Microbial communities in the native habitats of Agaricus sinodeliciosus from Xinjiang Province revealed by amplicon sequencing

Agaricus sinodeliciosus is an edible species described from China and has been successfully cultivated. However, no studies have yet reported the influence factors implicated in the process of fructification. To better know abiotic and biotic factors, physiochemical characteristics and microbial communities were investigated in five different soil samples collected in the native habitats of specimens from northern Xinjiang, southern Xinjiang, and Zhejiang Province, respectively. There are major differences in texture and morphology among different specimens of A. sinodeliciosus from Xinjiang Province. A. sinodeliciosus from southern Xinjiang was the largest. Concentrations of DOC and TN and C/N ratio are not the main reason for the differences. Microbial communities were analyzed to find out mushroom growth promoting microbes (MGPM), which may lead to the differences. Functional microbes were picked out and can be divided into two categories. Microbes in the first category may belong to MGPM. There may be symbiotic relationships between microbes in the second category and A. sinodeliciosus. Certain analyses of microbial communities support the hypothesis that interactions between microbes and mushrooms would be implicated in morphological variation of the collected mushrooms. Redundancy analysis results indicate that high DOC/NH4+-N ratio and NH4+-N concentration can improve the yield of A. sinodeliciosus.

were relatively lower. C/N ratio of soil samples of specimen ZRL20151244 from Zhejiang Province was higher than that of ZRL20152591. These results indicate that concentrations of DOC and TN and C/N ratio are not the main reason for the differences. It was assumed that there may be some MGPM in the native habitats of A. sinodeliciosus from southern Xinjiang, which may lead to the differences. Therefore, microbial communities in the native habitats of different specimens were analyzed.
Richness and diversity of bacterial and fungal communities in different samples. Due to many failures of PCR amplifications, fungal communities of soil samples in the native habitats of specimen ZRL20152590 was not successfully investigated. A total of 4438934 paired-end reads for bacterial communities and 12190961 paired-end reads for fungal communities were produced. After filtering, 2498745 and 1983159 clean tags were obtained, respectively. These clean tags were assigned to 26140 and 5325 OTUs at a 97% similarity, respectively. However, most of rarefaction curves cannot reach saturation ( Supplementary Fig. 1), which means that further sequencing is valuable to detect more species. For bacterial communities, the OTU number (Supplementary Table 1) in A06 (838) was the largest among the 45 samples. For fungal communities, the OTU number in D07 (223) was the largest among the 36 samples.
To better understand the differences among the communities, it is important to calculate the richness, evenness, and diversity 22 . Community richness can be demonstrated by Chao1 and ACE 23 . Simpson and Shannon diversity index were used to show community diversity, which demonstrate not only the species richness but the evenness of the species [24][25][26] . The patterns of Chao1 and ACE were very similar to the OTU numbers 27 . For bacterial communities, on the basis of OTU number, Chao 1 and ACE (Supplementary Table 1), soil sample A had the richest diversity, followed by soil sample D, whereas soil sample E showed the least richness. The Shannon diversity indices of soil sample A and D were higher than those of other specimens. The results of Simpson index were in contrast to those of Shannon diversity index. For fungal communities, soil samples A, D, and E had higher richness and diversity.
Comparative analysis of bacterial and fungal communities. Hierarchical cluster analysis of communities at genus level was used to demonstrate the different compositions of the microbial community structures (Fig. 3). For bacterial communities, the A-D group was separated from B, C, and E group. For fungal communities, in general, the A-D group was separated from B and E group. These results indicate that there are obvious differences in bacterial and fungal communities among the different samples. Based on considerations of differences in territory among different sampling sites, which can cause the differences in microbial community, there was a hypothesis that A-B, C-D, and E should be clustered together, respectively. Principal Coordinate Analysis (PCoA) was calculated (Fig. 4) and previous studies showed that the results from hierarchical cluster analysis were supported by PCoA 24,25,28 . In this study, for bacterial communities, A-B and C-D were clustered together, respectively and were well separated from E. For fungal communities, A and B were clustered and were well separated from D and E. These results were consistent with the hypothesis.
Unique and shared OTUs in the same sampling sites (01, 02, and 03) of topsoil from different specimens were summarized (Fig. 5). For bacterial communities, there were three main phyla in the shared OTUs: Actinobacteria, Firmicutes, and Proteobacteria, and Firmicutes were highly enriched with relative abundance of 59.64% in the shared OTUs of A03B03C03D03E03 (origin of mushroom) (Fig. 6). Some microbes belonging to Actinobacteria, Gram-negative Proteobacteria or Gram-positive Firmicutes could promote spore germination and hyphal  [29][30][31] . For fungal communities, the main phyla were Ascomycota and Basidiomycota (Fig. 6). However, the relative abundance of Basidiomycota in the shared OTUs of A03B03D03E03 was very high. This was due to the contamination of mushroom mycelium. If this reason was taken into account, the dominant phylum was Ascomycota in the shared OTUs of A03B03D03E03. Fungal decomposers of cellulose included Ascomycota and Basidiomycota 32 , which play an important role in the degradation of cellulose, the main polysaccharide in the soil, and were beneficial to mycelial growth 9 improving the mushroom productivity 12 . These results indicate that some microbes belonging to Firmicutes or Ascomycota may have the ability to improve the yield of mushrooms.

Microbial compositions.
To identify the phylogenetic diversity of microbial communities in different soil samples, qualified tags were assigned to phyla, classes, and genera. The phylum level identification of bacterial and fugal communities is illustrated in Supplementary Fig. 2. In total, 28 and 4 identified phyla were observed, respectively. For bacterial communities, a major difference in phylum level identification of bacterial communities between samples ABCD and E was the relative abundance of Chloroflexi, which played an important role in carbon cycling 33 . This difference may be attributed to regional divergence and differences in DOC concentration. It was found that the relative abundance of Bacteroidetes in samples C and D was higher than that in samples A, B, and E, especially in sample C. There was a hypothesis that bacteria in Bacteroidetes belonged to MGPM. To prove this hypothesis, community compositions at genus level should be detailedly summarized. Some isolates belonging to Actinobacteria, Gram-negative Proteobacteria or Gram-positive Firmicutes were MGPM [29][30][31] . Proteobacteria, Firmicutes, and Actinobacteria were highly enriched in all samples. The total relative abundance of these three phyla ranged from 19.2% (B02) to 97.2% (B04), only four samples of which were under 50%. These results indicate that the presence of these microbes may be beneficial for the growth of mushrooms. For fungal communities, if contamination of mushroom mycelium was taken into account, the dominant phylum was Ascomycota. However, Zygomycota was highly enriched in some samples D and E, and Chytridiomycota was only detected in D01 and D03.
The taxonomic breakdown at class level is shown in Supplementary Fig. 3. 58 bacterial classes and 11 fungal classes were detected. For bacterial communities, an obvious difference was that Acidobacteria and Actinobacteria were highly enriched in different soil samples of B, respectively. The relative abundances of Alphaproteobacteria and Acidimicrobiia in samples A, C, and E were much higher than those in sample B. Flavobacteriia was detected in samples C and D with high relative abundance. The relative abundance of Anaerolineae in sample D was much higher than that in other samples. Gemmatimonadetes was hardly detected in sample E. For fungal communities, the obvious difference was that the relative abundances of Dothideomycetes and Sordariomycetes in samples A, B, C, and D were higher than those in sample E. On the contrary, Tremellomycetes was detected in sample E with high relative abundance.  To prove the hypothesis, community compositions at genus level were detailedly summarized (Fig. 7). In total, 219 bacterial genera and 82 fungal genera were detected. There were great differences between ABCD and E. A lot of clean tags in each sample were not classified at genus level, especially in samples D and E.

Discussions
Community evenness was found to play an important role in resisting environmental stress 26 , which can be demonstrated by Shannon diversity index. Soil samples from Xinjiang Province were characterized by high salinity and pH. However, for bacterial and fungal communities, the Shannon diversity indices of soil sample E from Zhejiang Province were higher than those of soil sample B and C. Soil in Zhejiang Province was characterized by low salinity and pH. The relatively lower pH may lead to the higher values of Shannon diversity indices of soil sample E.
To find out functional microbes which were beneficial for the growth of A. sinodeliciosus, some special microbes were picked out. These special microbes can be divided into two categories. The first category is microbes highly enriched in the topsoil of sample C or D and hardly detected in other samples (Supplementary Table 2). The second category is microbes only detected in the soil of B, C, and D, and not detected in the soil of A and E (Supplementary Table 3). It was assumed that some microbes in the first category belong to MGPM, which have the ability to improve the productivity and reduce the harvesting time 6,9,11,12 . The microbes can be divided into five kinds. The first one was petroleum degraders. Alcanivorax was detected in sample C with high relative abundance (Fig. 7), which was considered as obligate hydrocarbonoclastic bacteria and took an important role in biological removal of petroleum hydrocarbons from petroleum-contaminated marine environments [34][35][36] .
Altererythrobacter was an important petroleum-aromatic degrader in marine environments 37 , which was also detected in sample C. These two genera were also detected in other samples from Xinjiang Province and not detected in sample E. Based on these results, there was an assumption that long long ago, the desert of Xinjiang Province may be a lake with high salinity and pH, which was polluted by petroleum.
The second one was microbes associated with nitrogen metabolism. Achromobacter highly enriched in sample C03 can grow anaerobically with KNO 3 38 . Parapedobacter, only detected in sample C and highly enriched in sample C03, can reduce nitrate into nitrite 39 . Filomicrobium, also only detected in sample C, was reported to be isolated from oil-polluted saline soil and positive for nitrate reduction activity 40 . The finding of Filomicrobium was consistent with the assumption. Halomonas can convert nitrate into nitrogen at high salinity and pH 28,[41][42][43] , which was detected in samples A, C, and D, and highly enriched in sample C01. Thiobacillus was only detected in sample D, which was able to reduce nitrate and nitrite 44 . The finding of these microbes indicates that there is nitrate in the soil of Xinjiang, and data of ion chromatograph shows that nitrate concentration was very high in the soil, especially in the topsoil of sample C (Supplementary Table 4). Paired-end sequencing detected not only denitrifying bacteria but also ammonia-oxidizing bacteria and nitrite-oxidizing bacteria. Nitrosococcus detected in sample D was reported to be able to utilize ammonia as energy source and reducing power for growth with nitrite as end product 45 . Nitrolancea detected in sample D can use nitrite or formate as energy source and CO 2 as carbon source 46 . The finding of these microbes indicates that nitrogen cycle existed in soil from sample D.
The third one was microbes associated with sulfur metabolism. Thiobacillus only detected in sample D was reported to be able to oxidize sulfide into elemental sulfur and convert nitrate into nitrogen simultaneously 47,48 . Sulfide was the product of sulfate reduction 41,42,49 . Desulfobacca also only detected in sample D was acetate-degrading sulfate reducer 50 . Although data of ion chromatograph shows that sulfate concentration was also very high in the soil of Xinjiang Province, bacteria associated with sulfur metabolism was not detected in other samples. It was found that sulfate concentration in the topsoil of sample D was the highest (Supplementary  Table 4), which may lead to this difference. In samples A, B, and C, there may be some rare microbes associated with sulfur metabolism, which were not detected under the current sequencing depth.
The fourth one was cellulose decomposers. It was reported that Owenweeksia producing oxidase, catalase, and alkaline phosphatase under high salinity conditions cannot use cellulose 51 , which were mainly detected in samples C and D. Iamia is positive for oxidase and catalase and can reduce nitrate into N 2 52 , which was highly enriched in sample C and hardly detected in other samples. Constrictibacter and Algoriphagus, highly enriched in sample C, can produce acid and alkaline phosphatase, esterase (C4), esterase lipase (C8), and β-glucosidase 53,54 , which plays an important role in cellulose degradation. Aspergillus can produce endo-glucanase and β-glucosidase 55 , which was detected in the topsoil of A, B, and D, and was highly enriched in sample D. Cellulose is the main polysaccharide in the soil 32 . Cellulose decomposers play an important role in the degradation of cellulose, which was beneficial to mycelial growth 9 improving mushroom productivity 12 . It was reported that in the acidic topsoil, cellulolytic bacteria included Betaproteobacteria, Bacteroidetes, and Acidobacteria, and fungal decomposers included Ascomycota and Basidiomycota, which were represented by Trichosporon and Cryptococcus 32 . Topsoil of Xinjiang and Zhejiang Province was alkaline and acidic, respectively. The bacterial decomposer Constrictibacter found in the topsoil of Xinjiang Province belongs to Alphaproteobacteria 53 and Algoriphagus belongs to Bacteroidetes 54 , respectively. The fungal decomposers found belong to Ascomycota 56 , which can be used as evidence of result that some microbes belonging to Ascomycota may have the ability to improve the yield of mushrooms (Fig. 6). The finding of this study provided some evidence for the hypothesis that bacteria in Bacteroidetes belong to MGPM, which was in agreement with the literature 32 . However, Alphaproteobacteria was not detected in the literature, which was mainly due to the difference in pH value between the literature and our study. The composition of decomposers in the topsoil of Zhejiang Province was in agreement with the literature.
The fifth one was hormones producers which can secrete bioactive growth regulators 57 . Promicromonospora can produce gibberellins promoting plant growth and development 57 . In this study, Promicromonospora highly enriched in sample C03 was detected in samples A, B, C, and D and not detected in sample E. This result indicates that Promicromonospora may produce hormones promoting mushrooms growth, which was related to the fact that mushrooms from Xinjiang Province were much larger than those from Zhejiang Province and mushrooms from southern Xinjiang was larger than those from northern Xinjiang.
There must be some microbes in the first category belonging to MGPM. It was reported that for A. bisporus cultivation, P. putida was found to be the best MGPM 6 . 1-octen-3-ol produced by conidia of Penicillium paneum can inhibit the germination process 58 , which can be consumed by P. putida. Therefore the yield of mushroom was increased 6 . However, the species of MGPM and the promoting mechanism for A. sinodeliciosus were still unknown, which need further studies. The microbes in the second category can be regarded as typical microbes of A. sinodeliciosus and there may be symbiotic relationships between the typical microbes and A. sinodeliciosus. It was reported that land plants and soil fungi of the phylum Glomeromycota can form arbuscular mycorrhizal (AM) symbiosis 59 . Bacteria was reported to be component of the lichen symbiosis 17 . Whether the symbiotic relationships existed or not still need further studies. Microbial community analyses indicate that interactions between functional microbes and mushrooms have something to do with the differences in texture and morphology among different specimens.
RDA biplots (Fig. 8) were drawn to reveal the relationships between microbial community compositions of samples or microbial groups and environmental variables. Different environmental variables made great influences. It was found that high C/N ratio, DOC concentration, NO 3 − -N concentration, and TN concentration were related to the bacterial communities living in samples C01, C02, and C03. High nutrient concentrations were also related to the bacterial community compositions in the intertidal wetland 60 . However, the bacterial community compositions of samples D01, D02, and D03 also collected from the southern Xinjiang were not related to those high nutrient concentrations, which were related to the high sulfate concentration. This may be due to the high sulfate concentration in the topsoil of sample D, which turned to be main influence factor among the different environmental variables and this may be able to explain the relationship between high NO 3 − -N concentration and the bacterial community compositions of samples C01, C02, and C03, too. Relative abundances of Iamia, Aequorivita, and Pelagibacterium belonging to the functional microbes were associated with high nutrient concentrations, and Algoriphagus and Parapedobacter were associated with high sulfate concentration. High concentrations of sulfate, NO 3 − -N, DOC, TIC, and TN were related to the fungal communities living in samples D01, D02, and D03. AM fungal community compositions were also related to soil NO 3 − -N content 61 . Relative abundance of Alternaria was associated with high concentrations of sulfate, NO 3 − -N, DOC, TIC, and TN, which were negatively correlated with relative abundances of Acremonium and Mortierella. It was interesting that relative abundance of Agaricus was associated with high C/N ratio and NH 4 + -N concentration, which was inconsistent with the previous results that concentrations of DOC and TN and C/N ratio are not the main reason for the differences in texture and morphology among different specimens. This was due to the fact that for RDA analysis, Agaricus was in the form of mycelium not fruiting body. High C/N ratio and NH 4 + -N concentration can enhance the mycelial growth of Agaricus. According to the determination of NO 3 − -N and NH 4 + -N (Supplementary Table 4

Methods
Study site and sampling. Five different mushroom specimens were collected from northern Xinjiang (ZRL20152585 and ZRL20152589), southern Xinjiang (ZRL20152590 and ZRL20152591), and Zhejiang Province (ZRL20151244) of China, respectively. As shown in Supplementary Fig. 4, soil samples of different horizontal (0, 10 cm, and 20 cm) and vertical (0-5 cm, 5-10 cm, and 10-15 cm) directions were symmetrically collected from the native habitats of different specimens. In total, 27 soil samples were collected from native habitats of each specimens and a total of 135 soil samples were obtained, which were stored at −20 °C in a car refrigerator. To identify mushroom species, DNA was extracted form specimen, and PCR reactions and sequencing were performed using primers ITS1F and ITS4B. Phylogeny of five specimens based on ITS sequences was shown in Supplementary  Fig. 5. Specimen ZRL20152585 was clustered with A. padanus, and specimens ZRL20152589, ZRL20152590, and ZRL20152591 were clustered with A. sinodeliciosus. Specimen ZRL20151244 was clustered with A. planipileus. To find out MGPM and typical microbes for A. sinodeliciosus, it is necessary to investigate the microbial communities between A. sinodeliciosus and other species. In this study, soil samples in the native habitats of specimens ZRL20152585 and ZRL20151244 were used as territorial and interspecific control, ZRL20152589 as territorial and intraspecific control. After being transported to laboratory, soil samples were kept at −80 °C before DNA extraction.
Analytical methods. Concentrations of sulfate and NO 3 − -N were quantified according to previous studies 41,42,49 . Dissolved organic carbon (DOC) and total soluble nitrogen (TN) were measured using a TOC/TNb analyzer (Elementar vario TOC, Elementar Co., Germany) 62 . C/N ratio was calculated, dividing DOC by TN. NH 4 + -N was measured by indo phenol blue method 63,64 . DNA extraction. Before DNA extraction, soil samples were treated using a 2 mm sieve to remove stone, plant roots, and tissues 16 . Total DNA was extracted from 0.25 g (wet weight) of soil sample using a PowerSoil DNA kit (MoBio Laboratories, CA, USA) following the protocol of manufacturer 28,41 . At the same time, DNA extraction of 135 soil samples was separately performed, and DNA solution from symmetric locations of each specimen was pooled. Finally, for each specimen, 9 DNA solution was obtained (Supplementary Table 5), which was used for sequencing.

Amplicon sequencing.
To determine the diversity and structure of microbial communities, the protocol as previously described was used 65 . PCR amplifications were performed with different primers. For bacterial communities, the primers were 338 F (5′-ACTCCTACGGGAGGCAGCA-3′) and 806 R (5′-GGACTACHVGGGTWTCTAAT-3′) 66 . For fungal communities, the primers were ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2 (5′-GCTGCGTTCTTCATCGATGC-3′) 67 . The primer contains an error-correcting barcode unique to each sample. To minimize the impact of potential early round PCR errors, twenty independent PCR products of each sample were quantified using a Qubit 2.0 fluorometer (Invitrogen, Carlsbad, CA) and then mixed accordingly to achieve equal concentration in the final mixture, which was used to construct PCR amplicon libraries. Sequencing was performed on an Illumina HiSeq platform. Raw sequencing data obtained from this study were deposited to the NCBI Sequence Read Archive database with accession no. SRP093673. Data analysis. FLASH was used to merge pairs of reads from the original DNA fragments to produce raw tags 68 . To obtain clean tags, raw tags were filtered strictly according to previous study 69 . First, QIIME (Quantitative Insights Into Microbial Ecology) 70 quality filters was used to filter raw tags. Then operational taxonomic units (OTUs) were picked using UPARSE pipeline 71 . Sequences were assigned to OTUs at 97% similarities. Alpha diversity and beta diversity were calculated. Venn diagrams were used to describe the similarity and difference among the same sampling sites of topsoil from different specimens. Hierarchical cluster analysis was performed using gplots package of R 24 , and distance algorithm and clustering method were "euclidean" and SCientifiC REPORTS | 7: 15719 | DOI:10.1038/s41598-017-16082-1 "complete", respectively. For PCoA of bacterial and fungal community, distance algorithm was "binary jaccard". In order to reveal the relationships between microbial community compositions of samples or microbial groups and environmental variables, redundancy analysis (RDA) was performed using CANOCO 72 .