Introduction

Global warming will modify fish distribution and abundance around the world, with tropical and subtropical species expanding poleward (Perry et al. 2005; Hiddink and Ter Hofstede 2008; Simpson et al. 2011), while those inhabiting colder regions are expected to change their depth range (Perry et al. 2005; Simpson et al. 2011). As a result, food web structure and dynamics are also expected to change (Hoegh-Guldberg and Bruno 2010; Simpson et al. 2011; Bas et al. 2019). Making precise predictions, however, about the consequences of global warming is challenging without a broad historical perspective (Swetnam et al. 1999; Jackson et al. 2001; Lotze et al. 2011; Friedlander et al. 2014).

The South-Western Atlantic Ocean is inhabited by two species of hake: the Argentine hake Merluccius hubbsi (Marini, 1933) and the Southern hake Merluccius australis (Hutton, 1872), which both support important commercial fisheries (Bezzi et al. 1995; Cousseau and Perrotta 1998; Bertolotti et al. 2001; Lloris et al. 2005). Commercial hake fishing began in Argentina in the 1960s and targeted mainly Argentine hake. Commercial fishing for Southern hake started in the 1990s, when shallow water stocks of Argentine hake were declining (Lloris et al. 2005). Both species are similar in morphology and biology, but the Southern hake is tightly linked to the cold Malvinas current, whereas the Argentine hake prevails on warmer waters over the continental shelf (Cousseau and Perrotta 1998; Lloris et al. 2005). Furthermore, the abundance of Argentine hake declines sharply south of 52 °S and coastal summer spawning aggregations do not exist south of 47 °S (Bezzi et al. 1995; Díaz de Astarloa et al. 2011), where annual average SST is 11 °C (Rivas 2010). This suggests that in a warmer world, the Argentine hake could expand southward and replace the Southern hake off Tierra del Fuego.

The zooarchaeological record of ancient fishing societies offers an opportunity to explore changes through time on issues such as age-length relationships (Leach and Davidson 2001; Bolle et al. 2004), geographic distribution (Enghoff et al. 2007; Scartascini and Volpedo 2013; Bas et al. 2019) and trophic position (Zenteno et al. 2015; Braje et al. 2017; Szpak et al. 2018; Bas et al. 2019). The fish remains left by societies living in the warmer periods of the Holocene are particularly interesting since they offer a glimpse to a plausible future in the context of global warming (Bas et al. 2019).

The zooarchaeological record of Tierra del Fuego dates to the Early Holocene and hake (Merluccius sp.) abound in many archaeological sites since the Middle Holocene (Torres 2009; Santiago 2013; Zangrando et al. 2016). According to a diversity of proxies, climate and SST during the Middle Holocene in the southernmost tip of South America were warmer than today (Bujalesky 2007; Shevenell et al. 2011; Caniupán et al. 2014). Currently, annual average SST off Tierra del Fuego is 7 °C (Rivas 2010), but was as high as 11–12 °C at 53 °S during the Middle Holocene (Caniupán et al. 2014). Therefore, data on the biology of hake during the Middle Holocene can inform predictions of the future distributions of both species.

The recovery of 9000 skeletal elements from an unknown number of hake specimens from Río Chico 1 archaeological site (Santiago 2013), clearly demonstrates the existence of a huge coastal population of at least one hake species off north-eastern Tierra del Fuego during the Middle Holocene. That population has since disappeared and morphological analysis cannot distinguish the skeletal elements of Argentine and Southern hake other than hyomandibular and urohyal bones (Lloris et al. 2005). Therefore, the species of hake recovered from Río Chico 1 remains unknown and, nothing is known about the habitat where the Río Chico 1 hake were captured. The strong winds and currents in this region, coupled with the absence of sailing technology during the Middle Holocene suggest that aboriginal hunter-gatherer-fisher groups likely captured hake onshore, but hard evidence is missing. Stable isotope analysis can be informative about the habitat used by ancient hake, but detailed studies comparing the stable isotope ratios from sympatric Argentine and Southern hakes have not yet been published and the comparison of the existing data is hindered by very small sample sizes (Table 1).

Table 1 Stable isotope ratios (δ13C and δ15N) in the muscle and bone of modern specimens of marine fish species from the Patagonian Shelf (PS) and the Beagle Channel (BC)
Full size table

Here, we analyse mitochondrial DNA of a subsample of hake skeletal elements recovered from Río Chico 1 to separate the species, reconstruct the body size of those specimens through osteometry, and track changes in trophic position and habitat over time by analysing stable isotope ratios of C and N of modern and ancient hake.

Materials and methods

Study area and sample collection

The study area is located at the Atlantic Coast of Isla Grande de Tierra del Fuego, Argentina. The samples were collected from two distinct archaeological sites from the Middle Holocene (Online Resource 1). Two radiocarbon dates are available for the shell midden called Río Chico 1: 5828 ± 46 BP and 5856 ± 44 BP, corresponding to 6558 cal BP and 6260 cal BP (lab codes AA75285 and AA65165, respectively; Santiago 2013). They are comparable to those for the nearby archaeological site known as La Arcillosa 2: 5508 ± 48 BP and 5068 ± 66 BP, corresponding to 5868 cal BP and 5776 cal BP (lab codes AA60934 and AA102166; Salemme et al. 2007, 2014).

The faunal assemblage at the shell midden Río Chico 1 is largely dominated by fish elements, followed by bird and mammal bones (Santiago 2013). Conversely, La Arcillosa 2 is a shell midden dominated by mammal remains, followed by bird and fish bones (Salemme et al. 2014). Limpets [Nacella magellanica (Gmelin, 1791)] and mussels [Mytilus chilensis (Hupé, 1854)] also occur at both sites. The latter dominates the malacological samples by more than 90% in both archaeological sites (Santiago et al. 2014).

Bones from ancient hake of unknown species identity, pink cusk-eel [Genypterus blacodes (Forster, 1901)] and Patagonian blenny [Eleginops maclovinus (Cuvier, 1830)], as well as shells of limpet and mussel, were recovered from Río Chico 1 and La Arcillosa 2 (Table 2); all these samples were stored dry until further analysis. To avoid pseudoreplication, only the first vertebra was selected for hake, and bones from the neurocranium with the same laterality were used for the other fish species. Modern adult Argentine hake, Southern hake, pink cusk-eel and Patagonian blenny specimens were captured at the adjoining Atlantic Ocean and then bones were collected for stable isotope analysis (Online Resource 1). Shells from modern limpets and mussels were collected from the open-water beach, Punta María, on the adjoining Atlantic Ocean (Online Resource 1). All the specimens were sampled between 2016 and 2017 and stored at − 20 °C until further analysis.

Table 2 Stable isotope ratios (δ13C and δ15N) of ancient and modern fish from the Atlantic coast of Tierra del Fuego
Full size table

Pink cusk-eel and Patagonian blenny provide benchmarks to interpret the stable isotope ratios of C in ancient hake. Previous research has demonstrated no significant differences in stable isotope ratios of C and N across skeletal elements in fish with acellular bone (Bas and Cardona 2018), so the stable isotope ratio of hake, pink cusk-eel and Patagonian blenny can be compared directly although different skeletal elements have been analysed. The trophic positions of the pink cusk-eel and the Patagonian blenny are similar to those of the two hake species (Table 1), but the two former species differ in habitat selection. The pink cusk-eel is broadly distributed across the continental shelf and the upper slope (Villarino 1998), although is rare in very shallow coastal habitats (Pequeño et al. 1995; Cousseau and Perrotta 1998). Conversely, the Patagonian blenny is a coastal species highly associated with shallow areas influenced by freshwater runoff (Pequeño et al. 1995; Cousseau and Perrotta 1998; Quiñones and Montes 2001; Licandeo et al. 2006; Riccialdelli et al. 2017). As a result, the δ13C value of the pink cusk-eel is much lower than that of sympatric Patagonian blenny, and similar to that of both hake species (Table 1). Thus, if ancient hake recovered from Río Chico 1 inhabited more onshore areas than they do currently, their δ13C values were expected to be significantly higher than those of contemporary pink cusk-eel and modern hake and closer to those of contemporary Patagonian blenny.

Hake identification and taxonomy

The species identity of the ancient hake specimens was identified through ancient mitochondrial DNA (mtDNA) analysis. DNA was extracted in the Ancient DNA Laboratory at the University of York, which follows strict protocols for contamination control and detection, including positive pressure, the use of protective clothing, UV sources for workspace decontamination, and laminar flow hoods for extraction and PCR-set-up. The 45 hake bones were subsampled prior to isotopic analysis, providing between 2.7 and 35 mg of bone for analysis (see Online Resource 2). The bone samples were crushed using a micropestle, and DNA was extracted using a silica spin column protocol (Yang et al. 1998) modified as reported in Yang et al. (2008); DNA was eluted in a 50 µL volume. Blank extractions and negative controls were included within all extraction batches and PCR amplifications.

Primers were designed to target a 295 bp fragment of the mtDNA cytochrome b (cytb) gene, which can distinguish species within the genus Merluccius (Campo et al. 2007) bracketing positions 14,556–14,851 of the M. merluccius mitochondrial genome, Genbank accession NC015120; Forward primer—F14556 5′-ACCGCAAACGTCGAAATAGC-3′ and reviser primer R14851 5′-GGTACGGCAGACATTAAGTTTGTG-3′. PCR reactions were prepared and amplified following Speller et al. (2012) using an annealing temperature of 55 ºC, and successfully amplified PCR products were sequenced using forward and/or reverse primers at Eurofins Genomics, Ebersberg, Germany. The obtained sequences were visually edited using the ChromasPro software (www.technelysium.com.au), truncated to remove primer sequences, resulting in a final sequence length of 232 bp [positions 14,596–14827]; haplotypes were assigned using this longer fragment. Forty-two sequences were uploaded to the Genetic Sequence Database at the National Center for Biotechnical Information (NCBI) (GenBank ID: MK882658-MK882699).

Edited sequences were initially compared with published references through the GenBank BLAST application (https://www.ncbi.nlm.nih.gov/BLAST/) to ensure they matched with Merluccius species. Multiple alignments of the ancient sequences with 98 published Merluccius sequences were conducted using ClustalW (Thompson et al. 1994), through BioEdit (Hall 2001). Species identifications were assigned through phylogenetic analysis of a 175 bp fragment comparable to available reference sequences. The ModelTest (Version 2.3) software (Posada and Crandall 1998) was employed to determine the best-fit model (GTR + G, selected by AIC), implemented in MrBayes 3.2.5 (Ronquist and Huelsenbeck 2003). Ten million generations of analyses were performed to produce the phylogeny and clade credibility scores, with a burnin of one million generations. Phylogenetic trees were created using FigTree 1.4.0 (Rambaut 2007).

Size distribution analysis

The caudal fin of most of the modern specimens of Argentine and Southern hake sampled for this study was damaged, so fish size was measured as standard length (SL). The overall sample size was larger and the size interval was broader for Southern hake (N = 30; SL = 43.3–69.6 cm) than for the Argentine hake (N = 20; SL = 55.0–67.0 cm). Fish were slightly boiled, flesh removed, the skeleton was disarticulated and stored dry. The morphometric analysis focused on vertebrae because they are the most common skeletal element recovered from Río Chico 1. The axial skeleton of hake can be differentiated into three different types of vertebrae (Online Resource 3), but only the first thoracic vertebra can be individually recognized. Accordingly, the analysis focused on the first vertebra and three measurements were considered: the dorso-ventral height of the centrum (M1), the medium-horizontal width of the centrum (M2) and craniocaudal length of the centrum (M3) (Online Resource 3; Morales and Rosenlund 1979).

The morphometric analysis was conducted on the two species, but only the equation relating the dorso-ventral height of the centrum and the SL of Southern hake was used to assess the size of the ancient hake (see below). This is because the dorso-ventral height of most of the centra recovered from Río Chico 1 was outside the range of the centra measured from modern Argentine hake, thus indicating that the specimens recovered from Río Chico 1 were much smaller than those in our sample of Argentine hake. Conversely, the range of the centra recovered from Río Chico 1 overlapped broadly with that of the centra from the sample of modern Southern hake. Furthermore, the equations derived for both hake species yield the same results for the overlapping range of SL values (55.0–67.0 cm).

Stable isotope analysis

As previously reported, all modern samples were stored in a freezer at − 20 °C until analysis. Soft tissues were removed from limpets and mussels and the shells were rinsed with water, dried at room temperature and lightly scraped with sand paper to remove epibionts. Fishes were thawed at room temperature, boiled between 5 and 10 min and dissected to remove the selected bones. Shells and bones were latter dried in a stove at 60 °C for 24 h. Once dry, each sample was ground to fine powder and divided into two subsamples. This is because calcium carbonate and lipids have to be removed to obtain unbiased δ13C values (Newsome et al. 2006; Guiry et al. 2016; Bas and Cardona 2018), but demineralization consistently increases the δ15N values of the organic matrix (Bas and Cardona 2018). One subsample (“bulk” hereafter) was ground to fine powder with mortar and pestle and approximately 0.7 mg of bone powder were weighed into 3.3 × 5 mm tin cups, 7 mg of modern shell powder were weighed into 5 × 8 mm tin cups and 14 mg of ancient shell powder were weighed into 5 × 8 mm tin cups. The other bone subsample (“dml” hereafter) was ground to fine powder with mortar and pestle, dried again for 24 h at 60 °C and rinsed with a 2:1 chloroform:methanol solution to remove lipids (Folch et al. 1957). The chloroform:methanol solution was changed overnight until it was transparent. Bonedml subsamples were dried again for 24 h at 60 °C and demineralized with 0.5 N hydrochloric acid (HCl) until no more CO2 bubbles were released (Newsome et al. 2006; Bas and Cardona 2018). Shelldml subsamples were first demineralised by soaking in 1 N HCl until no more CO2 was released (Saporiti et al. 2014a). After demineralization, shelldml subsamples were rinsed with distilled water for 24 h, dried again for 24 h at 60 °C and mixed with a 2:1 chloroform:methanol solution to remove lipids. The chloroform:methanol solution was changed overnight until it was transparent. Then, samples were dried again for 24 h at 60 °C and 0.5 mg were weighed into 3.3 × 5 mm tin cups.

All tin cups were combusted at 900 °C and analysed in a continuous flow isotope ratio mass spectrometer (Flash 1112 IRMS Delta C Series EA, Thermo Finnigan; www.thermofisher.com) at Centres Científics i Tecnològics de la Universitat de Barcelona (www.ccit.ub.edu) in Barcelona, Spain. Gases from the combustion of bulk shell samples passed through a CO2 absorbent column for elemental analysis, containing CaO/NaOH. This was to avoid spectrometer saturation with CO2, because CaCO3 constitutes over 90% of the shells samples and large amount of shell had to be combusted to obtain enough N to measure δ15N values.

Abundance of stable isotopes is expressed using the δ notation, where the relative variations of stable isotope ratios is expressed as per mil (‰) deviations from predefined reference scales: Vienna Pee Dee Belemnite (VPDB) calcium carbonate for δ13C and atmospheric nitrogen (AIR) for δ15N. Due to limited supplies, however, isotopic reference materials, which included known isotopic compositions relative to international measurement standards, were analysed instead. All these isotopic reference materials were employed to recalibrate the system once every 12 samples and were analysed to compensate for any measurement drift over time. The raw data were recalculated taking into account a linear regression previously calculated for isotopic reference materials (Skrzypek 2013). Following Bas and Cardona (2018), only δ13Cdml and δ15Nbulk values were used for latter analysis. Furthermore, the carbon to nitrogen (C:N) atomic ratio of each dml subsample was used to assess the efficiency of lipid extraction (DeNiro 1985). The stable isotope ratios and the C:N ratios of all the samples are available as Online Resource 4.

Data analysis

First, linear regression was used the identify which of the three dimensions of vertebra centrum (M1, M2, and M3) best predicted the SL of modern hake specimens and that parameter was used to reconstruct the size of the ancient hake.

Secondly, the stable isotope ratios of modern and ancient organisms cannot be compared directly, because the isotopic baseline may vary temporally (Casey and Post 2011). Nonetheless, the proteins that make up the organic matrix of mollusc shells are preserved and unaffected by diagenetic changes (Misarti et al. 2017), hence offering material suitable to reconstruct the changes in the isotopic baseline (Casey and Post 2011; Drago et al. 2017; Misarti et al. 2017; Vales et al. 2017). Accordingly, values of δ13C and δ15N of ancient limpets and mussels from Río Chico 1 and La Arcillosa 2 and modern conspecifics were compared independently using General Linear Models (GLM) as run in IBM SPSS Statistics (Version 23.0.0.2 for Mac) with two fixed factors (period and species) followed by a Tukey’s (HSD) post-hoc test to assess the temporal variation of the δ13C and δ15N values in shells. This approach was not possible for fish, because all the ancient Argentine hake and pink cusk-eel samples come from Río Chico 1 and all the ancient samples of the Patagonian blenny come from La Arcillosa 2.

Third, Pearson correlation coefficients were computed to assess linear relationships between δ13C, δ15N and trophic position (TP) values and fish size for both hake species and for both periods. According to correlation results, GLM was run with one fixed factor (species or period) and SL as a covariate to compare the δ13C, δ15N and TP average values of each species in the two periods considered, after correcting for any baseline shift following mollusc stable isotope ratios. In cases for which correlations were not significant, a Student’s t-test was performed instead of GLM with the SL as a covariate. Normality and homoscedasticity assumptions were checked by means of Lilliefors test and Levene test, respectively.

The trophic position of each species (TPp) was calculated as:

$${\mathrm{T}\mathrm{P}}_{p}=\left[\left({\delta }^{15}{\mathrm{N}}_{p}-{\delta }^{15}{\mathrm{N}}_{m}\right)/3\right]+2$$

where δ15Np is the δ15N average values of each species; δ15Nm is the δ15N average value of molluscs; three corresponds to Trophic Discrimination Factor (TDF); and mussels and limpets were considered herbivores at TP = 2 (Caut et al. 2009). Baseline corrections are necessary to compare the stable isotope ratios of consumers from different systems, as differences in the stable isotope ratios of primary producers propagate to consumers. On the contrary, no correction is needed to compare the TP of the consumers from different systems, because TP is calculated independently for each food web and hence accounts for any difference in the isotopic baseline.

Fourth, SIBER (Stable Isotope Bayesian Ellipses in R; Jackson et al. 2011), was used to calculate standard ellipses to compare the size of the isotopic niche of the Argentine hake population for each period (Layman et al. 2007). The area of the convex hull and standard ellipse are independent from any difference in the isotopic baseline. For Argentine hake, both in the ancient and the modern specimens, the total area of the convex hull (TA) and two estimates of the ellipse area (SEAC and SEAB) were calculated. SEAC is the area of the standard ellipse corrected for small sample size but has no information about the associated error (calculated with p.interval = 0.95). SEAB is the Bayesian estimate of the standard ellipse area and is reported as median values and 95% credible intervals, as calculated by SIBER. All codes for SIBER analyses are contained in the package SIBER (Jackson et al. 2011).

Results

Hake identification and taxonomy

Mitochondrial DNA was successfully extracted and amplified from 42 of the 45 samples, identifying six cytb haplotypes (see Online Resource 2); phylogenetic analysis assigned all six haplotypes to M. hubbsi (Fig. 1). The majority (N = 30) of the samples carried the same cytb haplotype (Mhub2). Seven samples were assigned to haplotype Mhub1, two samples were assigned to Mhub5 and three samples carried unique cytb haplotypes (Mhub3, Mhub4, and Mhub6). No amplifications were observed within the blank extractions and negative controls; all unique haplotypes underwent repeat amplification and sequencing to ensure that polymorphisms were not the result of DNA damage or sequencing error.

Fig. 1
figure 1

Phylogenetic tree displaying the relationships between obtained ancient haplotypes (denoted by bold type) and published Merluccius cytochrome b sequences from GenBank. The Bayesian (Monte Carlo-Markov chain) consensus tree was composed using MrBayes 3.2.5 with Atlantic cod (Gadus morhua NC002081) as the outgroup. Model parameters (GTR + G) were identified through MrModelTest 2.3 and consensus trees were generated from two runs of MrBayes using 10 million generations each. Posterior probabilities of the major nodes are listed for each of the branches

Full size image

Size distribution analysis

The dorso-ventral height of the centrum (M1) was the best predictor of fish body size (F1,28 = 422.13; P < 0.001; r2 = 0.938) and hence was used to calculate the size of the ancient hake recovered from Río Chico 1 following this equation:

$$\mathrm{S}\mathrm{L}= \left(2.13+\mathrm{M}1\right)/0.016$$

The recovered hake ranged from 37.2 to 58.1 cm SL.

Stable isotope analysis

Collagen yield from fish bones ranged 12.3–22.5% and the C:N ratio ranged 2.3–3.7 (Online resource 4).

Body size (SL) was weakly correlated with δ15N and TP values in modern Southern hake (r = 0.404, P = 0.013; r = 0.404, P = 0.013; respectively), but was uncorrelated with δ13C values (r = 0.197, P = 0.149). On the other hand, body size (SL) of modern Argentine hake was uncorrelated with δ13C, δ15N or TP (r = 0.011, P = 0.483; r = 0.264, P = 0.145; r = 0.263, P = 0.146; respectively), but there was a weak, positive correlation between δ13C values and SL in ancient Argentine hake (r = 0.417, P = 0.003). There was no correlation between δ15N and TP values and SL in ancient Argentine hake (r = 0.197, P = 0.109; r = 0.195, P = 0.111; respectively).

The stable isotope ratios of C and N from mollusc shells of each period were normally distributed and were homoscedastic. Statistically significant differences existed between the δ13C values of limpets and mussels, as well as throughout time, but there was a statistically significant period × species interaction term (Table 3). Tukey’s post-hoc tests revealed that a significant interaction term emerged because the δ13C values of modern limpets were higher than those of conspecifics from both ancient sites and this was also true for the δ13C values of modern mussels and their conspecifics from La Arcillosa 2, but modern mussels did not differ from the ancient mussels from Río Chico 1 (Fig. 2). Conversely, differences in the δ15N values of contemporary mussels and limpets were not statistically significant, but they changed over time, although there was a significant period x species interaction term (Table 3). Tukey’s post-hoc tests revealed that the significant interaction term resulted because the δ15N values of modern mussels were lower than those of conspecifics from both ancient sites and the same was true for the δ15N values of modern limpets compared to ancient conspecifics from La Arcillosa 2 but not compared to ancient limpets from Río Chico 1 (Fig. 2). Nevertheless, it should be noted that mean δ13C and δ15N values of both limpets and mussels from Río Chico 1 were always in between those of La Arcillosa 2 and Punta María (Fig. 2). This suggests that a baseline shift certainly existed between ancient (Río Chico 1 and La Arcillosa 2) and modern (Punta María) δ13C and δ15N values and that the inconsistent differences between Río Chico 1 and Punta María likely result from a small sample size and a low statistical power. Because of this, we calculated correction factors to account for baseline shifts and allow the comparison of ancient and modern fish stable isotope ratios. To do so, we averaged the stable isotope ratios of ancient limpets and mussels from each site (Río Chico 1 δ13C = 2.64‰ and δ15N = 0.93‰; La Arcillosa 2: δ13C = 4.65‰ and δ15N = 2.61‰) and calculated the offset with the average stable isotope ratios from modern limpets and mussels. Later, we subtracted the offset from the δ13C and δ15N values of ancient fish samples, to correct for the baseline shift and allow the comparison with modern values. Baseline corrected values are shown in Table 2 as δ13Ccorr and δ15Ncorr. The δ13C and δ15N values from Argentine hake and pink cusk-eel that came from Río Chico 1 were corrected taking into account the correction factor calculated for this archaeological site (see above), and the same for he δ13C and δ15N values from Patagonian blenny that came from La Arcillosa 2.

Table 3 Summary statistics of GLMs to assess the effect of sampling period and species identity (fixed factors) on the temporal variation of the δ13C and δ15N values in shells from ancient (Río Chico 1 and La Arcillosa 2; Online Resource 1) and modern (Punta María; Online Resource 1) samples
Full size table
Fig. 2
figure 2

Stable isotope ratios (mean ± standard deviation) of ancient and modern mollusc shells from the Atlantic coast of Tierra del Fuego. Modern mollusc shells were collected in Punta María (PM; Online Resource 1). Ancient samples were recovered from Río Chico 1 (RC1) and La Arcillosa 2 (LA2) archaeological sites (Online Resource 1). Homogenous subsets were identified according to the results of post-hoc Tuckey’s tests. Letters denote statistically significant differences between δ13C and δ15N values from different periods and species, respectively. Sample size is N = 5 for each species and location. Consult Online Resource 4 for the raw data of stable isotope ratios

Full size image

Currently, Argentine and Southern hake differed significantly in δ15N, TP (Table 4) and δ13C values (t48 = − 2.317, P = 0.025). On the other hand, the δ13Ccorr values of ancient Argentine hake were significantly higher than those of modern ones (Table 4). Nonetheless, there were no significant differences between modern and ancient Argentine hake for δ15Ncorr or TP (t59 = 0.887, P = 0.379; t59 = 0.077, P = 0.939, respectively). This result suggests that adult Argentine hake have not changed their trophic position over time, but have moved offshore. Related to these results, the credible intervals of the SEAB from ancient and modern Argentine hake overlap (Table 2), and something similar occurred with the SEAC, which overlaps but in different ways (Fig. 3). Overlapping area of the modern specimens correspond to 58.6%, whereas the ancient specimens correspond to 29.7%. In general, values of SEAC, SEAB and TA are higher in ancient Argentine hake than in modern ones suggesting a higher diversity of individual foraging strategies in the past.

Table 4 Summary statistics of GLMs to assess the effect of species identity (fixed factor) and body size (covariate) on δ15N and TP of modern Argentine and Southern hake and the effect of period (fixed factor) and body size (covariate) on the δ13C values of modern and ancient Argentine hake
Full size table
Fig. 3
figure 3

Isotopic niches of modern Argentine hake from the Atlantic coast of Tierra del Fuego (N = 20) and ancient Argentine hake recovered from Río Chico 1 (N = 42), as revealed by standard ellipse areas corrected for small sample size (SEAC)

Full size image

The topologies of ancient and modern fish in the δ13C–δ15N isospace revealed two major differences at species level (Fig. 4). The first and most relevant was the dissimilar position of ancient Argentine hake compared to the pink cusk-eel and the Patagonian blenny, suggesting a more coastal habitat for Argentine hake during the Middle Holocene. Secondly, Patagonian blenny had lower δ15N values in the past, thus revealing differences in trophic position between ancient and modern samples.

Fig. 4
figure 4

Scatterplot of pelagic and benthic fishes into the δ13C–δ15N space. Top panel: ancient samples (6000–5000 cal BP). Bottom panel: modern samples. Error bars show standard deviation

Full size image

Discussion

The results of this study demonstrate that ancient Argentine hake inhabiting the Atlantic coast of Tierra del Fuego during the Middle Holocene had a broader isotopic niche and foraged on average in more coastal habitats than currently. This suggests that ancient Argentine hake might have gathered at least seasonally so close to the shore to be captured by ancient hunter-gatherer-fishers devoid of sailing technology. Currently, dense coastal aggregations of hake occur only north of 47 °S during summer spawning (Bezzi et al. 1995; Díaz de Astarloa et al. 2011) and annual average SST is at least 11 °C in those areas (Rivas 2010). On the other hand, SST off north-eastern Tierra del Fuego is 7 °C (Rivas 2010) and spawning aggregations do not exist (Bezzi et al. 1995; Díaz de Astarloa et al. 2011). If environmental conditions in a warmer world would match those prevailing in the Middle Holocene, the results reported here indicate that Argentine hake might become more abundant in the foreseeable future and form summer spawning aggregations off north-eastern Tierra del Fuego. Nevertheless, this interpretation of the results presented here relies on our capacity for accurate hake species identification, body length assessment and habitat use reconstruction.

Body size can be inferred confidently from fish skeletal elements (Casteel 1976; Smith 1995; Leach and Davidson 2001; Gabriel et al. 2012; Lernau and Ben-Horin 2016) and our results confirm that the dorso-ventral height of the centrum of the first vertebra of hake allow an accurate estimation of body size. On the other hand, the morphology of the first vertebra is of little use to tell apart closely-related species of hake, which is not a surprising result, as often the skeletal elements of species from the same genus are rather similar and have little diagnostic value (Cannon 1988; Yang et al. 2004; Lloris et al. 2005). In this scenario, only molecular methods allow species identification, as far as DNA is preserved in the bone tissue. This is not a problem in the cold environment of Tierra del Fuego, as demonstrated by the high recovery rate for mtDNA reported here and in Evans et al. (2016), but it could be more difficult in warmer regions.

Interpreting changes in stable isotope ratios is more challenging, because potential confounding factors. The most important is the likely historic change in the isotopic baseline, which can be addressed only through adequate correction (Casey and Post 2011). There is growing evidence that the organic matrix of mollusc shells offers a good record of changes in the isotopic baseline (Hill et al. 2006; Casey and Post 2011; Misarti et al. 2017) and is not affected by diagenetic changes (Misarti et al. 2017). Thus, the stable isotope ratios of samples from different periods can be compared directly after an appropriate correction for baseline shifts. Nevertheless, it should be noted that the organic matrix of mollusc shells is a mixture of proteins and chitin, a polysaccharide containing N (Furuhashi et al. 2009). As a result, the C:N ratio of the organic matrix of mollusc shells including equal amounts of protein and chitin is close to 5.5 and hence differs from that of collagen. This does not reduce the suitability of mollusc shells for correcting changes in the isotopic values, although a different benchmark is needed to assess correctly sample preservation and removal of carbonates and lipids.

Here we used the stable isotope ratios of C and N in the organic matter of mollusc shells to detect temporal shifts in the isotopic baseline. We averaged the stable isotope ratios in the shells of one species of suspension feeder (mussel) and one species of grazer (limpet), as in previous studies (Saporiti et al. 2014a, b; Zenteno et al. 2015; Bas et al. 2019). The coastal habitat of mussels and limpets is an obvious limitation, because the baseline changes recorded in such coastal species do not necessarily parallel those operating in the offshore habitats used by hake. Unfortunately, no offshore molluscs occur in the middens from Tierra del Fuego, so we assessed changes in the habitat use of hake using two different approaches, one of them dependent and the other independent from any baseline correction (see below).

The δ15N values of contemporary limpets and mussels did not differ, as they have the same trophic position (herbivores). Nevertheless, their δ15N values changed throughout time, except that of limpets from Río Chico 1, which is likely the consequence of small sample size and low statistical power. A declining pattern in the δ15N values of coastal molluscs had been previously reported for the South-Western Atlantic Ocean since the Middle Holocene and interpreted as indicative of a steady decline in marine primary productivity (Saporiti et al. 2014b; Bas et al. 2019). The reason for such decline is unknown, but more intense upwelling has been predicted in eastern boundary currents at high latitudes as a result of global warming (Bakun 1990; Sydeman et al. 2014).

The δ13C values of mollusc shells also revealed a baseline change over time but the reasons remain poorly understood. Again, one species from Río Chico 1 did not differ from modern conspecifics, which is likely the consequence of small sample size and low statistical power, but the overall evidence strongly suggests an increase in δ13C values. The massive burning of fossil fuels since the Industrial Revolution has resulted in a drop of the δ13C values in the atmosphere and the ocean during the past 150 years. Such drop, known as the Suess effect, has been some − 0.5‰ in the top 100 m of the water column off Tierra del Fuego (Eide et al. 2017) and should be recorded in the mollusc shells. However, the δ13C values of limpets and mussels are currently higher than in the Middle Holocene, not lower. This means that other processes have operated to alter the isotopic baseline, obscuring the Suess effect. Changes in the δ13C values of the primary producers associated to declining marine primary productivity, changes in the relative contribution of different types of primary producers to the organic carbon pool fuelling the food web or both might have operated. Those changes propagated to fish, as the δ13C values of all the fish species analysed are currently higher than during the Middle Holocene. Interestingly, changes in the δ13C values of benthic consumers (limpets, pink cusk-eel and Patagonian blenny; 4.2‰, 3.5‰ and 5.9‰ respectively) were more intense than those in species supported ultimately by phytoplankton (mussels and hake; 1.0 ‰ and 0.6‰, respectively). The δ13C values of another pelagic forager, the Fuegian sprat (Sprattus fuegensis) from the nearby Beagle Channel, have also increased approximately 0.4‰ since 1100 years BP (Bas et al. 2019), thus reinforcing the pattern of a less intense enrichment in 13C in phytoplankton-dependent species.

Results reported in this study also revealed significantly lower average δ13C values in modern Southern hake than in contemporary Argentine hake, although their ranges overlap. Previous studies also reported large, overlapping ranges of δ13C values in the muscle and bone tissue of these species, but formal statistical comparisons were hindered by very small sample sizes (Ciancio et al. 2008; Zangrando et al. 2016). For δ15N and TP, the overall stable isotope evidence (Ciancio et al. 2008; this study) confirms that adult Southern hake have a lower TP than Argentine hake of similar size. Crustaceans dominate the diet of Argentine hake which are less than 30 cm and fish and cephalopods become more important as hake grow larger, although little change in diet is observed for specimens larger than 50 cm (Angelescu and Prensky 1987; Belleggia et al. 2014; Botto et al. 2019). This explains why the δ15N and TP values of the adult Argentine hake studied here are uncorrelated with size. The diet of the Southern hake is poorly studied, but a delayed ontogenetic shift compared to Argentine hake might explain why δ15N and TP values are correlated with size even in adult fish. Further research is required on this topic.

The most important finding of this study is the variation of the δ13C values of Argentine hake since the Middle Holocene and the resulting shift in the topology of the fish community. Not only ancient and modern Argentine hake differ in average δ13Ccorr values. Modern Argentine hake also have a smaller standard ellipse area than ancient ones, independent of any correction and thus reveals a much narrower isotopic niche. Furthermore, a large fraction of the isotopic niche of modern Argentine hake is encompassed by that of ancient conspecifics. Differences between average δ13Ccorr values and the actual overlap between the standard ellipses of ancient and modern Argentine hake are sensitive to the correction factor used to account for baseline shifts, but differences in standard ellipse area are independent of any correction. Thus, there is little doubt that modern Argentine hake have a narrower isotopic niche than ancient ones. A similar reduction has been reported for other species from the South-Western Atlantic Ocean which have been intensely exploited since the arrival of the European societies to South America (Drago et al. 2017; Bas et al. 2019).

Changes in the topology of ancient and modern Argentine hake within the δ13C–δ15N biplot in relation to pink cusk-eel and Patagonian blenny are also independent of any baseline correction. Pink cusk-eel and Patagonian blenny differ largely in current habitat use (Pequeño et al. 1995; Cousseau and Perrotta 1998; Villarino 1998; Quiñones and Montes 2001; Licandeo et al. 2006; Riccialdelli et al. 2017) and δ13C values (Ciancio et al. 2008; this study) and they also differed largely in their δ13C values in the Middle Holocene (this study). The δ13C values of ancient Argentine hake were between those of ancient pink cusk-eel and Patagonian blenny from the same water mass, but those of modern Argentine hake are more depleted in 13C than both contemporary pink cusk-eel and Patagonian blenny. This is strong evidence that a large fraction of the ancient hake population foraged in shallower areas than currently. Furthermore, the standard deviation of δ13C was broader in ancient Argentine hake, thus revealing a higher diversity of foraging strategies, according to the broader isotopic niche above reported.

Conversely, there has been no change in the trophic position of Argentine hake off Tierra del Fuego since the Middle Holocene. Note that differences in the trophic position of ancient and modern Patagonian blenny are likely because of large differences in body size, as this species exhibits a strong ontogenetic dietary shift (Lloris and Rucabado 1991; Cousseau and Perrotta 1998; Martin and Bastida 2008; Bas and Cardona 2018) and the premaxilla of the Patagonian blennies recovered from La Arcillosa 2 were much bigger than those of the modern blennies used for reference.

Changes in SST might have contributed substantially to the habitat shift of Argentine hake reported here. Currently, the population density of Argentine hake is higher in the northern part of Argentine shelf (Boltovskoy 1981; Cousseau and Perrotta 1998) and the southernmost coastal spawning aggregations are reported from areas less than 50 m deep off Comodoro Rivadavia (46 °S) (Bezzi et al. 1995; Díaz de Astarloa et al. 2011; Botto et al. 2019), where the average SST is 11 °C (Rivas 2010). South of that latitude, Argentine hake are distributed in deeper and colder waters and no coastal spawning aggregation is known (Bezzi et al. 1995; Díaz de Astarloa et al. 2011). Currently, the average SST off north-eastern Tierra del Fuego is 7 °C (Rivas 2010), but the average SST at that latitude during the Middle Holocene has been reported to be 11–12 °C (Nielsen et al. 2004; Bentley et al. 2009; Caniupán et al. 2014). Hence, SST values off Río Chico 1 archaeological site during the Middle Holocene likely matched those currently observed off Comodoro Rivadavia. This suggests that Argentine hake would have reached the coast off north-eastern Tierra del Fuego during the summer of the Middle Holocene for spawning, thus becoming vulnerable to hunter-gatherer-fisher people devoid of sailing technology.

In conclusion, increasing SST resulting from global warming could lead to an increase in the abundance of adult Argentine hake onshore Tierra del Fuego and the development of onshore spawning aggregations. Combined with increased primary productivity, this could result into major changes in the fishing industry of the region.