Influence of maternal effects and temperature on fecundity of Sebastes fasciatus on the Flemish Cap

Summary: The conservation of a sufficient reproductive potential of an exploited stock is one of the goals of fisheries management, as it ensures sustainable productivity. However, there is evidence that spawning stock biomass (SSB) does not represent well the variation in stock reproductive potential, often leading to impaired stock-recruitment relationships. In this study we show that fecundity of Sebastes fasciatus on Flemish Cap is not proportional to SSB and shows temporal fluctuation influenced by maternal effects. Females were collected in 23 research surveys between 1996 and 2020. An auto-diametric calibration model was developed for S. fasciatus for the first time to estimate fecundity. Mean potential fecundity was estimated as 36000 oocytes and mean relative fecundity as 79 oocytes g –1 . Potential fecundity varied significantly with female length, age, condition index, gonadosomatic index and environmental variability. Mixed-effect linear models were fitted to assess the effect of maternal traits and bottom temperature on fecundity. Fecundity increased significantly with condition factor and sea bottom temperature. Relative fecundity also increased significantly with length, age and gona-dosomatic index, indicating that older, larger and better-conditioned females produce more eggs per female gram. This suggests that SSB is not a good proxy to stock reproductive potential so it is unsuitable for use in stock assessment and scientific advice. Considering that S. fasciatus is a viviparous species, future research should focus on maternal effects on offspring and on building time series of reproductive potential indexes that take into account maternal effects.


INTRODUCTION
Changes in spawning dynamics, size or age at maturity, size structure and poor condition can increase the variability of recruitment (Marteinsdottir and Thorarinsson 1998, Blanchard et al. 2003, Anderson et al. 2008, reduce the resilience and capacity of populations to dampen environmental changes (Hsieh et al. 2006) and increase the impact of climate change (Cheung et al. 2009). Fisheries management will considerably benefit from a better understanding of how maternal features affect offspring phenotypes (the socalled maternal effect) and hence of how stock reproductive potential determines population productivity and recruitment.
Consequently, fecundity studies are critical for understanding the reproductive potential of fish populations (Tomkiewicz et al. 2003, Lambert et al. 2003, Saborido-Rey and Trippel 2013 and how maternal effects can interact with fecundity (Thorsen and Kjesbu 2006). Fecundity is a highly temporal and geographically sensitive variable that changes drastically with attributes of the individual spawners, including length, age and condition factor Saborido-Rey 2003, Rideout andMorgan 2010). In consequence, the population's egg production is highly dependent on adult stock demography and factors affecting demography, such as growth, maturation schedules, fishing pressure, environmental conditions and disease (McElroy et al. 2013, Chang et al. 2021. Moreover, in many teleosts, significant differences have revealed disproportionally positive relationships between potential fecundity and fish length (Stafford et al. 2014, Love et al. 2002, age and condition (Thorsen et al. 2006, Lambert 2008, highlighting the importance of maternal effects. Several studies have shown that in Pacific rockfish species maternal effects are determined by release offspring date and seasonal changes in the productivity of the California current, so offspring quality is directly affected (Fisher et al. 2007).
Monitoring of fecundity, as reported in the literature, can be used in stock assessment and fisheries management (Yoneda and Wright 2004, Lambert 2008, McElroy et al. 2013, especially under the climate change scenario in species with a strong maternal influence, such as those showing viviparity. However, long time series of fecundity are usually not available, as reported by Tomkiewicz et al. (2003), and the situation has not improved over time. The difficulty of estimating fecundity is likely the main hindrance to regular and routine estimation. In this regard, the autodiametric method developed by Thorsen and Kjesbu (2001) must facilitate fecundity estimations.
In this study, for the first time we applied the autodiametric method to estimate fecundity in S. fasciatus on the Flemish Cap bank to build a unique long time series of fecundity data of 20 years from 1996 to 2020. We analysed the maternal influence on several reproductive traits and tested whether water temperature influences fecundity. Our overall aim was to improve our understanding of the effects of maternal influence and climate variability on the productivity of S. fasciatus, following the hypothesis that spawning stock biomass (SSB) and other stock indexes do not represent well the variation in stock reproductive potential, often leading to impaired stock-recruitment relationships. Our results highlight the importance of building time series of reproductive potential variables other than SSB, such as fecundity.

Study area
The study was carried out on the Flemish Cap in the northwest Atlantic, between 46°N and 49°N and 44°W and 46°W (Fig. 1). It is separated from the Newfoundland shelf by the Flemish Pass, a channel with depths in excess of 1100 m, which hinders the migration to and from the Grand Bank for most of the fish species inhabiting the Flemish Cap, including S. fasciatus. The Flemish Cap is a dome-shaped, deep-water mountain, with a total area of 17.000 square miles up to 1460 m and 10.555 square miles up to 730 m, with the shallowest part of the bank (120 m depth) located in the southeastern quadrant.

Data collection, histology and ovarian processing
Ovaries of S. fasciatus were collected from the EU Flemish Cap survey conducted annually in June/ July since 1988 as part of the European Union sampling programme with the participation of Spain and Portugal. For each fish, fork length (FK), total weight (TW), gutted weight (GW) and maturity stage were recorded on board. Otoliths were removed for further age determination.
Ovaries were preserved in 4% buffered formaldehyde and then weighed in the laboratory. Ovary sections of 0.5 cm thickness from the central portion of the gonad were embedded in paraffin based on conventional histological processing. Sections of 3 µm were stained with haematoxylin and eosin protocol. The ovarian developmental phase, as described in Brown-Peterson et al. (2011), was determined under a microscope.
Bottom temperature was obtained from the Copernicus Marine Environment Monitoring service (https:// doi.org/10.48670/moi-00021) for the Flemish Cap area (between 45°N to 49°N and 47°W to 43°W) for July within the period 1996-2020 and for sampling depths between 300 and 600 m.

Fecundity estimation and image analyses
Fecundity was estimated in ovaries with the presence of advanced vitellogenic oocytes and no signs of postovulatory follicles and/or fertilization. Because Sebastes species are group-synchronous with a determinate oocyte recruitment mode, this single leading cohort of oocytes is considered representative of the potential fecundity (Murua and Saborido-Rey 2003). A total of 281 ovaries were selected according to the above criteria between 1996 and 2020 (Table 1). Ovary weight was recorded and subsamples of about 0.15 g were taken from the central ovary section. Several studies have shown no significant differences in mean follicle diameter and abundance between and within ovaries (Nichol and Acuna 2001, Kennedy et al. 2007, McElroy et al. 2013. Then, oocytes of each subsample were washed and separated from the connective tissue throughout sieving (150 µm, 300 µm and 600 µm).
Potential fecundity was estimated using the autodiametric method (Thorsen and Kjesbu 2001). This method relies on a relationship between mean vitellogenic oocyte diameter (OD) and oocyte packing density (OPD). Once this relationship is attained, fecundity is obtained by estimating the mean OD of an ovarian subsample and then converted to OPD to scale up to the weight of the ovary.
To build the autodiametric calibration curve, 115 ovaries were used. Oocyte counts and measurements were carried out using the software Leica LAS and the images were taken with a Leica Z6 APOA macroscope using a Leica DFC 490 camera. Each subsample was divided into 2 to 3 portions, and each one was analysed separately to facilitate the image analysis. The oocytes were counted and measured using a macro developed by Lucia Sánchez-Ruiloba (IIM-CSIC) in Image J software (https://imagej.nih.gov/ij/). Oocytes depart-ing from sphericity were not considered for estimating average OD in each ovary, but they were counted to determine the final oocyte density (number of oocytes/g of ovary tissue) in each ovary.
The autodiametric calibration curve is based on the principle that OPD is inversely proportional to OD with a power relationship (Thorsen and Kjesbu 2001): where a and b are equation constants.
To improve the fit of the autodiametric calibration curve, OPD and OD data of S. mentella and S. norvegicus collected in the Irminger Sea and Iceland (Witthames et al. 2009, Saborido-Rey et al. 2015 were also used, and results among species were compared before the data were merged. Finally, the potential fecundity of 281 S. fasciatus females on the Flemish Cap was estimated by obtaining the OD of each ovary using the image analysis described above and applying the autodiametric calibration curve: where OW () is the ovary weight (g) of each female analysed.

Maternal traits
The Fulton condition index (K) and the gonadosomatic index (GSI) were calculated as follows: (3) where GW represents gutted weight (g) and FL is fork length (cm) recorded for each female. Age (yr) and GW (g) were also recorded for each female for further analyses as an explanatory variable in the models and to estimate relative fecundity.

Statistical analysis
Generalized linear models (GLM) were fitted to examine the relationships between the reproductive investment (absolute and relative potential fecundity) and maternal traits (length, age and fish condition).
When bottom water temperature was included, generalized linear mixed models (GLMM) were used to analyse the effect of female traits on their reproductive output. Models were fitted using length, age, K and GSI as fixed effects, haul as a random effect and water temperature as a random slope to allow the relationship with bottom temperature to differ by year. Water temperature data were unavailable for numerous coordinate-year combinations, hindering possible water bottom temperature effects on potential fecundity relationships. However, a dataset covering 15 years was obtained. The reproductive output was analysed as follows: Reproductive output t,i = α + FL or A + K + GSI + TBTM + a i + ε t,i where reproductive output is the absolute and relative potential fecundity in year t and haul i , α is the intercept, FL is the fork length, A is age, K is Fulton's condition factor, GSI is the gonadosomatic index, TBTM is the bottom temperature at deep habitat range (300-600 m), a i is the random intercept allowing for variation between years, and b i is the random intercept allowing for variation between hauls. The residuals ɛt,i are a normally distributed random error with mean 0 representing the within-year and haul variation. To avoid collinearity due to fish length and age correlation, they were used in separate models: the models were fitted for absolute and relative fecundity using age and length separately. Haul and Year were included as random effects to correct for the non-independence of reproductive output from the same year and haul. Thus, we evaluated the effects of how these maternal traits and water bottom temperature affect potential fecundity. GLMM were fitted using negative binomial mean variance with a "log" link function. Diagnostic plots testing residual homogeneity, independence and normality and the Akaike information criterion (AIC) were used for model validation (Supplementary material, Tables S1, S2). We avoided transforming the response variable as long as possible using a negative binomial distribution. First, Poisson distribution was used in all the models because of the nature of the response variable (count data). However, high overdispersion values were obtained, so negative binomial distribution was used to avoid overdispersion problems (Zuur and Ieno 2013). Variance inflation factor was calculated in each model to test for collinearity between independent covariates. All statistical analyses were performed with the statistical software R4.0.1 (R Core Team 2020) and using the glmmTMB package (Brooks et al. 2017).

The autodiametric method
The estimated autodiametric relationship between S. fasciatus oocyte density (n/g) and mean OD was significant (p<0.001, r 2 =0.80, n=108). No significant difference was detected (df=255, P=0.133) between the autodiametric curves of S. fasciatus on the Flemish Cap, S. norvegicus in Iceland and S. mentella in Iceland and the Irminger Sea (Fig. 2). The autodiametric curve with all species combined (p<0.001, r 2 =0.88, n=256) was the following: We then used this curve to estimate the potential fecundity from OD and ovary weight for S. fasciatus on the Flemish Cap.

Influence of female traits on fecundity
Four maternal traits (fork length, age, GSI and K) were used to study their influence on fecundity. To avoid using age and length together, two separate models were built. The resulting two GLM models explained 75% and 62% of potential fecundity (Table 2) using length and age, respectively. In the case of relative fecundity, the two models explained 47% and 50% of the variation (Table 2). Our results show that potential fecundity increased significantly with age and size. Interestingly, relative fecundity also increased with those female traits, indicating a disproportionally higher fecundity at larger sizes and older ages (Fig. 3).
Females with higher K and constant GSI (1.58, the average value of the time series) had a higher potential fecundity than fish with lower K (Fig. 3). For example, for a length of 34 cm (the average of the mature stock), the potential fecundity varied between 33707 oocytes with a K=1.2 compared with 47512 oocytes with a K=1.6 and 56409 oocytes with a K=1.8, i.e. an increase of 41% and 67%, respectively (Fig. 3A). Similarly, for a female at 15 years old, the predicted potential fecundity for all three scenarios of K was 34544 oocytes for K=1.2, 40739 oocytes for K=1.6 and 44242 oocytes for K=1.8 (Fig.  3B). Thus, potential fecundity of females in poorer condition was notably lower.
However, relative fecundity did not increase significantly with condition. For a fixed length of 34 cm, the relative fecundity was 71 oocytes g -1 body weight for K=1.2, 76 oocytes g -1 body weight for K=1.6 and 79 oocytes g -1 body weight for K=1.8, i.e. a difference of 7% and 11%, respectively. Moreover, for a fixed age of 15 years, relative fecundity was 72 oocytes g -1 body weight for K=1.2, 76 oocytes g -1 body weight for K=1.6 and 78 oocytes g -1 body weight for K=1.8, i.e. a difference of barely 5% and 8% ( Fig. 3C and D).

Interannual variation of fecundity
Interannual variation in potential fecundity was examined by comparing potential relationships between fecundity and maternal traits between 1996 and 2020 (Fig. 4). Fork length and age showed a significant effect on potential fecundity in all years analysed (p<0.001), and the optimal model showed a significant year effect (p<0.001). However, the post hoc Tukey test showed that the fecundity variation between years was caused by only a few years (Tables S3, S4), mostly 2010, a year with a low sample size. Figure 5 shows the fecundity variation for a 34 cm female. Potential fecundity showed generally higher values at the beginning of the time series, an average of 48500 oocytes between 1996 and 2001 and four years with fecundity above 50 thousand oocytes. Later, fecundity decreased to an average of 42000 oocytes for the rest of the times series (except 2010). During this period, fecundity was below 45000 oocytes in all years except 2010, with particularly low values in the latest years (2015)(2016)(2017)(2018)(2019). The year with the highest (2000) fecundity for a fixed size of 34 cm and age of 15 years showed 1.8-fold greater fecundity rates on average than the year with the lowest fecundity (2015) (not considering 2010).
The analyses with relative fecundity yielded similar results to those with absolute potential fecundity. Optimal models included length, age and year, which explained 22% and 34%, respectively (Supplementary Table 5 and Supplementary Table 6). For a 34 cm female, relative fecundity ranged between 62 and 139 oocytes g -1 , showing a very similar pattern to potential fecundity, with higher values before 2002 (mostly above 90 oocytes g -1 ) and lower values thereafter (mostly below 80 oocytes g -1 ).

The role of bottom water temperature in fecundity
The range of bottom temperature in which females were sampled varied between 3°C and 5°C, with the highest frequencies in a narrow range between 3.5°C and 4°C, i.e. 70% of females were sampled in a range of 0.5°C (Fig. 6). Because samples were randomly taken during the survey, this result likely reflects the distribution of females in advanced stage of vitellogenesis.
The potential and relative fecundity increased per degree of bottom temperature water (Fig. 6 A, B). The median of potential fecundity increased from 30743 oocytes in 3°C to 45000 oocytes in 4.5°C, i.e. by 31%. Similarly, relative fecundity increased from 64 oocytes g -1 body weight at 3ºC to 85 oocytes g -1 body weight at 4°C, i.e. an increase of 24%. The two GLMMs fitted (using length and age) showed that fecundity-at-length and at-age increased with temperature (Fig. 6C, D). However, only the model using age showed a significant positive relationship between bottom temperature and potential fecundity in the age model (Table 3 and  Table 4).

DISCUSSION
Our findings provide empirical evidence of autodiametric curve stability, indicating that the autodiametric method for estimating fecundity originally developed in cod (Thorsen and Kjesbu 2001) can be applied in North Atlantic Sebastes species.
This study demonstrated no significant differences in autodiametric curves between three species of Sebastes on the Flemish Cap, in Iceland and in the Irminger Sea. Likewise, no significant differences were obtained between autodiametric curves from different stocks in the northeast Arctic, the northern Gulf of St. Lawrence and Georges Bank (Thorsen and Kjesbu 2001, Lambert 2008, Alonso-Fernández et al. 2009). For example, for a fixed diameter size of 800 µm, the oocyte density varied between 3174 oocytes in the Flemish Cap autodiametric curve and 3303 oocytes in the Irminger Sea autodiametric curve, a difference of 4%. Lambert (2008) found a similar difference of less  Table 3. -Parameters of the optimal GLMM using potential fecundity as the response variable and including length, condition factor (K), gonadosomatic index (GSI) and bottom temperature from 21 years (N=280 observations) as explanatory variables. SD, standard deviation; SE, standard error. R2LMM(m) describes the proportion of variance explained by fixed effects alone and R2LMM(c) describes the proportion of variance explained by fixed and random effects combined.   0.778 than 6.5% in oocyte density estimated with different calibration curves for two cod stocks, concluding that they were essentially the same curve. The use of OPD and the success of the autodiametric method could vary between areas, stocks and species (Dominguez-Petit et al. 2018)spatial differences in the autodiametric calibration curve were observed in the Northwest Atlantic, but did not translate into differences in fecundity at length. This is the first time that spatial differences between ACCs of the same species have been reported, what could be the result of (i, for reasons such as energy allocation and preservation techniques (Friedland et al. 2005).Thus, fecundity estimations could be inaccurate when published calibra-tion curves not estimated for the species or a stock of interest are used (Witthames et al. 2009).
In this paper, we have studied the fecundity of S. fasciatus on the Flemish Cap for the first time, building a twenty-year time series between1996 and 2020; such long time series in fecundity are rarely seen in the literature. Mean potential fecundity and mean relative fecundity were 36000 oocytes per female and 78.17 oocytes/gram female respectively. These results are in accordance with the fecundity reported for S. mentella in the Irminger Sea (Saborido-Rey et al. 2015). Our study shows annual changes in potential fecundity between several years of the time series, as other reported in species of the genus Sebastes (Beyer et al. 2015).
We have shown that larger, older and better-conditioned fish produced more offspring in both absolute and relative terms than smaller individuals. Therefore, SSB may not be an accurate metric for the reproductive potential of stocks with a different demographic composition. The relative fecundity-age relationship suggests that there is a significant effect of repeat spawners in S. fasciatus stocks and highlights the importance of maintaining a strong length/age population structure. Similar results have been reported in several species, such as cod (Blanchard et al. 2003, Yoneda and Wrigth. 2004, Mion et al. 2018. We have also shown significant variation in fecundity between years. It is well known that fecundity, like many other life-history traits, is highly variable between stocks, geographic areas and/or years (Kraus et al. 2000, Marteinsdottir and Begg 2002, McElroy et al. 2013. Nevertheless, fecundity is still mostly ignored in the monitoring programmes. As a consequence, population egg production is rarely estimated for assessment purposes, or if estimated a constant fecundity-at-length or at-age relationship is used. An increase in potential fecundity with female size was observed in other Sebastes species, including Sebastes melanops, S. goodei, S. entomelas, S. flavidus and S. atrovirens (Berkeley et al. 2004, Sogard et al. 2008, Dick 2009). However, our results show an increase in reproductive potential with size and age and the importance of using indexes other than SSB to measure stock reproductive potential. This finding has ben reported in S. mentella and S. norvegicus (Saborido-Rey et al 2015), where the exponent of the fecundity-length power function differed significantly from 3. It is important to highlight that we used females with ovaries showing advanced vitellogenic stages, as down-regulation of fecundity has been shown to drastically modify fecundity during the course of vitellogenesis (Saborido-Rey et al 2015). This process is likely driven by fish condition and environment factors (Murua et al. 2003, Armstrong andWitthames 2012).
In line with length and age, Fulton's condition factor and the GSI were only significantly related to potential but not to relative fecundity in our study. In addition, other studies have demonstrated that fish condition has a high influence on potential fecundity, with the result that fish in better nutritional status had a higher fecundity than fish in poorer conditions (Thorsen et al. 2006, Kennedy et al. 2007, Lambert 2008. In this paper, explained variance of fecundity was high when K and GSI were included. The GLM model using fish length, condition factor and GSI as a dependent variable explained 75% of the variability in fecundity, in agreement with an earlier study carried out in cod (Lambert et al. 2008). Considering that the effect of the condition factor can be related to the fact that it intervenes in the final part of oocyte recruitment, i.e. during this phase fish will feed and therefore the condition factor will be a key maternal trait determining fish fecundity. However, a recent study (Beyer et al. 2015) showed that the hepatosomatic index (HSI) was significantly related in four studied species, whereas K was significant in one species. This finding suggests that a more accurate index of fish condition, such as HSI, lipid concentration or muscle water content and prey availability index (Kraus et al. 2002), should be included in future research into maternal effects on fecundity.
In this study, we found a positive relation between potential fecundity and bottom water temperature. Several studies have described water temperature as an important factor that can play a direct or indirect key role in fecundity variation in fish (Kjesbu et al. 1998, Kraus et al. 2000, Lambert et al. 2008. Moreover, bottom temperature, which has been increasing on the Flemish Cap since the 1990s (Colbourne et al. 2018), could generate changes in the way in which S. fasciatus allocates energy to reproduction during the whole time series. For example, Yoneda and Wright (2004) describe spatial and temporal fecundity variation as changes in energy allocation that influence maternal condition. The increasing temperature reported on the Flemish Cap may be one of the causes of the sharp increase in S. fasciatus abundance after several strong year-classes in (González-Troncoso et al. 2022. Although recruitment was poor thereafter, it produced a shift in dominance on the Flemish Cap, where the traditionally more abundant S. mentella declined in favour of S. fasciatus, traditionally considerably less abundant. It is important to highlight that S. mentella has a distribution towards more northern and colder waters than S. fasciatus. Reproduction of other aquatic species can also be affected by variability of environmental factors such as sea surface temperature, which plays an important role in regulating brooding activity in crustaceans (Chang et al. 2021) and barnacles (Román et al. 2022) through the primary productivity.

Potential implications and future directions
Firstly, our findings provide for the first time an autodiametric calibration curve between oocyte mean diameter and ovarian oocyte density in S. fasciatus, which can be applied to estimate potential fecundity in North Atlantic for species of the genus Sebastes. Secondly, our study shows that potential fecundity varies interannually in S. fasciatus, probably a response of maternal effects of individual females to varying combinations of biological and environmental factors. Because maternal effects have been reported in a number of exploited species, we suggest that annual variations in fecundity should be monitored regularly. This would improve stock reproductive indexes and increase our understanding of the processes affecting reproductive success. Our results suggest that developing a better understanding of how maternal effects impact on offspring quality may help to understand recruitment processes, enhance stock assessment models, and ultimately improve our capacity to achieve a sustainable fisheries management. Table S1. -Selection of random effects for GLMM fitted with potential fecundity as a dependent variable. First, optimal random effects were tested. The covariates in all models (i.e. fixed structure) are the maternal traits length/age, , condition factor (K), gonadosomatic index (GSI) and sea bottom temperature.     Table S6. -Summary of the multiple regression models for relative fecundity against age and interannual variation (years) of S. fasciatus on the Flemish Cap bank. GLM NB, negative binomial generalized linear model.