Multiscale Spatial Variability and Stability in the Structure and Diversity of Bacterial Communities Associated with the Kelp Eisenia cokeri in Peru

Ecological communities are structured by a range of processes that operate over a range of spatial scales. While our understanding of such biodiversity patterns in macro-communities is well studied, our understanding at the microbial level is still lacking. Bacteria can be free living or associated with host eukaryotes, forming part of a wider “microbiome,” which is fundamental for host performance and health. For habitat forming foundation-species, host-bacteria relationships likely play disproportionate roles in mediating processes for the wider ecosystem. Here, we describe host-bacteria communities across multiple spatial scales (i.e., from 10s of m to 100s of km) in the understudied kelp, Eisenia cokeri, in Peru. We found that E. cokeri supports a distinct bacterial community compared to the surrounding seawater, but the structure of these communities varied markedly at the regional (~480 km), site (1–10 km), and individual (10s of m) scale. The marked regional-scale differences we observed may be driven by a range of processes, including temperature, upwelling intensity, or regional connectivity patterns. However, despite this variability, we observed consistency in the form of a persistent core community at the genus level. Here, the genera Arenicella, Blastopirellula, Granulosicoccus, and Litorimonas were found in >80% of samples and comprised ~53% of total sample abundance. These genera have been documented within bacterial communities associated with kelps and other seaweed species from around the world and may be important for host function and wider ecosystem health in general.


Introduction
Marine communities are structured by a range of processes that operate over a range of spatial scales.For example, over large latitudinal scales (100 s-1000 s km), temperature has been repeatedly shown to be the key determinant for biogeographic distributions [1,2].At small to moderate scales (10s m-10 s km) other factors (e.g., wave exposure, local connectivity patterns, predation) may also exert strong influence.Therefore, by examining how community structure changes over multiple spatial scales, insight can be gained into the relative importance of underlying processes that vary across similar scales.Indeed, the consideration of multiple spatial scales to better our understanding of biodiversity patterns has become a central goal of modern ecology [3,4].
Microbes can be found either free-living or in association with host eukaryotes, forming part of a wider "microbiome," comprising bacteria, fungi, viruses, and micro-eukaryotes. 1 3 Within this microbiome, bacteria in particular are known to play fundamental roles in healthy host functioning.Foundation species (e.g., kelps, seagrass, corals) underpin wider ecosystem functioning, and as such, the host-bacteria relationships of foundations species may play disproportionate roles in mediating processes for the wider ecosystem [5,6].However, while determining the spatial scales at which macro fauna associated with foundation species are structured has a rich history [7,8], our understanding at the microbial level has only recently been made possible, following technological advances in sequencing technologies [9].
In marine systems, direct contact between hosts and the surrounding seawater can lead to thousands of host-bacteria interactions, making them complex and challenging to understand [10].However, the host itself may exert a strong selective pressure for specific taxa that can promote bacterial community stability and persistence, despite a dynamic external environment [11][12][13][14].By identifying this stable "core community," it may be possible to define components of the bacterial community that are important for host function.Therefore, examining bacterial community structure across multiple spatiotemporal scales is fundamental to advancing our wider understanding of microbial ecology and identifying functionally important host-bacteria relationships [15].
Extending along the coasts of Chile and Peru, the Humboldt Current System (HCS) is one of the most productive marine ecosystems in the world and supports important fisheries for regional economies [40].While the coastlines of Chile and southern Peru are fringed by kelp forests dominated by Lessonia sp. and Macrocystis pyrifera, central and northern Peruvian forests are characterised by monospecific stands of the endemic kelp Eisenia cokeri [41].Despite serving as an important foundation species and playing a fundamental role in local food webs [42], patterns of variability in the structure of E. cokeri populations and their associated communities remain almost entirely unknown (but see [42,43]).Moreover, the dynamics of microbial communities associated with marine foundation species within the HCS remain unexplored.
Here, we characterised host-bacteria relationships for E. cokeri populations at the regional (~ 480 km), site (1-10 km) and individual (10s of m) level in Peru.In doing so, we aimed to determine (i) the scales at which the bacterial communities associated with E. cokeri are structured and (ii) determine whether there are spatially consistent host-bacteria associations.

Sampling Approach
Sampling followed a spatially-nested hierarchical design with three study sites nested within each of two regions (Fig. 1).Regions were separated by ~480 km, while distances between sites (within each region) ranged from ~1 to ~10 km.Sites were sampled over a 10-day period in February 2022.At each site, 15 adult E. cokeri > 1 m apart were haphazardly sampled from an area < 500 m 2 at 1-7 m depth by SCUBA divers.While underwater a segment of blade tissue (~20 cm 2 ) per individual was excised (approximately one third of the way along the blade, above the meristematic area; Figure S1) and placed inside an individual sterile plastic bag.In the laboratory, the blade tissue was rinsed with sterilized seawater for 30 s to avoid contamination by seawater DNA, and then scraped with a sterile razor blade.The sampled biofilm was placed in a 1.5-ml Eppendorf containing 300 μl RNA Later solution.At each site, three 1 L replicates of seawater were collected in sterile Nalgene bottles and concentrated by filtering through a 0.22 μm nitrocellulose filter.Filters were folded into quarters and placed a 1.5-ml Eppendorf containing 1000 μl RNA Later solution.
At one site (R1 S3), additional sampling was conducted to examine intra-specific variation in bacterial community structure and diversity across different parts of the blade.Here, five whole kelps were randomly selected and biofilm samples were collected from the meristem, basal and distal areas following the same procedure as above (Figure S1).
DNA was extracted from kelp biofilms and filters using Qiagen DNeasy PowerSoil kits following the manufacturer's instructions.Library preparation and sequencing (MiSeq, Illumina, San Diego, CA, USA) of the V4 region of the 16S rDNA gene using primers (515f-GTG CCA GCMGCC GCG GTAA + 806r-GGA CTA CHVHHHTWT CTA AT) was conducted by StarSEQ (StarSEQ GmbH, Mainz, DE) following an optimized protocol [44].At least one negative PCR control was run on each plate, and demonstrated runs were free from contamination.

Sequence Processing
Paired-end reads were processed according to the BIOCONDUCTER workflow [45].Sequences were trimmed and truncated using the "filterAndTrim" function in DADA2 with the following parameters: truncLen, f = 240, r = 160; truncQ = 2; trimLeft, f = 20, r = 19, to remove primers and low-quality reads.Amplicon Sequence Variants (ASVs) were resolved using DADA2 [45].Chimeric sequences were removed using the "removeBimeraDenovo" function in DADA2.Sequence taxonomy was assigned using the RDP naïve Bayesian classifier against the SILVA release database [46] using the "assignTaxonomy" function in DADA2.Sequence read counts, taxonomic assignments and metadata were assembled as an object in the R package PHYLOSEQ and was used in downstream analysis [47].Samples containing <10,000 reads, taxa contributing <0.01% of the reads in the dataset and ASVs identified as mitochondria or chloroplast were then removed from the PHYLOSEQ object.Sequence counts were then expressed as relative abundance (in proportion to the total sample count).ASV table and metadata are available at (https:// doi.org/ 10. 6084/ m9.figsh are.22182 457.v1).

Statistical Analysis
As the focus of the study was to track the shifting bacterial diversity and structure of communities associated with the kelp themselves, we made initial comparisons between kelp and seawater samples as a single dataset to determine overall differences between the two sample types.Variability in alpha diversity was determined using univariate Permutational Analysis of Variance (PERMANOVA), using the PERMANOVA module [48] within Primer 7 software [49].A similarity matrix was generated based on Euclidean distances and significance determined with 9999 permutations of untransformed data under a reduced model.To account for differences in sequence depth between samples in alpha diversity estimates, the dataset was rarefied to the minimum sample depth, using the "rarefy_even_depth" function in PHYLOSEQ.Alpha diversity for each sample was estimated through the Chao1 index [50] implemented through the "estimate_richness" function in PHYLOSEQ.The Chao1 index estimates ASV richness, and the standard error surrounding this estimate, based on the observed number of ASVs, the observed number of ASVs occurring only once and the observed number of ASVs occurring only twice  [50].Differences in community structure between sample types was determined using multivariate PERMANOVA (Anderson, 2001).This was conducted on the unrarefied relative abundance dataset and based on a Bray-Curtis dissimilarity index and determined with 9999 permutations of untransformed data under a reduced model.For community structure, the outcomes of all analyses between the relative abundance dataset and a dataset of rarefied sequence counts were conducted and found consistent patterns in the results.
Subsequent analyses of alpha diversity (univariate PERMANOVA) and community structure (multivariate PERMANOVA) between Regions and Sites was based solely on the basal (mid-blade) kelp samples (extra intraspecific samples removed) and followed the same parameters as above.Here, model factors consisted of region (fixed factor; two levels: R1 and R2) and site (nested in region; random factor; three levels: S1, S2, S3).Analyses were repeated at all taxonomic ranks.Non metric multidimensional scaling (nMDS) ordinations were constructed to visualize multivariate patterns using the "metaMDS" function in the R package VEGAN based on the unrarefied relative abundance dataset.A similarity of percentage (SIMPER) procedure was conducted in VEGAN to determine which taxa contributed the most to any observed dissimilarities in community structure.Detection of a "core community" was set at 80% prevalence [30].All means are presented ± 1 standard error.

Community Structure
Initial comparisons between kelp and seawater samples showed bacterial communities to be clearly differentiated (Pseudo-F (1,92) = 19.3,p = 0.001) (Figure S3, Figure S4), and further analyses were conducted solely on kelp-associated communities.Kelp-associated bacterial community structure Fig. 2 Relative abundance of bacterial phyla associated with the kelp, Eisenia cokeri, from two study regions (three sites within each region) in Peru.Site locations can be seen in Fig. 1.Phyla that made contributions of < 5% were collapsed into a separate category (dark purple) exhibited a Region and Site nested within Region effect (Table 1).Pairwise comparisons revealed all sites nested with Region were significantly different from one another apart from R1 S1 and R1 S3, which were similar to one another.Site nested within Region effects were consistent when analyses were repeated at all higher taxonomic ranks, but Region effects broke down at the level of Class (Table S1).nMDS ordination showed clear separation between regions and some structuring at the site level (Fig. 4).Moreover, large differences in community structure were also observed between individual kelps from the same site, separated by only a few metres (Fig. 4).Regional effects became decreasingly apparent with decreasing taxonomic resolution (Figure S5).SIMPER analysis showed that regional differences were largely driven by some of the most dominant taxa and ASVs from within the same genus (Table S2).For example, four Blastopirellula ASVs accounted for 12% of variation, while three Granulosicoccus ASVs accounted for 7% of the dissimilarity between the regions.
Our examination of within-kelp variability in bacterial communities (conduted at R1 S3 only), showed that overall richness (i.e., alpha diversity) was similar across all parts of the blade (Pseudo-F (2,12) = 2.68, p = 0.11) (Figure S6).However, bacterial communities persisting towards the distal tips of the blade area exhibited dissimilar community structure (Pseudo-F (2,12) = 1.76, p = 0.031) (Figure S6).This perhaps suggests that the older kelp tissue has been available for colonisation by a range of bacterial taxa for a longer period of time, or that older senescing tissue has lower anti-microbial properties or a wider diversity of substrate types.

Discussion
Kelp species function as foundation organisms that alter the environment and often support rich and abundant associated communities.As well as macroflora and fauna, kelps also harbor rich microbial communities, which in turn may influence the health and functioning of the host.Our study shows E. cokeri supports complex bacterial communities that are distinct from the surrounding seawater, and that the structure of these communities varies at the site and regional scale.However, despite pronounced spatial structuring, consistency was also evident, with the presence of a small "core community," which persisted across individuals, sites, and regions.We observed clear partitioning in community structure between our two study regions (~480 km), as well as between the majority of sites within them (1-10 km).While the specific mechanisms driving this variability remain unclear, it is evident that a range of processes operate over both regional (480 km) and site scales (1-10 km) and may be important drivers of the observed variability.These processes may be deterministic or neutral in nature.Deterministic processes include salinity [28], wave exposure [34], turbidity [51], nutrient concentrations [29], while neutral processes associated with dispersal may lead to ecological drift between sites/regions.Alternatively, the development of different host traits, may in turn drive differences in the associated bacterial community [31].Regional structuring in bacterial communities associated with seaweed hosts seems commonplace [27,29,31], whereas site level structuring seems more system specific.For example, no such site level (1-10 km) variability was observed in the kelps Laminaria hyperborea and Saccharina latissima in the UK [30] or Ecklonia radiata within regions in east and West Australia [29], but has been observed in seven kelps in the NE Pacific [28,33].Indeed, even within our study regions, sites were not always different from one another and highlights the importance of local biogeographic/environmental context in structuring communities at this scale.
In the current study area, the HCS is characterized by marked spatial variability in temperature, nutrients, light and oxygen driven by variation in the strength and persistence of upwelling [52].Our southernmost region (R2: Ica) is subjected to intense upwelling, which is active year-round but stronger during winter/spring [53], and is on average ~1.5 °C cooler than the northernmost region (R1: Ancash).Regional scale differences in temperature, upwelling intensity, coupled with regional scale connectivity pathways, are likely to influence the bacterial species pool Fig. 5 Relative abundance of the core community (defined at genera present in > 80%) associated with the blade of the kelp, Eisenia cokeri.Abundance is expressed as proportion of entire sample and the environmental context for community succession and development.At smaller spatial scales, betweensite variability may have been driven by differences in environmental drivers (e.g., wave fetch, grazing pressure, light) or local oceanographic processes leading to different levels of connectivity between E. cokeri populations.Ultimately, resolving the underlying processes responsible for overlying differences in community structure will only be achieved with a detailed understanding of environmental conditions, local oceanography and connectivity patterns in the HCS and detailed knowledge at the ecological scales at which E. cokeri populations are structured.We also recorded pronounced small-scale variability between kelp individuals separated by a few meters.This high-level variability is likely reflective of both stochastic and deterministic factors (either imposed by the wider environment or host itself) that dominate at the individual kelp level and is common in kelp-associated microbiota [29,30], as well as other compartments of kelp forest biodiversity [22][23][24].
Despite variability in community structure across spatial scales, we also observed consistent features in the bacterial community.While no individual ASV met the core criteria, a number of the most abundant ASVs were congeneric and contributed to a substantial proportion of observed dissimilarity between the two regions.When analyzed at the genus level, a small abundant "core community" was observed.These genera have been identified in core communities in both Australia [29] and the UK [30], and have been associated with kelp in different systems across the world [28,[54][55][56][57].While the role they play for host function remains largely unknown, many have been associated with healthy rather than degraded host communities, suggesting a role for host health [54,56].Moreover, recent functional insights have been provided for Granulosicoccus sp.(a core genus) in different populations of the bull kelp Nereocystis luetkeana in Washington, USA.Specifically, Metagenome-Assembled Genomes (MAGs) showed this genus to be important for vitamin B 12 synthesis and nutrient acquisition [58], which may in turn be provided to the host.Importantly, geographically separated ASVs within Granulosicoccus were almost functionally indistinguishable, suggesting identification of a core community at this taxonomic resolution is more appropriate than universal ASVs.Ultimately, further functional insight will be provided as these techniques become increasingly accessible, particularly at the scales required to look for universal patterns between hosts and to determine the nature and importance of host-bacteria relationships.
In conclusion, we identified both consistency and variability in bacterial communities associated with a critical yet understudied forest-forming kelp within a productivity hotspot.However, kelp are often distributed over vast spatial scales (often over 1000s of km), where other large scale processes may be more important in driving shifts in community structure.Therefore, range wide studies, that incorporate larger distances and increase replication at the regional level are needed to determine if our observed patterns hold true over even larger spatial scales and across different environmental/ geographical contexts.Nonetheless, our findings are consistent with other studies conducted in vastly different systems and on divergent seaweed hosts and suggests these patterns may be evident at a global scale.The identification of such universal patterns will greatly help our understanding of the processes and scales that influence host microbiomes.

Fig. 1
Fig. 1 Map of study area showing the position of the two study regions, Ancash (R1) and Ica (R2) in Peru and the approximate latitudinal distribution of Eisenia cokeri and the overlapping Macrocystis pyrifera (left).Inset maps indicating positions of sites (S1, S2 & S3) within each region (right).Representative individuals of the host kelp E. cokeri (bottom right)

Fig. 3
Fig.3Box plots representing alpha diversity (Chao1 index) for bacterial communities associated with the kelp, Eisenia cokeri, from two study regions (three sites nested within each region) in Peru.Site locations can be seen in Fig.1

Fig. 4
Fig.4nMDS plot depicting Bray-Curtis dissimilarity between bacterial communities associated with the kelp, Eisenia cokeri, from two study regions (three sites nested within each region) in Peru.Ellipses represent differences between regions.Site locations can be seen in Fig.1

Table 1
Results of univariate PERMANOVA for alpha diversity and multivariate PERMANOVA for community structure between Region, and Site nested within Region