Bacterial communities in habitats of Dongchuan, China: Their role in slate weathering in triggering and owing areas of debris ow

Debris ow is one of the most severe disasters in China’s mountainous areas. Landslide soil materials are deposited on steep hill terrains or in channels and, once triggered by heavy rainfall, they can evolve into lethal debris ows. Some studies have suggested a prominent role of rock weathering in the formation of large landslides. In the present study, the a-diversity, composition, and weathering potential of bacterial communities inhabiting surfaces of slate rocks from the triggering area (altered less or more) and slate rocks from the owing area of debris ows were studied. We found that a-diversity indices were positively correlated with the organic matter content, and Firmicutes accounted for about 66% of the total bacterial community. The variation in bacterial community composition was majorly driven by available Ca. The highest relative abundances of genetic families associated with organic acid production, agellar assembly and carbonic anhydrase were found in surfaces of less-altered slate rocks from the triggering area by using PICRUSt. All the bacterial strains isolated from rock surface have the ability to accelerate the weathering of slate, among which Pantoea and Pseudomonas genera showing the highest ability to enhance the dissolutions of Fe, Si, Al, and Ca from slate. These results indicated an important role of the rock surface-inhabiting bacterial communities in the weathering of slate in both triggering and owing areas of debris ow. This study also provides new insights into the biological characterizations of rocks in debris ow areas of Dongchuan, China.


Introduction
Debris ow is one of the most serious disasters in mountainous areas of China [1][2]. Due to the awareness of the risk of debris ows, studies concerning on the phenomenon and its potential consequences have received increasing attention [3]. However, there is hitherto no published study focusing on the biological characterization of rocks in debris-ow areas.
Considered as a natural museum of debris ow disasters, the Dongchuan district of Kunming city is one of the areas that suffered the most frequent and intense debris ows in China [4]. Within Dongchuan area, the exposures lithology is mainly composed of slate rocks. Being affected by tectonic movement and climatic factors for a long time, rocks in this region have suffered both chemical and physical weathering. Chemical weathering can change the chemical composition, textural, and mineralogical of rocks through oxidation and hydrolysis processes, reducing the strength of rocks [5]. So chemical weathering of rock may be more prone to slope failures, erosion, and landslides [6]. Landslide was considered one of the major triggers of the hill-slope debris ow in this area [7], where the weathered rock pro les provide abundant solid materials for the occurrence and development of debris ows [8].
Rocks can be considered primary ecosystems, on which only a few adapted microorganisms with rockweathering abilities can survive and grow [9]. Microbial communities have been documented to enhance rock and mineral weathering reactions [10][11] through oxidoreduction reactions, acidolysis and chelation reactions [12]. Furthermore, microbial adhesion, where bacterial agella play an important role [13], contributes great contribute to biological weathering of minerals [14]. Carbonic anhydrase secreted by bacteria also plays a key role in mineral weathering [15].
Microbial communities have also been identi ed in various rocks [16], where they may facilitate the extraction of elements through bio-weathering [17][18]. Previous studies have suggested that microbial colonization on mineral surface was driven by surface charge, chemical property of rocks and mineral particles, and the local ecological niche [19][20]. Up to now, there is no published study investigating the diversity, composition, and ecological functions of bacterial communities inhabiting rocks with different alteration levels in debris ow areas.
Explaining the effects of rock mineralogy on bacterial communities' diversity, structure composition, and ecological functions related to rock condition greatly contributes to enhance the understanding of the role of bacterial communities in rock weathering [21]. In this study, rocks with different degrees of alteration from the triggering area and rocks from the owing area of debris ow areas in Dongchuan district were collected for analyses of bacterial communities' diversity, structure composition, and rock-weathering functional potential by combining culture-independent and -dependent methods. We hypothesized that i) bacterial communities inhabiting rock surfaces in the debris ow areas are distinct from those inhabiting other rock ecosystems; ii) the α-diversity and composition of bacterial communities and the rockweathering potential vary among different groups; iii) the variation of bacterial communities may be driven by geochemical properties of rock samples. Our results add a new insight into the microbial communities inhabiting rock surfaces in debris ow areas and the speci c functional groups involved in rock weathering, offering a perspective to advance the current understanding of the biological characteristics of debris ow.

Site description and sample collection
The rock samples were collected from the metamorphic slate deposit area (20°14 N, 103°08 E) in Jiangjiagou valley (Dongchuan district, Kunming city, Yuannan province, China) in October 2014. Less altered and more altered rock samples (considered as LT group, and MT group, respectively) from triggering area, and rock samples from owing area (considered as F group) were collected with sterilized shovels for comparison (Fig. S1). The weathering degrees of rocks were judged by visual observation. In particular, rock samples of LT group were obtained by striking the rock with sterilized hammers. Samples from each rock group were collected in triplicate (nine samples totally, each of which 600 g; rock samples were < 2 cm in diameter). Then they were taken to the laboratory with dry ice and sieved (2 mm) within 48 h after collection. Samples for microbial DNA extraction were frozen at -80°C, while samples for chemical analysis and bacterial isolation were stored at 4°C within 24 h.

Geochemical analysis of rock samples
Each rock sample was air-dried to measure pH and content of organic matter (OM) [22]. The contents of available elements (including Al, Si, Ca, Cu, Fe, K, Mg, Mn, and Na) were determined by ICP-OES (Optimal 2100 DV, Perkin Elmer, USA) according to the method described by Huang et al. [22]. For details, see Supplementary Methods. For the analysis of mineral composition, the rock samples were ground to powder (< 150 µm) and detected by X-ray Diffraction (XRD) as described by Wang et al. [23]. In addition, analysis of particle size separation was conducted by applying wet sieving and centrifugation [23].

DNA extraction and MiSeq sequencing
Genomic DNA was extracted from 1 g rock samples with the Fast DNA® Spin kit soil (MP Biomedicals, Santa Ana, CA, USA). A bacteria-based primer set 515F/806R [24] was used to amplify about 300-bp fragments of the V4 hypervariable region of the bacterial 16S rRNA genes. PCR was performed according to the method described by Wang et al. [25]. Then, the high-throughput paired-end sequencing was performed with the Illumina MiSeq PE250 platform (Illumina, San Diego, CA).

Isolation of bacterial strains and phylogenetic assignment
The bacterial counts on rock surfaces were analyzed by using the sucrose-salts medium (SSM) according to the method described by Zhao et al. [21]. DNA extraction, PCR, and sequencing analysis of the obtained bacterial strains were conducted according to the method described by Huang et al. [22]. The resulting nucleotide sequences were analyzed by availing of the NCBI database to obtain the closest species match. Phylogenetic trees were constructed by the neighbor-joining method using the MEGA 4.0 software, and bacterial strains belong to the same genus was clustered into one branch.

Rock weathering assays
The rock-weathering potential experiment was performed according to the method described by Huang et al. [22]. The elemental composition of the slate rock was as follow: SiO 2 68.80%, Al 2 O 3 16.20%, K 2 O 3.89%, Fe 2 O 3 5.14%, CaO 2.72%, MgO 1.94%, and TiO 2 0.76%. Bushnell-Haas medium (BHm) was used to evaluate whether the bacteria could release Si, Al, Fe, and Ca from slate under low-nutrient condition [23]. Sterilized BHm (containing slate) without incubation of bacterial strain was used as control to monitor the range of abiotic dissolution. After 7 days of incubation, pH, and concentrations of Si, Al, Fe, and Ca were determined by a pH meter and an ICP-OES, separately. Further details of the method are given in Supporting Materials.

Statistical analyses
The raw sequences of 16S rRNA gene MiSeq sequencing were processed on the i-Sanger platform (http://www.i-sanger.com/) as described by Wang et al. [25]. In order to avoid heterogeneity related to different sequencing depths, a randomly-selected subset of 11,000 sequences per sample was performed for down-stream analysis. Alpha diversity metrics studied were Shannon index [26], Chao 1 index [27], and Faith's PD [28]. PCoAwas employed to assess the variation in microbial community composition between samples These were followed by signi cance tests using ANOSIM [29] and PER-MANOVA [30]. In addition, bootstrapped trees were constructed using the UPGMA [31]. RDA was carried out to determine the effects of rock properties on the bacterial community in R using vegan. Possible correlation between bacterial community diversity, and rock properties was investigated by using the Spearman's correlation coe cient [32]. T-test was applied for signi cance testing. Potential differences in the rock properties, the α-diversity indices of bacterial communities, the pH, and concentrations of elements in the cultural medium with incubation of bacterial strains from different groups were tested by using Tukey's HSD posthoc test by using SPSS v.21. The effects of bacterial genus on rock weathering were determined by ANOVA and Fisher's test at a threshold value of α = 0.05.
PICRUSt, a computational approach [33] to predict the functional composition of a metagenome using marker gene data and a database of reference genomes, was also conducted on the i-Sanger platform to compare the potential ecological functions of bacterial communities among different groups. For details, see Supplementary Methods. In addition, functions associated with siderophore biosynthesis, carbonic anhydrase (CA) [15], agellar assembly [13], and organic acid production [34] were selected as indicators of rock or mineral weathering potentials. Potential differences in the relative abundance of each rockweathering related functional category were also tested with Tukey's HSD post-hoc test.

Characterization of the rock samples
As shown in Table 1, different patterns in available element content were found among different groups (Table 1). A signi cantly lower content of available Ca was found in group LT than in groups MT and F. All the rock samples had a relative alkaline pH (7.79~8.40). Based on the phase analysis of XRD, rock collected was found to be slate, and similar mineral constituents were found among these rock samples.
The possible major mineral phases included 2:1 type clay minerals, kaolinite, biotite, and quartz ( Fig S2). The proportion of gravel (> 2 mm) was signi cantly higher in group LT than in groups MT and F, suggesting a lower alternation degree of rock samples of group LT (Supplementary Table S1). Overall bacterial α-diversity indices exhibited signi cant differences (P < 0.05 in all cases, Tukey's test) (Fig. 1a). Collectively, the values of all the indices were highest in group LT, followed by groups F and then MT, suggesting that the less alerted slate harbored a higherα-diversity of bacterial communities UPGMA was used to compare the relatedness of bacterial communities based on the distribution of OTUs. The nine samples under investigation were divided into three groups, with group MT closely relating with those of group F (Fig. 1c). These results are supported by the results of two-dimensional PCoA using Bray-Curtis similarity distances (Fig. 1b), unweighted unifrac distance (Fig. S3a), and weighted unifrac distance (Fig. S3b). The results were further con rmed by ADONIS and ANOSIM (Table  S2).
At the phylum level, the relative abundances of Acidobacteria and Proteobacteria were highest in group LT, followed by groups F and then MT. The highest relative abundances of Chloro exi and Firmicutes were found in group MT, followed by groups F and then LT. The relative abundance of dominant taxa at class level also varied among rock samples from different sampling sites (Fig. 1c).The highest relative abundance of Bacilli was found in group MT, followed by groups F and then LT (Fig. 1d).
At genus level, 190, 125, and 172 genera were obtained from LT, MT, and F, respectively. The distribution of major bacterial genera (accounting for more than 0.25 % of the total sequences) were presented in Table S3. Especially the ammonia oxidizing-related genus Nitrospira appeared signi cantly more abundant in group LT than in groups MT and F. Furthermore, Rhodobacter, Hydrogenophaga, and Limnobacter were only observed in group F (Table S3).
Regarding OTU level, the exclusive and shared species-level OTUs among rock samples were inspected.
The results showed that 281 out of 3,114 total OTUs (9.5%) were shared among groups LT, MT, and F (Fig. 1e, left panel), accounting for 84.7 % of all the sequences, indicating that these shared OTUs are mostly more abundant species. While OTUs unique to each group were mostly species with low abundance (Fig. 1e, right panel).

Correlations between rock properties and bacterial community composition
Rock properities (including contents of Ca, Mg, Si, and Mn) signi cantly correlated with the RDA model (P < 0.05) were selected to analyze their respective in uences on bacterial community composition (Fig. 2). RDA explained nearly 99 % of the total variation in the rock bacterial community structure, with the rst two axes explaining 93.13% of the variation (Fig. 2a). Rock bacterial communities formed similar clusters on the RDA plot (Fig. 2a), in agreement with PCoA analysis (Fig. 1b). Canonical variation partitioning revealed that the available Ca content was the major contributor of bacterial community variation, explaining 36.6% of the variation (Fig. 2b). The bacterial diversity indices were correlated negatively with available Ca and Mg contents (r = -0.694~-0.907) and positively with Si and Mn contents (r = 0.760~0.973) of rock samples (Table S4). In addition, the diversity indices correlated positively with OM content (r = 0.795~0.913) and negatively correlated with pH (r = -0.932~-0.913) of rock samples.

Predicted microbial metabolic pro les
Using PICRUSt, 28 of 43 level 2 KEGG Orthology groups (KOs) were represented by the data set, and 23 genetic families showed signi cant differences (P < 0.05) among groups LT, MT, and F (Fig. S4). As for the mineral weathering-related functional potential, the relative abundances of genetic families associated with organic acid production and agellar assembly were highest in group LT, followed by groups F and then MT (Fig. 3). The highest relative abundance of CA was found in LT group. However, the relative abundance of genetic families involved in siderophore biosynthesis in groups MT and F was signi cantly higher than that of group LT. Collectively, these results indicate that the bacterial communities may in uence the rock weathering through a variety of mechanisms in the rock environments.

Rock weathering assays
On the rock surfaces, 10 5 -10 6 bacterial counts per gram of rock were obtained. Furthermore, cultivable bacterial counts in group LT were signi cantly higher than those in groups MT and F (  Fig. 4, the ratios of active Fe, Al, and Si solubilizers were signi cantly higher in group LT than in groups MT and F. The proportion of active Fe solubilizer was signi cantly higher in group F than group MT (Fig. 4b). No signi cant difference in the ratio of active Ca was found among different rock samples (Fig. 4f). Based on the distribution of the concentrations of Fe, Si, Al, and Ca in cultural medium in the presence or absence of bacterial strains (Table S6), these strains could be grouped into the following three categories: strains with poor rock-weathering potential (< 5 µM Fe, < 8 µM Al, < 110 µM Si, and < 800 µM Ca in the culture), moderate rock-weathering potential (5-10 µM Fe, 8-10 µM Al, 110-120 µM Si, and 800-1000 µM Ca), and high rock-weathering potential (> 10 µM Fe, > 1 µM Al, > 120 µM Si, and > 1000 µM Ca). As shown in Fig. 4, the proportion of highly active Fe, Al, and Si solubilizers was highest in group F, followed by groups LT and then MT. The proportion of highly active Ca solubilizer in groups LT and F was signi cantly higher than that in group MT, and no signi cant difference was found between groups E and F (Fig. 4f). Collectively, the rock weathering potential of slate-weathering bacteria form groups LT and F was higher than mineral-weathering bacteria from group MT.
The pH in the culture medium ranged from 3.21 to 8.76 in the presence of different bacterial strains ( Table 2, Table S6). Most of the bacterial strains (about 78 %) showed high alkaline pH values. The ratio of highly (pH < 4) active acid-producing strains was highest in group F, followed by groups MT and LT. In addition, the proportion of the alkaline-producing (pH > 7) strains in groups LT and MT was higher than group F.
3.6 Linking of culturable bacterial taxonomy with their rock weathering potential As shown in Fig. 5, the rock-weathering bacterial strains were a liated with 14 genera (8, 7 and 8 genera were obtained from groups LT, MT and F, respectively). Bacterial strains assigned to Pseudarthrobacter (n = 129), Arthrobacter (n = 11), and Bacillus (n = 59) were common in the rock samples. Strains belonging to Microbacterium (n = 2) and Sphingomonas (n = 4) were detected in both LT and MT groups, and Pseudomonas (n = 18) was found in both MT and F groups. Furthermore, each group had its unique genera (Fig. 5a) The relative activity of solubilizing Fe, Al, Si, and Ca from slate were compared among different genera (with a minimum of 3 strains for each genus). As shown in Fig. 5b, representatives of Pantoea and Pseudomonas exhibited considerably higher activities promoting the release of Fe, Al, Si, and Ca from slate, compared with other genera. The medium pH of Pantoea (4.41 ± 1.96) and Pseudomonas (4.95 ± 1.94) strains was signi cantly lower than that of other genera. In addition, pH showed negative correlations with Fe, Al, and Ca released from slate (r = -0.824 ~ -0.876).
When inter-comparing the taxonomic data obtained from the cultivation-dependent and -independent methods, 4 of the 25 most abundant bacterial genera identi ed through 16S rRNA gene Miseq sequencing (19 % of total sequences) were also identi ed by the cultivation-dependent method (Table   S3). Most of the cultured genera with rock-weathering potential were also detected by the cultureindependent method, except for Pesudarthrobacter, Micrococcus, Terrabacter, Ensifer, Pantoea, and Enterobacter (Fig. 5a). Generally, the cultivable rock-weathering bacterial genera took up a small proportion of the bacterial community detected by cultivable-independent method.

Discussion
Phylogenetic analyses revealed diverse and distinct bacterial communities in diverse ecosystems [19].
Especially, Firmicutes was the most dominant phylum identi ed in this study, ranging from 34-83% of the total bacterial communities. According to previous studies, abundant Firmicutes was found in soil and aquatic environments, as well as in the anaerobic digestion processes where they play an important role in the decomposition and recycling of organic matters [38]. In the present study, Bacillus, belonging to Firmicutes, was found to be one of the most abundant genera by both cultivation-dependent andindependent methods in the studied area. Previous studies detected Bacillus in various rocks and soils [23,36,39], possibly be due to their key role in biological control of pathogenic fungal growth, enzymes used in agriculture and industry, mineralization of organic matter, plant growth promotion, and rock weathering. Based on the investigation of bacterial communities in rock cores of common limestonemudstone alterations, Lazar et al. [40] suggested that Bacillus originate from the forest soils and are most likely feeding labile organic matter inputs in the rock matrix.
Previous studies have documented that changes in limited nutrient substrates (including Ca, Mg, P, K, Si, and Al) can cause variation of bacterial communities inhabiting rock surface [35,37,39]. Especially, the presence of limited nutrients in rocks can render rocks attractive to preferential colonization by microbiota with different composition [41]. Consistent with previous studies, available Ca content in the slate rocks was found to be the major driver of variations in bacterial community structure. As an essential nutrient for living organisms, Ca is involved in bacterial growth and metabolism processes, including heat shock, chemotaxis, differentiation, and cell cycle [37]. Sridevi et al. [42] also found that the addition of Ca signi cantly shifted the composition of soil bacterial communities in a deciduous secondgrowth forest, and affected the relative abundance of hundreds of bacterial taxa. In this study, the available Ca content correlated negatively with the α-diversity indices. Similarly, Groffman et al. [43] found that Ca addition led to a lower microbial N biomass in a northern hardwood forest. Besides, Kalwasińska et al. [37] found that the number of OTUs in the salino-alkaline lime was positively correlated with the Ca concentration. The discrepancy in ndings may be contributed to the unique microhabitats the bacterial communities occupied. In addition, the α-diversity indices also positively correlated with the content of OM in this study. Kirtzel et al. [44] found that the weathering of black slate resulted in the degradation of OM and the release of organic chemical compounds when exposed to oxygen. A similar phenomenon was found inour study, where the OM content in the less altered slates was signi cantly higher than that in more altered slate rocks.
In order to evaluate the rock weathering potential and the involved mechanisms of the bacterial communities, we analyzed both the relative abundance of the weathering-related gene functional category with PICRUST and the effects of bacterial strains on the release of structural elements from slate from different groups with cultivation method. The PICRUST results suggest that the bacterial communities may in uence the rock weathering through a variety of mechanisms among different sites.
Based on the cultivation-dependent method, it was found the surface bacterial communities may play an important role in slate weathering. Furthermore, distinct bacterial rock-weathering effectiveness and phenotypes were observed among the different groups based on the element release patterns of the bacteria ( Table 2, Table S6), which might be affected by the sample-speci c characteristics (Tables 1, Fig.  S2). Previous studies have suggested that the weathering effectiveness and phenotypes were dependent on the bacterial populations and their origin [22].
In the present study, acidolysis was considered as the major mechanism involved in slate weathering for both Pantoea and Pseudomonas genera, which have also been documented to accelerate the weathering of K feldspar, biotite, apatite, smectite, and granite through the production of organic acids and siderophores, as well as pilus formation [34,[45][46]. However, in this study about 78 % of the slateweathering bacterial strains could alkalize the medium. Alkalization is one of the mechanisms involved in mineral weathering [47]. Kisiel et al. [48] found that glauconite weathered to form pseudomorphs of goethite, which was possibly due to the alkaline pH of the environment caused by the high calcite content. During the microbial weathering of calcite, Dou and Lian [49] found that in the presence of rockinhabiting fungal strain pH in the cultural medium decreased from 7 to 6.8 within 24 h and then increased to 7.8 within 32 to 56 h, and production of acetic acid was considered as the major mechanism involved in the weathering of calcite. Considering the practical constrains of time and resources for inducing the release of the elements from the slate by 248 bacterial strains, a deeper insight into the analysis of metabolites of bacterial strains during their bio-weathering of slate needs to be further considered in the future.

Conclusion
By combining cultivation-dependent and -independent approaches, this study demonstrates for the rst time that slate rock surfaces were inhabited by diverse and distinct bacterial communities with slateweathering potential in both triggering and owing areas of debris ow. The αdiversity indices positively correlated with the content of OM of rock samples, and the less altered slated collected from the triggering area had the highest diversity indices. Our study found that the content of available Ca was the major driver of the variation in bacterial community structures among the slate rocks of different groups, with Firmicutes being the most abundant bacterial phylum. The culture-dependent analysis suggested that all the bacterial strains could accelerate the weathering of slate, among which Pantoea and Pseudomonas genera showed the highest effectively ability to enhance the dissolution of Fe, Si, Al, and Ca from slate.     In uence of bacteria isolated from rock surface on the elements released from potash trachyte. The raio of bacterial strain that signi cantly enhance the release of Fe, Si, Al, and Ca from slate isolated from groups LT, MT, and F, respectively (A). Proportion of the highly, moderately, and poorly effective Fe (B), Si (C), Al (D) and Ca (E) solubilizers isolated from groups LT, MT, and F respectively. Error bars indicate standard errors (n = 3), while different lowercase letters above error bars represent signi cant differences according to Tukey's Honestly Signi cant Difference (HSD) post-hoc test (P < 0.05). A total of 248 bacterial strains were tested for their ability to weathering slate.

Figure 5
Comparison of the distribution and their rock-weathering ability of bacterial genera. A. Taxa and lineages of cultivable rock-weathering bacterial populations clustering, and the relative abundances of these genera in different groups based on both cultivation-dependent and -independent analyses. Bootstrap