From all locations presented above, and due to its proximity to the laboratory, a permanent sampling site was selected at La Herradura Bay, Coquimbo, northern-central Chile (29°S; Figure 1a). In this site, fifteen subtidal surveys were carried out from May 2016 to March 2020, and whenever possible, egg masses were hand-collected by SCUBA diving at depths between 6 and 30 m. Samples were transported in seawater to the laboratory and each egg mass placed in a 20 L plastic container with constantly flowing fresh seawater at ambient temperature (range: 12–15°C in winter, and 16–20°C in summer). For each sampling date, sea surface temperature (SST, °C) was also obtained from the Hydrographic and Oceanographic Service of the Chilean Navy¹ databases, allowing further contrasts with the early life stages analyzed. Morphometric traits were evaluated in fresh samples, including number of capsules per egg mass, capsule size (mm), number of encapsulated embryos, egg size (mm), and paralarvae size (mm). Only eggs in early developmental stages (i.e., stage 16–17; see Guerra et al., 2001) and recently hatched paralarvae [i.e., 48 h post-hatch; mantle length (ML, mm)] were considered for morphometry (see Carrasco and Pérez-Matus, 2016). All measurements were performed by using a manual caliper and a dissecting scope at 20X magnification (Zeiss Stemi 2000-C) at Sala de Colecciones Biológicas, Universidad Católica del Norte (SCBUCN). At this time, egg capsules and paralarvae were frozen at −20°C for further stable isotope analyses (see below sections). In order to assess differences in early life-stages (i.e., capsules, eggs, and paralarvae size) among sampling temperatures (SST), independent Generalized Least Square Models (GLS) were performed for the evaluated traits. In each case, normality and homoscedasticity were previously checked by using Shapiro–Wilk and Levene’s tests, respectively.

Collections of juvenile and adult stages were conducted during four nocturnal surveys (20:30 PM to 01:30 AM) carried out between October 2018 and March 2020 in different points of the same bay (La Herradura). Squids were attracted to the boat by light fishing, carefully collected with a hand net, and individualized into 10 or 20 L plastic containers depending on the specimens’ size. In the case of collecting a considerable fraction of the school (i.e., up to 60 individuals), specimens were maintained at a maximum density of 3–4 individuals per 20 L container, and carefully transported to the laboratory located nearby within the same bay. In general, squids from low density collections remained individualized into the same 20 L containers with running seawater at ambient temperature (16.05 ± 1.4°C; mean ± SD). However, at high density collections, individuals were separated into two size classes (20–40 and 50–70 mm ML) and placed in two 3,000 L flow-through tanks, allowing them to maintain schools of around 30 individuals each. In all cases, squids were fed at a daily basis with amphipods, decapod megalopae, shrimps, and fish (depending on availability). Survival was also evaluated at that time, with any dead individual being removed from the tank to be sexed, measured with a manual caliper to the nearest millimeter (ML, mm), and weighed (g) using a digital balance (Radwag WTC 2000). All specimens were frozen at −20°C. Stomach content was also visually inspected by ventrally dissecting six adult specimens (i.e., 1–2 survival days in laboratory); however, considering that all samples evidenced highly digested food with no recognizable prey remains, no further analyses were carried out. Cannibalized squids from the laboratory (n = 20) were excluded from the analysis as only the mantles (i.e., no head or arms/tentacles) were recovered. With this information, a length-weight relationship was calculated using most specimens collected (46 out of 76) regardless the survival time in captivity (from 1 to 110 days). A power function model was used to evaluate the relationship between ML and weight. Additionally, ML of males, females, and undetermined juveniles, was contrasted with an ANOVA test after normality and homoscedasticity were evaluated using Shapiro–Wilk and Levene’s tests, respectively. Frequency distributions of macroscopic maturity stages (immature, maturing, and mature) between male and females were analyzed using a 2-sample test for equality of proportions with Yate’s correction (Proportion test). Accordingly, the maturing stage was characterized by the absence of spermatophores in males and the development of the nidamental glands in females.

As described above, benthic egg capsules of D. gahi were collected during subtidal surveys, whereas juvenile/adult squids, small juvenile fishes, zooplankton (e.g., copepods, euphausiids, brachyuran zoeae), and emergent benthic fauna (e.g., amphipods, nereid polychaetes) were collected during nocturnal surveys. For squids’ egg capsules (post-spawning capsules without embryos) and newly hatched paralarvae, complete individuals were considered in the analyses. In the case of juvenile and adult squids, mantle, arms, and lower beaks were dissected to compare potential differences in stable isotopes values. Samplings from lower beaks were collected as suggested for several species of squids (Cherel and Hobson, 2005; Cherel et al., 2009a, b, 2019; Xavier et al., 2015) and benthic octopuses (Guerreiro et al., 2015; Matias et al., 2019). For juvenile fish, muscle pieces were sampled, whereas for euphausiids, amphipods and polychaetes, complete individuals were used. Due to their small sizes and weights, samples of copepods and brachyuran zoeae were obtained by pooling ∼ 10 individuals to complete the necessary mass. In all cases, around 10 mg of wet tissue (in order to get ∼1 mg dry mass) was obtained, washed with Milli-Q water, placed in pre-combusted vials, and dried in an oven (60°C) for 48 h to be grounded into a fine powder with an agate mortar. Small amounts (∼0.5 mg) were placed in pre-weighed tin capsules and stored in a desiccator until the shipment for stable isotope analyses. When possible, voucher specimens were preserved in 95% ethanol and deposited at Sala de Colecciones Biológicas Universidad Católica del Norte (SCBUCN) for further taxonomic corroboration and genetic analyses. These ethanol-preserved specimens are also suitable for future isotopic analyses (see Pauli et al., 2017).

Analyses of carbon (δ¹³C) and nitrogen (δ¹⁵N) stable isotope ratios were conducted at UC Davis Stable Isotope Facility, using a PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (IRMS, Sercon Ltd., Cheshire, United Kingdom). Samples were combusted at 1000 °C in a reactor packed with chromium oxide and silvered copper oxide. Following combustion, oxides were removed in a reduction reactor (reduced copper at 650°C). The helium carrier then flowed through a water trap (magnesium perchlorate and phosphorous pentoxide). N2 and CO2 were separated on a Carbosieve GC column (65°C, 65 mL/min) before entering the IRMS. Stable isotope ratios were reported in the δ notation as the deviation relative to international standards (Vienna Pee Dee Belemnite for δ¹³C and atmospheric N² for δ¹⁵N), so δ¹³C or δ¹⁵N = [(R _sample /R _standard) – 1], where R is ¹³C/¹²C or ¹⁵N/¹⁴N, respectively. Typical precision of the analyses was ±0.2‰ for δ¹³C and ±0.3‰ for δ¹⁵N.

Since lipids are ¹³C-depleted relative to protein (DeNiro and Epstein, 1977), lipid normalization was applied for invertebrate and fish using C:N ratios (molar) [Eq. (3) in Post et al., 2007]. This correction was applied to samples with C:N values > 3.5 (>5% lipid content) according to the following equation:

In order to determine differences in mean values of δ¹³C and δ¹⁵N among different squid’s body parts (beaks, mantle, and arms), stable isotope values were independently compared through analyses of variance (ANOVA). Previous to the analyses, normality and homoscedasticity were tested with Shapiro-Wilk and Levene’s tests, respectively. Parametric post hoc Tukey’s tests were used to determine statistical differences in isotopic values among tissues.

To detect potential changes in prey selection and trophic position of D. gahi through ontogeny, Pearson’s correlation coefficient was used to determine the relationship between mantle length (i.e., 2 to 70 mm ML) with δ¹³C and δ¹⁵N values from mantle tissues, respectively. Mantle tissue was selected because it is more stable through ontogeny than arm tissue, which could be damaged and regenerated due agonistic and/or predatory-prey interactions. Data normality was tested through the Shapiro-Wilk test.

To assess the status of D. gahi within the pelagic food web, we estimated the trophic position for different consumers (e.g., squids, invertebrates, and fish) using the “oneBaseline” model, incorporating uncertainty in the trophic discrimination factor (TDF) for muscle nitrogen (mean ± SD Δ¹⁵N = 3.4 ± 0.1‰) (Post, 2002), as well as baseline and consumer δ¹⁵N values through Bayesian inference with tRophicPosition package (Quezada-Romegialli et al., 2018). Assuming that pelagic species of squids, invertebrates, and fish, could be supported mainly by pelagic carbon and nitrogen (Thiel et al., 2007), copepods were used as baseline with a mean trophic position of 2.2 (see Chen et al., 2018). Models were run with 2 chains, 20,000 adapting samplings, and 20,000 iterations.

To evaluate the relative contribution of six potential preys (n ≥ 3 individuals), including four invertebrates (i.e., Euphausia mucronata, Heterophoxus oculatus, Pseudonereis gallapagensis, and cyclopoida copepods) and two fishes (Odontesthes regia and Hypsoblennius sordidus) to the assimilated diet of D. gahi population (estimated for both mantle and arm tissue), Bayesian stable isotope mixing models were used (MixSiar; Stock et al., 2018) employing δ¹³C and δ¹⁵N values as tracers and elemental concentration from preys (concentration-dependence model). The TDF values for muscle (see trophic position methods above) were taken from a global meta-analyses (Post, 2002). The error term “Process^∗ Residual” was selected for the mixture (i.e., consumers). The residual error incorporates the potential variation associated with consumers (e.g., different metabolic rate, assimilation efficiency or digestibility), whereas the process error incorporates the variation associated to sampling process and consumer specialization (Stock and Semmens, 2016). Uninformative priors (all values between 0 and 1 are likely equally) were used in the models. MixSiar were run using a “normal” chain length through Markov Chain Monte Carlo (MCMC) and assessed through Gelman-Rubin and Geweke diagnostic test (Stock and Semmens, 2016).

All statistical analyses were performed in RStudio (R Studio Team, 2020).

As in latitudinal observations above, when considering a particular spawning site (i.e., La Herradura bay, Coquimbo), egg capsules were obtained from field and laboratory conditions during different months of the 4-year observation period. Overall, successful collections were recorded in January, March, April, May, June, July, August, October, and December (Table 2), with early life history traits (i.e., capsule, eggs, and paralarvae) evidencing distinct patterns depending on SST at the specific collection date. Egg capsules showed a significant positive association with sampling temperatures (β = 6.51; t = 11.41; p < 0.001), with larger capsules being recorded during warmer periods of around 18°C experienced during 2020 (range: 19–96 mm; n = 62; Table 2; and Figure 2a), compared with smaller capsules recorded during colder periods of around ∼15.5 during years 2016, 2018, and 2019 (range: 15–38 mm; n = 87; Table 2; and Figure 2a). Conversely, egg size evidenced a significant negative association with sampling temperatures (β = −0.10; t = −34.20; p < 0.001), with smaller eggs being observed during warmer periods of 2020 (range: 1.33–1.52; n = 87), in comparison with egg sizes observed during the colder years 2018 (range: 1.00–2.14 mm; n = 509) and 2019 (range: 1.00–2.00 mm; n = 559; Table 2; and Figure 2b). In the other hand, paralarvae size showed a weak but significant negative association with sampling temperatures (β = −0.032; t = −3.60; p < 0.001), although small individuals (<2 mm ML) were spread over the 4-year sampling period (n = 172; Table 2; and Figure 2c).

In general, egg capsules were attached to a variety of available substrata, including the typical rope moorings (from 6 to 20 m depth), the red alga Sarcodiotheca gaudichaudii in shallow waters up to 10 m, and a mix of unidentified porifera and erect bryozoans in a deep collection of around 30 m (Table 2). Egg capsules (range: 15–96 mm) and paralarvae (range: 1.8–2.9 mm ML) presented similar morphological traits compared to previous collections from central Chile (Table 1), suggesting that similar adult sizes would be responsible for reproductive interactions and egg laying in these coastal environments.

Samplings of older ontogenetic stages evidenced the constant presence of juvenile/adult squids through the year. In sum, 76 squids were collected with sizes ranging from around 19 to 77 mm ML and total wet weights from 0.3 to 15 g (Table 3). When combining all specimens collected (regardless survival time in laboratory), the power function model evidenced a strong positive relationship between squids’ mantle length (ML, mm) and weight (gr) (TW = 0.0014 ML^2.642; t_1,44 = 19.256, p < 0.001; Figure 3a).

From the forty-six specimens evaluated, 43% corresponded to females (n = 20), 33% to males (n = 15), and 24% to small juveniles of undetermined sex (n = 11). Mantle lengths (mean ± SD) evidenced significant differences (ANOVA, F₂,₄₁ = 8.912; p < 0.001), with males showing slightly larger sizes (54 ± 11 mm ML) compared to females (44 ± 11 ML mm ML; Tukey’s test: p = 0.045) and undetermined specimens (35 ± 9 mm) (Tukey’s test: p < 0.001) (Figure 3b). Frequency distribution of maturity stages evidenced that a high proportion of squids (i.e., 60% males and 45% females) corresponded to “mature” individuals, with sex evidencing no significant differences (Proportion test, chi² = 0.288, df = 1, p = 0.59; Figure 3c). Similar results were observed between “maturing” (20% males and 30% females; Proportion test, chi² = 0.077, df = 1, p = 0.78), and “immature” stages (20% males and 25% females; Proportion test, chi² < 0.001, df = 1, p = 1) (Figure 3c).

In captivity, individualized squids from first collections (2018–2019) did not survive longer than 10 days (Table 3); however, two females of different size (45 and 64 mm ML) laid eggs in the tank’s walls and in the artificial structures provided. Morphometrics on those egg masses and paralarvae have been provided in Table 2 (see egg masses 3 and 4, respectively). In the other hand, squids maintained in schools (20–40 and 50–70 mm ML) survived for up to 110 days (Table 3), which to our knowledge correspond to the longest captivity time recorded for D. gahi to date. Females of both size classes laid eggs between 2 and 3 months under these controlled conditions, although larger specimens produced more and larger egg capsules (and so more eggs per capsule) compared with smaller squids (see egg masses 12 and 13, respectively; Table 2). Within each of the two size classes, nearly 20% of the mortality was due to cannibalism over smallest sizes, suggesting that similar behaviors (reproductive and trophic) could be structuring schools in the wild.

The mean (±SD) δ¹³C values evidenced significant differences among the analyzed structures (ANOVA, F_2,33 = 4.708, p = 0.0159), with beaks being significantly more ¹³C-enriched than arms (−16.1 ± 1.2‰ vs. −17.6 ± 1.2‰; Tukey’s test, p = 0.019). Mantle tissue had intermediate values (−17.4 ± 1.5‰), but did not show significant differences with arms or beaks (Tukey’s test, p > 0.05 in both cases; Figure 4a). The δ¹⁵N evidenced significant differences in the mean values (± SD) of the structures analyzed (ANOVA, F_2,33 = 186.9, p < 0.001), with beaks being significantly lower (12.2 ± 0.7‰) in contrast to arms (16.9 ± 0.8‰) and mantle (17.5 ± 0.6‰) (Tukey’s test, p < 0.001 in both cases; Figure 4b). Intra-individual differences of δ¹³C and δ¹⁵N values between lower beak and soft tissues (i.e., mantle, arms) (correction factor) evidenced low differences for mean δ¹³C values (1.1 ± 0.6‰ vs. 1.4 ± 0.7‰, respectively; Table 4) and higher for δ¹⁵N values (5.3 ± 0.5‰ vs. 4.7 ± 0.4‰, respectively; Table 4).

Pearson’s regressions evidenced a positive but weak relationship between squids’ mantle length (ML, mm) and both δ¹³C (y = 0.013x–18.27; r² = 0.029, df = 12, p = 0.55; Figure 5a) and δ¹⁵N values (y = 0.0197x+16.382; r² = 0.22, df = 12, p = 0.09; Figure 5b). These findings suggest that through ontogeny (e.g., 2, 40, and 60 mm ML; see Figures 5c–e, respectively), D. gahi may present similar carbon sources (i.e., pelagic origin) and only slight increments in their trophic position.

Nocturnal pelagic species evidenced a wide variation of δ¹³C (7.3‰) and δ¹⁵N values (6.4‰), with the most ¹³C-depleted values (mean ± SD) reported in the blenniid Hypsoblennius sordidus (−21.7 ± 0.7‰) and the common krill Euphausia mucronata (−20.3 ± 0.7‰), whereas the most ¹³C-enriched values were detected in the polychaete Pseudonereis gallapagensis (−15.5 ± 3.9‰) and the amphipod Heterophoxus oculatus (−14.5 ± 0.3‰) (Figure 6 and Table 5). The δ¹⁵N values were ¹⁵N-depleted in E. mucronata (12.2 ± 1.3‰) and brachyuran zoeae (12.5‰), and ¹⁵N-enriched in P. gallapagensis (16.5 ± 1.0‰) and H. oculatus (18.6 ± 0.4‰) (Figure 6 and Table 5).

Estimations of trophic position indicated that primary consumers included E. mucronata (TP_mode = 2.2; 95% credible interval [CI] = 1.1-3.8), and H. sordidus (TP_mode = 2.7; 95% CI = 1.8-3.6; Table 5), followed by secondary consumers that included O. regia juveniles (TP_mode = 3.3; 95% CI = 2.4-4.4), P. gallapagensis (TP_mode = 3.4; 95% CI = 2.6-4.6) and D. gahi (mantle: TP_mode = 3.6; 95% CI = 2.8-4.5, arms: TP_mode = 3.7; 95% CI = 2.7-4.7; Table 5), and tertiary consumers such as H. oculatus (TP_mode = 4.2; 95% CI = 2.6-5.5; Table 5).

The Bayesian mixing models indicated that among the putative preys included in this study, Cyclopoida copepods could be the most important prey item assimilated in the diet of D. gahi, with a median estimate of 37.9% in the mantle (95% CI = 5.5–62.4; Figure 7a; Table 6), and 42.5% in the arm tissue (95% CI = 5.8-73.3%; Figure 7b; Table 6). The euphausiid E. mucronata was also relatively important in the mantle (median = 16.8%; 95% CI = 1.0–46.3) and arm tissue (median = 22.2; 95% CI = 1.7–53). For both tissues (mantle and arms), the contribution of pelagic fishes was in general low, with O. regia showing a 7.9% and 5.4%, respectively (Table 6), followed by H. sordidus with 6.7 and 5.0%, respectively (Table 6).

There was no evidence of a marked relationship between ML and δ¹³C and δ¹⁵N values, suggesting that carbon sources and trophic position, respectively, may not change significantly through the size range analyzed (2–70 mm ML). In fact, δ¹⁵N values from hatched paralarvae were slightly ¹⁵N-depleted than arms and mantle tissue from larger squids, reflecting a relatively constant diet through the ontogenetic stages evaluated. Similar to these findings, recent research exploring trophic allometries in cephalopods have also suggested that there are some species or Families (i.e., Cranchiidae) that do not exhibit ontogenetic trophic position shifts across different size classes, differing from the general description of cephalopod as voracious predators. These low activity level cephalopods were characterized by having low trophic positions, feeding largely upon herbivorous zooplankton and do not shifting to higher levels of piscivory until larger sizes (see Murphy et al., 2020).

Accordingly, previous studies have suggested similar changes in the diet of D. gahi, shifting from crustaceans in small-sized individuals (5.1-15 cm ML) to fish in larger sizes (15.1-25 cm ML) (Rosas-Luis et al., 2014). Changes in δ¹³C and δ¹⁵N values through the gladius sections have also supported dietary observations, suggesting important trophic changes through ontogeny (Rosas-Luis et al., 2017). Therefore, our findings are consistent with both observations, demonstrating a common source of food and trophic position in small and large-sized squids (via stable isotope analyses), and directly corroborating in laboratory that larger specimens D. gahi are able to prey upon different fish species, as well as cannibalize other squids, if predator and prey sizes are appropriate (i.e., ∼50% of squid’s ML; SA. Carrasco, unpublished data).

The trophic position (TP) estimated here for D. gahi (TP = 3.6 to 3.7) corroborate the trophic role of the species as a zooplanktivore (secondary consumer) in pelagic food webs. When contrasting TP values with other pelagic species occurring in the Humboldt Current system (all through C and N stable isotopes), these values were comparable to other pelagic species such as the mackerel Trachurus murphyi (TP = 3.8; Pizarro et al., 2019), the anchovy Engraulis ringens (TP = 3.7), and the jumbo squid Dosidicus gigas (TP = 3.9) (Espinoza et al., 2017). These estimations were comparable among studies given the use of the same trophic discrimination factor (TDF: ± 3.4‰; Post, 2002), except the trophic position of copepods (2.2) that was slightly lower than the value 2.5 used by Espinoza et al. (2017). In the other hand, TP estimated for D. gahi through different sections of the gladius evidenced higher values for juveniles (TP = 4.2) and adult stages (TP = 4.3) (Rosas-Luis et al., 2017); however, these values matched the error measurements from the TP models of mantle and arms.

Despite stomach content in the six specimens analyzed here only evidenced a liquefied content where no further taxonomic identifications of the dietary items were possible, the importance of planktonic crustaceans in the diet of D. gahi has been confirmed through stomach content in other areas of the distribution range, including Peru (Cardoso et al., 1998), Falkland/Malvinas Islands (Guerra et al., 1991), and South West Atlantic (Arkhipkin et al., 2013; Rosas-Luis et al., 2014). Bayesian mixing models were consistent with those observations and agreed with preliminary expectations, suggesting that copepods and euphausiids would be the most important putative preys, followed in a minor proportion by amphipods and polychaetes. In this context, the Coquimbo Bay system is located within an important upwelling center (Moraga et al., 2001; Thiel et al., 2007), with copepods dominating the zooplankton community structure, while euphausiids are frequently observed in lower abundances (Mattos and Mujica, 2012; Torreblanca et al., 2016) as response to extensive diel vertical migration into the oxygen minimum zone (Thiel et al., 2007). The presence of amphipods and swimming nereid polychaetes in nocturnal pelagic assemblages is not rare, and has been associated to the incidence of artificial lights in shallow depths (<30 m) in the former (Carrasco et al., 2017), or to the presence of reproductive stages (i.e., epitokes) attracted by lunar phases in the latter (Fong, 1993). Consistently with the present records, the presence of the polychaete P. gallapagensis has been reported in stomach content of individuals D. gahi from Peru (Cardoso et al., 1998) as well as in other members of the Family Loliginidae (Rocha et al., 1994), whereas amphipods have been recorded in the Atlantic Ocean (Arkhipkin et al., 2013). Therefore, the potential contribution of polychaetes and amphipods in the diet of D. gahi may provide temporal trophic subsidize to juvenile and adult squids in these coastal waters via emergent macrobenthic fauna (see Carrasco et al., 2017).

In the other hand, the trophic role of cephalopods as preys has been better understood given the presence of chitinous beaks within the stomach contents of upper consumers, including fishes, birds, mammals and other squids (reviewed in Arkhipkin et al., 2013). Differences in stable isotope composition between beaks and soft tissues (mantle and arms) observed in the present study show the same pattern reported in other cephalopods, with beaks tending to be slightly ¹³C-enriched and significantly ¹⁵N-depleted compared with soft tissues (Cherel and Hobson, 2005; Hobson and Cherel, 2006). The lower δ¹⁵N values in beaks compared to soft tissues could be explained by the presence of ¹⁵N-depleted chitin molecules (Cherel et al., 2009a). Therefore, differences in the present values for D. gahi could be useful for obtaining corrections factors for beaks found in different predators’ stomach contents, improving our understanding on the trophic role of the species along the HCS.

Overall, the present study unveils the permanent occurrence of a dwarf reproductive morphotype of D. gahi in shallow coastal habitats along the Humboldt Current System, providing the first insights for understanding potential adaptations to heterogeneous conditions in different sections of the distribution range (e.g., temperature, food availability), but also through the life cycle of the species (e.g., egg-capsules, paralarvae, adults). The combination of direct underwater surveys, isotopic composition, and behavioral components of alive organisms in laboratory, provided a strong approach for teasing out some of the ecological mechanisms underlying spatial connectivity patterns and the species’ role in a crucial (but unexplored) section of its geographic range, which connect the two genetically distinct centers of abundance, the Chilean – Falklands/Malvinas and the Peruvian. The incorporation of pelagic/benthic nutrients from small invertebrates (e.g., copepods, amphipods, euphausiids, and nereid polychaetes) and its transmission to upper trophic levels (fish, mammals, seabirds, and humans), enhance the role of D. gahi as an important nutritional link in these coastal environments and potentially through the complete distributional range.