Sampling was carried out at the beginning (D0) and at the end of the 6 months experiment. For each farm (A, D, F and P), the sediment was sampled in the 42 L pots on D0 (before the culture of S. australis) and at the end of the experiment in the dry condition (Dry) and in sediment with S. australis cultivation. For the chemical analyses, sediment of each farm (A, D, F and P), were sampled in triplicates in different 42 L pots from each modality (D0, dry, S. australis). To ensure sampling homogeneity, each sediment samples consisted of a pool of 6 samples from different pots, homogenized in a clean bucket and then stored in aluminum trays and 50 mL tubes. The collected sediment in aluminum tray was then oven-dried at 35°C for several days for analysis of the pH, total and available forms of phosphorus, total sulfur and organic carbon. The other part of the sediment collected in the 50 mL tubes were stored at −20°C for the analysis of available nitrogen forms (NO₃⁻ and NH₄⁺).

Several publications have used cDNA metabarcoding and proved that RNA is a useful tool to identify living organisms and to perform biological survey and monitoring (Laroche et al., 2018; Amarasiri et al., 2021; Miyata et al., 2021; Veilleux et al., 2021). Indeed, the high turnover of RNA molecules in the environment (from days to weeks) compared to DNA (from months to years) reflect better the metabolically active lineages at the sampling time. Thus, in our research, we opted to extract RNA to investigate the active microbiota in the rhizosphere and sediment, minimizing the chances of detecting microorganisms that may be inactive or dead in the samples. In order to explore the active microbial diversity in the sediment, the top 2–3 cm of sediment from the 42 L pots were also collected aseptically using RNA/DNA free gloves and spatula. For each farm (A, D, F, and P), the sampling was carried out at the same time as sampling for chemical analyses at D0 and in the Dry and S. australis conditions. We collected sample in triplicate and one replicate consist of 3 to 4 sediment samples from the same pot. Then, sampled sediment was transferred into RNA/DNA free 15 mL tubes. The collecting sediment were stored at 4°C during transport to the laboratory and then frozen at - 80°C until further processing.

For each sediment sample, RNA was extracted using RNA PowerSoil Total RNA Isolation Kit (MoBio Laboratories, Inc.) and then reverse-transcripted into complementary DNA (cDNA) as described in our previous study (Colette et al., 2023) using Second Strand cDNA Synthesis Kit (Invitrogen). All cDNAs were sent to MrDNA (Shallowater, Texas, United States) where PCR using the 515f-806R primers couple (Caporaso et al., 2011), barcode indexing and sequencing of the V4 hypervariable region of the reverse-transcripted procaryotic 16S rRNA molecule were carried out. The sequencing was done with an average of 20 k raw reads per sample.

The amplicon analysis was performed with DADA2 version 1.6 package¹ on R software as described in Colette et al. (2023). The chimeras were removed using the consensus method, and the taxonomy was assigned using the Silva 138 SSU Ref NR99 database (Quast et al., 2012). Sequences with no affiliation or affiliated to the Eukaryota, Mitochondria or Chloroplasts were removed from the ASV table, prior to further analysis.

The alpha diversity of each sediment sample was calculated on R software with the microeco (v0.20.0) package (Liu et al., 2021). Then, data were normalized with the Counts Per Million (CPM) as described in Callac et al. (2022, 2023). The beta diversity was investigated by a PCoA (principal coordinate analysis) plot using the Bray distance with the microeco package on R software. Then, a permutation test (PERMANOVA) was performed to highlight Bray distances significantly different between farms (A, D, F and P) and modalities (D0, S. australis and Dry). Venn diagrams were then made to exhibit both shared and specific ASVs between the farm sediments for each of the three different modalities (D0, dry and S. australis). The Venn diagrams were built using the open-source component for the web environment Jvenn². For each modality, stacked bar charts of the relative abundance of microbial communities were made to display the composition of the sediment microbial communities according to the farms sampled. Stacked bar plot were performed on R software with microeco and ggplot (v3.4.2) packages. For the S. australis modality, we performed a Functional Annotation of the Prokaryotic Taxa to predict the putative functions of the specific sedimentary microbial communities on each farm. The functional annotation was based on FAPROTAX v1.2.4 database and was done on R software with microeco package.

At the beginning of the experiment (D0), the chemical parameters of the sediments varied from farm to farm (Table 1). With the exception of sediment from farm D, which had a pH of 7.1, the pH of farms A, F and P was around 8 (Table 1). Sediment CEC values ranged from 23.7 to 30.4, with the highest value observed on farm D.

Farms A and D both had a lower percentage of calcium carbonate (CaCO₃) in their sediments (1.5 to 2.3%) than farms F and P (27.2 to 34.67%). The C/N ratio of farm sediments ranged from 7 to 12 (Table 1). Sediment organic carbon concentration did not vary significantly between farms, unlike NO₃⁻, NH₄⁺, phosphorus and sulfur concentrations. Sediments from farm P had significantly higher concentrations of sulfur, total and available phosphorus, and NO₃⁻ (Table 1). Sediments from farms A and D had significantly lower sulfur concentrations than those from farms F and P.

After 150 days of S. australis cultivation, sediment pH was still significantly lower in farm D than in the other farms (Table 1). For farms A, F and P, S. australis cultivation reduced the NO₃⁻ concentrations by 50 to 99% compared with D0. For farm D only, S. australis cultivation increased sediment NO₃⁻ concentration by 136% (Table 1). For all farms, S. australis cultivation increased the concentration of organic carbon in sediment compared to D0 and decreased the concentrations of phosphorus (total and available forms). Farm D showed the greatest increase in organic carbon (+ 46%) compared with the others farms (+ 15 to 25%). The greatest reduction in phosphorus under S. australis cultivation was found in the sediments of the farms F and P (Table 1). In terms of sulfur concentration, S. australis cultivation increased concentrations on farms D (+36%) and P (+16%) whereas it decreased on farms A (−23%) and F (−62%).

Sediment drying resulted in a pH increase of one unit compared with D0 in all farms except farm D (Table 1). Sediment pH values were around 9 on farms A, F, and P, and 7.3 on farm D (Table 1). NO₃⁻ concentration was reduced by 80% in farms F and P and by 17% in farm A. Conversely, NO₃⁻ and total sulfur concentration in dried sediments from farm D increased by 110 and 45%, respectively, compared to D0 (Table 1). The dried sediments from farm D showed higher NO₃⁻ and total sulfur concentrations than those from the other farms, which were not detected at D0 (Table 1). For all farms, sediment drying reduced the concentration of available forms of phosphorus compared to D0, within a range quite similar to that of the S. australis cultivation (Table 1). Sediment organic carbon increased in dry sediments from farms D, F, and P compared to D0, but the increase was smaller than under S. australis cultivation.

On D0, farms F and P had higher values of richness indices (Chao1 and observed ASV) and evenness indices of Shannon compared to the farms A and D (Table 2). After 150 days of S. australis cultivation, the average of all alpha diversity indices decreased in farms F and P compared with D0, while they increased in farm A and F (Table 2). The highest values of alpha diversity indices (Chao1, Observed ASV, Shannon and Simpson) with S. australis cultivation were found in farm A sediment. For all farms, average of richness indices decreased in the dry sediment compared with D0. In addition, for all farms, the richness indices values in dry sediments were lower than in sediment with S. australis cultivation (Table 2). In the dry sediments, the highest of Chao1 and Observed ASV values were found in farm P.

Our dataset was composed of 20,467 ASVs. The PCoA diagram displayed that the sediment microbiota differed between D0, dry and S. australis, as the points representing these different modalities are clearly separated in the graph (Figure 1). In addition, a differential test of distances among groups performed with a PERMANOVA evidenced that those three modalities were statistically different with a value of p at 0.001 (Table 3).

All modalities confounded, the PERMANOVA also evidenced that farm D sediment microbiota was significantly different (value of p < 0.05) from farms A, F, and P (Table 3). This can be evidenced in the plot for the modalities D0 where samples from farm D were distant from farms F and P (Figure 1). Furthermore, when considering the dry modality in the PCoA diagram, sediment samples from farm D were isolated from other farms.

For the S. australis modality, sediment samples were randomly distributed across the graph (Figure 1). For this modality, there was more distance between replicates of sediment samples, highlighting a greater variability of microbiota within sediments from the same farm. For farm F only, replicates of sediment microbiota samples with the S. australis modality were close to each other (Figure 1).

At the family level, the specific microbiota was varying between the farms (A, D, F and P) (Figures 3A,B). In the farm A, specific microbiota was composed of Balneolaceae (10%), Micromonosporaceae (5%) and Bradymonadaceae (3%). Specific microbiota of farm D was mainly composed of Terasakiellaceae (36%) family which was absent in the other farms (Figure 3B and Supplementary Table S3). In the farm F, the specific microbiota contained members Oscillatoriaceae family (7%) which was totally absent in the other sediments, and Sphingomonadaceae (3%) which was less than 1% t in the other farms. The specific microbiota of the farm F was also composed by lineages related to the Haliangiaceae (6%) and A4b (Fermentative organism found inside of anammox granules) (5%) which accounted for less than 2% in the specific microbiota of the others farms. The specific microbiota of the farm P was composed of Nodosilineaceae (8%) in greater proportion than in other farms, and of Nannocystaceae (5%) which was also present in the farm F (4%) and in lesser proportion (< 1.5%) in the farms A and D. The specific microbiota of the farm P also included the Geothermobacteraceae family (6%), which was absent from the other farms.

Rhodobacteraceae was found in the specific microbiota of all farms, with relative abundance of 3% in the farm D, of approximately 12% in the farms A and F and about 6% in the farm P (Figure 3B and Supplementary Table S3).

In the farm D, the putative functions of the specific microbiota of the sediment with S. australis were significantly and positively correlated with the sulfur cycle (sulfide and sulfur oxidation), ammonia oxidation and aromatics degradation (Figure 4). For farm A sediment microbiota, correlations were significantly positive with hydrocarbon degradation and methanol oxidation.

In the farm P, the putative functions were significantly positively correlated with nitrate ammonification, arsenate respiration and cellulolysis. Other lysis functions, such as chitinolysis, were also found as putative functions in farm F (Figure 4). However, in the sediment microbiota from Farm F, significant positive correlations were also found.

We could observe that at D0 and for all the farms, the Gammaproteobacteria was the main class that dominated the sediment microbiota (39%) followed by the Bacteroidia (15%) (Figure 5A). However, in the dry conditions (Figure 5B), the abundance of those two bacterial classes decreased by a factor two. In the dry conditions the relative abundance of the Polyangia increased compared to D0 (8% compared to 15%) and the Cyanobacteria rise in great proportion (14%) (Figure 5B). The Myxococcia and Anaerolineae found in the dry sediment were absent at D0. In the rhizosphere of S. australis, the relative abundance of Alphaproteobacteria had significantly increased compared to the others conditions and constituted the main bacterial class of the rhizosphere microbiota (32%) (Figure 5C). Cyanobacteria, Polyangia, Bacteroidia and Myxococcia were also present in the S. australis rhizosphere but in a lesser proportion than in the dry sediment.

The growth of Suaeda australis in sediment had two opposite effects on alpha diversity according to the farms. In farms A and D, the alpha diversity index had increased compared to D0 whereas it decreased in farms F and P while they had the highest alpha diversity indices at the start of the experiment (Table 2). A previous study has also reported that cottons roots have different effect on alpha diversity of the rhizosphere bacterial community in two types of soils (Qiao et al., 2017). They showed that in nutrient-rich soils, bacterial diversity in the rhizosphere was lower, because some bulk soil microorganisms were unable to adapt to variations in chemical and physical properties of the soil and to adapt to variation of root exudates. In our study, the decrease in alpha diversity may reflect the plant’s selection of microorganisms from the bulk soil into its rhizosphere. In contrast, in sediments with lower alpha diversity (farms A and D), the release of root exudate may have enhanced the diversity of microorganisms.

The Alphaprotebacteria dominated the specific rhizosphere microbiota of Suaeda australis (Figure 5C). Bacteria of this class have the ability to interact with plants as pathogens, symbionts or non-symbionts (Pini et al., 2011). Other studies have also found that Alphaproteobacteria was the dominant class in the rhizosphere soil (Jorquera et al., 2016). Thus, Suaeda australis may had promoted the occurrence of Alphaproteobacteria in the rhizosphere. However, at the family level, Suaeda australis rhizosphere specific microbial communities’ composition and abundance were varying according to the farms (Figure 3B). For instance, farm D was marked by the predominance of nitrogen-fixing Terasakiellaceae (Alphaproteobacteria class) which was absent in others farms sediment (Figure 3B; Weiler et al., 2018; Espín et al., 2021). In the farms A and F, the rhizosphere microbiota of Suaeda australis had higher relative abundance of Rhodobacteraceae family (Alphaproteobacteria class) compared to the farms D and P. The Rhodobacteraceae family was also reported to be deeply involved in sulfur and carbon biogeochemical cycles (Pujalte et al., 2014). The higher occurrence of this family in the farms A and F may have favored the reduction of sulfur level in sediment compared to D0 (Table 1). A previous study has reported that members of Rhodobacteraceae family are key player in hydrocarbon degradation in Mexico beach sand (Kostka et al., 2011). Thus, higher abundance of taxa from this family in farm A may explained its significant link to hydrocarbon degradation function (Figure 4).

Functions linked to biopolymer lysis were found significantly enriched in the farms F and P with celluloysis and chitinolysis functions. The chitinolysis function is relevant because chitin makes up the shrimp exoskeleton and this biopolymer accumulates in the sediments during shrimp rearing as the shrimps molt or die (Philip and Antony, 2006; Della Patrona and Brun, 2009). Those functions may be linked to the higher abundance of several families related to Myxobacteria (Haliangiaceae, Nannocystaceae, Sandaracinaceae) in farms F and P (Mohr, 2018). Myxobacteria were also found in farms A and D but in a lesser abundance. However, the occurrence of Myxobacteria in the specific rhizosphere microbiota may not be due to the plant influence as those families were found without plant occurrence on D0 and in dry sediment.

In farm F, the specific microbiota of Suaeda australis was significantly correlated with pathogens functions (human pathogens, animal parasites or symbionts) which may be explained by the presence of Alphaproteobacteria from the Sphingomonadaceae family. In fact, some genera of Sphingomonadaceae are known as plant pathogens or to cause human infection (Glaeser and Kämpfer, 2014). The functional assignment of prokaryotic taxa with FAPROTAX is based on information on taxa using standard references (Louca et al., 2016a,b). The FAPROTAX tool is relevant for predicting the functions of prokaryotes related to biogeochemical cycles. It can be applied to the analysis of the microbiome (terrestrial, human, animal), however, it is not specialized in the microbiome of shrimp sediment (Sansupa et al., 2021).

The Desulfuromonadia class was evidenced in the rhizosphere of Suaeda australis in farms A and P. It belongs to the Desulfurobacterota phylum that encompasses sulfate-reducing and fermentative taxa (Murphy et al., 2021). However, the abundance of the families affiliated to this phylum varied within farms, with Geothermobacteraceae and Bradymonadaceae found only in farms P and A, respectively. This evidenced that sediment types had influence the structuration of rhizosphere microbial composition. Previous study has also reported differences in the composition of the rhizosphere microbial communities of Arabidopsis thaliana, lettuces plants or also cotton grown in different types of terrestrial soil (Lundberg et al., 2012; Schreiter et al., 2014; Qiao et al., 2017). In their study, Qiao et al. (2017) reported that the soil background microorganisms were the main cause of the variation of the microbial community in the rhizosphere between different soils. We have highlighted in this study, that at the beginning of the experiment, the microbial diversity and composition varied between the four farms. Thus, as the rhizosphere communities were a subset of soil microbial communities, differences in the initial pool of soil prokaryotic taxa between farms may result in different recruitment of taxa by Suaeda australis. Also, variation of rhizosphere microbial communities between farms may also result from the growth and nutrition of Suaeda australis that could influence the chemical parameters of the sediments. Indeed, we have evidenced that under Suaeda australis cultivation, the pH, nitrogen and sulfur concentrations varied in different ways between the farms (Table 1). For instance, Marschner et al. (2004), evidenced that variation of soil pH value of one unit, can significantly affect the structure of the bacterial communities of Sudan grass rhizosphere, grown in sandy loam soil. In addition, variations in the amount of root exudation according to the farms sediments could also be a factor that influencing the microbial composition of the Suaeda australis rhizosphere (Neumann et al., 2014; Maurer et al., 2021).

Cultivation of Suaeda australis in the sediments seemed effective in removing nitrogen, sulfur and phosphorus but this varied according to the farms (Table 1). The nitrogen concentration was reduced in farms A, F and P whereas the total sulfur was reduced only in farms A and F. For all the farms, the cultivation of Suaeda australis has reduced both the total and available forms of phosphorus (Table 1). In all farms, the cultivation of Suaeda australis had increased the concentration of organic carbon in the sediment compared to D0 (Table 1). This increase may be due to plant litter or to the exudates released by the roots (Yan et al., 2018).

The sediment of the farm D was the only one for which the NO₃⁻ level had increased significantly under both Suaeda australis cultivation and sediment drying compared to D0 (Table 1). In contrast, in other sediments from shrimp farms, nitrogen levels were reduced by 55–99% under Suaeda australis cultivation. In addition, for farm D, the practice of drying or growing halophytes increased total sulfur and organic carbon concentrations by at least 36 and 40% respectively, which was significantly higher than the concentrations observed in the others farms (Table 1). Thus, theses experimental conditions have selected microbial communities that appeared to be not effective for sediment nitrogen, carbon and sulfur removal. Farm D also stood out from the others for its lower pH values, higher ammonia concentration and lower alpha diversity index (Tables 1, 2). The lower pH value observed in farm D at the initial state and in dry conditions compared to the other farms may have not favor nutrient removal by rhizosphere microbial communities. It was reported that soil acidity reduces the intrinsic activities of microbial communities (Kemmitt et al., 2006), this could explain the significantly increased of organic carbon and nitrogen in the dry sediment and with the halophytes (Table 1). Under the cultivation of Suaeda australis the pH of the sediment of the farm D increased but it was still lower than the other farms (Table 1). Thus for farm D, liming the sediment accumulation beforehand halophyte cultivation or drying period may be a way to increase the sediment pH and to improve the organic matter decomposition by microbial communities (Li et al., 2015). In addition to control sediment acidity, the liming application in shrimp sediment can also be beneficial to improve the sediment texture as the porosity (Della Patrona and Brun, 2009). This sediment parameter could also shape the microbial diversity and composition (Chau et al., 2011; Seaton et al., 2020).

For the other farms, sediment drying appears more efficient than Suaeda australis cultivation for the reduction of total sulfur. This can be explained by the fact that in the greenhouse, pots containing halophytes were watered twice a week with seawater, which is a source of sulfate for the sedimentary matrix. In addition, plant residues can be sources of organic sulfur for the sediments (Schoenau and Malhi, 2015). It would be therefore interesting to determine the different forms of sulfur in the sediments, such as the quantity of sulfur in its organic form that is available to the plant, or the quantity of bound sulfur, for a better understanding of the effect of halophyte cultivation on the sulfur cycle (Zhou et al., 2009). For all the farms, the available forms of phosphorus were reduced almost as much in the dry sediment as in the halophyte rhizosphere. This may be explained as phosphorus is an essential nutrient for microorganisms growth (Oliverio et al., 2020) and therefore microbial communities have used the pool of available sediment phosphorus for their nutrient requirement. This highlighted a pivotal role of the prokaryotes in the phosphorus cycle.

In the farm A, the nitrogen removal under the halophyte cultivation was significantly higher than in the dry sediment. Thus, for this farm, Suaeda australis nutrition and its associated microbiota was more effective than in dry condition to reduce nitrogen level accumulated in the sediment. However, in the farms F and P, the practice of drying alone, had greatly reduced the level of nitrogen in the sediments (Table 1), which meant that for these two farms, the microbial community alone was almost as effective (farm F) or even more effective (farm P) than the cultivation of halophytes and its microbiota. Those two farms had similar microbial communities at the beginning of the experiment, higher alpha diversity indices and higher CaCO₃ concentration in their sediments (Figure 2C and Tables 1, 2). Those parameters may have favored a better nutrients removal in the sediment with and without halophyte cultivation. However, for those two farms F and P, the reduction of the total phosphorus was significantly higher under halophyte cultivation compared to dry sediment. This higher decrease under Suaeda australis cultivation may be linked to the phosphorus nutrition for the plant. In the soil, mineralization of organic P is catalyzed by hydrolytic enzymes such as phosphatase that can be released by plant or soil microorganisms (Marschner et al., 2011; Nannipieri et al., 2011). The mineralization of organic phosphorus is thereby controlled by the plant and microbial demand (Richardson and Simpson, 2011). Thus, a higher mineralization of sediment phosphorus to meet the plant needs could have explained the greater reduction of total phosphorus in the Suaeda australis rhizosphere compared to the dry sediment. This could be the case particularly for the farm F, where the total pool of phosphorus available to the plant was significantly lower than for all the other farms (Table 1). Thus, the effectiveness of nutrient removal by halophyte in shrimp sediment seemed to be strongly influenced by the chemical characteristics of the sediments at the initial state. The aquaculture sediments are formed on the initial soil of the pond as soon as aquaculture activities begin, and then evolve with successive rearing and zootechnical practices (Munsiri et al., 1995). It was also reported that in New Caledonia, the pond sediment pH and CaCO₃ concentration may change with the aging of farms (Lemonnier et al., 2021). Thus, the background of farms aquaculture practices, farms age or type of original pond soil may explain variation of sediment chemical parameters between farms.