3.1. Sequences analysis information
The total number of sequences before processing was 13 406 675. After processing, 7 369 012 sequences were kept (i.e. 54.97 %). After removing sequences affiliated to chloroplasts and samples with a number of sequences reads inferior to 3 000, 432 samples were kept: 160 DG’s samples, 177 EF’s samples, 41 sediment’s samples and 54 seawater’s samples (27 SW8 and 27 SW022). The swarm clustering produced 1 255 different OTUs, which were identified in GD (1152), EF (1 201), sediments (935) and seawater (1 057 and 972 into SW8 and SW022 fractions, respectively). The mean number of sequences reads by sample was 16 063 ± 332 for all samples: 15 297 ± 478 for the DG, 19 251 ± 446 for the EF, 5 487 ± 203 for the sediments, 15 917 ± 851 for the seawater (10 924 ± 601 for SW8 and 20 910 ± 824 for SW022).
3.2. α-diversity comparison for clam tissues and environmental samples
The effect of the number of sequence reads was significant for observed richness (F = 7.6, P < 0.01) justifying the use of a Rarefied richness for describing α-diversity patterns. Models applied to clam’s tissues microbiota α-diversity indexes (rarefied richness and Shannon indices) highlighted two main factors: Depuration and Tissue (including the seawater fractions), followed by the Period (Fig.2). The Depuration induced a significant decrease of Rarefied richness (F = 4.7, P < 0.01) and Shannon indices (F = 62.2, P < 0.001), while Tissue differences results in an average EF α-diversity that were always significantly higher than those for the GD (Rarefied richness: F = 29.4, P < 0.001; Shannon: F = 14.6, P < 0.001). The two seawater fractions exhibited two clearly different α-diversity, where the bacterial communities from SW8 fraction were always greater than those from the SW022 one (Rarefied richness: F = 211.9, P < 0.001). Nonetheless, on average α-diversity of microbiota associated with the two clam tissues was always significantly lower than those of bacterial communities from sediment and SW8 (post-hoc Tukey test, P < 0.001).
The Period factor was significant for all models, except for the Shannon indices of sediment (F = 1.4, P > 0.05). Overall regarding both indices, average α-diversity of bacterial communities from non-depurated DG and seawater samples, significantly decreased during the studied period, notably between T1 and T3 (post-hoc Tukey test, P < 0.001). In the other hand, non-depurated EF and sediment alpha diversities significantly increased between T1 and T3.
3.3. Order level bacterial diversity from clam tissues and environmental samples
Based on the 30 most abundant bacterial orders, DG, EF, sediment and seawater fractions showed clearly distinct bacterial communities in terms of both abundance and taxa composition (Fig.3). The two fractions of seawater (SW8 and SW022) were dominated by the orders Rhodobacterales (55-70 % in SW022 and 20-30 % in SW8) and Flavobacteriales (12-15% in SW022 and 10-22 % in SW8). However, their composition differed for less abundant orders (inferior to 6% of total abundance). Indeed, SW022’s bacterial community was mostly composed of Oceanospirillales and Cellvibrionales, whereas SW8’s bacterial community was more diverse with Cellvibrionales, Desulfobacterales, Campylobacterales, Pirellulale, Rhizobiales, Verrucomicrobiales, Clostridiales and Myxococcales present in low abundance (inferior to 5%). Sediment bacterial communities were mostly composed of numerous few abundant orders, accounting for more than 25% of the total abundance, while the most abundant orders were mainly affiliated to Desulfobacterales (13-19%), Actinomarinales (6-8%), Campylobacterales (5-7%), Microtrichales (5-7%) and Rhizobiales (4-6%).
The non-depurated clams, DG microbiota was mostly composed in a decreasing order of contribution, Rickettsiales, Mycoplasmatales, Diplorickettsiales, Spirochaetales, Pirellulales, Clostridales, Oceanospirillales and Flavobacteriales. DG microbiota from clams sampled in T1 was characterized by a lower relative abundance of Rickettsiales and a higher relative abundance of Spirochaetales and Clostridiales, while higher relative abundances of Campylobacterales and Flavobacteriales were observed respectively in T2 and T3. One of the most abundant order found in DG microbiota, the Diplorickettsiales, drastically decreased in T3 compared to T1 and T2. In depurated clams, the DG microbiota was globally composed of the same orders as non-depurated clams, but in different abundances. The main changes were observed for Spirochaetales, increasing from around 6% to 20% between the non-depurated and depurated clams, while Pirellulales abundances were divided by 7 in depurated clams.
Although it harbored some taxa (at the order level) in common, EF microbiota of non-depurated clams was clearly different from that of the DG’s. It was mainly composed of Oceanospirillales, Spirochaetales, Rickettsiales, Flavobacteriales, Sneathiellales, Campylobacterales, and also Rhodobacterales, Francisellales, Verrucomicrobiales, Cellvibrionales and Alteromonadales, those last clearly contributing to the differences with DG microbiota. In T1, the EF microbiota was characterized by a higher relative abundance of Rickettsiales and Oceanospirillales and a lower abundance of Alteromonadales, Sneathiallales and Rhodobacterales. A higher relative abundance of Campylobacterales and Flavobacteriales were observed in T2 and T3, respectively. In EF microbiota from depurated clams, the same orders were globally found than in non-depurated clams but in different abundances, notably for Alteromonadales, Francisellales and Pseudomonadales which were present in much higher abundances, whereas Flavobacteriales and Rhodobacterales were reduced.
3.4. Factors influencing Beta diversity
From this point on, results are those of analyses conducted at the OTU’s level. PCA (Fig.4) allow clearly discriminating sample matrix (clams vs sediment and seawater on PC1), seawater fractions (SW8 vs SW022 on PC2) and clam tissues (EF vs DG on PC2). The SW8 bacterial community was more similar to that of sediments, whereas the one from SW022 was more similar to that of the EF.
Homogeneity of variances was previously tested for each sample grouping, revealing that multivariate dispersion was not homogenous (Fig.5) for both DG (F = 17.333, P < 0.001, 999 permutations) and EF tissues (F = 6.893, P = 0.001, 999 permutations). Using the betadisper function, the DG microbiota had a higher heterogeneity of group dispersions than EF. The heterogeneity of dispersions occurred between the depurated samples of L56 in T3 and all the groups in T1 (Fig. 5A). It also occurred between the non-depurated samples of L20 in T2 and all the groups in T2. The dispersions were also heterogeneous between T1 and T3. In the case of the EF’s microbiota (Fig. 5B), the heterogeneity of dispersions implied three groups of clams: the depurated at L20 in T1, the non-depurated at L56 in T2 and all the depurated in T3. The L20 levels of T2 and T3 seemed more similar to T1 (L20 and L56) than to L56. For the depurated clams, all clusters were overlapping, indicating a more similar community over the three periods. However, L56 and L20 of the three periods tended to pull away from each other and the centroid, indicating differences between levels.
Regarding the DG bacterial microbiota (Fig.5a), a high heterogeneity of dispersion, which increased between T1 and T3, was observed. However, it did not prevent us from seeing the dynamic of the microbiota. A clear distinction between depurated and non-depurated clam’s microbiota was observed along the first axis of the PCA for both levels (L20 & L56) in T1 and T2, which was much less the case in T3. For the non-depurated clams, the DG bacterial microbiota in T1 and T2 were very close regardless of the levels, no period or level effects were observed. It is only in T3 that the DG bacterial microbiota detached (especially for L20) from T1 and T2, and that a slight level effect appeared. For the depurated clams, the DG microbiota were more similar between the three periods. Moreover, in that case the level effect was more pronounced in T2 than in T3.
For the EF bacterial microbiota (Fig.5b), compared to the DG microbiota, the distinction between depurated and non-depurated was clearer along the first axis, as clusters did not overlap in the PCA regardless of levels and periods. The EF microbiota behaved differently from the DG ones. In T1, no level effect was visible (as for the DG), but in T2 and T3 the level effect becomes more important than the period effect, with a clear distinction between levels regardless of the time period. The L56 levels were clearly distinct from the other ones in both T2 and T3.
The PCA representing the sediments communities (Fig.5c) showed a level effect but no period effect. Only the L56 samples in T1 appeared to have a distinct microbiota from other L56 samples. Multivariate analysis of variance on the Box-Cox transformed and standardized environmental data reveals a period effect (F = 1.77, P = 0.025, R² adj = 0.069), the level (F = 9.25, P = 0.001, R²adj = 0.18) and their interaction (F = 1.59, P = 0.041, R²adj = 0.062). The seawater’s bacterial community (Fig.5d) was clearly different between the three periods and the two filters. The heterogeneity of dispersion between the groups did not allow a multivariate analysis of variance (F = 3.6108, P = 0.037, 999 permutations).