Assessment of spawning of Atlantic bluefin tuna farmed in the western Mediterranean Sea

Mediterranean tuna farms account for >60% of the eastern Atlantic bluefin tuna (ABFT) catch quota. Besides the direct impact of purse seining on wild stocks, ABFT farming practices may have environmental implications that are still poorly known. An unexplored potential source of interactions of ABFT farms with wildlife is the release of eggs into the environment in places other than spawning grounds. Purse seine-caught ABFT schools are known to spawn in towing cages as they are transported to farms. We show here that farmed ABFT are also capable of spawning during at least 2 subsequent reproductive seasons following their capture. The reproductive potential of ABFT commercial stocks was investigated in a farm located in the western Mediterranean Sea from 2012 through 2014, using occurrence and number of postovulatory follicles as proxies of spawning fraction and realised batch fecundity, respectively. Although the spawning fraction among farmed fish was lower than that in the wild, the mean fecundity of captive spawners was similar to that of wild fish; consequently, the number of fertile eggs released from grow-out cages is thought to be significant. Larvae hatched from eggs spawned in farms are likely to grow and join wild-born ABFT juveniles that use nearshore areas of the western Mediterranean as foraging grounds. Depending on the volume of fish ranched for >1 yr and the larval survival rate in the region, the escape through spawning may have a significant impact on the ecosystem and could affect recruitment, thus influencing the population dynamics of ABFT in the Mediterranean Sea.


INTRODUCTION
Since the 1990s, farming of Atlantic bluefin tuna (ABFT) Thunnus thynnus (Linnaeus, 1758), in the Mediterranean Sea has become a profitable activity which relies on the capture of live fish from the wild (Ottolenghi 2008, Mylonas et al. 2010, Vitalini et al. 2010, Metian et al. 2014).ABFT are caught in spring and summer by purse-seine fleets and transported to offshore sea cages for fattening/farming over 4 mo to 1−2 yr.These capture-based aquaculture practices have received strong criticism, as they can negatively affect the natural resources (Sumaila & Huang 2012).
In addition to the direct impact of purse seining on wild populations, ABFT farming may have a suite of significant ecological implications that are still poorly understood.Most studies on environmental impacts of tuna farms have been focused on the effects of waste discharged from the cages (e.g.Vezzulli et al. 2008, Piedecausa et al. 2010, Sarà et al. 2011, Vizzini & Mazzola 2012, Moraitis et al. 2013, Mangion et al. 2014).However, very little data currently exist regarding other interactions between wildlife and ABFT farming activities.A widespread phenomenon that has been identified recently is the strong attracting effect of Mediterranean ABFT farms on wild individ-ABSTRACT: Mediterranean tuna farms account for > 60% of the eastern Atlantic bluefin tuna (ABFT) catch quota.Besides the direct impact of purse seining on wild stocks, ABFT farming practices may have environmental implications that are still poorly known.An unexplored potential source of interactions of ABFT farms with wildlife is the release of eggs into the environment in places other than spawning grounds.Purse seine-caught ABFT schools are known to spawn in towing cages as they are transported to farms.We show here that farmed ABFT are also capable of spawning during at least 2 subsequent reproductive seasons following their capture.The reproductive potential of ABFT commercial stocks was investigated in a farm located in the western Mediterranean Sea from 2012 through 2014, using occurrence and number of postovulatory follicles as proxies of spawning fraction and realised batch fecundity, respectively.Although the spawning fraction among farmed fish was lower than that in the wild, the mean fecundity of captive spawners was similar to that of wild fish; consequently, the number of fertile eggs released from grow-out cages is thought to be significant.Larvae hatched from eggs spawned in farms are likely to grow and join wild-born ABFT juveniles that use nearshore areas of the western Mediterranean as foraging grounds.Depending on the volume of fish ranched for >1 yr and the larval survival rate in the region, the escape through spawning may have a significant impact on the ecosystem and could affect recruitment, thus influencing the population dynamics of ABFT in the Mediterranean Sea.Another source of potential interactions with wildlife is the production of eggs by farmed ABFT.Although spawning has been observed in commercial cages, no investigation has been conducted to monitor and assess the reproductive performance of ABFT in farms.However, given the substantial proportion (> 60%) of the eastern ABFT total allowable catch eventually absorbed by purse-seine fleets that supply live fish to tuna farms (Ortiz 2015), it is likely that the number of eggs leaked from the cages to the environment is significant.Therefore, the assessment of the egg production capacity of farm stocks is worthy of consideration with a view to improving our perception of potential impacts of the ABFT fattening/farming industry.Bluefin tuna held in captivity undergo physiological impairment that results in significant reduction of their reproductive capacity (Mylonas et al. 2007).However, experiments carried out on cage-reared ABFT have shown that most of the individuals treated with implants loaded with gonadotropin releasing hormone agonist (GnRHa) recovered the capacity to mature and release fertile gametes (Corriero et al. 2007, 2009, Mylonas et al. 2007, de la Gándara et al. 2010, De Metrio et al. 2010, Aranda et al. 2011, Rosenfeld et al. 2012).A small proportion of the untreated fish in the experimental broodstock were likewise capable of spawning as early as 1 yr after their capture from the wild, showing fecundity rates that were similar to those of GnRHa-treated individuals (Corriero et al. 2007, Aranda et al. 2011).Uninduced spawning has also been observed in an experimental ABFT stock maintained in captivity for 1 yr and transported to the Balearic spawning grounds during the reproductive season (Gordoa & Carreras 2014).
In this study, we assessed the short-term reproductive performance of farmed ABFT by gonad histology analysis during the breeding season.The occurrence and number of postovulatory follicles (POFs) in the ovaries were estimated to determine the female spawning fraction and batch fecundity.Such estimates can be used to calculate the egg production capacity of commercial stocks.

Animals and farming conditions
ABFT were caught by purse seining around the Balearic Islands (Spain) in June 2010, 2012 and 2013.
They were transported to grow-out floating cages in the farming facilities of Grup Balfegó, located 4 km off L'Ametlla de Mar (Tarragona, NE Spain; Fig. 1).The holding cages were circular (50 m in diameter and 30 m deep) or elliptical (120 m long, 60 m wide, 30 m deep) and were moored in water of a total depth of 50 m.The initial stocking density was ~3 kg m −3 .After an adaptation period of a few weeks, the fish were fed to satiation 5 d wk −1 with defrosted baitfish, mostly mackerel.The initial ratio of bait supplied per total ABFT weight was ~8% in summer and decreased gradually to ~2% in winter.The sea surface water temperature (SST) was recorded daily from each cage.SST ranged from 11.6°C (February 2013) to 28.3°C (August 2012).
Systematic experimental sampling was impracticable, as the harvesting dates were fully dependent on the market demand.Harvesting took place between 06:00 and 08:00 h (UTC).Total body weight (BW, to the nearest 0.01 kg) and straight fork length (FL, to the nearest 1 cm) were recorded for each fish as they were harvested.Gonads were sampled from late May to early August in order to span the natural reproductive season of ABFT in the western Mediterranean Sea (Heinisch et al. 2008).Gonad weight (GW) was recorded to the nearest 0.01 kg, and the gonadosomatic index (GSI) was calculated as GSI = 100 GW BW −1 .The ovarian volume (OV) was estimated from the ovarian mass according to the equation: OV = 0.9174 GW (Medina et al. 2007).
All fish used in this study far exceeded 135 cm in FL, which is the size at which 100% of eastern ABFT are assumed to attain sexual maturity (Corriero et al. 2005).This, added to the fact that Mediterranean ABFT purse seiners primarily target schools of spawners, suggests that all the individuals examined were sexually mature.

Histology and functional classification of gonads
A piece of tissue was removed from the central part of one of the gonads and fixed in 10% phosphatebuffered formalin (4% formaldehyde in 0.1 M phosphate buffer, pH 7.2).The tissue samples were then washed in buffer, dehydrated in ethyl alcohol, cleared in xylene and embedded in paraffin.Serial 10 µm sections of the ovaries were stained with haematoxylin-VOF (Gutiérrez 1967) and photo graphed on a light microscope Nikon eclipse Ci® equipped with a Jenoptik ProgRes® CT5 digital camera.Histological sections of the testes were stained with haematoxylineosin.
Following Schaefer (1998), female ABTF were classified into 4 reproductive functional stages (Table 1), based on the most advanced group of ovarian follicles and the extent of atresia (for atretic states, see Hunter & Macewicz 1985a).The ovaries of resting (R) females contain only unyolked and/or early yolked oocytes and no sign of atresia.Active nonspawning (ANS) females show large yolked oocytes and minor (< 50%), if any, α atresia.The active spawning (AS) condition is characterized by signs of either imminent spawning (presence of migratory-nucleus and/or hydrated oocytes) or recent spawning (evidenced by postovulatory follicles).Inactive mature (IM) females have entered regression following the end of reproductive activity, hence the ovaries contain either previtellogenic or early yolked oocytes plus α and/or β atresia, or advanced yolked oocytes plus major (> 50%) α atresia (Fig. 2).Among females, the spawning fraction was calculated as the proportion of mature fish with POFs (Hunter & Macewicz 1985b).
Three or 4 distinct developmental stages have been identified from gonad histology in captive bluefin tuna males (e.g.Corriero et al. 2007, Sawada et al. 2007, Seoka et al. 2007).The male reproductive stages distinguished in our samples are shown in Table 2, which is based on Corriero et al. (2007).In fish at late spermatogenesis (LS), the germinal epithelium of the testes consists mainly of cysts composed of spermatids and spermatozoa, although spermatocyte and spermatid cysts are also present; the lumen of the seminiferous lobules and central ducts becomes filled with sperm.The testes of spent (S) males lack germinal cysts, and the lumen of the seminiferous lobules appears completely empty or shows scarce, loose residual sperm (Fig. 3).

Stereology
POFs were quantified by the physical disector method of Sterio (1984) adapted to fish ovarian samples (Aragón et al. 2010, Aranda et al. 2011, Ganias et al. 2014).The disector pairs consisted of 2 consecutive sections (referred to as reference and look-up sections) that were 40 or 60 µm apart (depending on the POF size in the histological sample).Three counting frames of 9.78 mm 2 were used per disector pair, and the total number of counting frames used per ovary was 18. POFs that appeared in the counting frame on the reference section, but not in the look-up section, were counted.When POFs touched the left or bottom lines of the frame, they were not counted.Counts were also made in the opposite direction.The volume number density of POFs was estimated ac cording to the formula: N V = ΣQ -(2 ΣP a/f h) −1 , where ΣQis the total number of POFs counted, ΣP is the number of disectors used per ovary (18), h is the disector thickness (40 or 60 µm), and a/f is the area of the working frame (9.78 mm 2 ).During histological processing, tissue

Statistical analysis
Differences of means between groups of data were analysed using either Student's t-test or ANOVA followed by Tukey's HSD post hoc test.In order to integrate data of Gregorian dates into statistical analysis, they were converted to days of the year, rang-

Animals and morphometry
Eighteen fish (all females) were sampled in 2012, of which 17 had been captured in 2010 and 1 in 2012.In 2013, 22 females and 7 males sourced from the wild in 2012 and 2013 were examined.All tuna sampled in 2014 (18 females and 17 males) were captured in 2013, with the exception of 1 male which had been caught in the previous month in 2014 (see Table S1 in the Supplement at www.int-res.com/ articles/ suppl/ q008p089_supp.pdf).

Histology
The majority (12) of the 18 tuna sampled in 2012 (all females) were classified as ANS; 4 of these fish (ANS* in Table S1) were entering the regression phase as they showed advanced yolked oocytes and the proportion of α atretic oocytes was close to 50% (> 40%).One of the fish was at the regenerating phase at the time of sampling (6 August); 2 individuals that were sacrificed in late July were IM, and the remaining 3 fish, which were sampled between 11 June and 12 July, were at the AS stage (Fig. 4).Spawning occurred at SSTs between 21.5 and 24.3°C (Table S1).The estimated spawning fraction was 0.2 (Fig. 4).
Five of the 22 females analysed in 2013 were sampled only 1 mo after their capture from the wild.One of them, sacrificed on 10 July, was IM, whereas the 4 others, which were sampled in late July, had completed ovarian regression and were therefore in the R stage.Among the fish that had spent over 1 yr in captivity, 7 were AS females, which were harvested between 10 June and 23 July.One of them had no POFs but did contain migratory-nucleus stage oocytes.Spawning occurred at SSTs between 21.47 and 27.0°C, with the exception of 1 individual sampled on 10 June 2013 at 17.2°C SST.Nine females were found at the ANS stage, 7 of which were nearing the regression phase (close to 50% α atresia).The remaining 2 females held in captivity for over 1 yr (sampled in late July and early August) were R (Fig. 4, Table S1).The estimated spawning fraction was 0.4 (Fig. 4).
The males harvested on 5 June and 10 July 2013, 1 yr after their capture, were found to be at the LS stage.The 4 males sampled in late July had been caught during the year's fishing season 1 mo earlier, and were all considered S (Table S1).
In 2014, the 8 females classified as AS were sampled between 9 June and 14 July.Spawning occurred at SST between 20.3 and 22.5°C.Another 8 females were classified as ANS; 1 of them was apparently in the transition towards the IM stage.One IM female was harvested on 31 July, and the only R individual was sampled on 4 August (Fig. 4, Table S1).The estimated spawning fraction was 0.5 (Fig. 4).
All males appeared to be reproductively active (LS stage) throughout the sampling period (30 May to 4 August; Table S1).
We found significant differences in GSI values among different histological categories in both males and females (ANOVA, F 3, 54 = 9.38, p < 0.001).GSI was significantly lower in R females (mean ± SD:

Stereology
The mean (± SD) realised batch fecundities estimated from the number of POFs (in millions of eggs) were 2.07 ± 1.69 (year 2012), 8.54 ± 5.28 (2013) and 12.71 ± 5.98 (2014).ANOVA indicated significant differences among the 3 groups of samples (F 2,14 = 4.481, p = 0.03).Even when the number of spawned eggs was calculated relative to the fish weight (mean [± SD] rela tive batch fecundities of 17.89 ± 10.40 g −1 , 39.92 ± 22.41 g −1 and 56.06 ± 17.72 g −1 , respectively, Fig. 4), the statistical analysis indicated significant differences, although the result of the test approached the limit of significance (F 2,14 = 3.90, p = 0.045).For both absolute and relative fecundities, Tukey's HSD post hoc analysis showed that the fish sampled in 2012 were significantly less fecund than those sampled in 2014, whereas significant differences were not de tected between the individuals from 2013 and the 2 other groups at a level of significance of 0.05.
Linear regressions revealed significant relationships between the realised batch fecundity and FL, BW and GW, but the predictive capacity of the models was relatively poor.The best predictor of the batch fecundity was BW through the equation: N POF (realised batch fecundity) = −4.51× 10 6 + 67.52 × 10 3 BW (F 1,15 = 27.44,p = 0.0001, r 2 = 0.62).Table 3 shows the values of Pearson's correlation coefficients of pairwise comparisons between variables.As seen in the table, relative batch fecundity was not sig nificantly correlated with any of the variables considered.

DISCUSSION
ABFT schools caught by purse seining in the Balearic Sea have been observed to spawn in towing cages during their transportation to commercial facilities (Gordoa & Carreras 2014).Our results show that farmed ABFT are also capable of spawning in commercial cages in subsequent years following their capture, even if the farm facility is relatively far from their natural spawning grounds.Histological evidence of spawning was observed between 9−11 June and 12−23 July.This period matches the reproductive peak reported previously for wild spawners in the western Mediterranean Sea (Heinisch et al. 2008, Gordoa & Carreras 2014).Substantial oocyte resorption (IM reproductive stage) was identified from mid-July, leading to spent ovaries (R stage) in late July to August.This temporal pattern is also consistent with the migratory dynamics revealed by electronic tagging in the region (Aranda et al. 2013a).Spawning occurred at SSTs ranging between 20 and 27°C, but POFs were unexpectedly found in the ovaries of a fish sampled on 10 June 2013 at SST slightly over 17°C, although the batch fecundity of this specimen was the lowest of all females examined.These observations support earlier results indicating that spawning may take place at SSTs well below the currently assumed minimum spawning temperature of ~24°C (Gordoa & Carreras 2014).
The fraction of spawning females appears to be lower in the captive environment than in the wild.A proportion of spawning females over 80% has been reported for ABFT breeding schools in the western Mediterranean Sea (Medina et al. 2002, 2007, Aranda et al. 2013a,b).In contrast, only 3 of the captive females sampled in 2012 were classified as AS, representing 20% of the total number of individuals found in the spawning-capable phase sensu Brown-Peterson et al. (2011), i.e.ANS plus AS fish.The spawning fraction was substantially greater in the samples of the 2 other years, where the percentage of AS females related to the total spawning-capable fish was 40% in the sample of 2013 and 50% in 2014.Such differences could be due to the smaller size of the fish in the 2012 group, but the effect of size/age on spawning fraction requires further investigation to draw reliable conclusions.The POF size and morphology varied among individuals, indicating they had different ages (see Aragón et al. 2010).Since all harvests were performed within a narrow time range early in the morning, distinct POF morphologies could reflect an irregular temporal spawning pattern in the cages.This is consistent with observations of spawning at different hours of the day (pers.comm.from the farm staff).In the wild, conversely, spawning appears to take place at nighttime within a narrow temporal window, approximately between 12:00 and 03:00 h UTC (Gordoa et al. 2009, Aranda et al. 2013a, Gordoa & Carreras 2014).
Preceding studies on experimental ABFT broodstocks (Corriero et al. 2007, Aranda et al. 2011) found that the spawning fraction in cages was low (similar to our 2012 data) even after a 3 yr period of adaptation to captivity.In ABFT, as in other teleosts, the stressful captive environment causes reproductive dysfunctions that primarily include oocyte maturation and ovulation failure in females, and reduced sperm quality/quantity in males (Mylonas et al. 2007).Although previous experimental trials have shown poor egg production in captive ABFT compared to wild populations, our observations suggest that a substantial proportion of ABFT held in farm grow-out cages are functional spawners during the species' natural reproductive season.
The increasing aquaculture of fish species, such as cod or sea bream, until the reproductive age results in escapes of large numbers of eggs into the environment (Uglem et al. 2012, Somarakis et al. 2013).The ABFT, like the Pacific bluefin tuna, is an example of those species where so-called 'escape through spawning' (Uglem et al. 2012) may be significant in commercial facilities, since > 60% of the ABFT eastern stock total allowable catch ends up in Mediterranean tuna farms (Ortiz 2015).We know that ABFT eggs collected from grow-out cages show a high hatching rate (75−87%) and produce apparently healthy yolk-sac larvae at temperatures between 21 and 26°C (Ortega 2015).However, the number and quality of the eggs produced in farms are difficult to determine, as it is virtually impossible to make sure that the entire spawns are collected from offshore cages.
Nevertheless, while reliable direct estimations of numbers of spawned eggs are unfeasible, the occurrence of POFs is useful to estimate the spawning fraction of females (Hunter & Macewicz 1985b, Hunter et al. 1986, Ganias 2012), and the quantification of these structures from histological samples provides accurate measures of fecundity.POFs are not detectable for more than 24 h in tuna species, such as the ABFT, that spawn in warm waters (Schaefer 2001).Thus, the number of eggs spawned within the past 24 h by ABFT can be estimated from unbiased stereological counts of POFs in ovarian samples (Aragón et al. 2010, Ganias et al. 2014).The average batch fecundity estimated for AS females in 2012 was ~2 million eggs (relative batch fecundity ~18 eggs g −1 ), which is far below the estimates for the 2 other years: 8.5 and 12.7 million eggs (~40 and ~53 eggs g −1 , respectively).The 2 latter figures are similar to the fecundities calculated elsewhere for captive and wild ABFT using the same counting technique (Aranda et al. 2011).
In agreement with a previous study conducted on wild ABFT (Aranda et al. 2013b), the absolute batch fecundity was strongly correlated with fish size, expressed as both FL and BW.However, the relative batch fecundity was not dependent on fish size, so that, generally speaking, all fish would contribute to the number of eggs produced by the broodstock in proportion to their weight.Consequently, the daily egg production of a stock could be easily estimated from data of the biomass stocked in the cages, the  Jensen et al. (2013) showed that Atlantic salmon escaped from farms at the smolt stage or early in the post-smolt stage may grow, migrate and survive to adulthood as wild specimens do.A similar situation may occur with ABFT larvae hatched in farms.The productive coast off the Ebro River delta, where the studied farm is located, is used as a feeding habitat by early stages of ABFT (about 3−4 mo old) coming from the breeding grounds around the Balearic archipelago (Medina et al. 2015).Although the survival rate from hatching to age 3 mo in this area is unknown, some of the larvae hatched from eggs spawned in the farm are likely to grow and join wild-born young ABFT in the nearby feeding ground.The same may hold true for other Mediterranean farms located close to age-0 ABFT feeding areas.This could have an impact on the recruitment and hence on the population dy namics of ABFT in the Mediterranean Sea.In order to assess the survivorship of ABFT larvae, ongoing experimental work is being conducted to investigate the temperature tolerances of ABFT larval and postlarval stages.
Depending on the number of fish that are ranched for over 1 yr and the larval survival rate in the zone, the 'escape through spawning' phenomenon could have significant ecological implications and, hence, warrant further investigation regarding population dynamics and stock assessment.

Fig. 1 .
Fig. 1.Approximate location of the Atlantic bluefin tuna Thunnus thynnus farm where the study was conducted (black dot) and the fishing grounds where the fish were caught (circle) samples experience a mean volume loss of 34.8%(Knapp et al. 2014).Therefore, a correction factor was applied to account for sample shrinkage.The total number of POFs (realised batch fecundity) was thus calculated as N = 0.652 • OV • N V (OV: ovarian volume), and the relative realised batch fecundity as the number of POFs g −1 BW.

Table 3 .
Output of bivariate correlation analysis between the continuous variables considered in the study showing values for Pearson's correlation coefficient r.BW: body weight; FL: straight fork length; GW: ovarian weight; GSI: gonadosomatic index; DOY: day of year; N POF: number of postovulatory follicles (proxy for realised batch fecundity); N POF g −1 : number of POFs per gram of total body weight (proxy for relative realised batch fecundity); SST: sea surface temperature.Correlations were significant at *p < 0.05 and **p < 0.01 sex ratio of the stock and the estimates of spawning fraction and relative realised batch fecundity.For instance, the proportion of spawning females found in this study was ~0.3 (17 fish with POFs out of 58 females sampled in all 3 years, Table1).