 years old, and therefore sub-samples from their vertebrae would reflect different time scales. Although such inter-individual differences are generally important in comparing isotopic patterns among individuals, they do not affect our conclusion that the freshwater signal was detected from the vertebral sections of all adult masu salmon sampled.
However, the age and growth rate of fish are critically important when using segmental analysis of vertebrae to compare multiple fish. The utility of segmental isotope analysis of vertebrae would be maximized if it were used together with age determination by using vertebral annuli, although the age estimate would be less accurate than that estimated from otolith annuli (Gunn et al., 2008).  & Kurle, 2016). Therefore, it may be more practical to chronologically reconstruct teleost fish habitat use with incremental δ 13 C and δ 15 N analysis of sequential vertebral sections using the method we outline here. We believe that our method is applicable to a broad range of fish species and isotopes. However, to ensure the utility of the method it is important to conduct more validation studies, ideally based on feeding experiments, which involve other species and isotopes.
In conclusion, our method provides a useful tool not only for investigating the ecology of teleost fishes, but also for the study of isoscapes (Bowen, 2010)