Environmental changes such as rising sea surface temperatures (Hoegh-Guldberg et al., 2007; Hughes et al., 2017), ocean acidification (Hoegh-Guldberg et al., 2007), and pollution (Silbiger et al., 2018) have resulted in substantial degradation to coral reefs worldwide. As reefs worldwide change in both structure and function (McFall-Ngai et al., 2013; Thompson et al., 2015; Ainsworth and Gates 2016; Apprill, 2017; Trevathan-Tackett et al., 2019), it is increasingly important to understand the contribution of the microbiome to the health, physiology, and environmental flexibility of coral species. Most studies investigating the coral microbiome have predominately focused on profiling microbial communities via next generation sequencing of 16S rRNA gene amplicons using whole fragments of individual corals, such as sections of branches collected from the colony of a branching coral species. Ultimately, the goal of these studies is to characterize coral-microbe interactions in different coral species (or morphotypes) in response to the combined challenges of environmental change (Stévenne et al., 2021). Considering the unique biotic interactions and dynamic relationships between the microbiome and the environment of morphologically, structurally, and functionally different coral hosts when assessing the response of the coral meta-organism may aid in meeting this goal.

Coral colonies harbor distinct and often complex microbial communities within the niche habitats of the colonial meta-organism (Rohwer et al., 2002; Sweet et al., 2011; Blackall et al., 2015), including the polyp tissue (Mouchka et al., 2010), surface mucus layer (Bythell and Wild, 2011; Glasl et al., 2016), and skeleton (Sweet et al., 2011). Whole coral fragments used for microbial analysis therefore include bacteria not only associated with the coral tissues but also a high number of taxa associated with coral mucus (Shnit-Orland and Kushmaro, 2009; Pereira et al., 2017), the calcium carbonate skeleton (e.g. nitrogen fixing bacteria, Williams et al., 1987; Shashar et al., 1994), and endolithic microbes (Marcelino and Verbruggen, 2016; Marcelino et al., 2017; Marcelino et al., 2018). The tissue and mucus harbor unique reservoirs of prokaryotes, with Apprill et al., 2016 finding 23-49% of sequences identified to be unique to tissues and 31-56% of sequences to be exclusive to mucus. The calcium carbonate skeleton of the coral also harbors a distinct microbial community with up to 80 taxonomic units associated with photosynthetic algae (usually Ostreobium spp: Le Campion-Alsumard et al., 1995; Massé et al., 2018; but see Marcelino and Verbruggen, 2016) and cyanobacteria within the endolithic layer (Ainsworth et al., 2010; Marcelino and Verbruggen, 2016; Marcelino et al., 2017; Marcelino et al., 2018). Interestingly, it has been suggested that in some species of coral the tissue is the microbial niche where the microbiota are responsive to different environmental conditions (Oculina patagonica, Rubio-Portillo et al., 2016; Astroides calcularis, Biagi et al., 2020), highlighting the importance of targeting the specific microbial community of the tissue in studies of coral response to anthropogenic stressors.

While there is growing awareness of the distinct microhabitats found in corals, studies often utilize whole fragments taken from the colonial coral structure to profile the coral microbiome (e.g. Pratte et al., 2018; Epstein et al., 2019). This approach results in a homogenization of the niche-specific microbial communities of the surface mucus layers, tissue, and skeleton (Rosenberg et al., 2007; Ainsworth et al., 2010). The inclusion of multiple microhabitats in whole coral fragments means that the microbial community composition of these samples includes the microbial communities of these microhabitats. Homogenization of microhabitats may therefore affect comparisons between studies describing and interpreting the role of the microbial community of the coral, depending on the objective of the study, with different methods of sample collection and processing capturing distinct compartments across the coral’s microbial landscape (Ainsworth et al., 2015, Neave et al., 2017, Weiler et al., 2018). The actual process of homogenization may also create variability depending on the method employed, e.g. between machine-programmed bead-beating and the manual variability of crushing samples with a mortar/pestle (Hernandez-Agreda et al., 2018), indicating a need for the standardization of homogenization methodology as well as targeting specific microhabitats for comparisons between species. When targeting coral tissues specifically, the effectiveness of tissue removal can vary amongst species, with tissue removal less successful for perforate coral species where the coral tissue penetrates throughout the coral skeleton (≥ 5 mm) (Edmunds, 1994; Edmunds and Gates, 2002). This can lead to potential differences in the microbial community profiles identified between species and subsequently require the application of distinct methodological approaches for different species.

Importantly, all studies comparing coral DNA extraction protocols have been conducted on homogenized whole fragments of coral, and to date there is little known about the importance of DNA extraction protocols on tissue samples. In the present study, we separated fixed coral tissue from whole coral fragments of both Pocillopora and Acropora, two morphologically distinct species, to investigate if methodological procedures for processing the microbial habitat (e.g. tissue) result in bias when interpreting the host microbiome. To further explore how different methods used in coral microbiome studies influence the microbial communities found in coral tissues versus homogenized fragments in two species, we also investigated three different sample processing approaches used in the DNA extraction steps on both coral tissue and coral whole fragments. Previous studies investigating methodological differences between DNA extraction protocols have found amplification protocols can influence the community composition (Santos et al., 2012; Baker and Kellogg, 2014; Weber et al., 2017; Bardenhorst et al., 2021), with some protocols resulting in greater microbial richness and coverage of dominant microbial groups (Weber et al., 2017) and others resulting in differing numbers of consistent amplifications (Baker and Kellogg, 2014). Of these studies, none have compared extraction methodologies on paraformaldehyde preserved tissues. A recent study also found all protocols tested suitable for comparing coral microbiomes (Pratte and Kellogg, 2021). The most common goal within coral microbiome studies is to determine if the composition of the microbiome differs significantly between two or more groups of samples, and these studies demonstrate that it is important to identify and account for sources of variation beyond variables that differ between the study groups. In the current study we aimed to determine how the microhabitat sample type and tissue processing steps, from tissue structures of differing coral genera and morphotypes, influence the resulting coral microbiome dataset. By investigating the influence of sample type/microhabitat, host species/morphotype, and DNA extraction protocols on the amplification of DNA from fixed coral tissues and the generation of microbial datasets, we found the greatest differences between the microbial communities of whole fragments versus tissue biopsies, as well as between the tissue biopsies of different species.

In total, 32,616 sequences from 36 samples and 4 negative controls were generated. Quality control and removal of chloroplasts, mitochondria, unassigned ASVs (classification absent at a phylum level), singletons, and potential contaminants resulted in the retention of 29,940 sequences with a mean of 785 ± 137 reads per sample, ranging from a minimum of 10 reads to a maximum of 4434 reads. Plotting species richness against sequence sample size for data filtered using decontam indicate that adequate depth of sequencing was achieved based on sequencing plateaus (Figure S2). Random subsampling of even numbers of sequences per sample has been criticized in studies with small sample sizes and where the number of sequences per sample vary by 2 or more orders of magnitude (McMurdie and Holmes, 2014), so rarefied data was not used in the present study.

A literature search was conducted to determine reported microbiome sequencing data for relevant coral taxa and methodological approaches used in the current study (Table 1). The total reads and average number of reads per sample were lower than what has been observed for P. damicornis and other species within its species complex (Schmidt-Roach et al., 2014), but still within the range of sequencing depth used for analysis of tissue and whole fragment samples of Pocillopora spp. (e.g. 412 – 22,900, Table 1). Clustering to 99% similarity and removing negative controls yielded 746 distinct amplicon sequence variants (ASVs) for analysis of the microbial communities present after extraction with each protocol or sample type (n = 36 samples). 11 ASVs were removed (from 22 of the 36 samples) as contaminants (Table S1; Figure S3), identified as all sequences more prevalent in negative controls than in samples and reducing total number of reads from 32,616 to 29,940. Between 1-6 ASVs identified as contaminants were removed from each sample. Contaminants were all assigned to the class Alphaproteobacteria and contained the families Caulobacteraceae, Hyphomicrobiaceae, Phyllobacteriaceae, Rhodobacteraceae, Rhodospirillaceae, Acetobacteraceae, Deinococcaceae, and Sphingomonadaceae.

This study aimed to identify sources of variation in microbial datasets generated from different sample types, species, and extraction protocols by comparing methodologies of coral microbiome data generation. In assessing yields, reads, taxa numbers, and diversity between protocols we find that no single extraction protocol is superior for DNA extractions when using whole coral samples, but only one protocol from the current study is suitable for tissue biopsy samples. We also find significant differences between whole fragments and tissue biopsies of Acropora, and between Acropora and Pocillopora tissue biopsies, suggesting that method of sample preparation (whole fragment versus tissue biopsy) does influence the coral microbiome based on the biology of the targeted species. We therefore suggest that future studies interpreting the response of the coral microbiome to change target communities of interest, for example by accounting for morphology when selecting species, and therefore conserve the role, function, and significance of the coral-microbe interactions across the coral meta-organism.

All extraction protocols tested met standard DNA quality standards for bacterial 16S amplification from whole fixed Pocillopora and Acropora fragments, a finding repeated in prior methodology studies (Baker and Kellogg, 2014; Weber et al., 2017; Pratte and Kellogg, 2021). However, in tissue samples from both species, low yields (< 20 ng/µL) and high 260/280 ratios (> 2.3) indicative of possible phenol contamination, DNA shredding, or poor sample quality limited extractions to the Micro protocol. It has been suggested that, in samples with low bacterial biomass in relation to overall sample size and DNA content (such as corals), contaminants may be incorrectly identified as novel experimental findings (Glassing et al., 2016; Olomu et al., 2020). Contaminants identified in blank protocol samples were therefore removed from study samples, reducing total number of reads from 32,616 to 29,940 (8% of reads associated with contaminants). Studies that have identified kit contaminants acknowledge that many of the contaminants detected can also be due to laboratory/cross-contamination or genuinely present in samples (Kellogg, 2019) so we conservatively removed only 11 taxa that were found to be more present in negative controls than in study samples. While read counts were relatively low for all samples, with 22% of samples (8 out of 36) having below 100 reads, these low reads were evenly distributed across protocols, species, and sample types and no clear pattern in read counts emerged. The number of read counts reported in this study includes samples with both large and small amounts of starting material, e.g. 1 cm diameter tissue biopsies (skeletal material removed prior to biopsy) and 3 cm branch fragments (from which 1 – 2 cm were used for DNA extraction). This is reported similarly to other studies where sample sizes with different amounts of starting material are analyzed, e.g. where surface mucus layer, tissue, and skeleton show significant differences from each other (Sweet et al., 2011). Consistently, tissue samples excluding skeletal material show lower read counts than whole fragments or mucus samples (Table 1, Meistertzheim et al., 2016; Zou et al., 2022). Specifically, for P. damicornis, read counts per sample as high as 96,134 counts have been recorded in the literature for whole fragments (Bergman et al., 2021), compared to 9,414 – 12,091 for tissue samples excluding coral skeleton (Li et al., 2021; Zou et al., 2022). Pollock et al. (2018) also found an average of 13,629 reads per sample in the P. damicornis mucus layer and 10,928 reads per sample in the P. damicornis skeleton, compared to only 7,317 reads per sample in the P. damicornis tissue, highlighting how sample size and read count per sample may bias the overall read count. In addition, similar to Clinton et al., 2021 study on fish gills, clear plateaus in the present study were seen in sequencing curves following filtration, suggesting that even where reads were < 1000 an adequate sequencing depth was reached. No significant differences were found between any alpha or beta diversity metrics across protocols for whole fragments, but one taxa differed in relative abundance in Pocillopora whole fragments across protocols. The only taxa observed to differ across protocols for Pocillopora samples, a Devosia spp. found enriched in samples using the Micro protocol, has been identified in the Clade C Symbiodinium core microbiome in corals previously (Lawson et al., 2018), suggesting that protocol choice has little bias on the bacterial community of whole Pocillopora samples but may be better suited to recover sequences of bacteria closely associated with Symbiodiniaceae.

Only the Micro protocol was found to be suitable for use with coral tissue samples. Similarly to Pratte and Kellogg, 2021, all protocols for whole fragments here used bead-beating, which has been found to result in higher DNA concentrations and higher degrees of microbial diversity than other cell lysis methods (Lim et al., 2018; Pollock et al., 2018) and for tissue samples bead beating was excluded. Pratte and Kellogg (2021) also hypothesized that the inclusion of a bead-beating step in all extraction methods contributed to very few significant differences between methods, and we therefore hypothesize that the exclusion of a bead-beating step for the tissue samples (chemical lysis only) could explain the differences in effectiveness between protocols for the tissue samples. However, the addition of a bead-beating step in preliminary trials (Table S2) did not increase yields, suggesting that the chemical lysis steps in the FFPE and RecoverAll protocols were not compatible with mechanical lysis steps for tissue samples in the present experiment. There is substantial evidence in the literature in both coral reefs (Pratte and Kellogg, 2021) and other systems that the effect of extraction protocol on bacterial sequencing result is small, and often nonexistent, when compared to variability across samples, species, and sample types (Wagner Mackenzie et al., 2015; Marotz et al., 2017). By comparing extraction protocols across different sample types, we find here that the effect of sample type (e.g. whole fragment versus tissue biopsy samples) had the greatest influence on extraction protocol success, and therefore more closely examined what influence sample type has on microbial community.

The microbiome of Acropora tissue biopsies was also found to differ from that of the whole coral fragment. The tissue and skeletal microbiomes of scleractinian corals have been found to be dominated by rare taxa and to differ in bacterial abundance (Ricci et al., 2022), so we expect to see some differences between samples including the coral skeleton and only including the tissue. Here we find a higher proportion of Rhodobacteraceae in whole fragments than tissue biopsy samples, possibly due to the inclusion of different microhabitats in the two sample types. Rhodobacteraceae have been shown to demonstrate environmental flexibility (Röthig et al., 2016) and have also been associated with thermally-stressed corals, including Porites (Pootakham et al., 2019), Montipora (van de Water et al., 2017), Acropora (Lee et al., 2017), and Pocillopora (Tout et al., 2015). However, without separating out the microhabitats from the whole fragment, it is impossible to specify what microhabitat certain bacteria are associated with or draw conclusions about the response of a specific microhabitat to possible environmental change. The advantage of tissue samples therefore is that the microbial community is limited to bacteria found in the coral tissue, not composed of a disproportionate number of bacteria from the SML, coral skeleton, and endoliths that may be included in whole fragment samples, and therefore studies can exclusively target the response of tissue-associated microbes to environmental change.

In the current study, tissue biopsies were standardized to a 1.5 mm wide biopsy punch and we find there were significantly more taxa and a higher diversity of taxa associated with Pocillopora tissue samples than Acropora. Distinct morphological, reproductive, and ecological traits between species can influence microbial associations in corals (Hernandez-Agreda et al., 2018; Ricci et al., 2022; Morrow et al., 2022). The species selected in the current study have different morphological structures: Acropora species are perforate, characterized by tissue integrated within the calcium carbonate skeletal matrix (Franzisket, 1970; Leuzinger et al., 2003; Muller-Parker et al., 2015), whereas Pocillopora are imperforate with tissues located superficially (Muller-Parker et al., 2015). Perforate corals are more likely to have concealed tissue areas shaded from a combination of temperature and light stress (Shashar et al., 1997) and have tissues up to five times thicker than imperforate corals (Yost et al., 2013). The functional traits of coral species (e.g. growth form, disease susceptibility) have also been found to have a stronger influence on tissue and skeletal microbiomes than mucus (Pollock et al., 2018), and we therefore hypothesize that distinct differences between host traits explain the variation between the taxa found in Pocillopora and Acropora tissue samples in the present study. In the present study, we observed significantly more taxa in Pocillopora tissues than in Acropora tissues. Pocillopora tissue samples were dominated by the family Rhizobiales, nitrogen-cycling bacteria capable of fixing and providing a source of organic nitrogen when in symbiosis with a plant host (Carvalho et al., 2014). On coral reefs, Rhizobiales have been reported to increase on reefs impacted by stressors such as climate change, water pollution, and overfishing (McDevitt-Irwin et al., 2017). As coral samples were collected from a shallow (1-3 m depth) reef flat during summer, one possibility in the present study is that surface-level Pocillopora tissue was more impacted by light and temperature (and therefore saw an increase in microbial taxa and diversity, e.g. Tout et al., 2015) than Acropora tissue located deeper in the skeleton. Other studies have also amplified Rhiziobiales in association with coral tissue, but not in the whole coral colony community (Ainsworth et al., 2015; Morrow et al., 2022), suggesting that characteristics of P. damicornis tissue results in higher abundances of Rhizobiales observed here. Additionally, it is possible that preservation method (e.g. fixed tissues) did not preserve communities from the surface mucus layer, which is known to be rich in Acroporid corals and has been found to host opportunistic pathogens and coral-associated commensal bacteria (Krediet et al., 2009; Lee et al., 2015; Hadaidi et al., 2017). For example, Ziegler et al. (2019) identified a mean number of 237 OTUs per sample associated with Acropora hemprichii tissues and 159 OTUs per sample associated with Pocillopora verrucosa tissues, but the airbrushed tissue slurry from frozen fragments used for16S rRNA amplicon sequencing included the mucus layer and likely accounted for the higher number of taxa in Acropora samples than Pocillopora. Liang et al. (2017) also found slightly higher mean taxa numbers associated with Acropora spp. tissues (558 OTUs) than Pocillopora spp. tissues (523 OTUs) in samples using airbrushed tissue like Ziegler et al. (2019). In the present study, we observed fewer taxa per sample, likely also due to the lower number of reads per sample. For example, in fixed tissue samples, we found an average of 11 ASVs associated with Acropora samples and an average of 24 ASVs associated with Pocillopora samples. The exclusion of the rich surface mucus layer of Acropora corals in the present study could therefore account for the lower taxa and read numbers observed overall in the present study, as well as the lower number of taxa and taxa diversity observed in Acropora corals.

The communities observed in Acropora and Pocillopora samples were likely also influenced by both preservation method and anthropogenic factors. While Alphaproteobacteria dominated the microbiome of both Acropora and Pocillopora in the present study, Bergman et al. (2021) previously found Gammaproteobacteria to dominate the microbiome of whole Pocillopora fragments. The samples in the present study were collected directly from the reef, whereas Bergman et al. (2021) collected corals and conducted an ex situ tank experiment using multiple warming trajectories, which likely influenced the differences from the microbial communities observed herein due to environmental effects. Additionally, both Acropora and Pocillopora samples were almost entirely lacking Endozoicomonas, a common bacterial associate in corals (Pogoreutz et al., 2018; Epstein et al., 2019; Glasl et al., 2019; Voolstra and Ziegler, 2020) that has been observed to be highly abundant in both Acropora and Pocillopora species (Ziegler et al., 2019). However, Ricci et al. (2022) characterized the microbiome of P. damicornis from Heron Island just a few months before the 2020 bleaching events and found a high abundance of Endozoicomonas. The difference between our results and Ricci et al. (2022) suggests that warming water can disrupt the coral-Endozoicomonas association and is also supported by Botté et al. (2022) results, who found low abundances of Endozoicomonas in P. acuta samples from the GBR. As a loss of Endozoicomonas is often recorded in bleached or diseased corals (Bayer et al., 2013; Glasl et al., 2016), one possibility for the lack of Endozoicomonas recorded herein is that repetitive and severe bleaching on the GBR has greatly reduced populations of coral tissue-associated Endozoicomonas over time.

Additionally, the present study investigated the microbiome of corals using fixed tissue, whereas many studies investigating the Pocillopora or Acropora microbiome analyze whole fragments snap-frozen in liquid nitrogen (e.g. Epstein et al., 2019, Ziegler et al., 2019, Bergman et al., 2021). While Hernandez-Agreda et al. (2018) found composition of microbial communities to be comparable between fixed and snap-frozen coral samples, some low abundance and low occurrence phylotypes were found to be variable between preservation methods. It is therefore possible that variation in the microbial community between studies, as well as number of reads and ASVs, is at least in part due to preservation method. Preservation method is a potential source of bias that should be considered when conducting experiments targeting the coral microbiome. Through investigations of the influence of sample type, host species, and DNA extraction protocols on the amplification of DNA from coral tissues, we find that differences in the microbiome are therefore potentially related to both the structure and preservation of the tissue of the tissue, skeleton, and distinct environmental niches within the two coral types.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, PRJNA802894.

JB, TS, and TA planned the experiments. JB and TS conducted the experiments. JB, TS, and TA contributed to the experiments. JB analyzed the data and wrote the manuscript with advice, contributions, review, and editing from TA, SE, and TS.

This research was supported by an Australian Research Council Discovery Project Grant (DP180103199) and UNSW Scientia Funding to TA. JB was also supported by the UNSW Scientia PhD Scholarship.

This research was conducted at Heron Island Research Station, Lord Howe Island, and the University of New South Wales. We thank A. Fordyce, C. Lantz, T. Moriarty, and C. Page for assistance in both the field and in the lab.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.