Red Mangrove Propagule Bacterial Communities Vary With Geographic, But Not Genetic Distance

Bacterial communities associated with plant propagules remain understudied, despite the opportunities that propagules represent as dispersal vectors for bacteria to new sites. These communities may be the product of a combination of environmental influence and inheritance from parent to offspring. The relative role of these mechanisms could have significant implications for our understanding of plant–microbe interactions. We studied the correlates of microbiome community similarities across an invasion front of red mangroves (Rhizophora mangle L.) in Florida, where the species is expanding northward. We collected georeferenced propagule samples from 110 individuals of red mangroves across 11 populations in Florida and used 16S rRNA gene (iTag) sequencing to describe their bacterial communities. We found no core community of bacterial amplicon sequence variants (ASVs) across the Florida range of red mangroves, though there were some ASVs shared among individuals within most populations. Populations differed significantly as measured by Bray–Curtis dissimilarity, but not Unifrac distance. We generated data from 6 microsatellite loci from 60 individuals across 9 of the 11 populations. Geographic distance was correlated with beta diversity, but genetic distance was not. We conclude that red mangrove propagule bacterial communities are likely influenced more by local environmental acquisition than by inheritance.


Introduction
Microbes are increasingly recognized for their intimate interactions with plants. Bacteria are ubiquitous on (epiphytes) and within (endophytes) plants where they may affect pathogen defense, nutrient acquisition, secondary metabolite production, and resistance to abiotic stressors [1]. Plant-associated communities are likely derived from a combination of environmental acquisition and inheritance from parent to offspring [2], and understanding the origin of these communities has important implications for our understanding of plants as holobionts [3]. Considering multicellular organisms as holobionts (units composed of the multicellular organism and its associated microbiota) may be more productive than considering the multicellular organism alone in certain contexts, especially those where the microbiota is inherited from parent to offspring. If microbes are inherited across generations, they may provide additional raw material for natural selection, and thus influence the evolution of their hosts [4]. In many cases, it appears that offspring are inoculated with microbes from their parents, but that their microbiota changes as the individual develops and interacts with its environment [5]. Here, we examine variation in the propagule-associated bacterial communities of red mangroves (Rhizophora mangle L.; Rhizophoraceae) across geographic and genetic distance to determine the relative influence of host environment and maternal relatedness on bacterial community composition.
The strongest environmental influence on plant microbiomes may be the substrate on which a plant grows. Plants may selectively enrich microbes on and near their roots by releasing carbon-rich exudates in exchange for protection from soil-borne pathogens and increased access to soil nutrients [1]. Soils contain immense microbial diversity, and some soil-borne microbes may cross the root endodermis and move throughout the plant via vascular tissue [6]. Additionally, physical (e.g., particle size) and chemical (e.g., nutrient content) characteristics of the substrate can influence plant physiology, and thus the niche space exploited by microbes and the composition of plant-associated microbial communities [1]. Surprisingly, leaf bacterial communities appear to be affected more by soil-borne bacteria than airborne bacteria, possibly due to the relative difficulty in establishment in the complex environment of the leaf surface [7]. In mangroves, the soil environment also includes seawater that periodically inundates the soil and lower portions of the plants. Seawater bacterial communities contain high abundances of the phyla Proteobacteria and Bacteroidetes, while sediment communities host distinct, significantly more diverse communities with high abundances of Bacteroidetes but lower abundances of Proteobacteria [8].
In plants, the typical pathway for inheritance is through seeds (the spermosphere) or fruits (the carposphere). Bacteria associated with the seeds of Chinese silver grass (Miscanthus sinensis) make significant contributions to the seedling microbiome, display adaptations to long-term survival within the seed, and respond rapidly to imbibition of water during germination by concentrating at the embryonic root tip [9]. A conserved assemblage of spermosphere bacteria has been identified across geographic and genetic distance in maize (Zea mays L.), and some of these bacteria may also inoculate the soil near the root system [10]. Seed endophytes in Elephant cactus (Pachycereus pringlei (S. Watson) Britton & Rose) contribute to the creation of suitable habitat by improving access to nutrients and promoting rock weathering [11,12]. Spermosphere bacteria in green bristlegrass (Setaria viridis (L.) P. Beauv) increase germination rate, though after growing several generations of plants in sterile soil, seed endophyte diversity, seed germination, and plant growth rates all declined [2], suggesting that the interaction of seed endophytes and soil microbes is important for longterm plant health. Some bacteria associated with propagules of the mangrove species Avicennia germinans L. improved growth of rice (Oryza sativa L.), barley (Hordeum vulgare l.), and Avicennia under salt stress, suggesting the importance of propagule-associated bacteria [13].
The carposphere is less well studied than the spermosphere, perhaps in part due to the increased complexity of fruits, or the immense variation in fruit anatomy and secondary chemistry. Studies to-date have largely focused on agricultural species. The microbiome of the carposphere is likely influenced by that of the anthosphere (the floral tissues). The bacterial microbiome of apple flowers (Malus domestica cultivar Gala (Suckow) Borkh.) undergoes successional changes during floral development, with different major clades rising and falling in abundance in predictable patterns over time [14]. Carposphere communities of grapes (Vitis vinifera L.) were distinct from those of the leaves and were influenced more strongly by local environmental differences than by host genotype [15]. Red mangroves produce viviparous propagules, which are single-seeded fruits that germinate prior to abscission from the parent plant [16] and have been shown to host bacterial communities distinct from their leaves, stems, and flowers that may vary across populations [17].
Mangroves are woody trees or shrubs found in the intertidal zones of the tropics and subtropics worldwide. There are approximately 20 evolutionary origins of mangrove habit [18], including three represented in the native flora of Florida. Red mangroves are distinguished from Florida's other two native mangrove species by the former's rhizophores (also known as prop roots) and elongated, viviparous propagules (Fig. 1). Red mangrove propagules include a prominent, waxy radicle (embryonic root) that ruptures the seed coat and fruit wall prior to germination. They have a longevity of at least 1 year in seawater and can establish rapidly after lodging in substrate following dispersal [16]. These adaptations, combined with a reduced frequency of hard winter freezes, are contributing to an ongoing poleward range expansion throughout its range [19]. This species also provides critical habitat to numerous organisms, including rare and endangered species such as goliath grouper (Epinephelus itajara [20]), smalltooth sawfish (Pristis pectinate [21]), and lemon shark (Negaprion brevirostris [22]).
Red mangroves have been the focus of several population genetics studies [23][24][25], through which numerous microsatellite loci have been developed. This species displays genetic variation throughout its Florida range, though previous work has not found significant evidence of population differentiation, suggesting that there is ongoing gene flow between populations [23,24]. This is likely due in part to the longdistance dispersal capabilities of its propagules. The most northwestern populations have lower genetic variation and higher population structure, probably due to founder effects [24].
We present data comparing the bacterial microbiome of 100 propagule samples of Rhizophora mangle from 11 populations throughout much of its Florida range, as well as microsatellite data from six loci across nine populations to investigate the relative influence of host geography and genetic distance on propagule bacterial community composition. Specifically, we address the following sets of questions: (1) two factors would tend to reinforce the distinctiveness of the bacteria within a population of R. mangle: inheritance of bacteria from parent to offspring when the offspring establishes at the same population as the parent and introduction of bacteria from the local environment. Given this, are there any bacterial ASVs shared among propagules from the same population or across the different populations of R. mangle in Florida? Are different populations of R. mangle associated with significantly different bacterial communities? (2) However, long-distance dispersal could frequently lead to establishment of propagules at new populations. Are the microbiomes of propagules found on maternal plants that are geographically near to one another most similar, perhaps due to significant influence from the local species pool and/or frequent establishment of propagules locally? Or are the microbiomes of propagules more similar when the maternal plants are more genetically similar, even when the propagules are separated by great distance, perhaps due to a strong influence of inheritance on the composition of the microbiome?

Sample Collection
Ten individuals of R. mangle were sampled from each of 11 sites across Florida for a total of 110 samples (Table 1, Fig. 2). Each sample consisted of 10 fruits from a single individual, pooled in one sample bag. This approach generated data that did not distinguish between endophytic and epiphytic microbes, but rather characterized the entire community. Sampled individuals were at least 5 m away from one another. Each site was visited once between October 1 and December 1, 2019, and GPS coordinates were recorded for each individual sampled. In a protocol like those commonly used in food microbiology [26], all samples were collected in 384-mL sterile Whirl-Pak sample bags (Nasco: Ft. Atkinson, WI) and stored on ice until they could be brought back to the lab and frozen at − 20 °C. Aseptic techniques were used during sample collection, and all utensils were sterilized with 95% ethanol between samples. Leaf samples consisting of 10 apparently healthy and mature leaves were also collected from each fruiting individual (i.e., the maternal plant) for microsatellite analysis. Leaf samples were stored in plastic bags with silica beads at room temperature until ready for DNA extraction.

Bacterial DNA Extraction
Sample collection and preparation followed previously published work [27]. Samples were prepared for DNA extraction by first adding 18 mL of phosphate buffer solution [PBS, containing (g L −1 ) Na 2 HPO 4 , 0.724; KH 2 PO 4 , 0.210; NaCl, 7.650; pH 7.4] to each bag, then pulverizing the sample in a Nasco Stomacher 400 Circulator Lab Blender (Seward Laboratory Systems: Bohemia, NY) at 260 RPM for 4 min and 30 s. The liquid was then pipetted into a sterile centrifuge tube and spun at 4200 RPM for 12 min at 10 °C. All but 250 μL of the solution and the pellet were then pipetted off, and the remaining 250 μL was vortexed briefly to mix the pellet back into the liquid. This 250 μL of liquid then served as the input for the Zymobiomics DNA Microprep Kit (Genessee Scientific: Irvine, CA), which was used to extract the bacterial DNA according to the kit protocol.

16S Libraries and Sequencing
We used the forward/reverse bacterial primer pair S-D-Bact-0341-b-S-17/S-D-Bact-0785-a-A-21 [28]. 16S rRNA gene sequencing amplicon libraries were prepared using a two-step PCR as described in the Illumina 16S Library Preparation guide [29] with the following modifications: intermediate and final amplicons were examined by agarose gel electrophoresis, quantified by Qubit dsDNA HS Assay (Invitrogen: Carlsbad, CA), and pooled at equimolar concentrations. The library pool was analyzed using Agilent's

Microsatellites
Leaf samples taken from the fruiting individuals represent the maternal parent of each propagule sample. Leaves were ground using liquid nitrogen and a ceramic mortar and pestle. Approximately 20 mg of ground tissue was used to extract plant DNA using a DNEasy Plant Mini kit (Qiagen, Valencia, CA, USA). Extractions were quantified using a Nanodrop 2000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and then concentrated to approximately 10 ng/μL. Microsatellite data was generated using a triple multiplex panel approach with five loci per panel (Table 2). Multiplex procedure and PCR conditions were adapted from previous work in R. mangle [25] and can be found in Supplementary File Two. Fragment analysis was performed on an Applied Biosystems 3730 Genetic Analyzer with capillary electrophoresis (Waltham, MA, USA). Multiplex one failed to amplify during PCR, so subsequent analyses were based on multiplexes two and three. Geneious Prime [31] was used to call peaks and identify alleles. Briefly, the steps involved were as follows: the trace document for each ladder was inspected, and ladder peaks were called manually as needed when Geneious failed to correctly identify them automatically. Loci were set by inputting basic information including expected size, repeat unit size, and the expected number of peaks for each locus. Alleles were binned using Geneious' binning algorithm. In cases where Geneious left alleles unbinned, we followed the Geneious tutorial and manually adjusted bins in order to include all alleles present [32]. We used GenAlEx [33,34] to calculate the pairwise geographic and genetic distances between each sample, visualize genetic distances between samples using principal coordinate analysis (PCoA), and calculate summary statistics for our microsatellite dataset. The PCoA was generated using a standardized covariance matrix of pairwise genetic distances between individuals. A sensitivity analysis was conducted by removing each of the 6 loci in turn and running the downstream analyses with each dataset of 5 loci.

Data Analysis
Raw 16S data in the form of forward and reverse demultiplexed FASTQ files were obtained from the Illumina MiSeq platform. These files were then analyzed using QIIME2 v. 2019.7.0 [30]. The DADA2 plugin [35] was used to join forward and reverse reads, and for quality control involving an assessment of the frequency of occurrence of nonplastid, high-quality, non-chimeric reads. The Silva v. 138 [36] database was used to assign taxonomy to each amplicon sequence variant (ASV). The core feature plugin in QIIME2 was used to identify "core" ASVs within populations and across all populations. These core community members were considered at thresholds ranging from 25 to 100% of samples of a given grouping (i.e., all sampled individuals of a population or all sampled individuals). The R package Wrench [37] was used to normalize and log-transform the data. The package GUnifrac [38] was used to calculate generalized Unifrac distances among samples. The package vegan [39] was used to calculate Bray-Curtis distances between samples and to perform pairwise PER-MANOVA tests comparing generalized Unifrac distances and Bray-Curtis distances across populations. We also conducted a dispersion check using the function betadisper which revealed heterogenous dispersion (F = 2.8706, p = 0.005).
However, PERMANOVA is robust to heterogenous dispersion when applied to balanced designs [40]. We also compared Bray-Curtis dissimilarity between populations after collapsing our ASV table to the genus level. Bray-Curtis measures community dissimilarity by comparing counts of community members (bacterial ASVs) among samples [41] while Unifrac distances measure the shared genetic distance of community members among samples [38]. The package vegan was also used to conduct Mantel tests and partial Mantel tests (permutations = 999) using our reduced dataset to compare generalized Unifrac distances and Bray-Curtis distances with genetic and geographic distances.

Results
One hundred of 110 propagule samples were successfully extracted, amplified, sequenced, and passed quality control (Table 1). Following DADA2 and the filtering out of all plastid, archaeal, and mitochondrial sequences, there were 177,732 reads associated with 2868 ASVs. The most common phylum represented in these samples was Proteobacteria (relative abundance (RA) = 60.4%), with substantial proportions of Actinobacteria (RA = 14.2%) and Bacteroidetes (RA = 11.8%) as well. The two most abundant classes were within Proteobacteria: Alphaproteobacteria (total RA = 48.6%) and Gammaproteobacteria (total RA = 9.9%).
The rarefaction curve (Supplementary File One) demonstrated that increasing sampling effort beyond 400 to 2000 reads per sample was unnecessary, suggesting that this approach provided a sufficiently accurate description of these bacterial communities.

Core Communities and Differences Between Populations
No ASVs were shared between 25% or more of all samples, suggesting that there is no core community of ASVs associated with red mangrove propagules along our sampling of the Florida coast. However, at the taxonomic level of phylum, most of our samples contained Proteobacteria (94% of samples), Actinobacteria (76%), and Bacteroidetes (51%). We found variation in the number of ASVs shared across individuals within each site (Table 3). One ASV from the Rhodobacteraceae was found in 77.5% of samples from Biscayne Bay, and this was the highest percentage of samples to contain the same ASV within any population. Two ASVs were found in 70% of samples from St. George Island, and 10 ASVs were found in 47.5% of samples from that population. This population had more shared ASVs in half its samples than any other population. (See Supplementary File Three for identities of core ASVs within each location).

Microsatellite Loci
We obtained data from 10 microsatellite loci, but this full dataset had a significant amount of missing data (41.5%).
To reduce bias associated with missing data, we used a reduced dataset of 6 loci (missing data = 7.3%). See Supplementary File Five for a comparison of the two datasets. Previous work in other organisms have used a similar number of loci to compare microbiome composition and host genotype [42,43]. This reduced dataset had 60 individuals across 9 populations. The sensitivity analysis did not find differences in significance between the systematically reduced datasets of 5 loci when compared to each other or the 6-locus dataset. We visualized genetic distances between maternal plants using principal coordinate analysis (PCoA) (Fig. 3), calculated the pairwise fixation index (F st ) between all populations (Table 5), and calculated summary statistics across microsatellite loci (Table 6) and across populations ( Table 7). The first axis of our PCoA explained 17.07% of the variation, the second axis explained 9.30%, and the third axis explained 8.60%.

Correlations Between Beta Diversity and Geographic and Genetic Distances
Geographic and genetic distance between maternal plants were positively correlated (r = 0.3977, p = 0.001). Using a Mantel test, we found a weak positive correlation between Bray-Curtis and geographic distance between samples (r = 0.07903, p = 0.033), but no significant correlation between Bray-Curtis and maternal genetic distance (r = 0.008878, p = 0.429). Unifrac distance was not significantly correlated with geographic distance (r = − 0.04915, p = 0.908) or genetic distance (r = − 0.006976, p = 0.502). Furthermore, partial Mantel tests did not find a significant correlation between either beta diversity metric and genetic distance after controlling for geographic distance (Bray-Curtis: Mantel statistic = − 0.01718, significance = 0.622; Unifrac: Mantel statistic = 0.01298, significance = 0.379). Genetic distances between individuals and populations were visualized using PCoA (Fig. 3). Pairwise F st between populations was correlated with geographic distance (Fig. 4), demonstrating that a significant proportion of the variation is due to isolation by distance.

Most Populations Are Significantly Different When Comparing Bray-Curtis, But Not Unifrac Distances
We did not find evidence for any core community of ASVs shared broadly across the Florida Gulf Coast range of R. mangle, though most samples had at least one ASV from each of the phyla Proteobacteria, Actinobacteria, and Bacteroidetes. These phyla may also be very abundant in seawater and coastal sediments [8,44], suggesting that the propagule   communities could be significantly influenced by these environmental sources. We found variation in the number of core ASVs shared between individuals within each population (Table 3). These results are consistent with patterns found across leaves, stems, flowers, and propagules from three populations of red mangroves in the Florida Panhandle in terms of the small number of shared ASVs [17]. Propaguleassociated bacteria from distantly related mangrove species are capable of plant growth promotion in agricultural species, suggesting that propagule-associated communities may merit further investigation for bioprospecting of beneficial microbes [13]. Most work exploring core communities within plants has focused on terrestrial species, where core communities sometimes occur across broad ranges. Bacterial communities associated with ponderosa pine (Pinus ponderosa P. Lawson & C. Lawson) were found to display lower intraspecies variability than interspecies variability when compared to those of a congener and two other tree species, even where samples of Pinus ponderosa originated on different continents [45]. Furthermore, four bacterial species were found in seeds from 14 cultivars of maize (Zea mays L.) as well as its wild ancestor teosinte (Zea diploperennis H.H.Iltis et al.); several additional bacterial species were found in most of the cultivars sampled [10].
The lack of a core propagule bacterial community in R. mangle could be related to their viviparous nature. Morphologically, anatomically, and physiologically, the propagules are similar to seedlings rather than seeds. As germinated seedlings, propagules interact with their dispersal environment in ways that seeds do not, suggesting the potential for environmental influence in ways not experienced by bacterial communities of seeds.
Carposphere bacterial community composition varied across mangrove populations. That we found Bray-Curtis to differ significantly between most pairwise comparisons of populations while no comparisons based on Unifrac distances were significantly different suggests that although the identities and relative abundances of ASVs associated with each population were different, the higher-level taxonomic diversity is relatively similar. This can be restated as bacterial communities differ in the precise ASVs found within them, but those ASVs are closely related to one another. These communities could contain similar taxonomic groups due to the similarity in the morphology, anatomy, and physiology of red mangrove propagules which are themselves a product of host evolution, even if the bacteria themselves are not faithfully inherited across generations. The seeds of terrestrial plants, and perhaps endophytic communities in general, tend to be dominated by Proteobacteria, Firmicutes, Bacterioidetes, and Actinobacteria [46]. Our results generally align with this pattern, though we found few ASVs from the phylum Firmicutes (RA = 2.1%). While there is ample evidence in the literature of inheritance of microbes in plants [9,10], we think it is unlikely that the patterns we describe here are the result of inheritance, for reasons explored below.

Host Geography, But Not Maternal Genotype, Is Correlated With Beta Diversity
We found Bray-Curtis dissimilarity, but not generalized Unifrac distance, to be correlated with geographic distance between samples. Samples located close together tended to be more similar in terms of the identities and relative abundances of their ASVs, but not in terms of the genetic similarity of the bacterial communities. Together, these results suggest that geography might affect propagule-associated bacterial diversity at fine taxonomic scales, but broader taxonomic groupings remain relatively unaffected across geographic distance. This is broadly consistent with our understanding of plant-associated microbial communities [10,45] and with the results of our pairwise PERMANOVA (Table 4).
Additionally, neither beta diversity metric was correlated with maternal genetic distance, suggesting that maternal genotype did not have a significant influence on bacterial community composition. These results are somewhat inconsistent with the literature. Bacterial endophytes of corn seeds were shown to be influenced by host genotype, and a core community of bacteria was identified across cultivars and even in the wild ancestor, Zea mays [10]. The bacterial and fungal leaf communities of grapes were shown to be influenced more heavily by geographic region (and the associated differences in environmental conditions) than by maternal genotype [15], suggesting that microbial communities found across different plant systems may differ in the relative influence of environment, geography, and genotype. Our sampling methods captured bacteria from the propagule surface and within the plant tissue. Bacterial communities associated with plant surfaces may be more readily influenced by differences in environmental conditions such as UV radiation, humidity, exposure to seawater, and ambient temperature than those found within plant tissues [47], and therefore may vary more strongly over geographic distances.
Genetic and geographic distance among mangroves were correlated, suggesting that more closely located individuals tended to be more closely related. We found evidence for population differentiation, with geographic distance accounting for nearly 67% of genetic variation  (Fig. 4). This is broadly consistent with previous work which found significant genetic variation between populations using 8 and 7 microsatellite loci, respectively [23,24]. However, additional analysis with larger RADseq datasets found lower F st between populations [48].

Factors that Influence the Propagule Microbiome
Plant microbiomes are derived via a combination of inheritance and environmental acquisition [1]. There is fine-scale variation in plant microbiomes across tissue types within individuals, perhaps in part due to differences in the relative importance of these two assembly mechanisms [49]. Specifically, internal and external tissues differ in the degree to which they interact directly with environmental microbes, and thus may differ in the degree to which they assimilate environmental microbes with internal tissues being more heavily influenced by the plant itself [47]. Like those of other plant tissues, the mechanisms that structure red mangrove propagule-associated bacterial communities are complex and difficult to disentangle. The propagule surface is exposed to environmental conditions that may significantly impact epiphytic bacterial populations including high salinity and substantial daily temperature fluctuations. Their waxy exterior provides protection from the elements, but it is not entirely clear how these surfaces may be colonized or utilized by microbial life. The exterior layers may also limit assimilation of microbes into the internal tissues, which might then be more likely to be colonized by microbes from the maternal plant. Seeds of Zea mays harbor a conserved set of bacteria across host genetic distance, bacteria which may be derived from the maternal plant and which may play important roles in seedling success [10]. In Setaria viridis, the internal seed environment is colonized by bacteria from the maternal plant, bacteria which respond to germination and provide valuable benefits during establishment [2]. This suggests that plant diaspores (whether seeds, fruits, or propagules) may obtain endophytic microbes from their maternal plant, though this may only be detectable if endophytic and epiphytic microbes are considered separately.

Conclusion
This work provides, to our knowledge, the first investigation of the influence of geography and maternal genotype on the bacterial communities of red mangrove propagules. We did not find a Florida-wide core community of species-level bacterial ASVs associated with red mangrove propagules, though the phyla Proteobacteria, Actinobacteria, and Bacteroidetes were found in most samples. Different populations of mangroves were associated with different ASVs but were not significantly different in the relatedness of their communities. We found variation in red mangrove propagule bacterial communities across geographic distance, but not across maternal genetic distance. This variation suggests that differences in host environment or local bacterial species pools influence bacterial community composition more than maternal genotype does. Taken together, we propose that red mangrove propagule bacterial communities are not inherited with fidelity across generations, though an approach that described endophytic and epiphytic bacteria separately could yield additional insight. Experiments mimicking dispersal could provide valuable insight into whether propagule communities change during the seawater dispersal of red mangroves.