In July 2010, five cold springs were sampled in Wuli Area (Figure 1), Qinghai Province, China, that is adjacent to the Daha coal mine (Zhou, 2004) and located in the Fenghuo Mountain-Wuli gas hydrate zone (Zhu et al., 2011). The Wuli area is located at the elevation of ∼4600 m. Water pH and temperature were measured in the field using a digital soil pH meter (Ferrymorse-Seed Company) and a mercury thermometer, respectively. During sample collection (around noon), the ambient temperature was 15–17°C, whereas the water temperature of the sampled cold springs was around 1–3°C. Sediments from five cold springs (named as QCS1, QCS3, QCS4, QCS5, and QCS6, respectively) were collected into 50 mL sterile Falcon tubes using a sterile spatula. The collected samples were stored at -20°C in the field as well as during transportation and subsequently at -80°C in the laboratory until further analyses.

Cation composition of pore water was analyzed by using inductively coupled plasma-optical emission spectrometry (ICP-OES; Varian Vista MPX, Varian, Palo Alto, CA, USA). Anion composition was analyzed using ionic chromatography (IC) on a Dionex ISC90 equipped with a conductivity detector and an AS14A column (eluent, 10 μM Na₂CO₃/NaHCO₃; flow rate, 1.0 mL/min; Jiang et al., 2010a). The sediment mineralogy was analyzed by using powder X-ray diffraction (XRD) on a Rigaku D/Max 2550/PC X-ray diffractometer with Cu Ka radiation (40 kV; 100 mA; Zhang et al., 2009).

DNA of the sediment samples was extracted using FastDNA^® SPIN Kit for Soil (MP Biomedicals, LLC, Solon, OH, USA) according to the manufacturer’s protocols. The actinobacterial 16S rRNA gene from the extracted DNA samples was amplified using the actinobacterial 16S rRNA gene-specific forward primer S-C-Act-0235-a-S-20 (5′-CGC GGC CTA TCA GCT TGT TG-3′) and reverse primer S-C-Act-0878-a-A-19 (5′-CCG TAC TCC CCA GGC GGG G-3′; Stach et al., 2003) with the same PCR conditions as described previously (Wu et al., 2009). PCR products were purified using Agarose Gel DNA Fragment Recovery Kit Ver. 2.0 (TaKaRa, Dalian, China) according to the manufacturer’s instructions. 16S rRNA gene clone libraries were constructed by ligating the purified PCR products into pGEM^®-T Easy Vector system (Promega, Madison, WI, USA) and transformed into competent Escherichia coli JM109 cells according to the manufacturer’s protocols. Positive clones were randomly picked for sequencing with an ABI 3730 XL DNA Sequencer (Applied BioSystems, Foster City, CA, USA). Rarefaction analysis was performed to evaluate the saturation of the sampled clones using the PAST software package¹ (see Supplementary Figure S1).

All the obtained clone sequences were assembled and edited by using Sequencher v.4.1 (GeneCodes, Ann Arbor, MI, USA) and then checked by BLAST function in NCBI (National Center of Biotechnology Information²). Potential chimeric sequences were removed from further analyses. Operational taxonomic units (OTUs) were identified at a 97% cutoff by using Mothur v1.36.1 with furthest neighbor method (Schloss et al., 2009). One sequence from each OTU was selected and the closest references were picked up from the GenBank database for phylogenetic analyses (see Supplementary Table S1). The representative sequences of OTUs and references were combined and aligned using ClustalW in MEGA (molecular evolutionary genetics analysis) program, version 6.06. Maximum likelihood phylogenetic trees were constructed using the above aligned sequences. Bootstrap replications of 1000 were assessed. The unique clone sequences determined in this study were deposited in the GenBank database under accession numbers JX667788–JX667977, JF712624–JF712648, and KU052203–KU052216.

Alpha-diversity indices, such as Simpson, Shannon, Equitability and Chao 1, were calculated by using the PAST software package (Hammer et al., 2001). Coverage values of the clone libraries were calculated with the equation C = 1-n/N, where n was the number of phylotypes that occurred only once in the clone library and N was the total number of sequenced clones (Jiang et al., 2010b). All obtained environmental variables were normalized (values ranged between 1 and 100) to improve normality and homoscedasticity for statistical analyses. Clustering analysis were performed by using PAST software package with unweighted pair group method with arithmatic mean. Mantel tests were performed to assess the correlation between actinobacterial community composition and environmental variables by using the PAST software package. Briefly, the biotic matrices were constructed on the basis of Bray-Curtis dissimilarity of actinobacterial community compositions. The abiotic matrices were constructed on the basis of the Euclidean distances of normalized environmental variables.

In order to compare the actinobacterial community composition difference between the QTP cold springs and other related habitats, reference actinobacterial clone sequences from Tibetan hot springs (Jiang et al., 2012a), Tibetan (hyper-)saline lakes (Jiang et al., 2010a), freshwater sample of Daotang river (Jiang et al., 2010a), Atlantic ocean deep-sea sediment (Stach et al., 2003), the Three Gorges Dam of the Yangtze River (Jiang et al., 2012b) and Tengchong hot springs (Song et al., 2009) were downloaded from the GenBank database and combined with the ones obtained in this study. In order to avoid any bias resulting from different primers, only actinobacterial 16S rRNA sequences amplified from the same primer set and PCR protocol as this study were included in subsequent analysis. The combined actinobacterial 16S rRNA sequences were aligned using ClustalW in MEGA and then were subjected to OTU identification at the 97% cutoff using Mothur v1.36.1 with furthest neighbor method (Schloss et al., 2009). Clustering analysis was performed to discern the difference of actinobacterial community composition among habitats based on Jaccard similarity using the PAST software package.

The pH of the sampled cold springs were neutral, and the temperature ranged 1.5–2.5° (Table 1). The concentration of Si⁴⁺ and total Fe were 0.6–5.1 and 0.0–6.9 mg/L, respectively. Heavy metals Mn and Sr only occurred in the QCS1 sample. The sediment samples were mainly composed of quartz, plagioclase, calcite, montmorillonite, illite, and kaolinite.

Five clone libraries (QCS1, QCS3, QCS4, QCS5, and QCS6) were constructed. A total of 484 actinobacterial 16S rRNA gene clone sequences were obtained: 117, 85, 76, 103, and 103 clone sequences for QCS1, QCS3, QCS4, QCS5, and QCS6, respectively. The number of sequenced clones represented 76–91% coverage for each clone library (Table 2). Out of these clone sequences, one hundred and twenty OTUs (29, 27, 32, 27, 31 for QCS1, QCS3, QCS4, QCS5, and QCS6, respectively) were identified (Table 2). These identified OTUs could be classified into Acidimicrobiales, Corynebacteriales, Gaiellales, Geodermatophilales, Jiangellales, Kineosporiales, Micromonosporales, Micrococcales, Nakamurellales, Propionibacteriales, Pseudonocardiales, Streptomycetales, and unclassified Actinobacteria (Figure 2). The diversity indices such as Shannon (2.6–3.0), Chao 1 (34.3–46.2) varied among the studied cold springs (Table 2). Acidimicrobiales, Geodermatophilales, Micrococcales, Propionibacteriales, and Pseudonocardiales were dominant actinobacterial groups (Figure 3C). Among the studied samples, Acidimicrobiales, Micrococcales, Pseudonocardiales, and unclassified Actinobacteria were dominant (relative abundance > 10%) in the QCS1 sample; Acidimicrobiales, Micrococcales, Pseudonocardiales, and Propionibacteriales dominated in the QCS3 sample; Acidimicrobiales, Geodermatophilales, Micrococcales, and Propionibacteriales were dominant in the QCS4 and QCS5 samples; and Acidimicrobiales, Corynebacteriales, Kineosporiales, Micrococcales, and Propionibacteriales dominated in the QCS6 sample (Figure 3C).

The order of Micrococcales was the most dominant (average abundance 25.6%) group in the studied cold spring samples, and a large portion of clones affiliated with Micrococcales were closely related (identity: 95–99%) to cultured psychrophilic Actinobacteria, such as Arthrobacter sp. (Reddy et al., 2000; Fong et al., 2001; Wang et al., 2009) and Demequina sp. (Finster et al., 2009; Figure 2 and Supplementary Table S1). Furthermore, many clone sequences obtained in this study were affiliated with Acidimicrobiales, and they were related to clone sequences retrieved from cold habitats such as arctic soil exposed by glacier retreat (Quince et al., 2011), cold spring sediment in Shawan, Xinjiang, China (Zeng et al., 2010), and Shule River permafrost soils on the Tibetan Plateau (Figure 2). The remaining 5.9% (32 out of 484) of the clone sequences retrieved in this study belonged to unclassified Actinobacteria (Figure 2).

Cluster analysis showed that the cold spring geochemistry (Figure 3A) presented similar grouping patterns to actinobacterial community composition (Figure 3B) among the studied samples. Mantel tests showed that actinobacterial community composition of the studied cold springs was significantly correlated (r = 0.748, P = 0.021) with the combined environmental variables but not significantly (P > 0.05) with any single environmental variable measured in this study. Furthermore, cluster analysis showed that the actinobacterial communities in the QTP samples (including clod springs, hot springs and lakes) were grouped into one cluster, which has little similarity (Jaccard similarity < 0.05) with that of marine sediments from Atlantic ocean and Tengchong hot springs (Figure 4).

The actinobacterial community composition in cold springs on the QTP was similar to that of cold habitats in other locations. The actinobacterial communities of the studied QTP cold springs were composed of major groups related to psychrophilic Actinobacteria species (e.g., Arthrobacter psychrochitiniphilus, Demequina lutea) and environmental clone sequences retrieved from cold habitats, such as snow/ice and soils in Qinghai–Tibetan Plateau and Arctic/Antarctic. This indicated that low temperature was a major environmental factor for dominating actinobacterial distribution in cold habitats.

Excluding low-temperature property, actinobacterial community composition in the studied cold springs may be affected by environmental variable composition. For example, samples of QCS3, QCS4, and QCS5 had similar environmental variables composition, and thus possessed similar actinobacterial community compositions (Figures 3A,B); the environmental variable composition of QCS1 and QCS6 was different from the other studied samples (Figure 3A): QCS1 possess highest concentration of Na⁺ and heavy metal Mn and Sr (Table 1), and QCS6 sample has highest Ca²⁺ and total Fe (Table 1), thus it is reasonable to observe distinct actinobacterial community compositions in QCS1 and QCS6 samples from that in QCS3, QCS4, and QCS5 samples (Figure 3B). Previous studies have shown that microbial community composition could be affected by multiple environmental parameters, such as salinity (Lozupone and Knight, 2007), temperature (Lindh et al., 2013), and heavy metals (Gong et al., 2015). Therefore, it is not surprising to observe significant correlation between actinobacterial community composition and environmental variables in the studied cold springs.

It is notable that some of the retrieved actinobacterial clone sequences from the cold springs showed high identity with those obtained from petroleum- or coal-related environments. This observation is expected in that the sampling sites in this study was located in the Wuli-Daha coal-bearing belt (Zhou, 2004) and Fenghuo Mountain-Wuli gas hydrate-bearing belt (Zhu et al., 2010) in southern Qinghai Province. The underlying coal or gas hydrate might provide abundant nutrients, which support diverse actinobacterial communities in the studied cold springs (Santos et al., 2008; Jiang et al., 2010a).

The actinobacterial community in the investigated cold springs was more diverse than other cold environments. For example, the Actinobacteria sequences obtained in this study were distributed into 12 orders (Figures 2 and 3C). In contrast, the Actinobacteria-related clones retrieved in the snow of four glaciers on the Tibetan Plateau were mainly affiliated with the order Micrococcales and unclassified Actinobacteria (Liu et al., 2009b). This suggested Tibetan cold springs might contain more suitable growth conditions for Actinobacteria than glaciers.

Actinobacterial communities from different habitats possessed certain geographic characteristics. The actinobacterial clones from the studied cold springs (this study) were closely related to those from the QTP hot springs and saline lakes (Figure 4), this indicated that the actinobacterial communities in the studied cold springs were more similar to that in other QTP samples (including hot springs and lakes) than to those in the samples from other locations. For example, the majority of the retrieved actinobacterial 16S rRNA gene clone sequences in the investigated cold springs were affiliated with Micrococcales, Propionibacteriales, and Acidimicrobiales. Actinobacterial clones retrieved from Tibetan saline lakes were mainly classified with Micrococcales, Propionibacteriales, and Frankiales (Jiang et al., 2010a). In contrast, the actinobacterial communities in Tengchong hot springs were mainly affiliated with unclassified Actinobacteria, Rubrobacterales, and Frankiales (Song et al., 2009). Previous studies have shown that Actinobacteria in hot springs, soils and oceans possess geographic distributions (Ward and Bora, 2006; Wawrik et al., 2007; Valverde et al., 2012). In addition, the Actinobacteria communities in the studied QTP cold spring sediments were different from those in marine sediments (Stach et al., 2003; Goodfellow et al., 2012) and freshwater ecosystems (Hahn et al., 2003; Warnecke et al., 2004; Ghai et al., 2012). The observed geographic distribution of Actinobacteria in the QTP samples could be ascribed to the distinct conditions (e.g., dry climate, low pressure, high intensity of UV radiation) of the cold springs, hot springs, and saline lakes on the QTP from other ecosystems (Jiang et al., 2010a, 2012a). However, the underlying reasons still await further investigation.

In summary, the actinobacterial communities in the studied Tibetan cold springs possessed unique compositional characteristics and were mainly consisted of Acidimicrobiales, Corynebacteriales, Gaiellales, Geodermatophilales, Jiangellales, Kineosporiales, Micromonosporales, Micrococcales, Nakamurellales, Propionibacteriales, Pseudonocardiales, Streptomycetales, and unclassified Actinobacteria. Biogeographical isolation and unique environmental conditions might be predominant factors affecting the observed similarities and differences in the actinobacterial communities between the investigated cold springs and other habitats.