Epibiont communities on mussels in relation to parasitism and location in the rocky intertidal zone

Abstract The factors shaping host–parasite interactions and epibiont communities in the variable rocky intertidal zone are poorly understood. California mussels, Mytilus californianus, are colonized by endolithic cyanobacterial parasites that erode the host shell. These cyanobacteria become mutualistic under certain abiotic conditions because shell erosion can protect mussels from thermal stress. How parasitic shell erosion affects or is affected by epibiotic microbial communities on mussel shells and the context dependency of these interactions is unknown. We used transplant experiments to characterize assemblages of epibiotic bacteria and endolithic parasites on mussel shells across intertidal elevation gradients. We hypothesized that living mussels, and associated epibacterial communities, could limit colonization and erosion by endolithic cyanobacteria compared with empty mussel shells. We hypothesized that shell erosion would be associated with compositional shifts in the epibacterial community and tidal elevation. We found that living mussels experienced less shell erosion than empty shells, demonstrating potential biotic regulation of endolithic parasites. Increased shell erosion was not associated with a distinct epibacterial community and was decoupled from the relative abundance of putatively endolithic taxa. Our findings suggest that epibacterial community structure is not directly impacted by the dynamic symbiosis between endolithic cyanobacteria and mussels throughout the rocky intertidal zone.


Introduction
Mussels are important intertidal foundation species .T he microbial communities living on or in mussels play a central role in host health as well as nitrogen and carbon cycling (Pfister 2007, Heisterkamp et al. 2013, Pfister and Altabet 2019 ) in coastal ecosystems .T he intertidal zone where mussels live is a heterogeneous habitat with abiotic gradients of temper atur e, salinity, nutrients, ultra violet light, wa ve action and rainfall (Connell 1972, Helmuth and Hofmann 2001, Harley and Helmuth 2003 ).The position of a mussel within the intertidal zone can affect host physiology (Place et al. 2012 ) and biotic interactions (Paine 1974, Lubchenco 1980 ), all of which can influence the epibiotic microbial community.Characterizing the abiotic and biotic drivers of host-microbial associations in foundation species, like mussels, is critical for understanding the contribution of microbes to diverse physiological and ecological functions.It can also aid in predicting the potential outcomes of envir onmental c hange on host-micr obial associations (Wilkins et al. 2019 ).
Mussels have long been a model of ecological zonation in the intertidal (Paine 1966(Paine , 1974 ) ) and are increasingly susceptible to mass mortality e v ents (Harley 2008, Seur ont et al. 2019 ) and other climate c hange-r elated str essors (Frölic her et al. 2018 ).T hus , natural populations of intertidal mussels represent a valuable system for studying host-microbial associations across heterogenous envir onmental gr adients that ar e under heightened pr essur es fr om climate c hange.Pr e vious studies hav e inv estigated the bacterial diversity in the internal tissues and fluids of mussels (Li et al. 2018, Vezzulli et al. 2018 ), especially in relationship to disease-causing agents (Li et al. 2019 ), because of their economic value and impacts on human health (Rubiolo et al. 2019 ).Less is known about the factors shaping the epibiotic shell bacterial community, which serves as an interface or protective cover between mussels and the abiotic environment.
Some constituents of the mussel shell microbial community, endolithic c y anobacteria, hav e attr acted the attention of researchers because of the role these organisms play in host survivability.Endolithic c y anobacteria are parasitic microbes that bore into mussel shells causing decreased shell thickness and str ength.The ener getic costs to r epair shell dama ge can compr omise m ussel gr o wth, b yssal attac hment str ength and incr ease mortality (Kaehler 1999, Zardi et al. 2009 ).The boring activity by endolithic c y anobacteria r emov es the dark, outer periostr acum of mussel shells exposing the light gray prismatic la yer.T he lightercolor ed, er oded shells reflect solar radiation, and the porosity of eroded shells also helps mussels retain more water compared with mussel shells with little erosion (Gehman and Harley 2019 ).
Consequentl y, high le v els of endolithic c y anobacteria infestation and resultant erosion reduce stressful heat gain and mussel mortality during high temper atur e e v ents at low tide (Zardi et al. 2016 , Gehman andHarley 2019 ).T hus , the parasitic shell-boring c y anobacteria become mutualistic in a context-dependent manner during intense thermal stress (Gehman and Harley 2019 ).There is some evidence that infestation by endolithic cyanobacteria increases with elevation in the intertidal zone because photosynthesis is enhanced by the prolonged exposure to light (Marquet et al. 2013 ).The relationship betw een endolithic c y anobacteria abundance, shell erosion and intertidal elevation in the northeastern Pacific is not well understood and the composition of endolithic c y anobacteria assemblages on mussel shells in this region have not yet been examined using molecular tools (Bo w er et al. 2002 ).This study c har acterized the c y anobacterial assemblages on shells of the California mussel, Mytilus californianus , using 16S rRNA amplicon sequencing.It further examined the relationship between live hosts, endolithic parasite identity, shell erosion and the ov er all epibiotic bacterial comm unity acr oss ele v ation gr adients in the r oc ky intertidal zone at four sites in British Columbia, Canada.
Microbial colonization of mussel shells may be constrained by the selective filter of host biology.Host biology has been demonstrated to influence the microbial communities occupying internal compartments of diverse host species (Woodhams et al. 2020 ), as well as the pr otectiv e outer layers of marine hosts like corals (Glasl et al. 2016 ).For m ussels, the bioc hemical composition (Bers et al. 2006 ) and micr otopogr a phies (Bers et al. 2005 ) of an intact periostracum, or waste products excreted by the host (Pfister et al. 2014 ), can influence the microbial taxa that colonize the outer shell.For macroscopic epibionts such as barnacles, growth is significantly faster for individuals that settle on live mussels compared with empty shells, likely because of the increased nutrition deriv ed fr om association with a living host (Laihonen and Furman 1986 ). Alternativ el y, m ussel shells may be surfaces with limited microbial selectivity, colonized by communities whose taxonomic composition, gr owth and pr oductivity ar e strictl y constr ained by abiotic factors (Palinska et al. 2017 ).Experimental manipulation is a necessary tool for determining the role of host biology and abiotic variation in shaping the epibiotic microbial composition on mussel shells.
In this study, we experimentally tested the hypothesis that live mussels influence the composition of epibiotic bacterial assemblages and modulate shell erosion caused by endolithic cyanobacteria.We transplanted minimally eroded pairs of live mussels and shucked mussel shells across an intertidal elevation gradient at four sites and c har acterized the epibiotic community 3 to 5 months later using 16S rRNA gene amplicon sequencing.We used killed, shucked mussels to represent shell substrate in the absence of host biological filtering.We predicted (i) that the shells of tr ansplanted liv e m ussels would be less susceptible to erosion b y endolithic c y anobacteria; (ii) that high shell er osion, whic h can modify host physiology, shell micr otopogr a phy (Zardi et al. 2009 ) and abiotic conditions experienced by the host (Zardi et al. 2016 ) and epibiont communities, would influence which bacteria colonize a mussel shell, and therefore be associated with unique taxa compared with mussel shells with less erosion; and (iii) that the bacterial community composition on live mussels would be less variable than on empty shells when exposed to different abiotic conditions because of the potential buffering capacity provided by a live biogenic habitat.

Experimental conditions
Mussel transplant experiments were conducted in 2017, at four sites around coastal British Columbia, Canada, which are dominated by the California mussel, Mytilus californianus (Fig. 1 ).Two sites were on Calvert Island near the Hakai Institute Ecological Observatory: one on the north side of the island (Crazytown, Table 1 ) and one on the west side of the island (Platform 6 3 4 , Table 1 ).The other two sites were located on Vancouver Island, Otter Point in Sooke and Bluestone Point near the Bamfield Marine Sciences Centre (Table 1 ).
A total of 19 to 23 experimental manipulations were established per site.Small mussels (25-30 mm) with minimal erosion ( < 20%) were collected from low-elevation mussel beds (donor beds) near each transplant site ( Supplementary Fig. S1 ).Individual experimental manipulations contained two mussels: one live mussel set in e po xy in an orientation that allo w ed the shell to continue to open and close, and one mussel that was killed by shucking, which was placed with one half of the empty shell, the exterior of the shell facing outw ar d, set into the e po xy.Experimental manipulations were intentionally spaced over a continuous elev ation gr adient at eac h tr ansplant site, spanning the extent of the natur al m ussel bed, not at fixed interv als ( Supplementary Fig. S1 ).The intent was to ca ptur e natur al abiotic v ariation associated with ele v ation acr oss the extent of eac h m ussel bed.The experimental manipulations were established over a range of dates from April to June 2017 (Table 1 ) because of the logistical challenges in accessing the intertidal zone at remote sites.

Mussel and erosion data collection
To test the hypothesis that live mussels are less susceptible to erosion by endolithic cyanobacteria, we quantified shell erosion on e v ery liv e and empty m ussel shell pair at the start and end of the experiment.Pairs of transplanted shells were photographed during the establishment of manipulations (April-June 2017) and at the conclusion of the study (September 2017).The er oded ar ea of the upper mussel shell and the total area of the upper mussel shell were measured from photographs using ImageJ 1.51n (Schneider et al . 2012 ).We used this datum to calculate the total percentage of shell area eroded during the experiment (see Supplementary Fig. S2 ).The erosion rate was calculated by measuring the total change in eroded shell area from the start to the end of the experiment and then av er a ged to a weekl y r ate (c hange in pr oportion er oded * w eek −1 ).Mussel shell length w as measured in the field and used to calculate the gr owth r ate (mm * week −1 ).The gr owth r ates demonstr ated that liv e tr ansplanted m ussels maintained normal growth during the experiment (Gehman and Harley 2019 ).

Bacterial community sample collection
Samples for epibacterial comm unity anal ysis wer e taken in September 2017 (Table 1 ) at the conclusion of the study, after time for shell erosion by endoliths.We sampled shells in situ for 16S rRNA gene amplicon sequencing fr om eac h tr ansplanted m ussel pair and from unmanipulated mussels at the lo w est extent of the donor bed (live controls) at every site.Sampling involved rinsing each shell with 0.22 μm filtered, sterile seawater for 10 s to remov e tr ansient envir onmental micr obes and then rubbing with a Puritan ® sterile swab for 10 s.Swabs were deposited in individual 2-ml cryovials (VWR), placed in coolers on ice and, upon return to

Abiotic measurements
Ele v ation in the intertidal zone was calculated using tide charts, combined with measures of seawater height and the height of each manipulation against the stationary position of a laser level.On Calvert Island, elevation and tide height were measured using a combination of Real Time Kinematic (RTK) positioning surv ey and dr one-based digital surface ele v ation models.Using a geogr a phic information system, ele v ation data were extracted for all survey areas and temperature loggers not directly measured with the RTK surv ey.Ele v ation was calculated via the Canadian system, which uses the Lo w est Normal Tides as chart datum (Forrester 1983 ).
Roughly one-half of the experimental manipulations at each site were inlaid with iButton temper atur e loggers that recorded the temper atur e hourl y during the study.Temper atur e data wer e extr acted fr om eac h iButton and separ ated into measur ements of exposed substratum (subaerial) and seawater (immersed) based on whether the manipulation was submerged at the time of that measurement using tide charts and the ele v ation (tide height) of the manipulation.For each experimental manipulation, the mean and 90th quantile temper atur e v alues wer e calculated for subaerial and immersed time periods ( Supplementary Fig. S3 ).

16S rRNA gene amplicon sequence processing
DN A w as extracted from swabs using the MoBio Po w erSoil ®-htp 96-w ell DN A extraction kit (Carlsbad, CA, USA) follo wing the man-ufactur er's r ecommended pr otocol.Extr acted DN A w as sent to Integr ated Micr obiome Resource (IMR), Centr e for Compar ativ e Genomics and Evolutionary Bioinformatics (CGEB) at Dalhousie University for PCR amplification and library construction.Primers targeted the V4-V5 region of the 16S rRNA gene for bacteria and arc haea, namel y, 515f: 5 -GTGYCAGCMGCCGCGGTAA-3 and 926r: 5 -CCGYCAATTYMTTTRAGTTT -3 (Comeau et al. 2011(Comeau et al. , P ar ada et al. 2016 ) ). Amplicon library preparation and sequencing with Illumina MiSeq using paired-end (2 × 300 bp) v3 chemistry was performed at the IMR at Dalhousie University, Halifax, Nova Scotia, Canada, according to published protocols (Comeau et al. 2017 ).Quality filtering, trimming, dereplication, chimera removal, inference of true amplicon sequence variants (ASVs) and taxonomic assignment against the SILV A database (v .1.3.2) was carried out with D AD A2 (Callahan et al. 2016 ).For D AD A2 processing, the filter and trim step was set to a minimum read length of 150 bp forw ar d and 120 bp r e v erse.Reads wer e truncated after a quality score of less than or equal to two.Reads with higher than eight forw ar d and 10 r e v erse maxEE "expected err ors" wer e discarded.Chimera detection was performed using the pooled method.Singletons and reads assigned as mitochondria were removed for downstream analyses.Of the experimental manipulations sampled, 10 to 12 transplanted pairs per site had sufficient amplicon sequencing read coverage for paired bacterial comm unity anal ysis.Data fr om manipulations wher e onl y one of the experimental pair had sufficient read coverage were included in all downstream statistical analyses, except those explicitl y comparing differ ences in the composition and r elativ e abundance of taxa between live and empty shells of each experimental pair.

Sta tistical anal yses
All statistical tests were conducted in R (R Core Team, 2021; version 4.1.2). 16S amplicon sequencing data were rarefied to 8500 reads per sample ( Supplementary Fig. S4 ) using the Phyloseq pac ka ge (McMurdie and Holmes 2013 ).To examine the comm unity structur e and r elativ e abundance of putativ e shellboring c y anobacterial taxa, w e anal yzed the r elativ e abundance of reads assigned to the phylum Cyanobacteria in experimental treatments and geographic sites, after excluding chloroplast sequences.16S rRNA gene amplicon sequences from the chloroplasts of microalgae and macroalgae are assigned to the phylum Cyanobacteria.We analyzed the subset of ASVs assigned as c hlor oplasts separ atel y in an attempt to c har acterize eukaryotic algal colonizers on mussel shells, which can also be endolithic (Palinska et al. 2017 ).The relative abundance of c y anobacteria r eads, or c hlor oplast r eads, was calculated as the percentage of total reads for each subset of rarefied data.To test for significant differences in the relative abundance of c y anobacterial or eukaryotic algal sequencing reads on live, empty or liv e contr ol m ussel shells, we used Kruskal-Wallis tests.We subsequently conducted these tests for eac h geogr a phic location independentl y to look for site-specific differences.We also analyzed the combined datasets (i.e.all ASVs assigned to the phylum Cyanobacteria), to look for patterns common to all potentially photosynthetic taxa.We used IndVal anal ysis fr om the indicspecies pac ka ge (Cáceres and Legendre 2009 ) to identify ASVs that were significantly associated with experimental transplants that differed in shell erosion.The IndVal analysis assesses the relationship between ASV occurrence or abundance values from a set of samples and the membership of those samples to gr oups, whic h may r epr esent habitat types, sampling points and experimental treatments , etc. T he method calculates an IndVal index value based on specificity, or the proportion of samples in a group to which the ASV belongs, and fidelity, or the proportion of the counts of that ASV (abundance) that is exclusive to the group (Dufrêne and Legendre 1997 ).An index value is calculated for every ASV, in every group, and the ASVs with the highest association value for a particular gr oup ar e identified as significant using permutation tests.We also used IndVal analysis to identify taxa significantly associated with transplanted versus live control hosts, to characterize any general effects of the experimental manipulation on the pr e v alent and abundant epibiotic shell bacteria.
To test the hypothesis that bacterial communities on live mussel shells are unique in taxonomic structure compared with empty mussel shells, we used PERMANOVA on Bray-Curtis dissimilarity with sampling location as strata.All PERMANOVA tests were conducted with 999 permutations using the vegan package (Oksanen et al. 2019 ).Differences in bacterial alpha diversity between live and empty mussel shells were quantified with paired Wilcoxon tests for each of the four sites.
To understand natural geographic variation in epibacterial comm unities, we compar ed the alpha div ersity between liv e contr ols fr om eac h of the sites with Kruskal-Wallis tests for both richness (total observed ASVs) and Shannon-Weaver index (H').We anal yzed the Br ay-Curtis dissimilarity of liv e contr ol m ussels using PERMANOVA to determine if epibacterial community structur e differ ed significantl y by geogr a phic location.We conducted pairwise comparisons between sites using the pairwise Adonis wr a pper function (Martinez Arbizu 2020 ).
T he en vfit function in the v egan pac ka ge was used to assess the relationship between geographic location, shell erosion and measured abiotic variables with the first two ordination axes of the PCoA plot of epibacterial beta diversity (Bra y-Curtis).T he significance of the relationship, based on m ultiple r egr ession between variables and ordination axes, was assessed with permutation tests .For this analysis , only the subset of the amplicon sequencing data with corresponding iButton temperature data was included.A Mantel test was used to confirm the statistical significance of correlations between continuous abiotic variables and epibacterial community dissimilarity.
Raw 16S rRNA sequence files are available from the European Nucleotide Arc hiv e EMBL-EBI database under project accession PRJEB51984.R code for sequence processing and statistical anal yses is av ailable on GitHub: https:// github.com/katherine-mdavis/Mussel _ shell _ microbiome/.

Results
We sampled the epibiotic bacterial assemblages on 45 live unmanipulated control and 135 experimentally transplanted mussels across four geographic sites in British Columbia, Canada (Table 1 ).A total of 1 530 000 sequence reads belonging to 5733 unique ASVs were k e pt after processing via the D AD A2 pipeline.

Li v e m ussels experience less erosion by endolithic cyanobacteria
We tested whether live mussels modulate erosion caused by endolithic c y anobacteria at four r oc ky intertidal sites .T he area of mussel shell eroded over the course of the experimental period was significantly lower for live transplants compared with transplanted empty shells (Wilcoxon P = 0.00065; Fig. 2 a).Variation in erosion between experimental treatments was driven by significant differences at Otter Point ( P < 0.001) and Crazytown ( P = 0.031) ( Supplementary Fig. S5 ).We used IndVal analysis to identify bacterial taxa differ entiall y associated with live and empty mussel transplants that experienced significant differences in erosion, using only samples from Otter Point and Crazytown.Empty mussel shells with increased erosion were more likely to be colonized by uncharacterized Rhodobacteraceae and Maribacter spp.( Supplementary Table S1 ).Live mussel transplants with significantly less erosion were associated with Maribius spp.( Supplementary Table S1 ).The r elativ e abundance of reads from photosynthetic taxa (c y anobacteria and eukary otic algae) did not differ between live and empty transplants ( P = 0.079; Fig. 2 b and Fig. 3 ), e v en among shells that differed in erosion.

A few taxa dominate cyanobacteria assemblages on mussel shells
We c har acterized the c y anobacterial assemblages on shells of Mytilus californianus to understand the composition and distribution of these organisms in coastal British Columbia ( Supplementary Fig. S6 ).We found the r elativ e abundance of c y anobacteria reads was higher on transplants compared with liv e contr ols at the two sites on Vancouver Island ( Supplementary Fig. S6 ), but there was no difference in the r elativ e abundance of c y anobacteria reads betw een live and empty shells across the dataset ( P = 0.31).There was no significant association between ov er all r elativ e abundance of reads assigned to c y anobacteria and m ussel shell er osion or ele v ation in the intertidal zone ( P = 0.93 and P = 0.98, r espectiv el y; Supplementary Fig. S7 ).The c y anobacteria assembla ges wer e dominated by r eads fr om two gener a, Pleurocapsa and Phormidesmis, which consistently co-occurred at r oughl y the same ratio across the dataset ( Supplementary Fig. S6 ).In a few samples, Pleurocapsa was detected but Phormidesmis was not.The r elativ e abundance of r eads assigned to Pleurocapsa , a confirmed endolithic c y anobacterium, did not differ betw een live and empty mussel transplants (Fig. 2 c) and was not significantly correlated with elevation in the intertidal zone (Pearson's R, R 2 = 0.059, P = 0.5).Phormidesmis is not endolithic.ASVs assigned to the gener a Leptoc hromothrix , Pseudophormidium and Leptol yngb y a (family Phormidesmiales; also a confirmed endolithic c y anobacteria) occurred at low relative abundance ( < 3% of the total community), but comprised most of the other c y anobacteria reads detected ( Supplementary Fig. S6 ).
The 16S rRNA gene primers amplify c hlor oplasts fr om eukaryotic algae, so we examined these communities on mussel shells because some marine eukaryotic algae are endolithic (Marcelino and V erbruggen 2016 ).W e found c hlor oplast r eads wer e significantly higher on experimental transplants compared with live controls ( Supplementary Fig. S6 ) and exhibited a significant negativ e corr elation with tidal ele v ation ( Supplementary Fig. S7 ), but we did not detect any known endolithic taxa.

Mussel shell bacteria are associated with abiotic factors more than host-parasite interactions
We examined patterns of beta diversity across all four sites to understand the contribution of geogr a phic location to variation in the mussel shell bacterial community.Site significantly structured the bacterial communities on shells of unmanipu-lated live control mussels (PERMANOVA: R 2 = 0.200, pseudo-F = 3.49, P = 0.001; Supplementary Fig. S8 ) and pairwise comparisons sho w ed significant differences in composition among all sites, except for betw een Crazyto wn and Platform 6 & 3/4 ( Supplementary Table S2 ).Alpha diversity, as quantified by ASV richness and Shannon-Weaver index (H'), was significantly higher  2 . in m ussels fr om the donor beds on Vancouver Island compared with those from Calvert Island (ASV richness: F = 43.22,P < 0.001; H': F = 27.19,P < 0.001).Site also explained significant variation in bacterial beta diversity of experimentally transplanted mussels (PERMANOVA: R 2 = 0.079, pseudo-F = 5.059, P = 0.005).Experimentall y tr ansplanted liv e and empty m ussel shells did not hav e significantl y differ ent bacterial comm unities in terms of alpha div ersity (Shannon-Weav er index (H'): P = 0.23; ASV richness: P = 0.81) (Fig. 3 ).Beta diversity based on Bray-Curtis dissimilarity was also not significantl y differ ent between tr ansplanted liv e mussels and empty mussel shells (Fig. 4 ; Table 2 ).The bacterial comm unity structur e on tr ansplanted shells, r egardless of host viability, was significantly different from shells of live control mussels (PERMANOVA: R 2 = 0.040, pseudo-F = 7.506, P = 0.001; Table 2 ; Fig. 3 and Fig. 4 ).These patterns of beta diversity were consistent with and without the inclusion of c hlor oplast sequences .T he same patterns were also present when considering only manipulations that were transplanted to low elevations, closer to the unmanipulated live controls ( Supplementary Fig. S9 ; Supplementary Table S3 ).IndVal analysis identified many ASVs, including those assigned to Litorimonas , Lewinella , undescribed Rhizobiaceae, Pleurocapsa and some c hlor oplast ASVs, as significantl y enric hed on the shells of tr ansplanted m ussels compar ed with live controls ( Supplementary Table S4 ).
To understand the effect of abiotic conditions that vary within the intertidal zone on the mussel shell bacterial community, we tested for differences in diversity metrics across tidal ele v ations of tr ansplanted m ussels .T here was a significant negativ e r elationship between alpha diversity and tidal elevation at Otter Point for both live and empty transplants ( Supplementary Fig. S10 ), where the tidal ele v ation r ange of transplants was greatest.There was no significant relationship between alpha diversity and elevation at any other sites .T he relationship between bacterial beta diversity and tidal ele v ation of tr ansplants was significant acr oss the dataset (PERMANOVA: R 2 = 0.047, pseudo-F = 6.64,P = 0.005), indicating that epibacterial beta diversity on mussels at the same tidal ele v ation was similar both within and among sites.Because of unequal replication and range of tidal elevations at each site, we also analyzed the relationship between bacterial beta diversity and tidal ele v ation binned into thr ee categories: low ( < 2 m), medium (2-3 m) and high ( > 3 m).After accounting for differences among sites, ele v ation category was also significantly associated with the bacterial beta diversity of transplants ( Supplementary Fig. S9 ; Supplementary Table S3 ).
To further explore the effects of abiotic variation on the mussel shell bacterial community, we examined the relationship between subaerial (substratum) and immersed (seawater) temper atur es and epibacterial community composition.Temperature data were collected for about one-half of the tr ansplanted m ussel pairs at each site, but there was a significant loss of temperature loggers during the study .Consequently , we examined the subset of transplant pairs with available iButton temperature data (n = 24 pairs; Bluestone Point, n = 6; Otter Point, n = 4; Crazytown, n = 7; Platform 6 & 3/4, n = 7; Supplementary Fig. S3 ).The variables site, intertidal ele v ation, mean immersed temper atur e and 90th quantile subaerial temper atur e explained significant variation in the first two axes of the multidimensional ordination space for Bray-Curtis dissimilarity of bacterial communities on transplanted mussel shells using envfit analysis (Fig. 5 ; Table 3 ).A Mantel test comparing Bray-Curtis dissimilarities and a dissimilarity matrix of ele v ation confirmed significant dissimilarity of shell bacterial communities between mussel transplants of increasing distance apart along vertical elevation gradients in the intertidal zone ( P < 0.001).Measured shell erosion was not significantly correlated to elevation of transplants in the intertidal zone (Pearson's R, R 2 = 0.028, P = 0.75), nor was it a significant predictor of epibacterial community composition in the envfit analysis (Table 3 ).

Erosion activity by endolithic cyanobacteria depends on host condition
We hypothesized that live mussels would be less susceptible to erosion by endolithic cyanobacteria because live hosts have the potential to influence c y anobacteria colonization and erosion activity by modulating the shell surface.Such modulation could come from active shell surface maintenance, alteration of the shell thermal environment via evaporative cooling, or other bioc hemical mec hanisms.In a gr eement with our hypothesis, we found significantly less erosion on live mussel shells compared with empty mussel shells from the same transplant pairs .T his pattern was driven by significant differences in the percentage of shell eroded between live and empty mussel shell transplants at two out of the four sites .T he site-specificity of these results is consistent with findings that endolithic erosion rates can depend on abiotic conditions (Kaehler 1999, Zardi et al. 2009 ).Alternativ el y, abiotic factors that varied among sites could have weakened the periostracum and facilitated erosion on empty mussel shells in a site-specific manner (Kaehler 1999 ).Still, differences in erosion between live and empty mussel shells in some abiotic contexts suggest that live mussels may have the capacity to influence the activity of shell-boring parasites.
Inter estingl y, we did not find a significant difference in the relative abundance of c y anobacterial reads between live and empty mussel shells at any site.We expected that increased shell erosion would be correlated with a more abundant c y anobacteria community.It is possible that our sampling method did not ca ptur e endolithic c y anobacteria cells that were localized inside small bore holes on highly eroded shells, causing us to miss a portion of the c y anobacteria community in the amplicon sequencing data.It is also possible endolithic c y anobacteria colonization and abundance is consistent, but endolithic metabolism differs depending on host condition or site.In such a case, the area of shell eroded would be higher on shells and at sites where cyanobacteria were mor e activ e, while the r elativ e abundance of c y anobacteria detected would not differ.

Known endolithic taxa dominate cyanobacteria assemblages on mussel shells in the northeastern Pacific
The c y anobacteria assembla ges on m ussel shells wer e dominated by Pleurocapsa spp .and Phormidesmis spp.Pleurocapsa have previously been described as members of endolithic assemblages and ar e closel y r elated to endolithic gener a, suc h as Hyella , that ar e responsible for significant shell erosion in other mussel species (Kaehler 1999 ;Brito et al . 2017 ;Ndhlovu et al. 2021 ).Phormidesmis are filamentous c y anobacteria with no record of endolithic activity in marine hosts.Co-occurrence of these two taxa at a relatively consistent ratio on mussel shells, and the infrequent observation of Pleurocapsa in the absence of Phormidesmis , may indicate that Pleurocapsa facilitates shell colonization by Phormidesmis.We could not explicitly test for such successional dynamics in our study de-sign.It would be an interesting avenue of future research to test this hypothesis using more intensive temporal sampling.
It is not clear why the r elativ e abundance of the putative shellboring taxa was not positiv el y corr elated with measur ed shell erosion or intertidal elevation as predicted (Kaehler 1999 ).It is possible that recruitment dynamics of endolithic c y anobacteria might be confounded with the abiotic drivers of erosion activity.Because we selected low ele v ation m ussels with little to no erosion for transplantation, experimental mussels could have had a v ery r educed endolithic comm unity at the start of the experiment.Cyanobacteria detected on transplanted mussels could have recruited to mussel shells over the course of the experiment.In this case, the microbial samples may have been taken before significant erosion took place under site-specific abiotic conditions.Abiotic factors including water movement and air temperatur e hav e pr e viousl y been shown to affect the distribution and activity of endolithic c y anobacteria (Kaehler 1999 ).Nonetheless, this stud y mak es an important contribution showing that Pleurocapsa , a known shell-boring cyanobacterium, dominates the 16S rRNA sequence reads of cyanobacterial assemblages on California mussels in British Columbia, Canada, which had not been pr e viousl y c har acterized using molecular tec hniques (Bo w er et al. 2002 ).
Inter estingl y, neither of the dominant c y anobacteria w e detected on Mytilus californianus are common to epibiotic and endolithic assemblages described from other mussel species .T he endolithic communities on M. galloprovincialis and Perna perna are fr equentl y comprised of three to seven species of c y anobacteria fr om the gener a Hormathonema , Hyella , Kyrthutrix , Plectonema ( Leptolyngb y a ), Mastigocoleus and Solentia (Ndhlovu et al. 2021, Kaehler 1999, Marquet et al. 2013 ).We found that sequences assigned to Leptolyngb y a spp .and Pseudophormidium spp .were prevalent, but low-abundance members of the cyanobacterial community on M. californianus shells in BC.Leptolyngb y a spp .are observed as early successional members of endolithic communities on mussels in South Africa (Kaehler 1999, Ndhlovu et al. 2021 ).They are also recognized for producing diverse secondary metabolites that could potentiall y influence micr obial comm unity assembl y on m ussel shells (Brito et al . 2015 ).
It is noteworthy that we detected a high r elativ e abundance of c hlor oplast r eads on m ussel shells acr oss the dataset.These include diatoms and macroalgae common at these sites, such as Ectocarpus and Pyropia spp ., and many ASVs with poor taxonomic r esolution.Giv en that marine eukaryotic algae can also be endolithic (Marcelino andVerbruggen 2016 , Pernice et al. 2019 ), it would be interesting to further characterize phototrophic eukaryotic taxa on mussels using more specific molecular markers (18S, ITS, tufA, rbcL) to determine if these or ganisms ar e euendoliths, contributing to shell erosion in the northeastern Pacific.It would also be interesting to determine if phototrophic eukaryotic comm unities on m ussel shells ar e facilitated by endolithic c y anobacteria or other factors (Zuykov et al. 2021 ).

Shell erosion and host condition do not structure epibiotic bacterial communities on mussel shells
We hypothesized that increased shell erosion, which can modify host physiology, abiotic conditions experienced by the host (Zardi et al. 2016 ) and shell micr otopogr a phies (Zardi et al. 2009 ), would influence which bacteria colonize a mussel shell.Our data reject this hypothesis as we did not find a significant impact of incr eased shell er osion on the composition or structure of mussel shell bacterial communities.It is possible that bacterial community functions, but not the taxa present, may be impacted by the physical habitat modification caused by shell erosion or by the presence of endolithic c y anobacteria (e.g. via photosynthetic exudates).Change in bacterial functions but not taxa has been shown for the Mytilus edulis gut micr obiome, wher e carbon metabolism pr ofiles c hange seasonall y, despite consistency in the taxonomic composition (Pierce and Ward 2019 ).Additional r esearc h is needed to understand how bacterial functions are impacted by changes in shell erosion and varying abiotic conditions throughout the intertidal zone.
We tested whether biological filters exerted by live mussels select for different epibiotic bacteria compared with on empty mussel shells.We found no significant difference between the shell bacteria on tr ansplanted liv e m ussels and empty shells.Instead, we found significant differences in the epibiotic bacteria on transplants (live mussels and empty shells) compared with live contr ols fr om the donor beds.We also found incr eased c hlor oplast r eads fr om micr oalgae and macr oalgae on tr ansplanted m ussels compar ed with liv e contr ols .T his suggests that tr ansplanted m ussels may create new available habitat for macroalgal spores and early bacterial colonizers, and may have experienced reduced competition or grazing compared with on undisturbed, live contr ol m ussels.We confirmed that liv e tr ansplants experienced normal growth (unpublished data) and thus active host biological filters were present in the experimental set-up, suggesting host filtering was not the primary factor shaping epibiotic bacterial assembla ges on m ussel shells.We note that liv e contr ol m ussels wer e al ways fr om the lo w est extent of eac h m ussel bed, while tr ansplanted m ussels v aried in ele v ation.Differ ences in ele v ation between treatments likely influenced the composition and structure of the bacterial communities sampled.
In a recent meta-analysis, the external microbiomes of marine or ganisms wer e found to be mor e str ongl y sha ped by abiotic conditions than host factors, consistent with our results (Woodhams et al. 2020 ).Effects of the experimental transplant, for both liv e m ussels and empty shells, also likely produced unintended changes in the local abiotic conditions of transplanted shells, with resulting impacts on the epibiotic community.Embedding transplanted mussel pairs in marine e po xy removed the experimental organisms from dense aggregates, like those in natural mussel beds.As a r esult, tr ansplants likel y r eceiv ed mor e dir ect solar exposur e and alter ed hydr odynamics compar ed with natur al, a ggr egated m ussels.Consequentl y, the temper atur e and r ate at whic h er oded ar eas of shell dried out, as well as the seaw ater cir culation and exposure to bacterial colonizers, likel y differ ed between tr ansplanted and contr ol m ussels.Further, the e po xy ma y ha ve tempor aril y exposed the epibiotic shell community and potential colonizers to chemical leachates .T herefore , the chemical, physical and thermal micr oenvir onments of m ussel shells likel y differed between transplants and live controls, supporting our finding that the epibacterial community is closely associated with abiotic conditions.

Correla tions betw een abiotic factors and the bacterial communities on mussel shells
We found differences among sites to be the most significant correlate of bacterial community composition on the shells of unmanipulated control mussels .T hese results are aligned with observ ations of str ong differ ences in the bacterial communities of mussels between marine lakes and open marine waters (Cleary and Polónia 2018 ) and increasing dissimilarity in the gill and shell micr obial commm unities of Mytilus californianus acr oss a latitudinal gradient (Neu et al. 2021 ).Community variation across spatial or geogr a phic locations indicates that pr e v ailing envir onmental conditions and microbial source pools strongly impact bacterial comm unity assembl y on m ussel shells.
By tr ansplanting m ussels acr oss ele v ations in the intertidal zone , we pro vide additional e vidence for the r ole of abiotic filtering in shaping the bacterial communities on mussel shells.In particular, w e sho w that ele v ation, subaerial substr atum temper atur e and immersed seawater temper atur e all are associated with the observ ed differ ences in the mussel shell bacterial community.Intertidal ele v ation has been shown to influence the digestive gland bacteria of oysters and clams in a transplant study (Offret et al. 2020 ) and the micr obial comm unities of sympatric macroalgae (Lemay et al. 2020, Quigley et al. 2020 ).Benthic microbial communities (Rothr oc k and Garcia-Pic hel 2005 ) ar e also sha ped by geogr a phic location and position in the intertidal zone.We suspect that thermal or other abiotic factors that v ary acr oss intertidal gradients can select for bacterial taxa with distinct abiotic preferences (Yung et al. 2015 ) and lead to spatial community variation in bacteria throughout the intertidal zone.Mussel and oyster gut bacteria have been shown to exhibit temper atur e-driv en compositional conv er gence at distinct geogr a phic locations (Pierce and Ward 2019 ).Likewise, the effect of intertidal abiotic gradients is recognized to influence the distribution of macroorganisms (Connell 1972 ).Our r esearc h pr ovides e vidence for similar abiotic structuring of bacterial communities on the external surface of an intertidal foundation species.Additional r esearc h with gr eater replication at specific elevations across multiple sites would bolster this finding.The shifts in transplanted mussel shell bacteria observ ed at differ ent ele v ations in the intertidal zone here may indicate the potential for futur e extr eme weather e v ents and pr edicted ocean warming to dr amaticall y impact the mussel shell community and its ecological functions.
In summary, this study provides experimental evidence of incr eased er osion b y endolithic c y anobacteria on empty mussel shells compared with on live hosts .T his suggests that host biology may moder ate er osion by shell-boring parasites.We also identify potential endolithic c y anobacteria taxa occurring br oadl y on mussel shells using 16S rRNA sequencing.In contrast to our hypothesis, w e sho w that variation in epibiotic bacterial communities is not associated with the extent of shell erosion or living hosts compared with empty shell.Instead, transplanting mussels along an ele v ation gr adient in the intertidal zone caused marked shifts in bacterial community composition on both live and empty shells.Our findings demonstrate that spatiall y v ariable abiotic factors correlate more strongly with epibacterial comm unity structur e on the California m ussel than biotic inter actions with the host or shell-boring parasites .T he r esults fr om our experimental transplants further indicate that alterations in abiotic conditions have the potential to significantly impact constituents of epibiotic communities on mussel shells, with unknown consequences for host health and ecosystem functions.

Figure 1 .
Figure 1.Map of experimental sites.

Figure 2 .
Figure 2. P air ed Wilcoxon tests examining differences between experimental pairs of transplanted live mussels and empty mussel shells in (A) percentage of shell area eroded, (B) relative abundance of 16S rRNA reads from all c y anobacteria and eukaryotic algae, and (C) r elativ e abundance of 16S rRNA reads from potentially endolithic taxa.

Figure 3 .
Figure 3. Bar plots of the r elativ e abundance of dominant bacterial families, c y anobacterial families and c hlor oplast sequences on mussel shells across sites and treatments.Each bar represents an individual mussel shell sample.Samples from live and empty transplants are arranged from left to right by incr easing ele v ation in the intertidal zone.Samples from live control mussels were all taken from the lo w est extent of the donor mussel bed at each site.

Figure 4 .
Figure 4. Non-metric multidimensional scaling analysis of Bray-Curtis dissimilarity among epibacterial communities on mussel shells by experimental treatment and site.Corresponding PERMANOVA results comparing treatment groups are presented in Table2.

Figure 5 .
Figure 5. Variables correlated with epibacterial diversity among tr ansplanted m ussels.Significant continuous explanatory v ariables (arr ows) fr om Table 3 ar e displayed in the PCoA plot based on the Bray-Curtis metric.Plotted points represent dissimilarity values for bacterial communities on transplanted mussels with available iButton temper atur e (T) data.

Table 1 .
Study locations, number of experimental transplants, key dates, and sample numbers.

Table 2 .
P airwise PERMANOVA r esults for Br ay-Curtis dissimilary between epibacterial comm unities on m ussel shells by experimental tr eatment.P -adjusted v alues in bold, corr ected by the Bonferr oni pr ocedur e, indicate statistical significance ( P -adjusted < 0.05).

Table 3 .
Envfit results of the explanatory variables correlated with e pibacterial di v ersity among tr ansplanted m ussels.Statisticall y significant values ( P < 0.05) are in bold.