Introduction

Aquaculture breeding programmes are widely applied as a response to societal demands for conservation and exploitation of economically important fish populations (Naish et al. 2007; Fisch et al. 2015). In supportive breeding, a primary and also the most challenging goal is the long-term preservation of genetic diversity possessed by wild founders and the avoidance of hatchery selection (Frost et al. 2006; Fraser 2008; Neff et al. 2011; Withler et al. 2014). The establishment phase of the captive broodstock determines the genetic characteristics of the parental fish that will be transferred to the subsequent hatchery generations. Here, appropriate mating practices play a major role.

For wild-caught spawners, fertilizations need to be typically performed randomly without information about the genetic relationships of the individuals (Fisch et al. 2015). Then, particularly for small and endangered populations, a (full-)factorial mating design is feasible, as it can maximize the amount of genetic diversity produced (Busack and Knudsen 2007; Neff et al. 2011). Owing to a typically small number of founders, however, genetic variation may be low already from the beginning of the hatchery-based propagation. The remaining diversity can be reduced even further by hatchery operations that impair fertilization success or that lead to unintentional among-family differences in embryo and juvenile survival.

When gametes are collected from wild fish, often from remote areas into a central hatchery, the logistics and practical constraints may not allow for immediate fertilizations of all the gametes, but they are typically conducted at planned intervals, for example, once per week (e.g. Huysman et al. 2019). Secondly, especially in small populations, the capture rate of spawners may be generally low and also asynchronous between the sexes. Thirdly, the readiness for spawning (ovulation) typically involves considerable temporal variation among females. For these reasons, multiple factorial fertilizations may require some individuals’ gametes to be stored in vitro for prolonged periods. Although the storage of gametes may enable simultaneous fertilizations for different females and males, it potentially has a negative consequence on the production of alevins and juveniles due to over-ripening of eggs and decreased quality of sperm. In salmonid eggs, for example, the time required for the significant impairment of fertility during post-stripping storage has been reported to vary from few hours to few weeks depending on the species and the storage temperature (Withler and Morley 1968; Poon and Johnson 1970; Niksirat et al. 2007a,b; Ginatullina et al. 2018).

The existence of critically endangered landlocked Atlantic salmon (Salmo salar m. sebago) in the Vuoksi water system, Finland, has been entirely dependent on supportive breeding and stockings of 2- and 3-year-old smolts since the 1970s when the last natural spawning areas were destroyed due to damming of rivers for hydroelectricity (Pursiainen et al. 1998; Hutchings et al. 2019). Since 1983, a new year class of captive broodstock has been annually established as a statutory responsibility from the gametes of spawners captured from the River Pielisjoki and, to a lesser extent, from the River Lieksanjoki (another original spawning area). In recent years, the gametes of wild-caught spawners have also been directly used for production of 2–3-year-old smolts that are released in these two rivers. During the years 1975–2021, the average effective population size observed in the River Pielisjoki has only been 33. Further, the sex ratio of the spawning stock has been strongly biased towards females (74% of all captured fish). To secure the preservation of extremely low genetic variation within the broodstock (Vuorinen 1982; Koljonen et al. 2002; Tonteri et al. 2005), several batches of full- and half-sib families are produced annually using factorial breeding designs. Depending on the availability of spawners, the fertilizations are performed as a full matrix (i.e. in all possible male–female combinations when the number of parents is small) or in factorial mating blocks of varying size. Due to different capture times of the spawners, their unequal sex ratio and variable ovulation rhythm of the females, the gametes are often collected over a long period of time during the spawning season. Generally, all fertilizations of the Finnish landlocked salmon are aimed to be performed within 2 weeks after the gamete collection.

To optimize the efficiency of the supportive breeding programme for the landlocked salmon and other endangered fish populations, consequences of the hatchery practices related to the extended post-stripping storage of gametes need to be clarified. In this study, we investigated whether and to what extent the simultaneous storage of salmon eggs and milt affects some of their ageing-related indicators (ovarian fluid pH and sperm motility) and decrease embryo viability. We performed factorial sperm activation trials (in ovarian fluids) and fertilizations for the same sets of parents at three time points: first immediately after the collection of gametes and then after 7 and 14 days. The applied mating design allowed us to evaluate the effects of female, male and female-male interaction on both sperm motility traits (percentage of static sperm cells and sperm swimming velocity) and embryo viability (proportion of hatched, normally developed offspring) in relation to the gamete storage time.

Materials and methods

Collection of gametes

Ten females (mean body length ± SD = 733 ± 51 mm, mean body mass ± SD = 3977 ± 817 g) and ten males (778 ± 69 mm, 4150 ± 961 g) were captured for this experiment below the Kuurna hydropower plant in River Pielisjoki, Eastern Finland (WGS84 coordinates: 62°42′N, 29°52′E) between 18 September and 14 October 2019 (Online Resource 1). During that period, water temperature decreased from 12.3 to 5.2 °C. The study fish were kept on the capture site (sexes separately) in round flow-through storage tanks (water volume approx. 3.3 m3) which received water from the river. Following routine operations, the readiness of females for spawning was checked once per week, and only the females that had ovulated shortly before stripping were used in the experiment. On 14 October, the gametes from all experimental fish were collected by stripping to round 2.5-l Orthex™ plastic bowl–like containers with airtight lids (eggs buried within ovarian fluid) and 0.5-l Minigrip® ziplock bags filled with 100% oxygen (milt). To avoid contact of the gametes with water or skin mucus, the abdomen of each fish was carefully dried before stripping. In addition, the urinary bladder of each male was emptied before the stripping to prevent the milt sample from contaminating with urine (Ciereszko et al. 2010). Because we aimed to collect most of the ovulated eggs from each female, the eggs formed several cell layers in each holding container. After collection of the gametes, egg containers and milt bags were placed into the styrofoam boxes filled with ice groats and transferred to the Saimaa Fisheries Research and Aquaculture Station (62°5′N, 28°54′E), Enonkoski, Finland, where the fertilizations and egg incubation took place.

In vitrofertilizations and egg incubation

The in vitro fertilizations were performed as two full-factorial mating blocks, where five females and five males were mated in all possible combinations (n = 25 fertilizations per mating block). In total, the design thus generated 50 families of maternal and paternal half-siblings. The fertilizations were repeated at three time points: first at the same day of collecting gametes (14 Oct; within 8 h after stripping), then after 7 days (21 Oct) and lastly after 14 days of gamete storage (28 Oct). According to the established protocol in the captive breeding programme, the egg containers and milt bags were kept on ice in a large quadrangular vessel (350 l) with an airtight lid during the entire storage time. The temperature within the vessel remained at 1.0 °C. A very low temperature is expected to promote the viability of gametes for a prolonged period. The milt bags were filled with fresh oxygen every 3 days, and concurrently the egg containers were gently shaken to prevent the eggs in the surface layer from drying out. To equalize the number of eggs per family for fertilizations, the mass of 50 eggs for each female was first weighed without ovarian fluid (with 0.1-g precision). Thereafter, similar batches of approx. 50 eggs per female were weighed into family-specific plastic cups (three replicates per full-sib family) and fertilized by pipetting 10-µl milt from each male onto the eggs and by adding approximately 2-dl water right after. Rather than equalizing sperm-to-egg ratio or number of sperm among the males, their milt volumes were kept constant and expectedly high enough for a complete fertilization rate. Then, the fertilized egg batches were placed into family-specific 12 × 5 cm incubation compartments with 2 × 20 mm rectangular holes on the bottom. The compartments were divided into nine incubation trays within five flow-through troughs (1–2 trays per trough; Fig. S1 in Online Resource 2). The placement of families was randomized among the compartments. In total, the study design comprised 450 incubation units (50 full-sib families × 3 replicates × 3 fertilization time points). Water temperature followed that of the ambient waterway throughout the study period (Fig S2 in Online Resource 2).

Survival of the developing embryos was followed from fertilization until hatching, and dead eggs and alevins were weekly counted and removed from the incubation compartments. At eyed embryo stage (cumulative degree days of water temperature approx. 300 °C), the embryos were mechanically disturbed by pouring the batches of eggs into a plastic cup, after which the eggs that contained unfertilized or dead embryos turned white and could be removed. When all batches of eggs had completely hatched, on 31 March 2020, the number of embryos with developmental disorders was counted from each compartment (see Fig S3 in Online Resource 2). The proportion of hatched and normally developed embryos out of the initial number of eggs per incubation compartment was defined as the final viability trait. Hence, the viability examined in this study comprises both fertilization success and development of healthy embryos until hatching.

Measurements of ovarian fluid pH and sperm motility

The pH of each female’s ovarian fluid was measured directly from the egg containers (without extra handling) at each fertilization time using calibrated PHM220 LAB pH meter (MeterLab™, Copenhagen, Denmark). Average pH values from two replicate measurements per female, within each time point, were used in further statistical analyses. Samples of milt and ovarian fluids were taken for sperm motility analyses at 0, 7 and 14-days’ post-stripping (each day of fertilization). For each male, the motility recordings of milt samples were separately conducted within each of the five ovarian fluid samples used in fertilizations, using a B/W CCD camera (capture rate of 60 frames s−1) mounted to a negative phase contrast microscope (100 × magnification). After vortexing the sperm samples for 5 s, 0.1 µl of milt from each of the 10 males was added to Leja 2-chamber (chamber height 20 µm) microscope slides (Leja, Nieuw-Vennep, the Netherlands), and then sperm motility was activated by adding 3 µl of ovarian fluid-water (1:1) mixture. Thereafter, percentage of static sperm cells and sperm curvilinear swimming velocity (VCL) were determined 10 s post-activation using computer-assisted sperm analysis (CASA) (Integrated Semen Analysis System, ISAS version 1.2 Proiser, Valencia, Spain). The sperm motility measurements involved three independent recordings (replicates) within each male–female combination (in all three time points).

Statistical analysis

All analyses were performed using linear mixed-effects model in SAS 9.4. (SAS Institute, Cary, NC, USA). Alternative models with different composition of (co)variance parameters were compared, and the final model was selected based on its convergence, estimability of variances and Akaike’s information criterion (AIC).

First, we applied a linear mixed model for repeated measurements (MIXED procedure) with restricted maximum likelihood estimation method to test for the effect of gamete storage time (categorical fixed effect) and female identity (random effect; n = 10) on variation in ovarian fluid pH. After comparing alternative residual structures in the model based on AIC values, the residuals were defined to follow heterogeneous first-order autoregressive correlative structure between the consecutive measurements.

Second, we tested whether the sperm motility traits, i.e. percentage of static spermatozoa and VCL, varied between the different gamete storage times and among the males, females (ovarian fluids) and their combinations (n = 50 × 3 replicates × 3 storage times = 450 measurements per sperm trait). The linear mixed models included a fixed effect of gamete storage time, random effects (intercepts) of male identity, female identity and female × male interaction (interaction term was not included in the model of percentage of static sperm due to estimation problem) and random slopes of time for males (time × male), females (time × female) and female × male interaction (time × female × male). In addition, the random effects of mating block and time × mating block interaction were tested, but these parameters were excluded from the final models as insignificant, based on the AIC values.

Third, we used generalized linear mixed models with logit link function and Laplace approximation method (GLIMMIX procedure) to evaluate the parental influences on the proportions of viable embryos. Unlike the sperm trait models, the viability models were run separately for each gamete storage times, as this allowed us to estimate all variations due to independent female and male effects as well as their interaction. In each model, individual incubation compartments (n = 150) were statistical units and followed a binomial response y/n, where y = number of viable embryos and n = total number of eggs at the beginning of the experiment. The final models included random intercepts for females (covariation among maternal half siblings), males (covariation among paternal half siblings), female × male combinations (i.e. full-sib families) and incubation positions (i.e. trough × tray interaction). The random mating block effect was excluded from the viability model as its inclusion led to estimation problems, due to observed zero variances. We also tested whether the sperm motility traits determined for each female × male combination at each fertilization day (mean of three replicated measurements) predicted embryo viability. Thus, each of the sperm motility traits and its interaction with time were separately tested in the models as fixed covariates. For each fertilization time, the predicted probabilities of viability for different full-sib families and fertilization times were calculated on the scale of the data as the best linear unbiased predictors (BLUPs). Then, the degree of re-ranking among full-sib families (i.e. their change of order in terms of viability) between different fertilization times was quantified using Spearman’s rank order correlation for average family predictions.

To statistically test for the effect of fertilization time on embryo viability as well as the interaction effects between fertilization time and different parental effect terms, we also ran an alternative model for the whole viability data by including fertilization time as a fixed factor. Random slopes of time for each female and female-male combination were defined in the model as interaction terms (time × female and time × female × male). The results from this model are given in Online Resource 3 (Table S3).

The statistical significance of each variance parameter was individually tested using the likelihood ratio test between the full model and a reduced model without the tested effect (Morrell 1998). Each p-value was further multiplied by 0.5, because the asymptotic distribution of the likelihood ratio statistics for a single variance component (constrained to be positive) should be a 50:50 mixture of the χ2(0) and χ2(1) distributions (Self and Liang 1987). Pairwise comparisons between the storage times were performed using Tukey–Kramer’s post hoc tests. A significance level of 0.05 was used for all analyses.

Results

Temporal changes in ovarian fluid pH and sperm motility

The pH of ovarian fluids showed a strong decrease during oocyte storage (F2, 7.92 = 59.64, p < 0.001), and the change involved pronounced heterogeneity among the females (variance: 0.005 ± 0.002; χ2(1) = 7.50, p = 0.003; Fig. 1). The model-corrected pH means were 8.41 ± 0.03 (SE), 8.24 ± 0.02 and 8.01 ± 0.08 at the first, second and third measurements, respectively. Pairwise comparisons revealed that the first pH measurements differed significantly from the second (p = 0.007) and the third measurements (p < 0.001), whereas the difference was only approached significance between the last two measurement times (p = 0.075).

Fig. 1
figure 1

Change of pH in ovarian fluids of landlocked salmon females (n = 10) at three measurement times (0, 7 and 14 days) after collection of eggs

Full size image

The percentage of static sperm cells, measured 10 s after the activation, increased with gamete storage time (F2,19.2 = 25.98, p < 0.001). The model-corrected mean proportion of static sperm was 2.5 ± 3.7% (SE) immediately after stripping, 19.7 ± 3.7% after 7 days and 28.9 ± 3.7% after 14 days of storage. According to pairwise comparisons, the mean percentage of static sperm at the first sampling time differed from both second and third sampling time (p ≤ 0.001, in both cases), whereas the difference between the second and third sampling was only close to significance (p = 0.056). Sperm swimming velocity (VCL) also decreased in respect of gamete storage time (F2,23.5 = 59.94, p < 0.001). Again, the mean motility was significantly higher for the first sampling time (117.9 ± 2.6 µm s−1) compared to the later time points (p < 0.001, in both cases), but no difference was observed between the second (98.4 ± 2.6) and third time point (100.5 ± 2.6) (p = 0.556).

Male and female identity independently affected the variation in VCL, but not in the percentage of static sperm (Table 1). In both sperm motility traits, however, the effect of storage time showed highly significant variation among the males (Table 1; Fig. 2). For VCL, the effect of storage time also differed among females and male–female (sperm-ovarian fluid) combinations.

Table 1 Variance parameter estimates (± SE) and their corresponding significance levels for percentage of static sperm cells and sperm curvilinear velocity (VCL) measured at three sampling times after the gamete collection (0, 7 and 14 days). For % of static sperm, variance parameter male × female became zero and was excluded from the final model
Full size table
Fig. 2
figure 2

Percentage of static sperm cells (upper graphs) and sperm curvilinear velocity (VCL; lower graphs) in different sperm-ovarian fluid combinations at three sampling times (0, 7 and 14 days after collection of gametes). Mean values (± SD) for males from different sperm activation (mating) blocks (A and B) are shown in separate graphs. Within each block, the sperm from all five males were separately activated with the ovarian fluids of all five females

Full size image

Embryo viability

Storage of landlocked salmon gametes had a considerable negative influence on embryo viability (Fig. 3; Online Resource 3). The model mean (± SE) of viable (i.e. hatched and normally developed) embryos was 96.4 ± 0.7% for the incubation groups fertilized immediately after gamete collection, 45.5 ± 6.0% after 7 days and 13.1 ± 4.7% after 14 days of gamete storage. Based on direct counts, the average proportion of deformed (non-viable) embryos out of all hatched ones was 0.4% for the first fertilization group, 3.5% within the second group and 6.9% within the last group.

Fig. 3
figure 3

Female percentage means ± SD (n = 5 half-sib families per female) of viable embryos fertilized at three times after collection of gametes (0, 7 and 14 days). Values for females from different mating blocks (A and B) are shown in separate graphs. Within each block, the eggs from all five females were separately fertilized with the milt of all five male

Full size image

Both female identity and female-male combination (full-sib family) contributed highly significantly to embryo viability at each fertilization time, whereas no male effect was detected in any of the three time points (Table 2). The decrease of viability in response to gamete storage was associated with increased among- and within-female variation, whereas the importance of female × male interaction was highest at the first fertilization time (Table 2; Fig. 3). Like male effect, incubation position was not negligible at the first fertilization time (i.e. inclusion of the parameter led to an estimation problem) but yielded a small yet significant additional variation on viability at the latter two fertilization times (Table 2). Neither of the sperm motility variables predicted embryo viability at any fertilization time and were thus excluded from the final models.

Table 2 Variance parameter estimates (± SE) for embryo viability in landlocked salmon following three different fertilization times (0, 7 and 14 days after collection of gametes)
Full size table

The correlation of full-sib family predictions (average viability probabilities, n = 50 families) was close to zero between the first and second fertilizations (rs =  − 0.034, p = 0.813), indicating nearly virtually complete re-ranking among the full-sib families occurred after 7 days of gamete storage. Instead, the positive correlation between the first and third fertilization times was high (0.589, p < 0.001) and even higher than between the latter two times (0.346, p = 0.014). The variances of average family predictions were notably larger for the second (0.030) and third (0.021) fertilization times, compared to the fertilizations without gamete storage (0.001).

Discussion

Genetic diversity of captive broodstocks is influenced by many management practices, including the operations related to establishment of new hatchery generations. Consequently, appropriate mating designs are of utmost importance in all aquaculture breeding schemes, regardless of whether fish material is propagated for restocking purposes or food production. In the present study, we studied a pronounced dilemma of pursuing genetic diversity by large number of factorial fertilizations in a supportive breeding programme for critically endangered landlocked salmon population, when the gametes of few wild-caught spawners are obtained over a relatively long period of time, and the varying ovulation rhythms of females cannot be controlled by environmental manipulations. We found that an extended period of gamete storage (up to 14 days) at very low temperature (1 °C) has a clear debilitating effect on both sexes’ gamete quality, yet the decrease of embryo viability was predominantly associated with increased variation among and within females. Overall, the observed changes in ageing-related indicators of gametes occurred most strongly during the first week of storage. When fertilizations took place immediately after gamete collection, the proportion of successfully developed embryos was close to 100%. The very high viability after the first fertilizations suggests that in vivo ageing of the gametes before stripping had not affected the results. Following 7 days of gamete storage, by contrast, the overall viability more than halved and also showed considerable re-ranking among the genotypes (full-sib families). Further, after 2 weeks of gamete storage, the viability dropped 83%-unit from the first fertilization time. These findings highlight the general need for establishing the captive broodstocks shortly after the gamete collection to minimize unintentional selection and the loss of genotypes associated with critical early stages of hatchery propagation.

Over-ripening of eggs is considered a limiting factor in artificial propagation of salmonids, and its detrimental influences on fertilization, eyeing, hatching and larval anomalies are widely documented (Barrett 1951; Komrakova and Holtz 2009; Bahabadi et al. 2011; Eide and Barnes 2020; Eronen et al. 2021). A decrease in ovarian fluid pH is known to be associated with post-ovulatory ageing of eggs in salmonids, including Atlantic salmon (Lahnsteiner 2000; Mommens et al. 2015). For example, Lahnsteiner et al. (1999) found that in brown trout (Salmo trutta) the embryo survival until eyed-egg stage was highest (> 80%) when the pH of the ovarian fluid ranged between 8.44 and 8.57. In our study females, the pH values measured right after egg collection supported high viability at the range of 8.24–8.52. Based on a series of experiments on rainbow trout (Oncorhynchus mykiss), Komrakova and Holtz (2009) suggested that the pH in the ovarian fluid should not drop below 7.8, and the oxygen content should be at least 5 mg l−1. Compared to many other studies investigating the effects of salmonid egg storage at low temperature (e.g. Piper et al. 1982; Ginatullina et al. 2018; Eide and Barnes 2019), our storage time of 7 and 14 days can be considered a very long period (although up to 14 days of storage is currently frequently used in practice). Moreover, we kept the eggs in closed containers in order to avoid evaporation but neither renewed their gas atmosphere between the fertilization times nor limited the number of egg layers. Considering these factors, including provision of supplemental oxygen, may promote gamete viability to some extent during the prolonged storage (Stoss and Holtz 1983; Jensen and Alderdice 1984; Komrakova and Holtz 2009). Interestingly, the viability within four half-sib families of one female (B1) increased between the second and third fertilization times, referring to the presence of a potential artefact. One plausible explanation for the finding could be the drying of the surface layer of eggs during the first week of storage, which may have resulted in weak fertilization of the sampled eggs at the second fertilization time. It is also noteworthy that the pH of this female’s ovarian fluid only dropped during the first week of storage, whereas no further change was observed after 14 days. Nevertheless, the ultimate reason for these exceptional results remained unknown.

Immediately after gamete collection, the sperm motility levels were high and only differed among males in respect of swimming velocity. However, both sperm motility traits showed a significant drop from their initial levels within the first week of gamete storage, but VCL no longer decreased between 7 and 14 days. Differential male responses in both sperm motility traits became apparent across gamete storage times. Further, the temporal change in VCL varied among females and male–female combinations, suggesting that impairment of sperm movement during gamete storage was also, and presumably for the most part, dependent on differences in ovarian fluid pH as well as on ovarian fluid-sperm interaction. It is well established that the ovarian fluid of fish can influence sperm performance by modifying the motility of spermatozoa and ultimately determining the fertilization outcomes (see Zadmajid et al. 2019 for a review). Some earlier studies on salmonids have also shown that males’ sperm swimming ability tends to be differentially affected by the ovarian fluids from different females (Urbach et al. 2005; Rosengrave et al. 2008; Butts et al. 2012). Yet, the insignificant overall effect of male × female interaction in our study does not directly support the presumption for such ovarian fluid-mediated regulation.

Neither of the motility traits examined independently predicted embryo viability at any fertilization time. Thus, the sperm motility differences among males did not seemingly translate to variation in the families’ reproductive output. Contrary to our observation, Rosengrave et al. (2016) found in Chinook salmon (Oncorhynchus tshawytscha) that a male’s sperm swimming velocity in the female’s ovarian fluid, together with his genetic heterozygosity level, was positively related to the proportion of viable embryos. In our study, the effect of ovarian fluid on sperm swimming velocity also differed among females, independently of the males’ identities. This is understandable as acidification of the ovarian fluid did involve marked variation among females and is also known to negatively affect sperm swimming ability (Wojtczak et al. 2007).

As expected, the overall parental variation in viability was mostly due to females, whereas the independent male effect remained negligible. The maternally derived viability responses also substantially diverged with fertilization times, presumably reflecting differential impairment among females in their egg quality (Rime et al. 2004; Aegerter et al. 2005; Bahrekazemi et al. 2009; Samarin et al. 2019). Supporting this reasoning, the decrease of ovarian fluid pH over the course of egg storage showed highly variable patterns and increased variation among the females. In eight out of ten females, however, the decrease in ovarian fluid pH occurred more strongly during the first week than during the second week. In the full-factorial breeding design, female effects represent a combination of environmental maternal and direct genetic variation. However, the absence of any male effects (i.e. near-zero variance estimates) on embryo viability suggests that the additive genetic effects were likely weak.

In addition to maternal variation, the embryo viability in our experiment involved a significant interaction effect of parents (full-sib families), irrespective of the gamete storage time. This suggests that the observed viability variation at different gamete storage times also involved non-additive genetic effects. In fact, the relative proportions of maternal and full-sib family variations were even of similar magnitude at the first fertilization time, whereas the maternal variation increased notably in importance at later fertilization times. Previous studies on externally fertilizing fish, including salmonids, have shown that non-additive genetic effects are typically more important determinants of embryo survival than are additive genetic effects (Rudolfsen et al. 2005; Rodríguez-Muñoz and Tregenza 2009; Janhunen et al. 2010; Houde et al. 2013, 2015). It is important to note that although the variance in average full-sib family predictions became more pronounced at later fertilization times, this was mostly due to the increased differences between maternal family groups. Following the first week of gamete storage, the zero correlation of family predictions together with their substantially increased variation indicates the presence of significant genotype-by-treatment interaction in incubation results. The most surprising feature of the present results is, however, that the ranking among full-sib families was relatively consistent (rs = 0.59) between the first and third fertilization times, despite the growing temporal differences of viability within families.

To conclude, our present study underlines the critical role of hatchery propagation practices in genetic maintenance of captive broodstocks. Accordingly, the results on critically endangered landlocked salmon show that post-stripping storage of gametes up to 2 weeks in a chilled state induces strong heterogeneity (selection) in offspring viability among genotypes. This potentially compromises the preservation of already low diversity within the population’s gene pool. The negative influence of gamete storage on embryo viability was strongly associated with maternal differences and to a lesser extent with variation in parental interaction. To secure both the maintenance of remaining genetic variation within the studied population and the production of sufficient fish material for restocking purposes in the future, it would be more important to minimize egg storage time, preferably for a few days at most, rather than to maximize the number of mating combinations in fertilizations.