A duo of fungi and complex and dynamic bacterial community networks contribute to shape the Ascophyllum nodosum holobiont

Abstract

Background

The brown alga Ascophyllum nodosum and its microbiota form a dynamic functional entity named holobiont. Some microbial partners may play a role in seaweed health through bioactive compounds crucial for normal morphology, development, and physiological acclimation. However, the full spectrum of the microbial diversity and its variations according to algal life stage, season, and location have not been comprehensively studied. This study uses 208 short-read metabarcoding samples to characterize the bacterial, archaeal, and microeukaryotic communities of A. nodosum across three nearby sites, four thallus parts, and a monthly survey, aiming to explore the dynamics of ecological interactions within the holobiont.

Results

Our results revealed that A. nodosum harbors a predominantly bacterial microbiota, varying significantly across all covariables, while archaea were virtually absent. An innovative normalization using the co-amplified host reads provided an estimation of bacterial abundance, revealing a drastic decline in May, potentially linked to epidermal shedding. In contrast, fungal communities were stable, dominated by Mycophycias ascophylli and Moheitospora sp., which remained closely associated with the host year-round. We identified a core microbiome of 22 ASVs, consistently found in all samples, including Granulosicoccus, a genus consistently abundant in other brown algal microbiota. Sequence clustering revealed multiple species which vary according seasons, even in the overall stable Granulosicoccus genus. Co-occurrence network analysis revealed putative interactions between microbial groups in response to ecological niches.

Conclusions

Overall, these findings highlight the dynamic of bacterial interactions and stable fungal associations within the A. nodosum holobiont, providing new insights into the ecology of its microbiota.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40793-025-00825-z.

Introduction

Brown seaweeds provide habitats for diverse communities of epi- and endosymbiotic microorganisms (i.e., bacteria, eukaryotes, archaea, and viruses) [1, 2]. The algal surface is suitable for microorganism development because of organic substances, which play an important nutritional role for the microbiota [3]. Some microbial partners are necessary for seaweed health by producing bioactive compounds essential for normal morphology, development, and physiological responses to environmental factors [4]. These host-microorganism associations constitute interacting and sometimes co-dependent units, so-called holobionts [2, 5–7].

Ascophyllum nodosum is abundant and perennial along the North Atlantic coasts, anchoring to rocky substrates in sheltered intertidal zones. It can live up to 15 years, reach lengths of 2 m, and exhibits growth through apical elongation [8–10]. This dioecious species has a slow recruitment process due to limited dispersal and low germling survival rates [11, 12], with young individuals requiring approximately five years to reach maturity [13]. Ascophyllum is a valuable resource due to its bioactive components used in agriculture, food ingredients, health, and cosmetics. However, increasing demand has raised concerns about overexploitation.

A. nodosum has been described as engaged in a tripartite symbiosis with the endophytic marine fungus Mycophycias (formerly Mycosphaerella) ascophylli and the epiphytic red alga Vertebrata (formerly Polysiphonia) lanosa. The ascomycete M. ascophylli is consistently present throughout the host, without apparent pathogenic effects [14, 15], and is thought to protect A. nodosum germlings from desiccation [16]. V. lanosa is an obligate epiphyte, largely restricted to A. nodosum, causing limited damage to the host despite rhizoid penetration into host tissues [17, 18]. These species have undergone considerable coevolution resulting in intricate morphological, physiological, and ecological interactions. However, it is unclear whether these interactions also involve other bacteria or fungi, and whether these fungi and bacteria can increase A. nodosum fitness, as described in other brown algae [19].

A. nodosum-associated bacteria were first described in 1969 by culturing [20] and later by electron microscopy [21]. Experiments suggest that bacteria shield A. nodosum from digestion by M. ascophylli [22]. Further studies investigated surface-attached bacteria degrading polysaccharides [23], the role of endobacteria in quorum-sensing [24], and epibacterial community responses to stress [25] and decay [26]. Recently, Brunet et al. [27] quantified the epibacterial communities by qPCR across a year-long survey. Despite recent increased scientific interest, the full diversity of microbial communities associated with A. nodosum remains understudied.

This study aims to fill this knowledge gap and answer three key questions: (i) Does the microbiota change across a small spatial scale, algal thallus parts, and over the seasons? (ii) Is there a core microbiota? (iii) How do microorganisms associate with each other? To comprehensively describe the holobiont, we used a short-read metabarcoding analyses of bacteria, archaea, and eukaryotes and used co-occurrence networks to infer potential ecological interactions between microorganisms.

Material and methods

• 1. Sample collection

A monthly survey was conducted from November 2021 to November 2022 at Pleubian (48.849411, − 3.078291, Brittany, France, see [28]), a location known for the commercial harvesting of seaweeds, including A. nodosum. The Ascophyllum-associated microbial community was characterized in three nearby sampling sites. At each site, four algae were selected arbitrarily and four parts were sampled (~ 50 mm long): apex, receptacle (during reproductive season only–January to May), medium, and basal parts as described in [28].

• 2. DNA extraction and sequencing

The algal samples were freeze-dried and finely ground to access the epi- and endophytic microbial communities. The DNA extraction was based on CTAB extraction buffer, chloroform/isoamyl separation, and purification with the NucleoSpin Plant II kit (MachereyNagel, Germany) as well as AMPure XP Beads (Beckman Coulter, Brea, CA, USA) [28]. To reduce sequencing costs, and facilitate the bioinformatics analysis, three amplicon datasets were built based on the following research questions by pooling at equal concentration the DNA:

• “Sites” dataset: The strategy involved pooling extracted DNA from three thallus parts to remove the potential thallus part effect (apex, medium and base—no receptacle as it was not always present). These pooled from different site were examined for two arbitrarily chosen sampling dates: November 2021 and March 2022 (3 sites * 4 individuals * 2 months = 24 samples)

• “Algal parts” dataset: Samples from four individuals across four thallus parts were analyzed from site 1 for two arbitrarily chosen months: January and April 2022 (4 parts * 4 individuals * 2 months = 32 samples).

• “Season” dataset: Here we pooled DNA from three thallus parts from one individual alga collected over 12 months at site 1 (12 months * 4 individuals = 48 samples).

The sequencing run was shared with the samples described in [28] and included three replicates each of a bacterial and fungal mock community containing 10 bacterial and 10 fungal DNAs, as well as 8 negative controls. PCR reactions and library preparation were carried out as described in [28], using the “Parada” primers 515-F/926-R [29] to target bacteria, archaea, and eukaryotes, and ITS2 nested PCR (ITS-1/ITS4 and 5,8S-FUN/ITS4-FUN, [30, 31]) to increase taxonomic resolution for fungi. While these primers are among the most universal available [29], there may still be specific bacterial groups that are underrepresented. Libraries were sequenced on a NovaSeq 6000 SP at the Genomics Core Facility GenoA platform (Nantes, France) using 500 cycles (kit ref. 20,028,402).

• 3. Bioinformatic analysis

Cleaning, taxonomy assignment, and decontamination steps

Bioinformatics analyses were carried out as described in [28]. Briefly, the 16S and 18S rRNA genes (i.e., SSU rRNA genes) and ITS sequences were quality-controlled, and primers were removed with Cutadapt [32]. For the SSU rRNA gene sequences, we followed the online pipeline (https://astrobiomike.github.io/amplicon/16S_and_18S_mixed) to separate 16S and 18S rRNA genes, with magicblast tool [33]. Each set of 16S, 18S, and ITS files was filtered independently where the forward and reverse reads of the 16S and ITS datasets were merged, and those of the 18S dataset were concatenated. They were denoised into Amplicon Sequence Variants (ASVs) with the DADA2 v1.28.0 [34] in R v.4.2.3. Taxonomy was assigned using SILVA NR99 v138.1, PR2 v.5.0.0, and UNITE v.9, respectively. Short unassigned ITS sequences were compared with full-length ITS consensus sequences from our previous study [28] to aid their assignment. SSU sequences shorter than < 350 bp and plastid sequences were removed. The remaining sequences were decontaminated using microDecon package [35]. All sequences containing homopolymers longer than 5 bases were excluded. Finally, remaining chimera were removed using vsearch v.2.22.1 [36] with the –uchime_ref option and the SILVA non-redundant SSU database v.138.1 as reference.

Dataset normalizations

As the “Parada” primers amplified the 18S rRNA gene of the Ascophyllum nodosum host, we could use these host reads as an internal reference to detect variations in the proportion of microbial reads: microbial read abundances were divided by the abundance of A. nodosum reads (Supplementary Fig. 1) in the same sample and multiplied by the median of all A. nodosum reads across all samples. As the Illumina sequencing is insufficient to discriminate sequences among the Fucaceae family, all ASVs identified as “Fucus”, “Silvetia” and “Phaeophyceae” were summarized as Ascophyllum for this purpose.

SSU ASVs accounting for < 0.005% of the total reads in all samples were removed from datasets after normalization. As the ITS primers did not amplify the host, the fungal dataset was normalized by the median and the rare ASVs with < 0.01% relative abundance were removed for the three datasets. This threshold was increased compared to the SSU dataset considering that initial number of sequenced reads for the ITS was lower.

Finally, we removed three samples from the dataset because they were considered outliers. Sample L1B4_APRIL from the “Algal parts” dataset was removed because it contained approximately four times fewer reads than the average sample after the chloroplast removal process (97% of the total reads were removed) (Supplementary Tables 1 and 2). The other replicates did not show this pattern, which could be explained by over-grinding (perhaps a longer grinding time than other samples, leading to the release of more chloroplasts). This extreme read loss resulted in an inconsistent pattern compared to other samples in the NMDS plot. Samples L1I1_MARCH from the “Sites” and “Season” datasets and L1R3_APRIL from “Algal parts” were also removed as they showed inconsistent behavior after host normalization. The number of Ascophyllum reads was unexpectedly low in these samples (8% vs 27–39% and 12% vs 26–43% in other replicates, Supplementary Table 2), resulting in over-representation of the bacterial component after host normalization (Supplementary Fig. 2). One explanation may be insufficient grinding of the algal tissue, leading to poor lysis of Ascophyllum cells (Supplementary Fig. 1, outlier sample in the season panel in March). These three samples are confirmed to be outliers (p-value < 0.05) with a Grubbs test on read abundances in Past4 v.4.15 [37]. The three datasets were then processed separately.

Beta and alpha diversities

Beta diversity was estimated from raw datasets using the Bray–Curtis dissimilarity index with non-metric multidimensional scaling (nMDS) ordinations and statistical analyses were based on a PERMANOVA test with 999 permutations. Alpha diversity was further calculated with the Shannon index on normalized data using the microbiome package v1.22.1 and compared using ANOVA tests (stats package v.4.3.3) and the Tukey post-hoc tests (rstatix package v.0.7.2).

Core microbiota

Intermediate core microbiota were identified separately for the three datasets with Venn diagrams (MicEco package v.0.9.19) and pie charts (ggplot2, package v.3.5.0). The intersection of both intermediate core microbiota then formed the global core community of A. nodosum. Two occupancy thresholds were considered; 100% and 80%, meaning that ASVs present in 100% or 80% of the samples were considered part of the core. We kept the 100% occupancy threshold for the final analysis, but we provide the results of the alternative 80% threshold in the Supplementary Fig. 9. Metabolic capabilities from the taxonomic affiliations of each occupancy threshold were inferred using EsMeCaTa v.0.6.8 [38].

Heatmap

To distinguish different fungal ecotypes, especially concerning Moheitospora sp. or Mycophycias ascophylli, all three ITS datasets were merged and the sequences were clustered at 99% identity (e.g., Operational Taxonomic Unit, OTU) with vsearch v.2.22.1 [36]. OTUs were filtered to remove rare OTUs with under 0.01% relative abundance over the entire dataset. Clustered sequences were aligned with Mafft v.7 using the iterative refinement method G-INS-i. The alignment was curated with Gblocks allowing smaller final blocks, gap positions within the final blocks, and less strict flanking positions and a phylogenetic tree was calculated on phylogeny.fr [39] using PhyML, the GTR + G + I substitution model, and an approximate Likelihood-Ratio test. A heatmap was generated with ggplot2 package v.3.5.0.

Co-occurrence network

For the co-occurrence network, both bacterial and fungal sequences were clustered at 99% identity into OTU as above. They were separately normalized by the median and merged. The merged dataset was filtered with a minimum occurrence larger than 45 reads in at least 3 samples [40]. Putative associations between covarying OTUs were inferred with SpiecEasi v.1.1.3 [41] using the “glasso” probabilistic inference method and a lambda.min.ratio of 0.01. Significant covariances were converted to correlations with the function cov2cor [40]. The Leiden module detection algorithm was chosen after comparing eight algorithms (Fast Greedy, Infomap, Label Propagation, Leading Eigenvector, Leiden, Louvain, Spinglass, and Walktrap) using the igraph R package v.2.0.3 [42] as reported in [43]. For module detection algorithm comparisons, only positive correlations were considered because most module algorithms only allow non-negative correlations [43]. To visualize the network, Gephi v.0.10.1 [44] with the Force Atlas2 algorithm was used. The betweenness centrality was calculated with the “Network Diameter” statistic program of the Gephi software. The negative correlations were added when Force Atlas2 algorithm was stopped, as only positive edges are allowed to determine the spatial distribution of the nodes. Metabolic capabilities from the taxonomic affiliations of each module were inferred using EsMeCaTa v.0.5.4 [38]. The metabolic enrichments, implemented in EsMeCaTa, were performed with GSEApy [45] and the resulting heatmap with Orsum [46].

Results

1. Shifts in seasonal bacterial diversity and global community patterns across datasets

The alpha diversity of the Ascophyllum-associated microbiome in the SSU dataset varied significantly over the year (Fig. 1A) but not between “sites” and “algal parts” (Supplementary Fig. 3). The diversity was the lowest in May 2022 and differed significantly from the samples in November 2021, March, June, September, and October 2022 (Fig. 1A). No significant differences were detected in any ITS dataset (Supplementary Fig. 4).

Fig. 1.

Variation of SSU diversity and communities across the three datasets. A Alpha diversity (Shannon index) of the season dataset. Significant differences according to an ANOVA with a Tukey post-hoc test are indicated by different letters next to the bars. B The NMDS plot concerns the beta diversity of the “Sites” dataset and shows significant differences for months (p-value = 0.015, F = 2.7557, R2 = 0.11034) but not for sites. C The NMDS concerns the “Algal parts” dataset and shows significant differences for the algal parts (p-value = 0.001, F = 4.6197, R2 = 0.30726) and month (p-value = 0.001, F = 5.2764, R2 = 0.11698). D The NMDS concerns the “Season” dataset and shows significant differences for season (p-value = 0.001, F = 4.5595, R2 = 0.18771) and month (p-value = 0.001, F = 2.3545, R2 = 0.29081). Statistical analyses were based on a PERMANOVA test. Winter corresponds to January–February, spring to March–May, summer to June–August, and fall to September–December

The SSU communities were stable at a small geographic scale in November 2021 and March 2022 (Fig. 1B). Along the algal thallus, different groups emerged, separated by months and thallus parts (Fig. 1C). In January, the base and medium parts were similar, except for one base from April, and distinct from the receptacle and apex. Seasonal variations were observed with distinct communities between fall/winter and spring/summer (Fig. 1D). The sampling dates of this study were based on astronomical and meteorological definitions where winter corresponds to January–February, spring to March–May, summer to June–August, and fall to September–December. ITS profiling revealed only one significant difference between January and April in the “algal parts” dataset (Supplementary Fig. 5).

• 2.Dynamic and highly diverse bacterial communities: dominance of seven classes

The A. nodosum bacterial communities are composed of 7 classes (Fig. 2), with Alphaproteobacteria and Gammaproteobacteria consistently dominating. The most abundant genera include an unassigned Alphaproteobacterial genus, Granulosicoccus (Gammaproteobacteria), Pleurocapsa PCC-7319 (Cyanobacteria), Blastopirellula (Planctomycetes), and Algisphaera (Phycisphaerae).

Fig. 2.

Variation of the 16S rRNA gene diversities of Ascophyllum nodosum at Pleubian sites (Brittany, France). The boxplot figures at the top show the number of total bacterial reads after normalization with A. nodosum nuclear 18S rDNA reads (see methods Sect. “Dataset normalizations”). Here significant differences are estimated with an ANOVA and a Tukey post-hoc test and indicated by different letters next to the bars. No letter means no significant differences. The barplots show the variation in relative abundance of bacteria at the genus levels for the “Site” (A St1 = Site 1, St2 = Site 2, St3 = Site 3), “Algal parts” (B Rp = Receptacle, Ap = Apex, Md = Medium, Bs = Base) and “Season” (C). The different shades of blue represent the Alphaproteobacteria, orange the Gammaproteobacteria, dark pink Bacteroidia, green Cyanobacteriia, violet Planctomycetes, yellow Verrucomicrobiae, pink Phycisphaerae, and grey other bacteria in low abundance, not included in the 7 major classes. “NA” corresponds to unassigned taxonomy

Site dataset

Bacterial abundances relative to the host remained relatively stable across sites and months (Fig. 2A, boxplot)—despite site-dependent variations in terms of data dispersion—except at site 2 in November, where the unassigned Alphaproteobacterial genus was less abundant than in other samples.

Algal parts dataset

No significant differences were detected in bacterial read abundance relative to the host, in the different algal parts. Regarding the community structure, Granulosicoccus was the dominant genus in every sample except for the medium and basal parts in January where the unassigned Alphaproteobacterium was found in high relative abundance (Fig. 2B). We also separately analyzed the abundance of host plastid reads relative to nuclear reads across algal parts, revealing a significantly lower proportion of plastid reads in the receptacles (Supplementary Fig. 6B).

Seasonal dataset

Normalized bacterial read abundance was relatively stable from November 2021 to March 2022 (Fig. 2C, boxplot) before it significantly dropped in April and May and then increased from July to October. From November 2021 to March 2022, the bacterial communities were dominated by one ASV affiliated to an unassigned Alphaproteobacterium, but this group was less abundant from April to November 2022. From August, the genus Pleurocapsa PCC-7319 increased in abundance to reach the highest levels between September and November 2022. The ratio of plastid to nuclear host reads was also highly variable across samples (up to fivefold difference), although no significant seasonal pattern was detected (Supplementary Fig. 6A).

• 3.Stable eukaryotic communities dominated by two fungi

Fungi belonging to Ascomycota class dominated the 18S sequences (> 70% of total reads), in particular representatives of the classes Dothideomycetes and Sordariomycetes (Fig. 3). Similarly, for the ITS sequences two dominant fungi identified as Mycophycias ascophylli (Dothideomycetes) and Moheitospora sp. (Sordariomycetes, recently renamed Juncigena sp. [47]) were observed in almost every sample (Fig. 3).

Fig. 3.

18S (1.) and ITS (2.) based-primers diversities of the Ascophyllum nodosum microbiome at Pleubian sites (Brittany, France). The boxplot figures at the top show the number of eukaryotic reads after normalization with A. nodosum nuclear rDNA reads (see methods Sect. “Dataset normalizations”). The barplots (1.A, 1.B, 1.C) show the variation of eukaryotes in relative abundance at the genus levels for the “Site” (A St1 = Site 1, St2 = Site 2, St3 = Site 3), “Algal parts” (B Rp = Receptacle, Ap = Apex, Md = Medium, Bs = Base) and “Season” (C) The different shades of green represent the Ascomycota class, blue the Phaeophyceae, orange the Arthropoda, dark pink non-assigned genera at the phylum level, and grey other eukaryotes found in small abundance, not included in 4 major families. “NA” corresponds to unassigned phylogeny. The heatmap shows the variation of fungal strains across the three datasets (2.A, 2.B, 2.C). ASV were clustered at 99% (OTUs) of identity to distinct strains. OTUs not included in the three abundant species (Mycophycias ascophylli, Moheitospora sp. and Lulworthiaceae sp.) were summed in the “Other fungi” category. The stronger the red, the more abundant is the OTUs. The phylogenetic tree on the left was based on a sequence alignment of 22 sequences with the MAFFT (G-INS-I iterative strategy), curation with Gblocks (329 positions remained), and was calculated using PhML (SH-LIKE, GTR + G + I substitution model). Support values under 0.70 have been removed and there is no support value for the ‘other’ category because it represents several fungi. The three most abundant fungal genera were highlighted in the tree: red for Moheitospora sp., blue for Mycophycias ascophylli, and yellow for Lulworthiaceae sp

The host-normalized eukaryotic read abundance remained stable across datasets (Fig. 3A, B, C, boxplots), following similar patterns as the bacterial communities, except that seasonal variations were less pronounced (Fig. 3C). In the algal parts dataset (Fig. 3B), however, Phaeophyceae_XX;NA were abundant relative to the host in January in the receptacles and then decreased from apex to basal parts. In April, this group was observed in low relative abundance, and two of three receptacles were dominated by the Diathrodes genus in the Arthropoda phylum. However, it should be noted that, in reproductive tissues such as receptacles, nuclear DNA content may vary, which may contribute to the apparent fluctuations in eukaryote/host ratios observed across tissue types.

• 4.The A. nodosum core microbiome

The four-thallus parts shared 28 ASVs and a total of 60 ASVs were consistently present only on specific algal parts: 16 ASVs were consistently observed only in receptacles, 4 in apex, 10 in medium, and 30 in basal parts (Fig. 4A.1). The core community specific to basal parts was the most taxonomically diverse with 7 different classes, followed by that of the receptacles, the medium and the apical parts, the latter comprising only 3 classes in the core community (Fig. 4A.3).

Fig. 4.

Core microbiome (Bacteria and Eukaryota, SSU) of Ascophyllum nodosum specifically associated with the four algal parts (A.1) and over the four seasons (A.2) (i.e., “intermediate core microbiota”). Numbers in the Venn diagrams represent the number of ASVs. The bubble plot (A.3) represents genera consistently associated specifically with the four algal parts and the four seasons. The global core microbiome is formed by the intersection of both intermediate core microbiota (B.1). The pie chart shows the genera these 22 ASVs belong to and their relative read numbers in the global core microbiome. The color shows the bacterial class. The phylogenetic tree (B.2) was based on a sequence alignment of 22 core sequences with the MAFFT (G-INS-I iterative strategy), curation with Gblocks (373 positions remained), and was calculated using PhML (SH-LIKE, GTR + G + I substitution model). Support values (SH-like as approximate Likelihood-Ratio test) under 0.75 have been removed. The ASV sharing more than 99% of identity according to the vsearch tool, were highlighted in blue, yellow and orange

In the seasonal dataset, 46 ASVs were present on every individual the entire year, and another 45 ASVs were found consistently only in one season (20 ASVs observed only in fall, 10 in winter, 10 in spring, and 5 in summer; Fig. 4A.2). The fall microbiome also exhibited the most diversified core community with 6 classes, followed by winter, spring and summer (Fig. 4A.3).

The intersection of the core of the algal parts and season comprised 22 ASVs with a dominance of Granulosicoccus, followed by an assigned Alphaproteobacteria ASV (Fig. 4B.1). Among the 22 ASVs, two pairs of Granulosicoccus and one Roseibacillus shared more than 99% of identity (i.e., 19 OTUs, Fig. 4B.2).

• 5.Putative microbial interactions: the co-occurrence network analysis

We investigated groups of co-varying OTUs that could share the same ecological niches via a co-occurrence network analysis. After merging SSU OTUs with ITS OTUs, the final matrix contained 12,090 OTUs. After filtering out low-abundance OTUs, 531 OTUs remained for co-occurrence analyses, representing 99.3% of the total reads in the dataset. A total of 520 OTUs were correlated with 4130 positive (72.6%) and 1552 negative (27.3%) edges. Only 11 OTUs were isolated in the network and not connected to other OTUs (i.e., nodes without any edges).

In the inferred network with positive correlations, 8 modules comprising OTUs with similar distribution patterns were identified (Fig. 5A, 5). Module 4 was dominant with 88.1% of reads (including Ascophyllum, Mycophycias, Moheitospora and Granulosicoccus), followed by seven modules representing between 6.5% and less than 1% of the reads. Among the 19 OTUs found in the core microbiome (Fig. 4B.2), 14 OTUs fell into module 4, followed by module 8 with 3 OTUs, and module 1 and 6 (1 OTU each) (Fig. 5A). Some of the core genera were represented in more than one module, such as Granulosicoccus in modules 4 and 6.

Fig. 5.

Co-occurrence analysis of the Ascophyllum nodosum microbiota (SSU and ITS). A Each node represented an OTU (ASVs clustered at 99% identity) while each edge represents a positive correlation obtained from the covariance matrix calculated with SPIEC-EASI. The nodes are colored according to the module they were assigned to, and node size illustrates the betweenness centrality. The network was represented using Gephi and the Force Atlas layout algorithm. Yellow and black stars represent OTUs found in the core microbiome. Node with indicated names seem important as high betweenness centrality or directly in association with Ascophyllum. The network with negative correlations is shown in Supplementary Fig. 8. B Specific interactions with Ascophyllum nodosum, Mycophycias ascophylli, and Moheitospora sp. are illustrated by selecting edges with weight < 0.2. C Presentation of the 8 modules detected with the Leiden algorithm, ordered by the trend observed in algal part abundance, with the number of OTUs (n) and the relative abundance of each module. The taxonomic composition of the module at class level is illustrated by the pie charts, and the relative microbial abundance across seasons and algal parts is shown in the plots to the right. The Y axis represent the average ASVs abundance found in each algal part or across each month (X axis), where each sample was normalized by the total sum of the ASV in all samples, so each ASV has the same weight. The functional predictions based on the taxonomy were estimated with EsMeCaTa and the percentage of enriched functions in each module is given in the last column

By adding the measure of the betweenness centrality, we identified key nodes that have an impact on the network structure. The larger the node, the more important the OTUs is likely to be within the putative interactions. Few larger nodes were observed in the modules 2 and 3, whereas the largest were found in the module 5, associated with the Maribacter and Roseobacter genera. Mycophycias ascophylli and other members of the core microbiome were amongst the largest nodes in module 4 (Fig. 5A and Supplementary Table 3). The strongest associations of the host Ascophyllum were with OTU_3 of Granulosicoccus, Mycophycias, Rubritalea, Yersinia, Algisphaera, and one unclassified bacterium (Fig. 5B).

We also investigated the correlation between module distribution and metadata such as seasons and algal tissues (Fig. 5C). Microbes associated with modules 1, 2, and 3 were linked to the basal part of the alga whereas module 5 was linked to the receptacles from November 2021 to March 2022. Modules 4, 6, and 7 were present all the year. We further investigated the metabolic enrichments based on the taxonomic composition of each module to understand why these microorganisms co-varied together. Among 23,585 predicted functional annotations (GO terms or EC numbers), few were enriched in specific modules (maximum 0.13% for module 1), for example module 5, which was enriched in the siderophore metabolic process (Supplementary Fig. 7 and Supplementary Table 4).

While the positive correlations were predominantly found within modules, some negative correlations tended to occur between modules (Supplementary Fig. 8). One OTU from the Verrucomicrobiota phylum (DEV007 family), representing 19.8% of the total read found the network, showed the most negative correlations with 34 negatives edges (Supplementary Table 5). Ascophyllum nodosum and the Mycophycias ascophylli shared strong negative correlations (weight <  − 0.2 out − 0.30) with a Woeseia (OTU_142) and with an unknown Gammaproteobacterium (OTU_173).

• 6.Seasonal dynamics of OTUs within the stable genera of the core microbiome

Prompted by the observation of OTUs corresponding to the same core genera in different modules, we also performed detailed investigation of seasonal variations among the OTUs belonging to genera of the core microbiome. Three groups of OTUs emerged (Fig. 6): one exhibiting a stable pattern over the year, one most abundant in colder months between November 2021 and March 2022, and another one mostly found between April and November 2022. Our analyses confirm that some of core genera had OTUs in the different groups. Notably, the Granulosicoccus OTU_3 was highly stable, while OTUs 42 and 44 were most abundant in winter, and OTU_51 was most abundant in summer. Similarly, for Rubripirellula we observed OTU_86 with a pattern of highest abundance in winter, OTU_57 was dominant in summer, and OTU_145 was relatively stable. This variability suggests the succession of different ecotypes within the core genera.

Fig. 6.

The heatmap shows the variation of genera found in the core microbiome of Ascophyllum nodosum across the monthly survey. The OTUs come from the co-occurrence network ordered by a Pearson correlation. The stronger the red, the more abundant is the OTUs (log10). OTU names in red colors indicate genera with different OTUs that were characteristic for warmer and colder months

Discussion

• 7.Bacterial versus fungal communities: complex and dynamic versus simple and stable

We explored the Ascophyllum microbiome from multiple angles, examining its composition under natural conditions across nearby sampling sites, different parts of the algal thallus, and season over a one-year in situ survey. We assessed the microbial diversity using primers targeting bacterial, archaeal, and eukaryotic partners in a single PCR [29]. These primers have proven suitable for capturing microbial biodiversity in diverse sample types with high resolution [48], and their overall performance was validated by correctly reconstructing the diversity in our mock communities [28].

Although the eukaryotic component was observed in A. nodosum, bacteria dominated (98% of the symbiont SSU reads, but extraction biases and SSU copy number differences are likely to impact this figure). Archaeal sequences (< 0.001%) were rare and mostly chimeras, suggesting that archaea are absent in A. nodosum, unlike in the brown alga Laminaria rodriguezii [49]. This supports the idea that bacteria are key microbial symbionts [2].

A. nodosum harbors a complex and dynamic bacterial community, in contrast to the stable and simple fungal communities. The bacterial communities comprised 7 different classes with multiple genera, which showed diverse dynamics between months. Microbial composition varied with season and tissues types, aligning with previous studies on macroalgal-associated bacteria [27, 50–53]. Microbial dynamics across tissue types could be related to the physiology and the age of the alga. Given Ascophyllum’s apical growth [10], the apex likely host an establishing microbiome, while the basal region likely harbors a more mature microbiome.

In contrast, eukaryotic analysis revealed the predominance of only two fungi: Mycophycias ascophylli, and Moheitospora sp., belonging to Dothideomycetes and Sordariomycetes, respectively. Our study confirms that Moheitospora sp. is a central component of the fungal community in A. nodosum as shown previously. [19]. Both of these two fungi could contribute to the “lichenization” of A. nodosum [14]. Unlike A. nodosum, other brown and red macroalgae usually exhibit higher fungal diversity [24, 54, 55]. This suggests a regulation of fungal colonization originating either from the host, the bacteria, or the fungi. Putative antifungal molecules could be synthetized by fungi, as reported in the marine fungi Aspergillus sp., and Penicillium sp. [56–58]. Bacteria may also control fungal presence, since antibiotic treatment of A. nodosum result in M. ascophylli hyphae penetrating algal cells and degrading the host [22]. Genome sequencing of these symbionts could provide further indication regarding these hypotheses. The role of Moheitospora sp. could also be tested experimentally using its available isolate [19] as shown for M. ascophylli [16]. Another perspective is to explore at the viral partners as they remain largely understudied, with only a few reports detailing seaweed-associated viruses [59–63]. This limited virus knowledge may limit constrain our capacity to understand the full spectrum of processes occurring within the A. nodosum holobiont.

• 8.Host normalization highlights epidermal shedding as a potential driver of bacterial dynamics over the year

In general, metabarcoding analysis only provides relative quantification of the microbiome, although qPCR has been used to mitigate this restriction [27, 64]. Here, we circumvent this without further experimentation, by leveraging our access to the A. nodosum 18S rRNA gene allowing us to calculate the proportion of microbial to host reads. We also evaluated the plastid 16S reads for this normalization, but found them to differ up to fivefold compared the nuclear marker, demonstrating that the two genes cannot be used interchangeably for normalization. This discrepancy likely reflects natural variation in chloroplast abundance, which can change markedly with tissue type, developmental stage, or environmental conditions such as light intensity and photoperiod [65]. Such physiological regulation introduces instability into chloroplast-based normalization making the 18S rRNA gene the likely better choice for normalization. However, it is important to note that even nuclear DNA content may vary between tissues and developmental stages, particularly in reproductive structures. Such variation naturally limits the precision of host-based normalization approaches and should be considered when interpreting differences in microbe/host read ratios across tissues or seasons.

Based on the proportion of microbial to nuclear host reads, we conclude that the abundance of Ascophyllum-associated bacteria relative to the host varies over the year. This is globally in line with a previous study by Brunet et al. [27], but our analysis also highlights a drastic drop in the relative bacterial abundance between March and May. Such changes may be related to seasonal temperature shifts, nutrition variations, or other abiotic factors, but they could also be explained by epidermal shedding. The latter process was first observed in 1981 where small pieces of A. nodosum were detached occasionally from the thallus surface [66] and considered an adaptation for long-lived macroalgae to control epiphytes [67, 68]. It occurs regularly throughout the warmer months of the year [69, 70]. However, we observe that the reduction in the ratio of microbial to host reads did not concern all microbes. A large proportion of the unassigned Alphaproteobacterial ASV seems to be impacted by the process, while, e.g., one Granulosicoccus OTUs or the two main fungi remained abundant over the year.

In the context of shedding, this would mean that these consistently abundant organisms are either fast colonizers rapidly taking hold on the free surface, or that these microbes also penetrate into deeper cell layers. The latter is likely true at least for the most abundant of the two fungi, M. ascophylli, which has even been found intracellularly [8] or in the intercellular spaces embedded in the thick cell walls of the medulla and cortex tissues [71]. Regardless of the reasons driving these different patterns, our observations suggest that even within a genus, distribution patterns and thus the underlying ecological niches and strategies of different OTUs differ. Fine taxonomic distinction will therefore be crucial in future investigations of these taxa, while microscopy and microbial quantification in shed tissues will be the key to validating the epidermal shedding hypothesis.

• 9.The core microbiome of A. nodosum comprises 22 key ASVs

Identifying core microbiomes has become widespread to identify potential key actors in host biology. To determine the threshold of the core, we compared two occupancy thresholds; 100% and 80% (i.e., an ASVs should be present in 100% or 80% of samples to be considered in the core). With the stringent definition [72] of the core microbiome, i.e. 100% occupancy, only 22 ASVs met that strict criterion, but these taxa accounted for 43% of all sequence reads, indicating that they represent a substantial portion of the community. Lowering the threshold to 80% increased the number of ASVs to 45 and the read coverage to 70%. However, that introduces taxa that may be conditionally present, like Dokdonia, Erythrobacter, Illumantobacter, Maxixanthomonas and Yersinia and unassigned genera like Rhodobacteraceae, KI89A clade and OM190 (Supplementary Fig. 9). No functional enrichment was observed comparing the functional prediction based on taxonomy between the two tested thresholds (data not shown). This, alongside the overall high stability of the A. nodosum microbiome and the high sequencing coverage in this study led us to choose 100% occupancy for our downstream analyses.

Among the 22 core ASVs present in all Ascophyllum individuals, several genera, in particular Granulosicoccus, Litorimonas, Hellea, Blastopirellula, and Roseibacillus, are generalists that have been observed on a diverse array of brown, green, and red macroalgae [27, 52, 53, 73–80]. The ubiquity of Granulosicoccus genus in several brown algae suggests that it is especially well-suited for life as a symbiont. Metagenome-assembled genomes of Granulosicoccus were annotated in the bull kelp Nereocystis luetkeana, and harbored genes for synthesizing cobalamin possibly providing this and other vitamins to the algal host [81]. However, evidence of metabolic exchanges between Granulosicoccus and the seaweed has not yet been completed.

Interestingly, our seasonal analyses revealed distinct patterns among the Granulosicoccus OTUs. Some were stable throughout the year, while others peaked in summer. Such patterns may reflect ecological specialization among closely related strains. Changes in the host’s surface chemistry or metabolite production linked to temperature, light or reproductive status could favor different Granulosicoccus ecotypes at different times of the year [82].

Our core microbiome analyses also consolidate the tight association of M. ascophylli with the Ascophyllum, as all samples contained the fungus in high abundance. The symbiotic association between Ascophyllum and this fungus has already been described and reviewed [8, 14, 55, 83–86]. Previous light microscopy analyses have shown M. ascophylli to be present at the highest concentration in receptacles [15], which is not consistent with our findings, but this likely points to a limitation of normalization method–indeed the male receptacles produce high numbers of sperm cells each with a complete copy of the host genome thus potentially biasing the number of host reads used as a reference in this case. Regardless of these differences, the constant presence of the fungus is consistent with its supposed role in host development, fitness, and protection against biotic and abiotic stresses [16, 87].

• 10.Co-occurrence network reveals microbial contributions to energy synthesis in receptacles

Our co-occurrence networks analysis revealed that A. nodosum was not the central node in the network, and few nodes directly in association with the seaweed. This suggests that Ascophyllum may be a surface for few key microbial colonizers (whether endophytic or epiphytic), as Mycophycias and Granulosicoccus, which may be the base for the biofilm formation. These analyses also highlight module 4, as it is the most abundant module in terms of read coverage and contains many species present in the core microbiome. Module 4 could be linked to early-colonizing or intracellular microorganisms, such as those not responding to shedding phenomenon, and may include functions that contribute to the holobiont’s overall stability, while module 5 may be linked to reproduction.

Because modules regroup highly connected OTUs that share the same environment and potentially exhibit coherent ecological strategies [88], we expected to find the modules to be enriched in functions related to niche adaptation. One example of such adaptations concerned the functioning of receptacles. In the reproductive period, energy should be allocated to reproduction and not to vegetative growth [13]. Iron, essential for ATP production in photosynthesis and respiration, is present in low concentration and poor solubility in seawater [89, 90]. To facilitate iron acquisition, many bacteria synthesize siderophores [91] and functions present in the module 5 were enriched in the siderophore metabolic processes (GO:0009237, Supplementary Fig. 8). The bacterium Roseobacter likely possesses genes involved in siderophore production [92] (Supplementary Table 4) and occupies a structuring position in the module 5. As these enriched metabolic functions are merely predictions based on the microbial taxonomy, they need to be confirmed by complementary approaches, such as metagenomic analyses. Overall, our analyses highlight that very few metabolic functions are enriched across modules, underscoring the functional stability of the holobiont.

Conclusions

Our study provides a comprehensive overview of the microbial communities associated with Ascophyllum nodosum, highlighting the complexity of its holobiont and the dynamic interactions between its bacterial and fungal components. Using metabarcoding analysis with universal primers and using the host 18S for normalization, we found that Ascophyllum harbors a predominantly bacterial microbiota, with no detectable archaea. Community composition and structure vary across algal tissues and seasons, including a notable drop in bacterial abundance in May, potentially linked to epidermal shedding. In contrast, the Ascophyllum’s fungal communities appear stable and simple, primarily consisting of Mycophycias ascophylli and Moheitospora sp., both of which appear to maintain a close association with the host throughout the year. The core microbiome with 22 ASVs suggests the importance of certain microbial taxa in maintaining host biology, such as the genus Granulosicoccus. While the latter demonstrates significant stability, our findings also reveal that within this genus, different OTUs are subject to seasonal variation, highlighting its intraspecific diversity and adaptability. The co-occurrence network and its module-based analyses add another layer of complexity to our understanding, showing how microbial groups interact in response to ecological niches, such as parts of the thallus or varying seasonal conditions. Further investigation using metagenomic and functional approaches will be key to uncovering the adaptations and the roles of key microorganisms and understanding their contributions to the overall health and stability of the A. nodosum holobiont.

Author contributions

CR, CL and SD designed and conceptualized the research. CR collected samples, performed laboratory work, analyzed data and wrote the manuscript. GT and EL performed sequencing preparations. SB and AB implemented the EsMeCaTa tool and contributed to the functional analyses. SR helped for sampling and DNA extractions. PP discussed around algal physiology, provided funding and reviewed the manuscript. CL and SD collected samples, supervised the research, reviewed the manuscript and provided funding. All authors have read and approved the manuscript.

Funding

This work was funded by the SEABIOZ ANR project (AAPG2020, ANR-20-CE43-0013) and has received financial support from the Centre National de la Recherche Scientifique (CNRS) through the MITI interdisciplinary programs (ALGOMETABIONTE PRIME project). It has also partially been supported by the CNRS and Sorbonne University through the UMR8227 annual financial supports.

Data availability

The raw sequence reads and metadata are available in the European Nucleotide Archive (ENA) database under BioProject PRJEB85975 and sample accession numbers ERS23809686-ERS23809790. ASV table with final abundances, taxonomy, and sequences are available in supplementary tables. Scripts are available on GitHub (https://github.com/rssco/novaseq_ascophyllum).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.