After a series of processing, a total of 21,997,761 high-quality sequences were obtained from 307 samples from three countries (USA: 24–26°N, 80–82°W, Australia: 23–24°S, 151–152°E, Fiji: 18.13°S, 177.42°E; Table S1 for detailed information): 62 samples of Montastraea cavernosa, 25 samples of Orbicella faveolata, 128 samples of Pocillopora damicornis, 36 samples of Porites astreoides, 34 samples of Porites and 22 samples of Pseudodiploria strigosa. A total of 71,182 amplicon sequence variants (ASV) were obtained by clustering the high-quality sequences using DADA2.

To understand the effects on variations in microbial communities from diverse coral taxa across a large geographical distance, the alpha- and beta-diversity of the microbiome were calculated based on the resampled ASV tables. Jaccard distance-based principal coordinate analysis (PCoA) showed that geolocation has a strong and significant impact on the microbial community composition of stony corals (PERMANOVA, permutations = 999, Pseudo-F = 21.46, R² = 0.1238, p = 0.001; Figure 1A). Significant differences in microbial community structures were also observed at the coral family and species levels (family level: PERMANOVA, permutations = 999, Pseudo-F = 5.701, R² = 0.0951, p = 0.001; species level: PERMANOVA, permutations = 999, Pseudo-F = 6.784, R² = 0.0828, p = 0.001). It should be noted that the large geographical distance had a stronger impact on the coral microbiota than the host taxon. The family- and species-level coral hosts had a similar impact degree. Moreover, the hierarchical clustering analysis revealed that all samples were classified into three prominent groups corresponding to the geolocations of Australia, Fiji and USA (Figure 1B). The above results showed that geolocation was the main driving force for the microbial communities of the six coral species across a large geographical distance. Additionally, the β-nearest taxon index (βNTI) analysis revealed that deterministic processes became increasingly influential in shaping the structure of the bacterial community with distance growth (Figure 2).

Furthermore, the Shannon index and observed features index in alpha diversity were calculated, as shown in Figure 3. Kruskal–Wallis pairwise analysis results showed that both Shannon and observed features index (Shannon, mean = 5.1494 ± 2.5049 s.d.; observed features, mean = 448.8101 ± 392.9550 s.d.) of the coral microbiome in the northern subtropical group (USA) was significantly higher than that in the tropical group (Fiji) (Shannon, mean = 3.6470 ± 0.3671 s.d.; observed features, mean = 47.0952 ± 21.1652 s.d.) and southern subtropical group (Australia) (Shannon, mean = 1.2665 ± 0.3655 s.d.; observed features, mean = 40.2045 ± 14.9675 s.d.) (p < 0.001). These results suggested that the species diversity, evenness and ASV numbers of the coral microbial communities in the northern subtropical group were significantly higher than those in the tropical and southern subtropical groups, while those in the tropical group were significantly higher than those in the southern subtropical group.

To research the variations in coral microbial communities in the same geographical region, microbiome samples collected from five coral species in the same sea area of USA with a distance of less than 230 kilometers were analyzed. The hierarchical clustering analysis suggested that the similarity of coral microbial community composition could reflect the phylogeny of coral hosts. Except for P. porites corals excluded in the phylogenetic tree of a previous report (Quek and Huang, 2019), the similarity relationships of coral microbiota from four coral species were completely consistent with the phylogenetic relationships of coral hosts (Figure 4). The above results revealed that corals with closer phylogenetic relationships could exhibit more similar microbiota compositions in the same geographical region.

Proteobacteria and Cyanobacteria were the dominant phyla in all five coral species, accounting for 52.74 and 14.21% of the relative abundance of total prokaryotes, respectively (Figure 5A). The three most abundant phyla in Porites coral were Proteobacteria, Cyanobacteria and unidentified phyla. The order Rickettsiales was most abundant in M. cavernosa, P. porites and P. astreoides, corresponding to relative abundances of 49.22, 19.08 and 6.17%, respectively, while it was no more than 2.2% in other coral species. Despite permeability to NAD, Rickettsiales was also suggested to be a pathogenic factor that could cause coral diseases in previous studies (Casas et al., 2004; Garcia et al., 2013; Angly et al., 2016; Driscoll et al., 2017; Klinges et al., 2019). Notably, the unidentified phylum was found to account for 12.98% of the total prokaryotic abundance in coral Porites.

The three most abundant phyla in corals O. faveolata and P. strigosa were Proteobacteria, Bacteroidetes and Cyanobacteria, respectively. It was interesting that the most abundant phylum in coral O. faveolata was Cyanobacteria (35.96%), followed by Proteobacteria (32.54%), which was obviously different from the other four coral species. Cyanobacteria were proven to be responsible for nitrogen fixation for corals (Wiebe et al., 1975; Wang et al., 2000; Lesser et al., 2004).

The three most abundant phyla in coral M. cavernosa were Proteobacteria, Cyanobacteria and Bacteroidetes. According to the linear discriminant analysis effect size (LEfSe) analysis (Supplementary Table S1), M. cavernosa harbored three characteristic bacteria, Proteobacteria, Crenarchaeota and Nitrospirae, at the phylum level compared with the other coral species (Figure 5B). The genus Nitrosopumilus belonging to Crenarchaeota was observed to account for 6.68% in coral M. cavernosa, while it is almost nonexistent in other corals. Nitrosopumilus is an aerobic bacterium, which can efficiently convert ammonia into nitrate (Könneke et al., 2005; Siboni et al., 2008; Jensen et al., 2012; Chekidhenkuzhiyil et al., 2021; Polónia et al., 2021). Nitrospirae is an aerobic bacterium that efficiently converts ammonia into nitrate and is reported to be a common aerobic symbiont in sponges (Hoffmann et al., 2009; Han et al., 2013; Karlińska-Batres and Wörheide, 2013).

To further understand the distance-based shifts in microbial communities, the microbial functional traits from stony corals in three countries were elucidated by LEfSe analysis (Supplementary Table S1; Supplementary Figure S1). A total of 65 differentially abundant features of functions in coral microbial communities were identified to be associated with three geolocations. In the southern subtropical group, 37 functional traits, especially several degradation characteristic functions, e.g., degradation of organic acids, nucleosides, acetylene and sucrose, were enriched. In the northern subtropical group, 21 functional traits involved in metabolic pathways, e.g., pentose phosphate pathway and glycolysis, were revealed to be more abundant. In the tropical group, 7 functional traits, such as biotin and organic acid synthesis, were more abundant. In addition, according to the Venn diagram analysis, 168 functions presented in all three groups could be suggested as common functions, accounting for more than 70.29% of all predicted functions (Figure 6). Six functional traits were shared by the northern and southern subtropical groups, including biosynthetic pathways. A total of 38 functional traits were common between the southern subtropical group and tropical group, such as superpathway for organic synthesis. Interestingly, there were no common functional traits between the northern subtropical group and the tropical group. Only 23, 3 and 1 functions were unique to the southern subtropical, northern subtropical and tropical groups, respectively.

16S rRNA gene amplicon samples involving six stony coral species collected from three previous publications were merged and reanalyzed (Apprill et al., 2016; Beatty et al., 2018; van Oppen et al., 2018). These original sequence data can be obtained from the NCBI Sequence Read Archive under BioProjects PRJNA324417, PRJNA380169, PRJNA382809, and PRJNA872539 of NCBI and the publication of original papers. The stony coral samples across these four studies to sequence microbial amplicons were taken from coral tissue, mucus and holobiont. The V4 hypervariable region of the 16S rRNA gene was amplified with primers, and then the samples were sequenced on the Illumina platform (Supplementary Table S1).

For the sequences converted to fastq format, Trim Galore v.0.6.6¹ software was used to filter the sequences with low scores and short lengths maintaining a Phred score threshold of 20 and a length threshold of 100 bp. The paired sequences after quality control were combined using FLASH v.1.2.11 (Magoč and Salzberg, 2011) with a merge rate of more than 80%, then the abiotic sequences were removed with VSEARCH v.2.7.0 (Rognes et al., 2016). FastQC v.0.11.9² software was used to evaluate the quality of data, and unqualified samples were removed. Later, samples with sequencing depth less than 5,000 were filtered. Downstream bioinformatics analysis was performed using QIIME2 v.2020.11.1 (Bolyen et al., 2019). High-quality sequences were clustered into ASV, and Greengenes release 13_8 was used to annotate the sequence as the reference database (DeSantis et al., 2006; McDonald et al., 2012).

Prior to calculating the diversity index, samples were randomly rarefied to the minimum library size (5,800 sequences) with QIIME2 v.2020.11.1 due to large differences in library sizes between different samples. Alpha diversity metrics, including the Shannon index and observed features index, were calculated using QIIME2 v.2020.11.1. The mean value and standard deviation of each metric group were calculated by SPSS v.25 software. Differences in beta diversity were computed using Jaccard dissimilarity matrices and tested via permutational multivariate analysis of variance (PERMANOVA). PERMANOVA analysis was performed in R v.4.2.1 using the vegan package. Variation in community composition among samples was visualized with PCoA. Hierarchical clustering analysis was obtained by using the unweighted pair group method with arithmetic mean (UPGMA) clustering in PAST v.4.0 software, and bootstrap values were based on percentage agreement with the resampled ASV table. All βNTI values were conducted using Picante in R v.4.2.1. If βNTI >2 or βNTI < − 2, deterministic processes were important in shaping the community composition, whereas if values were between −2 < βNTI <2, stochastic processes would play an important role. The zero model was used to quantify the deviation from absolute phylogenetic distance to stochastic phylogenetic distance in this method. The greater the degree of deviation, the more the community was affected by deterministic factors, and the smaller the degree of deviation, the more the community was affected by random factors (Kembel et al., 2010). The phylogenetic tree in Figure 4A was constructed with RAxML v.8.2.11 by Quek in previous study (Stamatakis, 2014; Quek and Huang, 2019).

To better understand the potential functional profiles of the specific microbial taxa in reef habitats, Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was applied (Douglas et al., 2020). PICRUSt2 can predict metagenomic functional composition from 16S rRNA gene using marker gene data and a reference genome database. The ASV table of microbial communities with large geographical distances was imported into the “Picrust2_pipelines.py” script in QIIME2 format. The enzyme commission (EC) metagenomes, KEGG orthology (KO) metagenomes and MetaCyc pathway abundances were predicted. The weighted nearest sequenced taxon index (NSTI) score for each sample was calculated (mean NSTI = 0.29 ± 0.16 s.d.). The LEfSe method was used to further analyze the metagenomic prediction count table to identify significantly different functional traits of microbial communities across a large geographical distance [Linear discriminant analysis (LDA) > 3.0 for individual KOs].³ A Venn diagram was drawn to highlight shared functions among the different geolocations.