Marchantia liverworts as a proxy to plants’ basal microbiomes

Microbiomes influence plant establishment, development, nutrient acquisition, pathogen defense, and health. Plant microbiomes are shaped by interactions between the microbes and a selection process of host plants that distinguishes between pathogens, commensals, symbionts and transient bacteria. In this work, we explore the microbiomes through massive sequencing of the 16S rRNA genes of microbiomes two Marchantia species of liverworts. We compared microbiomes from M. polymorpha and M. paleacea plants collected in the wild relative to their soils substrates and from plants grown in vitro that were established from gemmae obtained from the same populations of wild plants. Our experimental setup allowed identification of microbes found in both native and in vitro Marchantia species. The main OTUs (97% identity) in Marchantia microbiomes were assigned to the following genera: Methylobacterium, Rhizobium, Paenibacillus, Lysobacter, Pirellula, Steroidobacter, and Bryobacter. The assigned genera correspond to bacteria capable of plant-growth promotion, complex exudate degradation, nitrogen fixation, methylotrophs, and disease-suppressive bacteria, all hosted in the relatively simple anatomy of the plant. Based on their long evolutionary history Marchantia is a promising model to study not only long-term relationships between plants and their microbes but also the transgenerational contribution of microbiomes to plant development and their response to environmental changes.


Introduction
The evolution of life on Earth has certain milestones. Colonization of land by plants was one of the major evolutionary breakthroughs about 450 million years ago [1][2][3][4]. Land plants (embryophytes) evolved several adaptive traits that allowed them to colonize Earth's terrestrial surface efficiently and bryophytes (mosses, hornworts, and liverworts) represent 60 the earliest diverging land plants [5]. 1900s, the cellular nature of organisms and to the study of dorsoventral body plans and polarity in developmental biology [8,9].
Extensive parallel sequencing approaches have fostered recent studies on plant-microbe interactions, and plenty of plant microbiomes are available for the study of model plants like 75 Arabidopsis thaliana [10,11]. Additionally, there are available sequenced microbiomes from plants models or with agricultural relevance. The plant microbiomes have aided in the study of plant-microbe interactions in an unprecedented way [11][12][13][14][15].
There are even proposed models for the establishment of the plant root microbiome like the 80 two-step model, that considers the root microbiome as a product of plant-independent features. The two-step model considers edaphic factors, the general selection of the microorganisms for general plant cell wall composition, rhizodeposits and the host genotype which actively selects their microbial inhabitants [11]. Early land plants established symbiotic associations with mycorrhizae, and it is thought that Palaeozoic drops in CO2 reduced 85 phosphate (P) absorption in nonvascular plants which in addition to competition for light favored the vascular plants Earth dominance [16].
Liverworts have a long story of endophytic relationships with fungi, and it has been suggested this is the result of a common ancient origin with multiple recent losses. For 90 example, it has been reported that M. paleacea can maintain a population of endophytic fungi while some subspecies of M. polymorpha have lost their fungi interactor [17]. Little is known about the bacterial inhabitants of Marchantia other than a handful number of reports on culturable bacteria isolated from Marchantia thalli [18,19]. Interestingly, there are reports of M. polymorpha extracts able to inhibit bacterial growth (Mewari and Kumar 2008), 95 suggesting an active microbial selection of its microbial guests. In summary, microbiome information from Marchantia species is practically inexistent.
In this work, we contribute the first microbiome study using massively parallel sequencing of the 16S rRNA gene for two wild Marchantia species, M. polymorpha, and M. paleacea along 100 their soil substrates to inquire if there is a differential microbe selection from its surrounding habitat, considering its relevant role in plant microbiome structure. Additionally, we contribute Specimens were thoroughly collected with sterilized tweezers and contained into 50 ml sterile tubes that were frozen in place with liquid Nitrogen. Approximately 50 ml of soil volume were collected for each plant to study the role of soil as the inoculant for Marchantia species microbiomes. Landowners allowed us to sample within its terrain and there was no need for 115 special collect permissions. Neither Marchantia species are included in the Red List of the International Union for Conservation of Nature (IUCN; http://www.iucnredlist.org/search).
Gemmae from both Marchantia (female) species grown in the wild was surface disinfected and grown under in vitro conditions (following the detailed protocols described in Ishizaki et al. 2016) to generate thalli. We recovered gemmae from in vitro cultured thalli, and the 120 gemmae obtained were grown and propagated in vitro for three subsequent cycles (gemmaeto-gemma). The resulting thalli were processed for DNA isolation.

DNA extraction and sequencing library preparation
Soil and Marchantia species DNA was isolated using the PowerSoil DNA Isolation Kit (MoBio Laboratories, Solana Beach, CA). Briefly, five plants of each Marchantia species were washed with phosphate buffer as described previously [10], and the pellets were used as input for the 130 DNA extraction. The occurring natural populations of M. polymorpha and M. paleacea were gently washed away from the surrounding particles with phosphate buffer, and then washed by vortex mixing with Tween-20 1%, the pellets were used to extract the metagenomic DNA

Sequence processing
We used a previously reported pipeline [21] . Briefly, paired-end reads were merged using PANDASEQ [22], and quality control was done using fastx tools (http://hannonlab.cshl.edu/fastx_toolkit/). Briefly, all the reads were trimmed to the expected 160 amplicon size (250 bp), then assembled using the following parameters: a minimum probability threshold of 0.95 which accounts for the minimum probability cut-off to assembly; a minimum length of 250 bp, and a maximum of 470 bp. Clustering and OTU picking was done using cd-hit-est [23] with a 97% identity cut-off over a minimum of 97% of the read length. OTU table was built using make_otu_table.py of QIIME's suite [24], as well as the 165 picking of the representative OTUs. Taxonomy assignment was conducted with BLAST against Greengenes database (13_8 release; [25]). Chimeras were identified and removed with ChimeraSlayer [26]. Finally, identified mitochondrial and chloroplast sequences were removed from following analyses.

Diversity metrics and statistical analysis
All the diversity metrics were calculated with R (R Core Team 2014) and its phyloseq [27], and vegan [28] packages. Plots were done using R's phyloseq, ggplot2 [29], and RColorBrewer (www.ColorBrewer.org) libraries. For the alpha diversity metrics, we calculated a number of 175 observed species, nonparametric Chao1 index [30][31][32], Shannon's index [33], and Simpson's diversity index, for the Marchantia species, and for the comparative dataset. Venn diagrams were generated using the Draw Venn Diagram tool (http://bioinformatics.psb.ugent.be/webtools/Venn/). Communities distances were calculated using Bray-Curtis dissimilarities and ordered with non-metric multidimensional scaling (NMDS) 180 with phyloseq. Several normalization procedures were conducted on the data, first for bar plots for phyla abundances relative frequency transformation was done to compare across samples/species. To compare significant differences between Marchantia samples, regularized logarithmic transformation (rlog) was applied to the OTUs. Counts which are computed by fitting each OTU to a baseline abundance using generalized linear model (GLM) 185 to a baseline level, then estimating the logarithmic fold change (LFC) and dispersion for each OTU to the baseline, finally correcting for multiple testing, these comparisons were made using R's DESeq2 package [34].

Compared microbiomes
For comparison purposes we choose several available root microbiomes for plant species, 195 trying to get a sparse representation of land plants. We took one Sphagnum magellanicum moss microbiome sequences [12]; four samples from the Pinus muricata microbiome study [13], here we were able to select root microbiomes from individuals with and without arbuscular mycorrhizas; two samples from B73 maize root microbiomes [14]; two samples from rice's roots [15]; one sample from the bladder associated microbes of the carnivorous plant 200 Utricularia gibba [35]; and finally, two root associated microbiomes from Arabidopsis thaliana and its surrounding bulk soil [10].
Sequences for similar microbiomes were downloaded from the databases declared in their respective publications. All the sequences were processed as stated before. We performed 205 both QIIME's pick_closed_otus.py and individual study OUT picking, and clustering with cdhit-est and Greengenes taxonomy assignment for all the microbiomes and the Marchantia related samples.

Marchantia microbiome richness and diversity
Marchantia has emerged as an important model organism for plant developmental biology [8], and it is also a suitable evolutionary model to understand the particular adaptations of 215 early terrestrial plants to colonize and conquer the Earth's surface [4]. This work is the first attempt to understand the richness, and diversity of Marchantia's microbial inhabitants, under wild and community established under axenic conditions [36]. The plant's microbiome, specifically the root-associated microbiome has been shown to have dramatic effects on plants establishment, survival, and access to nutrients [11]. Given their anatomical structure, 220 profiling microbiomes from Marchantia's thalli (also containing single cell rhizoids) would be the equivalent of microbiome profiles obtained from both root and phyllosphere in vascular plants.  Taxonomic assignments of the OTUs were conducted, and here we show phyla affiliations for 300 each of the samples, the data sets were transformed to relative frequency and are presented in Proteobacteria is observable the case of M. polymorpha (Fig 2A).
To better understand Marchantia microbiome results and make sense of the microbial 315 communities described here, we performed a comparative study with other plant and soil microbiomes (Fig 2B). We compared Marchantia bacterial communities with microbiomes from the following plant species: The bryophyte moss Sphagnum magellanicum, the gymnosperm Pinus muricata (containing two datasets, one from roots containing arbuscular mycorrhiza and two mycorrhizae free), rice, maize, Arabidopsis thaliana, and Utricularia gibba 320 (see Methods). By comparing the most abundant phyla across all datasets (Fig 2B) Using Shannon's index we can sort the compared microbiomes based on their diversity (Table  340   2   Comparison of the microbiomes distances using NMDS and Bray-Curtis dissimilarities (Fig 3) showed that the in vitro Marchantia species cluster apart from every other microbiome and 365 that they cluster together. Interestingly, there is less distance between wild M. paleacea and its soil, than the distance between M. polymorpha's soil and microbiomes, but the two Marchantia species are closer in natural conditions than to any other compared plant microbiome. Additionally, the plant's microbiome tends to cluster together like is the case for rice, Arabidopsis, and even a sparse cluster is observed for the Pinus. Maize has a 370 considerable distance between their microbiomes, but still into the left side of the NMDS1 axis, along with the plant and soil microbiomes, all but the in vitro Marchantia microbiomes. We conducted both qualitative and quantitative analysis to determine the shared, unique, and significant differential features for Marchantia spp., and their soils. First, the qualitative approach using sets comparisons shows that 8,035 OTUs in M. paleacea are shared between M. paleacea and its soil; 7,180 OTUs are unique to that ground, and 9,578 are unique to the plant (Fig 4). The large number of particular plant OTUs strikes us. Some sources for these 415 unique bacteria could be directly from air and water. However, air and water microbiomes are less diverse than soil microbial communities: in rainwater, and river microbial richness (Chao1 index) is estimated under 1,000 OTUS, while richness for air is estimated below 600 OTUS [39,40]. This diversity of microbial sources suggests that plant microbiome composition is influenced by plant-environmental interactions with some bacteria probably being inherited 420 from plant to plant, and some bacteria pass fleetingly by soil, air, or water. There are 16 OTUs  [19]. Methylobacterium is a phytosymbiont for liverworts and mosses and within liverworts is proposed to consume methanol as the by-product of plant's cell wall 450 metabolism which is emitted through Marchantia's upper epidermis stomata-like pores [19,45,46]. Methylobacterium is probably accumulated within plant's air chambers.

460
Paenibacillus chondroitinus which was previously classified as Bacillus chondroitinus and named by its ability to degrade complex carbohydrates like chondroitin, and alginates [48] was found as a core Marchantia microbiome member. Other common soil bacteria such as Azotobacter vinelandii have the ability to produce alginates and is crucial for the development of desiccation resistant cysts. Bacteria produced alginates can also be used as virulence 465 factors, like Pseudomonas aeruginosa which happens to infect plants, insects, nematodes, and mammals [49,50]. Marchantia is widely known for its desiccation resistance properties, and it is known that they can transpire the equivalent of their total water content in very short times under extreme conditions [51]. The genera with significant (padj=0.01) log2 fold changes are shown. "Other" corresponds to OTUs whose genera could not be assigned. Each dot represents an OTU. The OTUs over-480 represented in Marchantia spp., are located on top of the plot while the OTUs overrepresented in soil are located at the bottom. The majority of the OTUs were not assigned to a known genus, and are shown as "Others" in the plots. Shared genera between both Marchantia spp., are highlighted in blue. 485 The quantitative approach was used to detect the strongest interactions between significant (padj=0.01) differential OTUs between plants and soil bacteria by comparing the log2 fold change (log2fc) of the OTU abundance ratio for wild Marchantia spp., vs. origin soil (Fig 5).
The utility of using DESeq2 [34] log2fc where the fold changes are lower than 1 become negative values, while fold changes larger than one will be positive values, making easy to 490 plot the changes symmetrically in a single plot. Then the comparison gets an adjusted p-value Proteobacteria: Steroidobacter. Steroidobacter has been isolated from soils and plant roots. It has been shown that some strains like Steroidobacter agariperforans can degrade complex polysaccharides derived from rhizospheres, and even able to degrade agar. Even more, S. agariperforans was isolated as a commensal strain to a Rhizobiales bacterium [56]. Another species, S. denitrificans is capable of denitrifying under anoxic conditions, using nitrate as an 520 electron acceptor and can degrade steroidal hormones as well [57]. Recently, Steroidobacter has been described as part of the core rhizosphere microbiome of the gymnosperm Welwitschia mirabilis described as a living 110 million-year-old living fossil [58].
A nutrient rich environment like plant derived photosynthates is certainly attractive also for 525 predators, and bacteria are no exception. Bdellovibrio bacteriovorus is the best studied of the Bdellovibrio genus; it is a sophisticated Gram-negative bacteria predator [59]. Bdellovibrio species has been isolated from a wide range of environments, and it is the case for plant rhizospheres. There is a report that establishes one order of magnitude major concentration of B. bacteriovorus in rhizospheres than circumventing soil [60] and it is very likely they are 530 preying bacterial cells feeding on the C-rich plant exudates. Marchantia enrichment in Bdellovibrio is interesting, because it is not usually a major player in soil and rhizospheres, and it is possible it might be preying other Gram-negative inhabitants potentially pathogens as it is the case of Agrobacterium which happens to be over-represented in M. polymorpha.

535
Lysobacter is a Gammaproteobacteria commonly found in soils and rhizospheres. It is a bacterium with gliding motility that belongs to the family Xanthomonadaceae. Lysobacter produces antibiotics and lytic enzymes like chitinases and glucanases which protect plants from pathogenic fungi and bacteria promoting plant health and growth. Lysobacter has been proposed as a biological control agent for crop protection [61][62][63]. 540 Pirellula is a Planctomycetes ubiquitous in marine and terrestrial anoxic environments.
However, there are few cultivated members for the Planctomycetes. One of the few sequenced strains of formerly known Pirellula sp. 1 (Rhodopirellula baltica) revealed a potential to degrading C1 compounds and pathways that were thought to be exclusive for 545 anaerobic methylotrophic bacteria and Archaea and is considered to be generating energy from the metabolism of mono or disaccharides of disaggregated algae polymers [64]. The presence of Pirellula reinforces the importance of methylotrophs within Marchantia microbiomes. Methylobacterium, which have been reported as capable of Nitrogen fixing, and plant hormones synthesis. We also found complex organic compound degrading bacteria such as Paenibacillus, Steroidobacter, and Lysobacter, which have been reported to use plantderived polymers and return plant hormones that can provide pathogen protection to their hosts. We recorded the presence of bacterial predators like Bdellovibrio that actively 560 attacks and parasite other Proteobacteria which suggest that negative interactions among Marchantia spp., inhabitants might be taking place. The enrichment for methylotrophic bacteria is likely to be the result of the bacteria niche opportunity and specialization found in the Marchantia air chambers. 565 Marchantia stomata-like pores (air chambers) are visible to the naked eye and remain open as they are not fine regulated. Air chambers allow gas exchange while controlling internal moisture in contact with photosynthetic cells [51]. The importance of methylotrophic bacteria in the two Marchantia species has been highlighted as only two OTUs belonging to this genus are shared in both in vitro and wild conditions. Methanol has been reported as a normal 570 emission of volatile organic compound (VOCs) from leaves and through plant stomata [46], and the release of plant methanol has been correlated with leaf growth [45]. The air chambers of Marchantia are a wonderful habitat for various bacteria which can make use of C1 sources like methanol and live within the air chambers, as it is probably the case for Methylobacterium.