Aquaculture rearing systems induce no legacy effects in Atlantic cod larvae or their rearing water bacterial communities

Bacterial density and growth potential in the rearing water were related to the microbial carrying capacity

Quantifying the bacterial density in each tank verified that we obtained a higher bacterial load in the systems with added organic material. The bacterial density was, on average, 7.8× higher in the systems with high compared to low bacterial carrying capacity. This difference was particularly evident at 2 (34.8×, Kruskal–Wallis p = 0.0008) and 9 DPH (9.1×, Kruskal–Wallis p = 0.0007) (Fig. 1). The bacterial density increased throughout the experiment for the tanks with low microbial carrying capacity (treatment group MMS−, FTS−), reflecting increased larval feeding and defecation. Contrastingly, the bacterial density was relatively stable over time in the MMS+ treatment and even decreased over time in the FTS+ treatment. When averaging the densities at 11 and 15 DPH within each rearing treatment, we observed that the ‘MMS+ to FTS+’ had a considerable difference in the bacterial density between incoming and rearing water (24.2×). In contrast, this difference was below 8.2× in all other treatment tanks. Such differences in density indicated that some communities were below the microbial carrying capacity of the systems. We thus investigated the growth potential to determine if carrying capacity was reached in the rearing water.

The bacterial net growth potential in the intake and rearing water was quantified as the number of cell doublings after incubation for 3 days¹¹. Generally, the FTS− and MMS− rearing water had net growth potential with an average of 0.2 and 0.1, respectively (Supplementary Fig. 2). In contrast, the rearing water of the FTS+ and MMS+ had a negative net growth potential with averages of −0.2 and −0.06, respectively. In the case of negative net growth potential, the bacterial density decreased during the incubation. A negative net growth potential suggested that the rearing water bacterial communities were at the tank’s microbial carrying capacity at the time of sampling. Thus, the bacterial communities were at the carrying capacity of the high (+) carrying capacity systems and below in the low (−) systems. To gain a deeper understanding of the bacterial community characteristics the 16S rRNA gene of the bacterial community was sequenced at 1 and 9 DPH.

Initial rearing condition did not leave a legacy effect on bacterial α-diversity

The bacterial α-diversity of the rearing water was investigated at 1 and 12 DPH (Fig. 2). At 1 DPH, the richness was comparable between the FTS−, FTS+ and MMS+ treatments, but on average, 1.5× higher for the MMS− treatment (307 vs 205 ASVs, Tukey’s test p < 0.006). The diversity of order 1 was, on average, 1.5× higher for the MMS+ and MMS− treatments than for the FTS+ and FTS− treatments (ANOVA p = 0.05).

We were interested in determining whether the initial rearing system had a legacy effect on α-diversity. We first evaluated whether there were differences between the unswitched treatments at 12 DPH. For the high carrying capacity treatments, the MMS+ had, on average more ASVs than the FTS+ treatment (275 vs 182 ASVs, Tukey’s test p = 0.04). For the low carrying capacity group, the MMS− group had, on average fewer ASVs than the FTS- treatment (330 vs 356 ASVs, Tukey’s test p = 0.9). Note that statistical tests with data from 12 DPH have low power (n = 2 replicates/group). Comparing the switched tanks to those that continued with the initial treatment showed that ‘FTS− to MMS−’, ‘FTS+ to MMS+’ and ‘MMS+ to FTS+’ had a more similar richness to their post-switch treatments. Only the ‘MMS− to FTS−’ treatment had a more similar richness to the initial treatment. However, only 25 ASVs, on average, differentiated MMS− and FTS−. We thus conclude that the initial rearing treatment did not leave a legacy effect on richness. Similarly, there was no indication that the initial rearing treatment had a legacy effect on the diversity of order 1.

However, we did observe that the richness had increased in all treatments, except in the tanks continuing with FTS+. The increase in richness was similar in the tanks with low carrying capacity (FTS− and MMS−) regardless of whether the tank changed water treatment system or not. However, for the tanks with FTS+ as the initial treatment, the richness decreased 0.88× in the tanks continuing with FTS+ but increased 1.2× for tanks that switched to the MMS+ system. Interestingly, the opposite was observed for the tanks starting with MMS+. For these, the richness increased 1.3× in the tanks continuing with MMS+ but was stable for those that switched to FTS+ (1.0×). There were few differences in diversity of order 1 between the switched and unswitched treatments at 12 DPH. However, the diversity of order 1 had increased in all treatments, except in the tanks starting with the MMS+ treatment.

We interpret the increases in α-diversity as indicating that the bacterial communities were unstable at 1 DPH, thus allowing the inlet bacteria to disperse and establish. Notably, the decrease in diversity of order 1 in the tanks starting with MMS+ suggests that these bacterial communities were stable, more even, and resisted the establishment of the microbiota from the new intake water source (e.g. ‘MMS+ to FTS+’ had stable richness, and decreased 0.7× in diversity of order 1). The stability of the MMS+ bacterial communities was also supported by the β-diversity.

The MMS+ rearing bacterial community was most stable over time

The differences in bacterial community composition between samples were quantified using Bray–Curtis and the weighted UniFrac distances and then ordinated using PCoA (Fig. 3). The PCoA ordinations indicated that most of the differences in community composition were explained by sampling day and rearing treatment (Fig. 3a,b). The MMS+ samples clustered oppositely from the other three rearing treatments along Axis 1 at 1 DPH. Axis 1 explained 39.2% (Bray–Curtis) and 53.9% (UniFrac) of the variation in the distance matrixes, indicating that there was a large difference in community composition between MMS+ and the other treatments. At 1 DPH, the FTS+, FTS− and MMS− clustered together in the Bray–Curtis ordination but were more spread out when using the weighted UniFrac distance. As UniFrac is based on phylogenetic community dissimilarity, this spreading indicates that the ASVs that contributed to community differences between treatments were more different phylogenetically.

At 12 DPH, the differences in the bacterial community composition were separated based on the microbial carrying capacity along Axis 2. This axis explained 20.8% (Bray–Curtis) and 19.8% (UniFrac) of the variation. Moreover, we observed that all 12 DPH samples clustered closer to the 1 DPH MMS+ samples regardless of rearing treatment. This pattern indicated that succession drove the communities toward a common bacterial community composition. The MMS+ samples had already obtained this composition at 1 DPH, highlighting the advantage of pre-feeding the biofilter to acquire a stable microbial community composition.

The stability of the bacterial community composition was investigated by quantifying the within-system Bray–Curtis similarity within and between sampling days (Fig. 3c). The tanks starting with the MMS+ treatment had the highest bacterial community similarity when comparing 1 and 12 DPH with an average Bray–Curtis similarity of 0.4. In comparison, the Bray Curtis similarity was, on average, 0.1 in tanks starting with the other treatments (Kruskal–Wallis p < 0.001).

Next, we evaluated if the initial rearing condition had left a legacy effect on community composition. We compared the Bray–Curtis similarity at 12 DPH between switched and unswitched communities. Unfortunately, we could not perform statistics on these observations due to low power within the groups. The ‘FTS− to MMS−’ bacterial communities had an average Bray–Curtis similarity (± SD) of 0.4 (± 0.08) and 0.6 (± 0.05) to the communities of the MMS− and FTS−, respectively. The ‘MMS− to FTS−’ samples showed a similar pattern, with slightly higher similarity to communities continuing with the same initial treatment with average Bray–Curtis similarities of 0.6 (± 0.03) and 0.5 (± 0.01) to the MMS- and FTS- treatments, respectively. Thus, some legacy effects on the bacterial community composition might have established in both the MMS- and the FTS- tanks, but these effects were minor. Clearer patterns were observed in the conditions with high carrying capacity.

The bacterial communities switching from ‘MMS+ to FTS+’ resisted a change toward the FTS+ community structure. Instead, these ‘MMS+ to FTS+’ communities had higher Bray–Curtis similarities to the communities continuing with the MMS+ treatment (0.5 ± 0.1) than tanks that initially got the FTS+ treatment (0.2 ± 0.05). This is an indication of a legacy effect in the MMS+ rearing tanks. However, we observed the opposite for the ‘FTS+ to MMS+’ communities, which had higher Bray–Curtis similarity to the MMS+ communities (0.7 ± 0.06) than those continuing with FTS+ (0.4 ± 0.07). Thus, there was no legacy effect in the FTS+ rearing tanks. Due to the inconsistent patterns, we conclude that the initial rearing condition does not leave a legacy effect on the bacterial community composition. Instead, the mature biofilter (MMS+) supplied a bacterial community that was able to establish quickly in the tanks that previously were FTS+. To evaluate if the MMS+ biofilter seeded a bacterial community, we investigated the taxonomic composition of the samples.

The bacterial community composition in the MMS+ rearing tanks differed taxonomically from those of the other treatments

The class Gammaproteobacteria dominated the rearing water in all treatments with an average relative abundance of 76 (± 11% SD). At the order level, we observed differences based on sampling day and rearing treatment (Fig. 4). At 1 DPH, the FTS−, FTS+ and MMS− were similar in bacterial composition, with a high abundance of Alteromonadales. The composition was different in the MMS+ rearing water, with substantially lower abundances of Alteromonadales and high abundances of Thiotrichales. At 12 DPH, the abundance of Thiotrichales had doubled in the MMS+ treatment from an average of 24% to 50%. Interestingly, this order also increased in the rearing tanks that switched from ‘FTS+ to MMS+’. Its abundance was 56% in the ‘FTS+ to MMS+’ tanks but only 17% in the FTS+ tanks. This noteworthy difference in abundance indicated that the biofilter community was effectively seeded to the rearing tanks. Next, we investigated if the rearing treatments affected larval viability.

The present rearing treatment had the largest effect on larval performance

Comparing the larval dry weight between the treatments at each sampling day did not indicate that the rearing conditions affected the growth (Supplementary Fig. 2). At 17 and 18 DPH, there was no statistically significant difference between the average weight in the different rearing treatments. However, differences were observed in larval robustness.

The robustness of the larvae was investigated in side experiments on 8, 11 and 17 DPH by inducing stress through transfer or exposing the larvae to rearing water invaded with a Pseudoalteromonas and a Polaribacter bacterial strain (Fig. 5). While Polaribacter has been identified as a commensal^16,45, Pseudoalteronomonas contains many pathogenic strains towards Atlantic cod⁴⁶. The invaded rearing water thus pose a threat both through an increased bacterial load and exposure to a potentially pathogenic bacterium. Larval mortality was recorded 24 h after the challenge. Not surprisingly, the survival was higher for the larvae only challenged by transfer (mean 68.1 ± 21.2%) compared to larvae transferred to invaded rearing water (mean 20.5 ± 24.8%).

For the larvae that only were subjected to the transfer challenge, differences were observed between the rearing treatments. On average, the survival of larvae was comparable between the FTS−, FTS+ and MMS− treatments but was 1.5× lower for the MMS+ (Fig. 5). Generally, there was no indication that the initial rearing condition affected the general stress level of the larvae. Instead, robustness appeared to be related to the present rearing regime. For example, on 17 DPH, the larvae that continued with MMS+ had 2.1× higher survival than those that switched to FTS+ (i.e. ‘MMS+ to FTS+’). Thus, the initial rearing condition left no legacy effect on the general stress level of the fish.

For the invasion challenge, the larvae from tanks with low carrying capacity were the least robust. For these larvae, the mean (± SD) survival was 6.3 (± 8.6)%, and some flasks had 0% survival. In comparison, the larvae from tanks with high carrying capacity had a mean survival of 39.4 (± 26.8)% after invasion stress (Fig. 5). The data from the challenge tests did not indicate that the initial rearing condition left legacy effects on the larval robustness. For example, larvae from tanks that continued in MMS+ challenged with invasion had high survival [mean 69.4 (± 20.2)%], whereas larvae from the tanks that switched from ‘MMS+ to FTS+’ had 3.5× lower survival [mean 19.8 (± 16.0)%]. Unfortunately, we do not have samples from the FTS+ rearing treatment after 8 DPH due to high mortality in the rearing tanks. If there was a legacy effect, one would expect improved robustness to invasion when switching to a rearing regime associated with higher survival. Furthermore, the larval survival after the challenges was comparable between the FTS− and ‘FTS− to MMS−’and between the MMS− and ‘MMS− to FTS−’. In conclusion, there was no indication of a legacy effect in the larvae. Instead, the post-switch rearing treatment had the largest impact.

Larval survival was very low in FTS+ tanks

Larval survival at the end of the experiment was comparable and relatively high for the MMS+ , MMS− and FTS− treatments. In these treatments, the survival ranged between 12 and 26%. However, survival was low for all tanks that at some point received FTS+ water, ranging from 0 to 7% (Fig. 6). It should be noted that the water quality was visually poorer in the FTS+ tank water. Nevertheless, we investigated if any ASVs were linked to survival.

We identified ASVs with significant log-fold changes between the bacterial communities in high and low survival tanks using a DeSeq2 analysis. Fifty-two ASVs had higher abundances in the communities from tanks with low survival, and 85 had higher abundances in those with high survival (FDR adjusted p-value < 0.05, Supplementary Fig. 3). An interesting pattern emerged when investigating the abundance of the identified ASVs in each rearing tank (Supplementary Fig. 4). At 1 DPH, the abundance of ASVs associated with low survival was over 40% in FTS+, FTS− and MMS− but below 20% in the MMS+ tanks.

When comparing switched and unswitched treatments at 12 DPH, it was apparent that the abundances of these low survival-associated ASVs were treatment dependent. For example, the abundances of these ASVs were 55% in the FTS+ treatment but 3.7× lower in the ‘FTS+ to MMS+’ treatment. The opposite was observed between MMS+ and ‘MMS+ to FTS+’. The low survival associated ASVs were only present at 1% in the MMS+ rearing tanks but increased to 15% in the ‘MMS+ to FTS+’ tanks. Furthermore, we found five ASVs classified as Moritella to be especially interesting. These five ASVs all had over a 7.5-log2 fold increase in the low survival tanks. Four of these ASVs were most similar to the type strain Moritella viscosa, a known fish pathogen (Supplementary Table 3, similarity > 92%). Our findings show that the rearing conditions can be used to select for a beneficial microbial environment for the larvae.

Source: Ecology - nature.com