The microbiomes of two Singaporean corals show site-specific differentiation and variability that correlates with the seasonal monsoons

Corals host abundant microbial communities, or microbiomes, that play essential roles in the function of the coral holobiont. We examined the mucus microbiome in corals within the port of Singapore, where corals persist despite intense anthropogenic impacts. The coral mucus microbiomes of Pectinia paeonia and Platygyra sinensis at three reef sites were tracked by 16S rRNA gene amplicon sequencing from January 2019 to January 2020. Both coral species displayed spatial and temporal differences in microbiome composition, suggesting site specificity and seasonality in microbiome composition consistent with the monsoons. The temporal shifts in relative abundance of dominant taxa were different between the two coral species. Nonetheless, Proteobacteria was the most abundant phylum in both coral species and was reduced during the southwest (SW) monsoon, while Cyanobacteria and Crenarchaeota increased. The presumptive beneficial endosymbiont Endozoicomonas was only associated with corals at the reef site located the farthest from the Singapore mainland. The coral microbiomes reflected seasonal changes, while the seawater displayed distinct temporal microbial compositions and site-specific differentiation within all sampling dates. The persistence of coral reefs within the port of Singapore highlights the adaptive ability of corals to respond to stressful environments, and this study provides further evidence that a flexible microbiome could be an important part of the strategy employed by corals to remain resilient.


Introduction
Corals host a diverse microbial consortium, or microbiome, consisting of dinoflagellate Symbiodiniaceae, Bacteria, Archaea, viruses and fungi (Blackall et al. 2015;Bourne et al. 2016). The coral microbiome is suggested to contribute a number of essential functions for its host, including nutrient cycling, the breakdown of dimethylsulfoniopropionate (DMSP) and immune responses (Bourne et al. 2016;Moynihan et al. 2022;Peixoto et al. 2021). As corals experience global declines, particularly due to climate change, there is increasing value in understanding the extent to which the microbiome supports coral health, or if microbiome manipulation, through inoculation with presumptively beneficial bacteria, could provide a mechanism of enhancing coral resilience to environmental stressors (van Oppen et al. 2017;Webster et al. 2018;Voolstra and Ziegler 2020;Peixoto et al. 2021).
Coral microbiomes are species-specific (Pollock et al. 2018;Speare et al. 2020), although intraspecific variation can result from biogeographic differences (Klaus et al. 2007;Ziegler et al. 2017;Wainwright et al. 2019;Deignan and McDougald 2022) and environmental perturbations (Klaus et al. 2007;Ziegler et al. 2016Ziegler et al. , 2019Leite et al. 2018;Claar et al. 2020). Within a coral, the microbiome can differ by microhabitat of the tissue, skeleton, gut cavity or mucus (Rohwer et al. 2002;Pollock et al. 2018) The mucus microbiome displays comparatively greater responsiveness to environmental changes than the tissue or skeleton but has been shown to shift in succession with the tissue microbiome to thermal stress (Lee et al. 2015;Pollock et al. 2018;Marchioro et al. 2020). Microbiome stability was first suggested to aid corals during stress exposure (Grottoli et al. 2018;Hadaidi et al. 2017), but more recently, the microbiome flexibility hypothesis proposed that the ability of the microbiome to respond to changes in the environment enhances coral resilience (Reshef et al. 2006;Voolstra and Ziegler 2020).
While the rapid response of the microbiome to environmental stress has been well-documented, long-term patterns in microbiome composition are less well-characterized. On tropical reefs, long-term monitoring has shown temporal stability in microbiome composition (Epstein et al. 2019a;Dunphy et al. 2019), with seasonal changes correlating to a small extent with rainfall (Li et al. 2014). On temperate reefs with contrasting summer and winter conditions, seasonal shifts in microbiome composition are more common Roder et al. 2015;Sharp et al. 2017;Palladino et al. 2021;Yu et al. 2021), although this pattern is not universally observed (van de Water et al. 2018). Understanding the natural fluctuations in coral microbiomes will help better diagnose stress in corals and can serve as a baseline of the microbial community before attempts at inoculation to improve coral resilience.
The port of Singapore is a highly urbanized environment with most reefs constrained to < 10 m depths due to high sedimentation rates and subsequent low light attenuation (Bauman et al. 2015;Guest et al. 2016;Sin et al. 2016;Chou et al. 2019). Singaporean waters are a confluence of regional waters merging between the Andaman Sea via the Malacca Strait to the west and the South China Sea to the east, with daily tidal currents mixing local waters (Sin et al. 2016). Singapore experiences two monsoon periods, a northeast (NE) monsoon from December to mid-March and a southwest (SW) monsoon from June through September, separated by two inter-monsoon periods (Sin et al. 2016) The monsoons influence local water currents by increasing the westward flow of water, particularly during the northeast Monsoon (Sin et al. 2016). The seasonality of the two monsoons results in a bimodal pattern for many water quality parameters in Singapore including temperature, salinity, light attenuation, dissolved organic matter (DOM), dissolved inorganic nitrogen (DIN) and phosphate (Chénard et al. 2019;Tanzil et al. 2019;Martin et al. 2021). The SW monsoon is generally characterized by higher seawater temperatures and nutrients than the NE monsoon. Seasonal changes in surface seawater microbial communities have been detected in the Singapore Straits consistent with the monsoons (Chénard et al. 2019). However, no studies have looked for seasonal patterns in coral microbiome compositions within Singapore.
Differences in coral microbiome community composition have been detected for the corals Pocillopora acuta and Merulina ampliata across reefs within Singapore (Wainwright et al. 2019;Fong et al. 2020;Deignan and McDougald 2022), but it is unknown if other coral species follow the same biogeographic pattern of microbiome differentiation or if the spatial differentiation is maintained over time. In this study, we monitored the mucus microbiomes of two coral species, Pectinia paeonia and Platygyra sinensis. We sampled coral colonies on seven occasions encompassing both monsoon seasons at three different reef sites to track the spatial and temporal composition of the microbial communities to better understand how coral microbiomes are adapted to the urban reef environments of Singapore.

Methods
Six colonies each of Pectinia and Platygyra were identified morphologically and selected at random from Pulau Hantu (1°13′34″ N, 103°44′46″ E), Kusu Island (1°13′32″ N, 103°51′35″ E), and Pulau Satumu, or Raffles Lighthouse (1°09′39″ N, 103°44′26″ E; see Supplementary Information SI Fig. 1). All colonies were marked with color-coded tags for repeatable sampling. Pectinia and Platygyra were selected for this study because they were readily abundant on the study reefs and consistently produced sufficient mucus to allow for collection without resorting to destructive sampling methods. From January 2019, each reef was visited every 2 months until January 2020 (Fig. 1). Each sampling event was split over two consecutive days. For coral microbiome collection, mucus was sampled from the colony surface by agitating ~ 5 cm 2 with a 5-mL syringe. The syringes were placed upright on ice for transport to the laboratory, where the mucus was centrifuged for 20 min at 13,000 RCF and stored at − 80 °C. Additionally, triplicate 1 L seawater samples were collected at each site 0.5 m above the reef benthos and placed on ice for transport to the laboratory for filtration through a 0.2 µm cellulose acetate filter (Sartorius Stedim Biotech). All filters were immediately stored at − 80 °C. Approximately, 10 mL of filtered seawater was collected from each site in an acid cleaned falcon tube and flash frozen in a liquid nitrogen dry shipper for nutrient analysis on a SEAL AA3 autoanalyzer at the Asian School of the Environment at Nanyang Technological University. Seawater temperature was recorded via HOBO data loggers deployed on the reef flat at each site (Jenny Fong, National University of Singapore).

Coral genotyping
Each coral colony (18 Pectinia and 18 Platygyra) from the three reef sites was included in the molecular analysis. DNA was extracted from the coral mucus samples with the Qiagen DNeasy PowerBiofilm Kit followed by the Zymo Genomic DNA Clean & Concentrator™ Kit. DNA purity was measured spectrophotometrically with NanoDrop (Thermo Scientific) and quantified with Qubit™ 2.0 fluorometer. Two molecular markers were selected for species identification, the mitochondrial noncoding intergenic region (hereon referred to as IGR) and nuclear ribosomal DNA containing internal transcribed spacers 1 and 2 and 5.8S rDNA (hereon referred to as ITS). IGR were amplified with Faviidae specific primers MNC1f (5′-GAG CTG GGC TTC TTT AGA GTG-3′) and MNCIr (5′-GTG AGA CTC GAA CTC ACT TTTC-3′) in a total PCR reaction volume of 50 µL, with 50 ng of genomic DNA, 1 X KAPA HiFi HotStart ReadyMix kit (KAPA BIOSYSTEMS) and 0.2 µM of forward and reverse primers. The PCR amplification cycle, modified from Huang et al. (2009), consisted of 5 min at 95 °C for 1 cycle, followed by 35 cycles of 94 °C for 45 s, 50 °C for 45 s and 72 °C for 1.5 min, with a final extension of 72 °C for 5 min. Coral-specific forward primer A18S (Takabayashi et al. 1998) and universal primer ITS-4 were used for amplification of ITS in a total PCR reaction volume of 50 µL, with 25 ng of genomic DNA, 1 X KAPA HiFi HotStart ReadyMix kit (KAPA BIOSYSTEMS) and 0.2 µM of forward and reverse primers. The amplification conditions were set at 5 min at 94 °C for 1 cycle, followed by 30 cycles of 95 °C for 60 s, 55 °C for 60 s and 72 °C for 60 s, with a final extension at 72 °C for 10 min. PCR products were purified with Zymo Clean-up and Concentrator kit and sent for DNA sequencing via Applied Biosystems 3730xl DNA Analyzer after processing with BigDye Terminator v3.1 Cycle Sequencing Kit. Sequences were edited, aligned and concatenated using Geneious Prime. The coral ITS is known to have intra-colonial (van Oppen et al. 2000(van Oppen et al. , 2002Fukami et al. 2004;Lam et al. 2006) and intragenomic variation (van Oppen et al. 2000;Lam et al. 2006;Huang et al. 2011). Therefore, three to six clonal sequences were subsequently aligned at 50% threshold for each ambiguous sample to derive a consensus sequence with no ambiguities. IGR and ITS sequences were trimmed after alignment to exclude any missing data and the final length for analysis was 928 bp and 586 bp for Pectinia, 764 bp and 674 bp for Platygyra, respectively. Identity of coral sequences were confirmed via BLAST. Sequence reads were uploaded to GenBank under Accession numbers OL913748-65 (Pectinia ITS), OL913769-86 (Platygyra ITS), OM154196-213 (Pectinia IGR), and OM154178-95 (Platygyra IGR).
Genetic diversity among sites was calculated using DnaSP v 6.0 (Rozas et al. 2017). Genetic differences among and between populations were analyzed by computing pairwise FST and analyses of molecular variance (AMOVA) implemented in Arlequin version 3.0 (Excoffier et al. 2005).
Considering that ITS was present in multiple copies in several samples and can sometimes have higher intra-than inter-colonial variation (Lam et al. 2006), only IGR was used for population genetic analysis. Monthly mean seawater temperature at each reef site with the sampling dates on the x-axis. The northwest monsoon is indicated by the light gray color block, and the southwest monsoon is indicated by the dark gray color block Spatial and temporal microbiome analysis DNA extraction from coral mucus and seawater was conducted using the Qiagen DNeasy PowerBiofilm Kit followed by the Zymo Genomic DNA Clean & Concentrator™ Kit for prokaryotic community analysis. Extracted DNA was stored at − 20 °C. For examining the prokaryotic portion of the microbiome, PCR was run with 10 µL HotStarTaq Plus Master Mix (Qiagen), 1 µL each of 10 µM forward and reverse primers, 4 µL of RNAse-Free water, 1 µL of 100% dimethyl sulfoxide (DMSO), 1 µL of 200 ng/µL BSA, and 2 µL template DNA (5 ng/µL). The 515F (Parada) and 806R (Apprill) primers were used to amplify the V4 region of the 16S rRNA gene (Caporaso et al. 2011;Apprill et al. 2015;Parada et al. 2016). Triplicate PCR reactions were run using the following conditions: initial denaturation at 95 °C for 5 min, followed by 35 cycles of 94 °C for 30 s, 53 °C for 40 s, and 72 °C for 1 min and a final extension of 10 min at 72 °C. Triplicate samples were pooled, and a bead clean-up was done using Agencourt Ampure XP (Beckman Coulter) to isolate the targeted V4 region. Finally, all samples were quantified using a Qubit 2.0 fluorometer and quality checked on an Agilent 2200 TapeStation before library preparation and amplicon sequencing on an Illumina MiSeq platform at the Singapore Centre for Environmental Life Sciences Engineering (Nanyang Technological University). The samples were multiplexed and run on three consecutive runs to ensure sufficient sampling depth across all samples. Raw sequencing reads were uploaded to the NCBI Sequence Read Archive (SRA) under BioProject PRJNA791535.
Amplicon sequencing data were processed using Dada2 version 1.16 (Callahan et al. 2016) to generate amplicon sequence variants (ASVs) for each sample replicate. Firstly, forward and reverse sequence reads were merged by sample to combine the data output from the three sequencing runs into a single set of forward and reverse sequences per sample. Sequence reads were trimmed to 240 for the forward read and 210 for the reverse read, then filtered with an expected error rate of 2. Error learning algorithms were applied to the forward and reverse reads before reads were merged, and chimeric sequences were removed (SI Table 1). Contaminating sequence reads based on comparison with blank extractions were removed using the decontam package (Davis et al. 2018). Taxonomy was assigned to the genus level using the built in Bayesian classifier assignTaxonomy based on the SILVA SSU r138.1 database. Sequences identified as mitochondria, chloroplast or unassigned to a Domain were removed. Finally, rarefaction curves were used to assess samples which reached their ASV maximum (SI Fig. 2). Samples were then rarefied to 29,184 sequence reads per sample to account for variation in sequencing depths for the calculation of diversity metrics (Weiss et al. 2017). Alpha diversity metrics, including observed richness, Chao1, Shannon diversity (H), and inverse Simpson, were calculated using phyloseq (McMurdie and Holmes 2013) and compared by site and date for each species with Kruskal-Wallis tests following Shapiro-Wilk tests for normality. Follow-up Dunn tests with Bonferroni corrections were used for pairwise comparisons.
To examine any patterns within the coral-associated prokaryotic communities, non-metric multidimensional scaling (nMDS) plots were created using Bray-Curtis similarity matrices of square root transformed data with vegan and plotted with ggplot2 in R v3.4.3. Permutational multivariate analyses of variance (PERMANOVA) were performed using the adonis2 function to examine differences in community composition in the samples. The full experimental design was tested for species, site, and date and modeled for repeated measures to account for resampling of the same coral colonies over time. PERMANOVA was also run for each species and the seawater by site and date. Significance values were based on 9999 permutations, with a Bonferroni correction for all pairwise comparisons. Permutational multivariate analysis of dispersion (PERMDISP) was used to test for homogeneity of dispersion among sites for Pectinia and Platygyra using betadisper. Two factor PERMANOVA was employed to test for within-site temporal differentiation for Pectinia and Platygyra, and within-date site differentiation for the seawater samples.
To further examine site differentiation, an indicator species analysis was applied to Pectinia and Platygyra using the multipatt function in the indicspecies package based on a point biserial correlation coefficient function and 9999 permutations (De Cáceres and Legendre 2009). This analysis identifies indicator species, or in this case ASVs, based on both ASV relative abundance and fidelity to determine ASVs associated with a given group or location. Indicator ASV analysis was also applied to Pectinia and within-site for Platygyra for the peak of the NE monsoon (January 2019 and January 2020) and the SW monsoon (July 2020). Results focused on ASVs identified as highly significant (p < 0.01). Alluvial plots of indicator ASV taxonomy were created using RAWGraphs (Mauri et al. 2017). Seasonal changes in the coral microbiomes were further examined by correlating microbiome composition with environmental changes. The envfit function from vegan was used to correlate environmental parameters with sampling date for Pectinia and the seawater and for sampling date withinsite for Platygyra. The environmental parameters tested included nitrate, nitrite, ammonia, phosphate, silicate, and mean monthly seawater temperature. Only significant correlations are reported for each group (P < 0.05).

Coral phylogenetics
Phylogenetic analysis was unable to draw clear species distinction due to paraphyletic clades in both corals. Pectinia and Platygyra samples both generated poorly supported clades with Pectinia lactuca and P. alcicornis (Fig. 2a), and Platygyra daedalea and P. acuta nested within each (Fig. 2b), respectively. This could be due to low interspecific distances within mt IGR in some corals, especially Platygyra spp. (Huang et al. 2009) and high intraspecific variability of the ITS rDNA (Takabayashi et al. 1998;Vollmer and Palumbi 2004;Wei et al. 2006). Although the interspecific distance has been shown to be high for Platygyra (Wei et al. 2006), our un-concatenated ITS dataset was also unable to yield a resolved tree. Further, our Pectinia samples, which were morphologically identified with high confidence as Pectinia paeonia, were the most different from the P. paeonia sequence obtained from GenBank and genetically closest Fig. 2 a Maximum likelihood tree for Pectinia generated with concatenated mt IGR and ITS rDNA sequences, including one representative sequence of Pectinia paeonia and two Pectinia species from GenBank (black circles), with Mycedium elephantotus as outgroup. b Maximum likelihood tree for Platygyra generated with concatenated mt IGR and ITS rDNA sequences, including one representa-tive sequence of Platygyra sinensis and two Platygyra species from GenBank (black circles), with Leptoria phrygia as outgroup. Specimen names starting with 'H' indicate Hantu, 'K' from Kusu, and 'R' from Raffles. Distances calculated with Tamura-Nei Neighbor-Joining tree with 1000 bootstraps. Bootstrap values > 50% are given above branches to P. lactuca, with most of the difference residing in ITS rDNA. However, the specimens used for this study were closely related to each other with variability within Pectinia at 1.1% and within Platygyra at 3.9%. Therefore, we continue to refer to each species as P. paeonia and Platygyra sinensis given the morphological identifications and the poor species distinction arising from the low resolution of the IGR and ITS genes.
There were five Pectinia and six Platygyra haplotypes. Haplotype diversity for both Pectinia and Platygyra were high at all reef sites (0.600-0.800) and among sites (0.614, 0.797), and nucleotide diversity was low (0.001-0.004). For both coral species, Kusu had the highest haplotype diversity, followed by Raffles then Hantu (SI Tables 2 and 3). Both coral species had one haplotype shared across all three reef sites. Kusu and Hantu each contained a unique Pectinia and Platygyra haplotype. AMOVA analysis revealed a well-connected population indicating that the different reef sites are not genetically differentiated. Pectinia and Platygyra both had higher within population (101.82, 88.99%) than among population (− 1.82, 11.01%) variation and nonsignificant (P > 0.05) F values between sites (SI Tables 4-9). While the results are consistent with expectation (Tay et al. 2015(Tay et al. , 2016, higher sample numbers would be necessary to resolve species distinctions and population connectivity for both coral species.

Spatial and temporal microbiome analysis
In total, 125 samples of each Pectinia and Platygyra and 63 Seawater samples were processed for microbiome analysis. Across all samples, 42,442 unique ASVs were identified (23,306 for Pectinia; 25,913 for Platygyra; 4725 for seawater). There were no differences in any alpha diversity metrics by site or date for Pectinia (SI Table 10). For Platygyra, there were no differences in richness, Chao 1, or Shannon diversity by site; however, the inverse Simpson diversity was lower at Hantu compared to Kusu and Raffles. When examining the Platygyra samples by date, the richness and Shannon diversity were lower in July compared to the January timepoints (SI Table 11). Overall, Platygyra had a higher alpha diversity than Pectinia for all metrics (P < 0.0001; SI Table 12). For the seawater, there were no differences in alpha diversity metrics by site. However, there were inconsistent differences by date (SI Table 13). For example, richness was higher in September, March had a lower Shannon diversity, and the inverse Simpson was lower in March and November, particularly compared to January and July.
In Pectinia, this was dominated by members of the order Rhodospirillales (SI Table 16), while in Platygyra, the more abundant orders within the Proteobacteria included SAR11 clade, Pseudomonadales, Rhodobacterales, Rhodospirillales, and Rhizobiales (SI Table 17). The second most abundant phylum for Pectinia was Bacteroidota, which was highest at Hantu, and was dominated by order Flavobacteriales. For Platygyra, the second most abundant phylum was Cyanobacteria, which was highest at Raffles, and dominated by both Synechococcales and Cyanobacteriales. Indicator species analysis for Pectinia found 46 indicator ASVs for Hantu, 34 for Kusu and 33 for Raffles ( Fig. 4; SI Table 18). For Platygyra, there were 62 indicator ASVs for Hantu, 73 for Kusu, and 39 for Raffles ( Fig. 5; SI Table 19) with the greatest number of different taxa identified at Kusu. Both Pectinia and Platygyra at Raffles had members of Endozoicomonadaceae identified as indicator ASVs, however the proportion was higher for Platygyra. Pectinia contained potentially pathogenic Simkaniaceae as indicator ASVs from Hantu and Raffles. Members of Terasakiellaceae were identified for both Pectinia and Platygyra as indicator ASVs at Hantu and to a lesser extent at Kusu but was not identified for either coral species at Raffles. For Platygyra, there was a significant interaction between site and date in the two-factor PERMANOVA, indicating temporal differentiation within each site (Pseudo-F = 1.1238, P = 0.0026). However, this within-site temporal effect was not observed for Pectinia (Pseudo-F = 0.8833, P = 0.9983). For Pectinia and within-site for Playgyra, January timepoints showed the greatest differentiation from July, suggesting that the coral microbiomes change consistently with the peak of the monsoon seasons ( Fig. 6; SI Table 20). Temporal shifts in the dominant phyla were not always consistent across sites (Fig. 7). However, general trends in the relative abundance of the dominant phyla included increases in Proteobacteria, Bacteroidota, Actinobacteria, Planctomycetota, and Verrucomicrobiota during the NE monsoon (peak in January) and increases in Cyanobacteria, Crenarchaeota and Firmicutes during the SW monsoon (peak in July). Indicator species analysis revealed a low number of ASVs representative of the monsoon peaks in January and July (SI Table 21), suggesting that the temporal changes in microbiome composition were the result of a shift in the relative abundance of the dominant taxa rather than the induction of new microbial members in the community.
As for the corals, seawater microbiomes were dominated by Proteobacteria and Cyanobacteria. Cyanobacteria abundance was lowest in January, inversely correlated to an increase in Proteobacteria (Fig. 8b). Crenarchaeota and Thermoplasmatota increased with the SW monsoon season. The seawater microbiome composition did not differ by site (Pseudo-F = 0.8091, P = 0.7467) but displayed strong temporal differentiation for all sampling dates (Pseudo-F = 11.483, P = 0.0001; SI Table 22). Two-factor PERMANOVA for date and site found significant differentiation by site within each sampling date (Pseudo-F = 1.5841, P = 0.0006; Fig. 8a).
Seawater nutrients varied throughout the year, which was consistent with the changing seasons (SI Fig. 3). The envfit analysis for Pectinia found no correlation of environmental parameters with sampling date (P = 0.3031). For Platygyra, microbiome differences between sampling dates were correlated with mean monthly temperature at Hantu (Fig. 6b), while at Kusu, differences were correlated with nitrate, nitrite, ammonia, phosphate and mean monthly temperature (Fig. 6c). Nitrite, ammonia, phosphate and mean monthly temperature were correlated with microbiome differences at Raffles (Fig. 6d). Mean monthly seawater temperature was associated with months during the SW monsoon, along with nitrite and phosphate at Kusu and Raffles and nitrite at Kusu. Ammonia was correlated with sampling dates during the NE monsoon at Kusu and Raffles. The seawater microbiome composition was significantly correlated with all environmental parameters. Notably, mean monthly temperature and phosphate were correlated with the SW monsoon, while ammonia and silicate were indicative of the NE monsoon (Fig. 8a).

Spatial and temporal variability in coral microbiomes
The mucus microbiomes of Pectinia and Platygyra exhibited spatial and temporal differentiation in prokaryotic Phyla with a relative abundance < 0.01 were grouped as Rare taxa community composition. Genotyping analysis of the two coral species gave assurance that the site differentiation did not arise by genetic differentiation, as was previously observed in Acropora tenuis (Glasl et al. 2019a) and Millepora corals (Dubé et al. 2021). Both Pectinia and Platygyra had the highest haplotype diversity at Kusu, which is hypothesized to receive more external larval input than the other sites (Tay et al. 2015;Afiq-Rosli et al. 2021). However, both corals are broadcast spawning hermaphrodites (Baird et al. 2009) and displayed high genetic connectivity consistent with expectations for marine broadcast spawners in Singapore (Tay et al. 2015(Tay et al. , 2016. The temporal variation in coral mucus microbiomes was consistent with the monsoon seasons, with sampling that occurred at the peak of each monsoon (January and July) displaying the greatest overall differentiation. Pectinia microbiomes collectively displayed significant temporal shifts; however, those temporal shifts were not significantly associated with any of the environmental parameters measured. Platygyra displayed more specificity in its response to seasonal changes, experiencing significant within-site temporal changes at each of the three reefs and association with specific environmental parameters. Thus, Platygyra reflected a greater degree of environmental responsiveness, while Pectinia reflected more stability.
Microbiome flexibility is proposed as a key mechanism for coral adaptive resilience to environmental changes (Voolstra and Ziegler 2020). The mucus fraction of the coral is more responsive to environmental changes than the tissue or skeletal fractions (Pollock et al. 2018;Marchioro et al. 2020), but it still maintains coral host specificity (Pollock et al. 2018) and can shift in succession with  (Lee et al. 2015) The mucus microbiome contributes to coral health by creating a selective environment, which supports a distinct microbial community and acts as a first line of defense against pathogens (Rohwer et al. 2002;Shnit-Orland and Kushmaro 2009;Glasl et al. 2016). Therefore, we suggest that the mucus communities presented here reflect a valuable indicator of the impacts of environmental conditions on the coral holobionts. Pectinia and Platygyra mucus microbiomes conformed to both spatial and temporal environmental fluctuations. Both species are common on shallow impacted reefs in the Pacific (Dikou and Woesik 2006;Browne et al. 2014;Ng et al. 2016;Chow et al. 2019), and a flexible microbiome could be one of their mechanisms to tolerate these harsh environments. In particular, Platygyra is a slow-growing, massive coral (Darling et al. 2012) and grows in large colonies (frequently > 50 cm diameter) across Singapore, a trait associated with high microbiome flexibility (Ziegler et al. 2019). Correspondingly, this species showed a high degree of specificity in fluctuations of its microbiome to the environment, reflecting within site temporal differentiation correlated with specific seawater parameters. The microbiome of Platygyra was also characterized as having a higher diversity than Pectinia. Pectinia is a fast-growing coral (Ng and Chou 2014), which has been associated with lower microbiome flexibility. The adaptive microbial community, particularly of Platygyra, could also be a contributing factor in the resilience of this species to tolerating biotic stressors present within impacted reef systems, like regular competitive interaction with algae (Fong and Todd 2021). However, in Singapore, Pectinia has been reported to have higher resilience to bleaching than Platygyra (Ng et al. 2020).
Of the three reefs examined in this study, Raffles is located farthest from the mainland of Singapore, ~ 14 km. For this reason, it is thought to experience reduced anthropogenic impacts compared to Hantu or Kusu, which are located ~ 7.5 km and 5 km away from the mainland, respectively. Indicator species analyses supports this assertion. Endozoicomonas is considered a beneficial coral endosymbiont, occupying over 90% of bacterial relative Temporal differences are reflected in the relative abundance of phyla in seawater at each sampling date. Phyla with a relative abundance < 0.01 were grouped as Rare taxa abundance in some corals (Bayer et al. 2013;Epstein et al. 2019b;Maher et al. 2020). However, Endozoicomonaceae are frequently reduced or outcompeted under environmental or anthropogenic stress (Roder et al. 2015;Ziegler et al. 2016;McDevitt-Irwin et al. 2017 (Kelly et al. 2014) and heat stressed corals (Gajigan et al. 2017), were found as indicator ASVs present in noticeably higher abundance at Hantu.

Spatial and temporal variability in seawater microbiomes
In addition to spatial differentiation, coral microbiomes shifted in relative abundance of the dominant taxa in response to environmental changes associated with the peak of the seasonal monsoons. During the SW monsoon in Singapore, dissolved inorganic nitrogen (DIN), phosphate and silicate are elevated (Sin et al. 2016;Chénard et al. 2019;Tanzil et al. 2019) and the nutrient parameters measured in this study were consistent with previous observations. Seawater temperatures peak during the inter-monsoon periods (Sin et al. 2016), suggesting that seawater temperatures alone are not enough to drive the greatest shifts in coral microbiomes. At both Raffles and Kusu, phosphate, ammonia and nitrite were correlated with seasonal monsoon differences in Platygyra microbiomes, with the additional parameter of nitrate correlating with the SW monsoon at Kusu. Differences in coral microbiomes between Kusu and Raffles were previously linked with their location on either the windward or leeward side of the islands (Wainwright et al. 2019), but this is the first study to correlate those differences with sitespecific microbiome responses to environmental parameters associated with the seasonal monsoons. The temporal shift in seawater microbes was more tightly linked to environmental parameters than those observed for the corals, as previously reported for Porites (Glasl et al. 2019b). The seawater maintained a distinct microbial community at each sampling timepoint and were correlated with all environmental parameters measured. Previously, Chenard et al. (2019) found that DIN, phosphate and silicate were correlated with differences in surface seawater microbial communities from the Singapore Straits during the SW monsoon. The seawater samples from this study were collected directly above the reef benthos, suggesting perhaps a greater interaction between environmental parameters and the benthic community in shaping the seawater microbiome. However, higher Cyanobacteria and Crenarchaeota relative abundances were detected in both corals and the seawater during the SW monsoon, showing some consistency in the effect of environmental changes on these systems. Increased sampling intervals could help establish whether there is a predictive link between changes in the seawater microbial communities and changes in the coral mucus microbiomes.

Conclusion
Corals around Singapore survive in a highly urbanized reef environment, yet they persist in the face of this anthropogenic disturbance. One source of their resilience could be a flexible microbiome, which was particularly evident for Platygyra. Further research is needed to investigate the functional shifts in the microbiome in response to seasonal changes and the subsequent impact on coral host health. As targeted microbiome manipulation grows in popularity as a potential mechanism for enhancing coral resilience (Epstein et al. 2019c;Rosado et al. 2019;Santoro et al. 2021), it is important to understand the spatial variability and temporal shifts in coral microbiomes to ensure that any manipulation is deployed in accordance with the natural microbiome fluctuations. otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.