Introduction

The Antarctic silverfish (Pleuragramma antarctica) is the most abundant pelagic fish inhabiting high-Antarctic waters (DeWitt et al. 1990). In the Ross Sea this species accounts for more than 90 % by number and biomass of midwater fishes (DeWitt 1970), and over 97 % of the total ichthyoplankton (Guglielmo et al. 1998; Vacchi et al. 1999; Granata et al. 2000, 2002). As key species in the Antarctic food web, the Antarctic silverfish occurs in the stomach of top predators ranging from large fish such as Antarctic toothfish to penguins, marine flying birds and seals [reviewed in Hubold (1985)]. It is predominantly a zooplankton feeder, but its diet ranges from phytoplankton and meso-zooplankton to copepods and euphausiids, and to other fish depending on size, location and season (Giraldo et al. 2011; Mayzaud et al. 2011; Pinkerton et al. 2013).

Given its abundance and ecologic role, the Antarctic silverfish has been extensively studied and many aspects of its biology have been elucidated [reviewed in La Mesa and Eastman (2012)]. Nevertheless, knowledge on important steps of its life cycle, including reproduction, is still poor and patchy.

The first research on the spawning period of the Antarctic silverfish (Andriashev 1965; Hubold 1984) leads to hypothesize that spawning occurred during winter and hatching in spring. A more extensive study in east Antarctica (Faleeva and Gerasimchuk 1990; Gerasimchuk 1992) allowed the determination of the length of sexual maturity for both sexes, confirmed winter or spring spawning, and provided the first details on gametogenesis including the description of two oocyte cohorts (pre-vitellogenic and yolked) in maturing ovaries. A recent study from the Antarctic Peninsula geographically extended the knowledge on the reproductive biology of this species, confirmed the occurrence of two oocyte cohorts, and that vitellogenesis occurred in the austral autumn (late March to early May; La Mesa et al. 2015).

Faleeva and Gerasimchuk (1990) reported the presence of follicles undergoing resorption (atretic) in the ovaries of all observed females. The presence of atretic follicles in two females was also reported for specimens from the Antarctic Peninsula (La Mesa et al. 2015). Ovarian follicular atresia has already been described for other notothenioid species (e.g., Everson et al. 1991; Calvo et al. 1999; Vanella et al. 2005).

In fish, the presence of many atretic follicles is considered a failure to achieve final maturation of oocytes, usually caused by poor nutritional condition (Hunter and Macewicz 1985). Regardless of the reasons for mass ovarian atresia, this phenomenon is suggested to be one of the features underlying “skip spawning” (the fact that adults may not spawn every year), described in detail for teleosts by Rideout et al. (2005). The increasing number of species recognized as skip spawners supports the acknowledgment that this is a widespread phenomenon in fish (Rideout and Tomkiewicz 2011). Among Antarctic teleosts, skip spawning has been reported for the demersal Patagonian toothfish Dissostichus eleginoides (Arana et al. 2009), Antarctic toothfish D. mawsoni (Parker and Grimes 2010), blackfin icefish Chaenocephalus aceratus (Vanella et al. 2005), mackerel icefish Champsocephalus gunnari (Kock and Kellermann 1991) and pike icefish C. esox (Calvo et al. 1999). In most of those cases, skipped spawning was deduced from high intensity of pre-vitellogenic follicular atresia, indicated by Rideout et al. (2005) as an accurate criterion to identify non-reproductive individuals.

Data on the reproduction of the Antarctic silverfish are fragmentary, and in particular, information from the Ross Sea is lacking. The fact that the only known nursery ground for this species is located in Terra Nova Bay in the Ross Sea (Vacchi et al. 2012) emphasizes the importance of this area for developing an understanding of the Antarctic silverfish reproductive cycle and life history.

To obtain information on the reproductive traits of Antarctic silverfish from the Ross Sea, we analyzed specimens from various locations on the continental shelf. Macroscopic and histological approaches were integrated allowing an accurate description of the gonadal structure and the assessment of reliable maturity stages of the specimens. Particular attention was devoted to the analysis of follicular atresia, considering the current interest in this process.

Materials and methods

Sampling and data collection

Sampling was carried out during the New Zealand research voyage in the western Ross Sea from the RV Tangaroa in February 2008 (Hanchet et al. 2008). The voyage carried out a number of demersal trawls using a rough bottom trawl with a mouth width of 25 m and cod-end mesh of 40 mm, as well as a fine-mesh midwater trawl with a mouth opening of 12 m diameter and a cod-end mesh of 10 mm. The locations of all bottom trawl stations and three midwater stations were randomly selected, while the remaining two midwater stations (49 and 103) were non-random, targeted trawls on midwater layers (see also O’Driscoll et al. 2011). Antarctic silverfish were sampled from all eight bottom trawl and midwater trawl stations where adult fish were caught (Table 1). Five to ten specimens were randomly subsampled from each station and promptly fixed in 10 % formalin for further analyses, except for the trawling station 79 where only three specimens were sampled. A total of 58 fish were sampled in this manner in depths ranging from 272 to 752 m. Further details of the survey area, sampling design and sampling methods can be found in O’Driscoll et al. (2011) and Hanchet et al. (2008).

Table 1 Summary of sampling information
Full size table

For each fish, the following variables were determined: standard length (SL) to the nearest millimeter below; total body weight (TW) to the nearest centigram below. After dissection, sex and stage of maturity were assessed on the basis of macroscopic appearance of gonads according to the five-point scale of maturity for notothenioid fish (Everson 1977; Kock and Kellermann 1991). Gonads and livers were removed, weighed with an accuracy of 0.01 g and stored in 70 % alcohol for further analyses. Somatic weight (SW) was determined to the nearest centigram below.

Indexes calculation

The gonadosomatic index (GSI) was calculated as the percentage of gonad weight (GW) per somatic body weight (Crim and Glebe 1990) in order to avoid biases due to the stomach fullness.

Absolute and relative fecundity were estimated in females at gonadal macroscopic stage III and IV using the gravimetric method (Murua et al. 2003) that relate the number of most advanced oocytes in the weighed subsample to the GW. A piece of ovarian tissue was taken from the median portion of a randomly selected ovarian lobe from each specimen. The subsample of each ovary, representing 1.5–8.2 % of the GW, was weighed and immersed in ethanol 70 % in a Petri dish. Oocytes were spaced and photographed by a Cell Pad E “all-in-one” microscope camera (TUCSEN). To exclude pre-vitellogenic oocytes a threshold size of 400 μm diameter was set. Oocytes larger than the threshold size were counted through the Adobe Photoshop Software.

Absolute fecundity (F abs) was estimated as:

$$F_{\text{abs}} = n \times GW/w$$

where n = number of oocytes larger than 400 μm in the subsample and w = weight of the subsample (g). Relative fecundity (F rel) was calculated as the number of oocytes per gram of fish body weight (TW). The hepato-somatic index (HSI) was obtained as the percentage of liver weight (LW) per TW.

Gonad histology

Histology was performed on the gonads of all 58 specimens. Fixed ovaries and testes were cut in half along the medial–lateral axis before embedding, and a subsample was taken from the middle to minimize potential variation due to section location. The subsamples were then dehydrated, cleared with xylene, embedded in Paraplast (McCormick Scientific, USA) and sectioned at 5-μm-thick sections. Hematoxylin–eosin staining (Bio-Optica, Milano Italy) was performed. Sections were examined using Leica DMRB light microscope, and images were acquired with a Leica CCD camera DFC420C (Leica, Switzerland).

Histological classification of the maturity stage for male is based on the progression of spermatogenesis from spermatogonia (Sg) to spermatozoa (Sz) and on the changes in the germinal epithelium. Testes were therefore staged based on morphological and histological criteria according to Brown-Peterson et al. (2011).

The ovarian follicles were classified as: primary growth oocytes (PG) including pre-follicular and early growth follicles, initial alveolus stage oocytes (IA) with at least one visible row of peripherally situated vacuoles, cortical alveolus stage oocytes (CA) characterized by multiple rows of membrane limited cytoplasmatic vesicles, vitellogenic oocytes (Vtg) with accumulation of yolk granules in the ooplasm. In addition, atretic oocytes (A) and post-ovulatory follicles (POF) were detected and counted. Each gonad was classified based on the most advanced follicle stage observed in the histological sections according to the five-point maturity scale of Brown-Peterson et al. (2011).

Correspondence between the macroscopic gonad maturity stages described by Kock and Kellermann (1991) and the histological stages described in Brown-Peterson et al. (2011) are summarized in Table 2.

Table 2 Maturity scales
Full size table

Analysis of follicular atresia

Follicular atresia is a degenerative process by which ovarian follicles lose their integrity and are reabsorbed prior to ovulation (Santos et al. 2008). Here we refer to the alpha-stage of the atretic process as detectable in the histological sections, characterized by breaks in the chorion, hypertrophy of the granulosa cells and disorganization of the oocyte structure (Üçüncü and Çakici 2009).

Following Kjesbu et al. (2010) the relative intensity of atresia (A RI) was expressed as:

$$\left( {{\text{Nr}}\;{\text{of}}\;{\text{atretic}}\;{\text{oocytes}}} \right)/\left( {{\text{Nr}}\;{\text{of}}\;{\text{normal}}\;{\text{and}}\;{\text{atretic}}\;{\text{oocytes}}} \right) \times 100$$

In mature females, where two cohorts of oocytes were clearly distinguishable, intensity of atresia pertaining to each oocyte population was calculated as the percentage of atretic oocytes in a given cohort over the total oocytes (normal and atretic) in that cohort.

Statistical analyses

Statistical analyses were performed using R 3.2.2 (R Development Core Team 2015). When dealing with indices and ratios, data were arcsin \(\sqrt {\frac{p}{100}}\) transformed. One-way ANOVA and post hoc Tukey HSD test were applied after having verified normality and homogeneity of variances through Shapiro–Wilk and Levene tests. Kruskal–Wallis nonparametric test and post hoc Wilcoxon test were used when data did not meet the normality and homogeneity of variances assumptions for ANOVA. Significance for all statistical analyses was determined at α = 0.05.

The effect of size on the gonadosomatic index (GSI) was investigated in both sexes separately (18 males and 34 females). The subsamples were grouped by length (small adults 152–178 mm, large adults 179–236 mm) based on the criteria of Pinkerton et al. (2013).

Variation of GSI values in response to ovarian maturation stage was tested on the 32 developing females (the two immature females were excluded). Oocyte maturation was classed in three levels according to the maturation stage of the most advanced follicle: initial alveolus (IA), cortical alveolus (CA) and vitellogenic (Vtg).

The Wilcoxon test was applied to test for differences in the incidence of atresia between the two oocyte cohorts (primary and secondary growth cohorts).

As potential factors influencing the intensity of atresia, oocyte maturation stage and condition level were taken into consideration. In order to gain a finer resolution, the analyses were performed on the two oocyte cohorts separately. Oocyte maturation was classed in the three previously described levels (IA, CA and Vtg). Specimens condition, evaluated through the HSI, was classed in three levels (HSI < 1.20, 1.20 ≤ HIS < 1.70, HSI ≥ 1.70).

To test the effect of geographic origin on the incidence of atresia, stations from which at least six specimens (replicates) were sampled were taken into consideration (stations 49, 77 and 92, see Table 1).

Results

The sample of Antarctic silverfish included 18 males, 34 females and six specimens of undetermined sex, ranging in size from 13 to 19 cm SL (Fig. 1). Males were homogeneous in size ranging between 16 and 18 cm SL, with average size 16.9 cm ± 0.69 SD. The size distribution of females ranged from 13 to 19 cm SL, but the average size (17.1 cm ± 1.28 SD) was comparable to the males. The specimens of undetermined sex were among the smallest, ranging between 13 and 16 cm SL.

Fig. 1
figure 1

Length frequency distribution of the Pleuragramma antarctica sample by sex

Full size image

Macroscopic staging, gonadosomatic index and fecundity estimate

Based on macroscopic examination of the gonads, most males (67 %) were developed (maturity stage III) with the remainder developing (stage II) or ripe (stage IV; Table 3). Male GSI ranged from 1.29 to 3.04 and the mean GSI increased progressively from stage II, III and IV (Fig. 2a). The effect of size on GSI was not able to be evaluated given the homogeneity in the size of males (all small adults, according to Pinkerton et al. 2013).

Table 3 Numbers of fish in each gonad maturity stage based on macroscopic features (following Kock and Kellermann 1991) and corresponding gonadosomatic index (GSI) range, mean and standard deviation
Full size table
Fig. 2
figure 2

Plot of gonadosomatic index (%GSI) versus standard length (SL). Individuals are labeled according to the macroscopic stage of gonad maturity in a males, b females

Full size image

Most females (62 %) were developing (macroscopic maturity stage III), and of the remainder two were immature (stage I), seven were either maturing virgin or resting (stage II) and four were gravid (stage IV; Table 3). The mean GSI increased progressively from maturity stage I to maturity stages II, III and IV (Fig. 2a), and was positively correlated with SL (r = 0.6). According to Kruskal–Wallis test, GSI values were not significantly influenced by size (χ2 = 2.3339, p > 0.1).

Fecundity was estimated for 25 females at macroscopic stages III and IV (size range 15.8–19 cm SL). The largest oocytes (diameter larger than 400 μm) were not homogeneous, among them two groups of oocytes were clearly distinguishable based on size: a first group including oocytes of approximately 460 μm diameter and a second one including oocytes of approximately 600 μm diameter. Absolute and relative fecundity was calculated based on the sum of oocytes from the two groups. F abs was found to range between 2360 and 6700 eggs per female with a mean of 4939 ± 1289 SD. The relative fecundity ranged between 65.8 and 188.4 eggs per gram of body weight, with a mean F rel of 111.3 eggs/g of body weight ±27.8 SD.

Histological characterization

Histological examination of testes revealed homogeneity at the level of microscopic organization. The tissues in the cross sections were organized in cysts homogeneously composed by primary spermatocytes (Fig. 3). All the males were classified as developing (stage II) despite different size and GSI (Fig. 4a).

Fig. 3
figure 3

Cross sections of testes of P. antarcticum (16.2 cm SL, IV macroscopic testis maturity stage). Male gonad is composed of cysts a each one composed only by primary spermatocytes (b detail of cyst). SC = spermatocyst; ScI = primary spermatocyte. Hematoxylin and eosin stain

Full size image
Fig. 4
figure 4

Plot of gonadosomatic index (%GSI) versus standard length (SL). Individuals are labeled according to the most advanced stage of spermatocyte/oocyte maturity found from histological analysis in a males, b females

Full size image

Histological examination of ovaries allowed the detection of two different gonad states. The ovaries of the two smallest females were exclusively composed of primary growth oocytes (PG), were homogeneous in size and were characterized by a large nucleus that occupied most of the cell surrounded by a thin strongly basophilic layer of cytoplasm (Fig. 5a). The gonads of all the remaining females were classified as developing (stage II). Based on size and cytological characteristics, two cohorts of oocytes were clearly distinguishable in the ovaries. The first cohort (primary growth cohort) was composed of pre-vitellogenic oocytes (PG), sometimes heterogeneous in size (Fig. 5a–c), and generally smaller than 100 μm diameter. The other cohort (secondary growth cohort) included maturing oocytes at a homogeneous developmental stage within each ovary. As a general rule, the secondary growth oocytes were larger in size than the primary growth cohort, and generally larger than 400 μm diameter. Based on the morphological properties, three distinct stages of vitellogenic cells were recognizable: initial alveolus (IA), cortical alveolus (CA), vitellogenic (Vtg). The cells were round with a thin acidophilic layer at the periphery of the cytoplasm (the future zona radiata) beneath the follicle. The nucleus was spherical in the IA cells (Fig. 5b) and became irregularly shaped, with niches along its wall, in the CA (Fig. 5c) and Vtg (Fig. 5d) cells. A large number of basophilic nucleoli were accumulated at the periphery of the nucleus; small grainy basophilic, not marginated, nucleoli were sometimes interspersed in the nucleoplasm or circularly displaced around the central area of the nucleus. Vesicles started to be accumulated at the periphery of the cytoplasm, in a few rows (IA), many rows (CA) or multiple rows occupying almost the entire cytoplasm (Vtg). In the vitellogenic oocytes, yolk granules started to be deposited from the periphery of the oocyte toward the centrally located nucleus.

Fig. 5
figure 5

Cross section of ovaries of P. antarcticum at different maturity stages: pre-vitellogenic (a), initial alveolus (b), cortical alveolus (c) and vitellogenic (d). PG = primary growth oocyte; IA = initial alveolus stage oocyte; CA = cortical alveolus stage oocyte; Vtg = vitellogenic oocyte; N = nucleus; n = nucleoli; a = alveoli; yg = yolk granules. Hematoxylin and eosin stain

Full size image

Based on the most advanced oocyte developmental stage, and according to the Brown-Peterson et al. (2011) five-point scale, the females were classified as: immature (2 specimens) and developing (32 specimens). This latter group included females with different maturation stages of the most advanced follicles, eight were at the IA stage, 15 were at the CA stage and nine were at Vtg stage (Fig. 4b). A significant effect of ovarian maturation stage on the GSI for IA, CA and Vtg ovaries was detected through one-way ANOVA (F = 12.54, p < 0.01). However, post hoc Tukey HSD test revealed that for IA stage the value of GSI was significantly lower than that for the other stages.

Follicular atresia

Atretic follicles were present in all sampling groups. In most cases, the atretic process was in its initial phase (alpha atresia) characterized by oocyte shrinkage, distortion of the follicle resulting in irregular oocyte shape, disorganization of the cytoplasm in the oocyte, fragmentation of the chorion leading to cytoplasm flow out of the cell, deformation and sometimes dissolution of the nucleus (Fig. 6a–c).

Fig. 6
figure 6

Alpha atretic follicles in cross sections of P. antarcticum ovaries. a Initial alpha atretic oocyte, granulosa cells (g) start to look hypertrophic, clearly show a break in the chorion and nuclear envelope (arrowhead). b Alpha atretic oocyte characterized by hypertrophy of the granulosa cells, several breaks in the chorion (arrowheads) and disorganization of the oocyte structure emphasized by the outflow of the nucleoplasm (N) from the oocyte wall. c Terminal alpha atretic follicle, the de-structured oocyte collapsed in the middle of the follicle, granulose cells are phagocyting the oocyte remnants. d Post-ovulatory follicle characterized by convoluted shape next to initial alveolus stage oocytes. αA = alpha atretic oocyte; N = nucleus (or nucleoplasm where the nuclear envelope is disrupted); g = follicular granulosa cells; POF = post-ovulatory follicle. Hematoxylin and eosin stain

Full size image

The number of atretic follicles was counted in each female specimen with a minimum of 74 and a maximum of 846 oocytes. The two immature females showed 90 and 27 % of the oocytes undergoing resorption.

In developing females, the intensity of atresia was separately calculated in the primary growth oocytes (thereafter referred to as cohort 1) and in the secondary growth oocytes (thereafter referred to as cohort 2). The incidence of atresia was significantly different in the two cohorts (Wilcoxon test V = 257, p = 0.01143), and differed according to the ovarian maturation stage of the individuals (Fig. 7). Specimens at the earliest stages of maturation had higher intensity of atresia in cohort 2 relative to cohort 1; conversely, specimens at the later vitellogenic stages showed higher atresia in the oocytes of cohort 1. The Kruskal–Wallis test showed no effect of ovarian stage on the intensity of atresia in cohort 1 (χ2 = 3.305, p > 0.10). Conversely, the incidence of atresia in cohort 2 was significantly different in ovaries at different maturity stages (χ2 = 9.6862, p < 0.01). Wilcoxon post hoc test highlighted that the incidence of atresia was higher in the IA stage compared to the others.

Fig. 7
figure 7

Relative intensity of atresia in cohort 1 (a) and cohort 2 (b) in ovaries at different microscopic maturation stages. IA = initial alveolus stage oocyte; CA = cortical alveolus stage oocyte; Vtg = vitellogenic oocyte

Full size image

One-way ANOVAs showed a significant effect of fish condition (HSI) on intensity of atresia in the oocytes of cohort 2 (F = 4.609, p < 0.05) and no effect in the oocytes of cohort 1 (F = 0.54, p > 0.10). Post hoc Tukey HSD test for cohort 2 showed a decreasing incidence of atresia with better fish condition.

One-way ANOVAs applied to investigate the influence of the geographic origin of the specimens on the intensity of follicular atresia revealed a highly significant effect of the variable in cohort 1 (F = 22.83, p < 0.01) and no significant effect on cohort 2 (F = 1.303, p > 0.10). Post hoc Tukey HSD test highlighted significant differences among all stations, with the highest incidence of atresia at station 77 and the lowest incidence of atresia at station 92 (Fig. 8).

Fig. 8
figure 8

Relative intensity of atresia in cohort 1 (a) and cohort 2 (b) in ovaries of specimens caught at three different stations

Full size image

Post-ovulatory follicles (POFs) were occasionally found in three fish. In all cases, the POFs were convoluted and folded with reduced lumen (Fig. 6d). In one of the specimens, POFs were still recognizable but almost completely reabsorbed.

Discussion

The reproductive traits of Antarctic silverfish inferred from our study are consistent with those reported from other geographic areas (Faleeva and Gerasimchuk 1990; Duhamel et al. 1993; La Mesa et al. 2015).

The average GSI values of maturing specimens were quite low (Table 3) and similar to that found at the same time in the Weddell Sea (Duhamel et al. 1993), slightly lower that those found in March–April from the Antarctic Peninsula (La Mesa et al. 2015), and lower than those recorded in April from the Cosmonaut, Commonwealth and Mawson Seas (Faleeva and Gerasimchuk 1990). The histological data presented showed that the testes were at a very initial developing state with only primary spermatocytes occurring in the cysts, while females showed early-vitellogenesis oocytes as the most advanced stage. Based on these results spawning in the Ross Sea is likely to occur in the late austral winter/early austral spring and at a similar time to that hypothesized for Antarctic silverfish in east Antarctica (Faleeva and Gerasimchuk 1990; Duhamel et al. 1993).

According to our analysis, the Antarctic silverfish from the Ross Sea have absolute fecundity values lower that those reported elsewhere for the species. Gerasimchuk (1987) reported absolute fecundity values ranging from 4315 to 17774 (average of 7499 ± 524) for Antarctic silverfish individuals collected at the end of March beginning of April in the Mawson Sea. In the same period of the year at the Antarctic Peninsula La Mesa et al. (2015) found values ranging from 3953 to 11613. Our results for the Ross Sea are from mid-February, where we recorded values ranging between 2360 and 6700 eggs per female. This might be an underestimate of the absolute fecundity because samples were collected early in the maturation period, and therefore, early vitellogenic oocytes might still be under the threshold size for fecundity count. However, we cannot exclude the possibility that low fecundity values might be typical of Antarctic silverfish in the Ross Sea. Samples of fish collected from later in the season in the Ross Sea would be required to resolve those alternatives.

From a histological perspective, the main traits of gametogenesis for specimens collected in the Ross Sea are consistent with those described for fish caught in other Antarctic areas and similar to those of other notothenioid species (e.g., Shandikov and Faleeva 1992; Calvo et al. 1999; Molens and Matallanas 2004; La Mesa et al. 2008). Females show group synchronous ovarian development, with two cohorts of oocytes: an advanced batch of large vitellogenic oocytes that will be ovulated in the current spawning season, and a set of smaller pre-vitellogenic oocytes, which form the reserve stock for the following spawning seasons. Follicles at advanced stages of reabsorption, residual from the previous spawning event, were detected in three out of the 34 females. Given that residual oocytes are thought to disappear completely within 6–8 months in other Antarctic notothenioid species (Butskaya and Faleeva 1987), the scarcity of POFs found in February, and their advanced state of phagocytosis, is consistent with the occurrence of spawning in late winter (July/August).

Is the Antarctic silverfish a skipped spawner?

The occurrence of atretic follicles in Antarctic silverfish has been reported from east Antarctica (Faleeva and Gerasimchuk 1990) and from the western Antarctic Peninsula (La Mesa et al. 2015). Although the capability of the Antarctic silverfish to skip spawn has already been suggested (Faleeva and Gerasimchuk 1990), analyses of the intensity and prevalence of follicular atresia, necessary to assess actual non-reproductive condition, were not performed prior to the present study.

We detected follicular atresia in all the examined ovaries. The determination of the intensity rate of atretic process individually, as well as the relative prevalence of atresia in groups of specimens, has provided clues to better understand the extent of the phenomenon.

The finding of atresia in immature females is not surprising since this process is normally occurring in ovaries as a means to regulate the physiological development of gonads. However, its role in mature specimens is less clear. A limited proportion of atresia is generally included among the mechanisms to down-regulate the ovarian maturation in early gametogenesis and to ensure vitellogenesis only occurs for the number of eggs that could be supported by the individual. By contrast, massive ovarian atresia should be interpreted as an anomalous status, possibly induced by external factors such as stress, fasting, biocidal agents, light, temperature, confinement and inadequate hormone levels (Miranda et al. 1999).

In the present study, there was a significant inverse relationship between the condition of the fish and the incidence of follicular atresia. Thus, condition (as a proxy of the physiological state of an individual) was a key factor influencing the intensity of reabsorption in the secondary growth oocytes.

The feeding opportunities during vitellogenesis are likely to play a critical role for the spawning success, and nutritional condition has proved to be an important measurement to estimate fecundity in fishes (Witthames et al. 2010). It is therefore reasonable to hypothesize that females would first try to maximize their spawning ability by producing as many pre-vitellogenic oocytes as possible (limited only by the size of the ovary), but subsequently they would be forced to adjust the threshold for vitellogenic eggs to the number of oocytes that they can effectively take to maturation. Condition is likely to play a key role in this second part of the gametogenesis acting as a fine-tuning factor.

If down-regulation is taken to its extreme in an individual, it would result in extensive oocyte reabsorption and, therefore, in diminished reproductive potential of that fish, whose contribution to the spawning of the following season would then be limited or nil (skip spawning).

This behavior is advantageous in the short term because it avoids the energy waste for specimens in poor nutritional status that therefore would be unable to sustain the reproductive effort, but it also represents an investment in the future. Indeed, the energy of a fish can be allocated to survival, growth or reproduction depending on the individual’s characteristics (age, size, stored energy, etc.) as well as on the state of the environment (food availability, pollution, etc.) (Jørgensen and Fiksen 2006). Therefore, by foregoing spawning an individual can incur survival and growth benefits. In this respect, skipping a spawning season represents a trade-off between current and future reproduction, and survival (Rideout et al. 2005). Such a strategy has already been reported for several Antarctic fish species including D. eleginoides (Arana et al. 2009), D. mawsoni (Parker and Grimes 2010), C. aceratus (Vanella et al. 2005), C. gunnari (Kock and Kellermann 1991) and C. esox (Calvo et al. 1999). It also seems plausible for the Antarctic silverfish in the Ross Sea given the dramatic seasonal and annual fluctuations in food availability, related to changes in the physical environmental factors (Smith et al. 2007).

On an individual basis, and on a short timescale, skip spawning following decrease in nutritional status would simply represent the biological response to physiological and ecological unfavorable conditions.

However, on an evolutionary timescale, such individual phenotypic plasticity of life history strategies could also represent an adaptation that guarantees a higher fitness to the species. If this holds true, skipped spawning might not be just an abnormal occurrence brought by exceptional environmental and physiological conditions, but rather an adaptive trait leading to increased lifetime reproductive output.

In this broader context, and to the best of our knowledge about its life history, the Antarctic silverfish provides some interesting clues that might support such a hypothesis.

Indeed, the Antarctic silverfish is specialized for pelagic life (Eastman 1997) and known to undertake vertical as well as horizontal migrations during its lifetime (La Mesa and Eastman 2012). Of particular interest is its supposed homing behavior (Koubbi et al. 2011) that would lead the fish to undertake breeding migrations (even long ones) moving from open waters to the coastal sites where they were born (natal homing, Papi 1992) or where environmental conditions are similar to those experienced at the larval stage (environmental homing, Cury 1994). Such a breeding strategy might be energetically costly. Therefore, taking into consideration that variability in the energetic state is usually present within populations, and that according to body size and condition some specimens could not be in the optimal shape to undertake the migration, skipped spawning on an individual basis would represent an advantage not only for the single specimen, but for the whole population.

Partial breeding migration, where some individuals migrate and reproduce while others in the same population do not migrate and forgo reproduction, has been reported for several pelagic and bathypelagic fish species (Shaw and Levin 2011; Chapman et al. 2012). It has also been inferred indirectly by the occurrence of ovarian massive atresia (Burton 1999; Rideout et al. 2005; Loher and Seitz 2008; van Damme et al. 2009; Skjæraasena et al. 2012). In the case of the Antarctic silverfish in the Ross Sea, the migration from open waters to the coastal spawning ground, potentially happening during the early winter, poses further challenges. The only known nursery ground for this species in the Ross Sea is located in a coastal embayment covered by sea ice for most of the year, thus implying that during migration the silverfish would encounter progressively harsher environmental conditions including ice crystals, besides exposure to a wider variety of coastal predators. In such an environmental scenario, the partial breeding migration strategy, sustained by the capability to down-regulate the oocytes development through follicular atresia, would acquire adaptive advantage.

To what extent our observations on specimens from the Ross Sea would allow generalization is an issue that deserves further investigations. However, the systematic analysis of ovarian atresia reported here lays the necessary basis to undertake those in-depth analyses aimed at clarifying various aspects of the Antarctic silverfish reproduction including the possible occurrence of skip spawning as a widespread adaptive breeding strategy for the species.