Host traits and evolution shape key coral-bacterial symbioses

10 The success of tropical scleractinian corals depends on their ability to establish symbioses with 11 microbial partners. Host traits and evolution are known to shape the coral microbiome, but to what 12 extent they affect its composition remains unclear. Here, by using twelve coral species representing the 13 complex and robust clades, we show that functional traits and host evolutionary history explain 14% of 14 the tissue and 13% of the skeletal microbiome composition, providing evidence that these predictors 15 contribute to shaping the holobiont in terms of the presence and abundance of key bacterial species. 16 Additionally, our study shows that the coral tissue and skeleton are dominated by rare bacteria and the 17 skeleton can function as a microbial reservoir. Together, we provide novel insights into the processes 18 driving coral-bacterial symbioses along with an improved understanding of the scleractinian tissue and 19 skeleton microbiome.


Reproductive mode promotes beneficial functional associations 156
Our data show that tissue and skeletal microbiomes of scleractinian corals were influenced by the 157 reproductive mode (Fig. 3) and the reproductive mode variables (broadcast spawners, brooders and 158 mixed mode) were associated with key holobiont members, including Endozoicomonas, Alteromonas,159 Pseudoalteromonas and Myxococcales ( Fig. 4a and 4b). These bacteria are all known for their CCA analysis showed that some bacterial groups highlighted by these studies (i.e. Acinetobacter spp., 173 Bacillus spp., Caulobacterales, Cryomorphaceae, Endozoicomonadaceae, Pseudomonas spp., 174 Rhizobiales, Rhodobacterales) were likely vertically transmitted in the species analyzed in our study. 175 The correspondence between our findings and past reports shows that coral reproductive mode 176 influences the microbiome composition predictably and that early host-symbiont associations persist 177 across a coral's lifetime. 178 Besides the microbial taxa reported to be vertically inherited in corals before, our results show 179 that reproductive mode correlated with a range of other bacteria, suggesting they may also get 180 transferred from parents to offspring and, based on their roles in other systems, they can be hypothesized 181 to play roles in host development as well. Among the genera we found to be associated with the reproductive modes mixed and brooders, Stenotrophomonas are known to cycle sulfur and nitrogen 183 compounds in plants (Ryan et al. 2009) and Massilia are known to stimulate plant growth by producing 184 compounds like siderophores (Ofek et al. 2012 A closer look at microbiome variability within species showed that even within closely related 195 hosts most bacterial ASVs were rare, leading to wide variability between samples. The skeletal microbiome 196 showed a higher proportion of core members than the tissue microbiome for 8 out of 12 coral species 197 although for some coral species we found a low similarity between their microbiome and that of 257 concurrently sampled seawater and sediment (Supp. Data 5), this by no means implies that 258 environmental acquisition is not an influential process in the establishment of these coral species' 259 microbiome. Given that our seawater and sediment samples were taken over a period of 1 month, it 260 seems likely that a more prolonged sampling would recover a higher fraction of the coral microbiome 261 in these and other potential environmental reservoirs. 262

Unusual suspects show persistent associations with corals 263
Our work identified several bacteria that were consistently associated with corals but are either new to 264 the coral microbiome field or understudied. Cyclobacteriaceae were present in at least one sample of Roseospira has also not been reported in the coral literature before, but we found these bacteria 282 in the tissue and skeleton of three P. lutea, two G. retiformis, one G. tenuidens and one P. daedalea 283 samples. These purple non-sulfur bacteria seem to be able to colonize a diverse range of environments 284 and optimally grow photoheterotrophically (Guyoneaud et al. 2002). Thus, given their ability of 285 utilizing substrates known to be present in corals like acetate (Patton et al. 1977) and glutamate (Su et al. 2018) and using near-infrared wavelengths not absorbed by Symbiodiniaceae, the coral colony could 287 offer an array of microniches where these bacteria's niche preferences are met. 288

289
Through a combination of a homogeneous experimental design that minimizes external biases affecting 290 the microbiome, use of innovative technologies like micro-CT scanning to quantify host traits and the 291 application of a range of statistical analyses, our study allowed us to unravel the structure of the coral 292 microbiome and quantify how it is influenced by host traits and evolution. 293 We showed a significant association of host evolution and traits with the bacterial community, 294 including a range of host-bacterial relationships known to affect holobiont health and functioning. 295 Based on our results, we hypothesize that reproductive mode could aid vertical transmission of bacteria 296 beneficial to corals' development, while skeletal architecture works like a filter affecting bacteria 297 abundance. Although our analysis accounted for some of the most influential processes known to affect 298 the microbiome composition, these could only marginally explain the microbiome variation of tissue 299 and skeleton. A holistic view of the mechanisms determining the holobiont composition will be gained 300 by incorporating the physicochemical and dynamic biochemical environment of the coral colony and 301 its influence on the structure of the microbiome and also by assessing whether the presence of some 302 bacteria (whether dominant or rare) may influence the overall structure of the microbiome. In this study, 303 we provided substantial evidence that coral tissue and skeletal microbiomes are dominated by rare taxa 304 and differ in compositions, but a consortium of bacteria can colonize both compartments and the 305 skeleton could be a microbial reservoir. 306 While our study answers several unsolved questions about the bacterial community structure 307 of scleractinian corals and the mechanisms driving its composition, it also exposes knowledge gaps. 308 Despite our study's focus on commonly studied coral species, we identified several abundant bacterial 309 groups that were not previously reported in coral literature. This highlights that the field of coral reef 310 microbial ecology still presents substantial hiatuses even at the level of characterizing the taxonomic 311 composition of the microbiome and substantial further work will be needed to fully characterize the microbiome, understand its functions in the coral holobiont, its fine-scale distribution in relation to 313 ecological micro-niches and the metabolic hand-offs that happen among microbiome members and with 314 the host. The use of putative beneficial microorganisms has been proposed as a tool to mitigate the 315 increasing pressure of anthropogenic activities on coral reefs (Peixoto et al. 2020), therefore we hope 316 that the detailed knowledge about community structure gained in our study can form the basis for further 317 advances in probiotic strategies to improve coral resilience in future climate scenarios. 318

Material and Methods 319
Sample collection, processing and statistics 320 this study, we took a range of precautions to avoid any potential cross contamination and 333 computationally removed potential contaminants. For instance, we sequenced SSW used to remove the 334 coral tissue contaminants; sampled skeletal fragments 5 mm from the tissue to prevent tissue-skeleton 335 microbiomes cross contaminations; and sequenced control samples taken during the DNA extraction 336 and amplification.

Library preparation, sequencing and initial quality control 338
The total DNA of each coral tissue, coral skeleton, seawater, sediment and control sample was extracted 339 using the Wizard Genomic DNA Purification Kit (Promega). Extractions were also performed on eight 340 blanks taken during both the extraction and amplification protocols. SSW and blanks served as controls. 341 We used a 2-step PCR amplification, the first amplifying the target marker and the second adding 342 Illumina adapters (underlined). The V5-V6 regions of the 16S rRNA were PCR amplified using the 343 were scanned at a 10 µm resolution. Preserved samples were scanned in air and secured within a 377 specimen jar with bubble wrap to prevent movement during scanning. Scans were collected using the 378 datos|x acquisition software (Waygate Technologies) and X-ray energy of 110kV and 300 mA with a 379 tungsten target and 0.1 mm copper filter to pre-harden the X-ray beam. A fast scan setting was used 380 collecting between 1199 to 1798 projections through a full 360° rotation of the specimens, depending 381 upon sample width on the instrument detector, with an integration time of 0.5 seconds per projection 382 leading to a 10 to 15 minute scan time. Large specimens were scanned twice to capture the full specimen 383 structure using a multiscan feature. 384 Micro-CT data was reconstructed using the datos|x reconstruction software (Waygate 385 Technologies) applying an ROI and inline median filter during the reconstruction of the data. 386 Reconstructed data was imported into Avizo version 2019.3 (Thermo Scientific) for analysis. The 387 structure of each coral specimen was evaluated by segmenting three different phases observed in scans, 388 the dense skeletal phase (bright white structure in Supp. Fig. 5a), a lower density organic phase 389 Supp. Fig. 5a) and trapped air within the structure phase (dark gray-black 390 space in Supp. Fig. 5a). The Auto Threshold algorithm of Avizo was used for segmentation of the 3 391 phases (Supp. Fig. 5b), which is based on a factorization method developed by Otsu (1979) and 392 determines the point for segmentation between phases in the grayscale histogram (Supp. Fig. 5c). To 393 determine the total porosity (Vporosity, Eq. 1a) of specimens a sample mask was created by using the 394

(intermediate gray values in
Closing and Fill Holes operations of Avizo on the segmented skeleton (Vskeleton, Eq. 1a) plus organic 395 matter (Vorganic, Eq. 1a) to produce a solid sample mask (Vsample, Eq. 1a) encompassing the boundaries 396 of the sample. The total volume of segmented air (Vpores, Eq. 1b) and organic phase (Vorganic, Eq. 1b) 397 within the sample mask is then taken as a measure of sample porosity (Eq. 1b). Segmented label 398 volumes were calculated using the Volume Fraction algorithm in Avizo, which also determines the 399 volume fraction of each phase relative to the segmented sample mask. To determine whether the 20 400 micrometer resolution scan sufficiently captured micro-porosity within the skeleton the 10 micrometer 401 scans were registered to the 20 micrometers scans using the Register Images algorithm of Avizo. The 402 same 12x12x12 mm 3 ROI was then extracted from the 20 micrometer dataset and trends in porosity 403 compared between the two datasets at different resolutions (Supp. Fig. 5).  . aspera, G. retiformis, G. tenuidens, M. digitata, P. annae, P. australensis, P. daedalea, P. lutea, P. 420 sinensis), brooders (I. palifera, S. pistillata) and mixed (P. damicornis). 421

Statistical analysis 422
The significance level for statistical analyses was 0.05 and unless stated otherwise all analyses were 423 conducted on rarefied ASVs tables (10,000 sequences per sample). Differences in community 424 composition (β-diversity) among coral skeleton, tissue, seawater and sediment microbiomes were 425 computed using center log-transformed Euclidean distance matrices of the rarefied ASV tables. 426 Differences among groups were tested using non-parametric multivariate analysis of variance 427