Influence of CO₂/HCO₃^- on Microbial Communities in Two Karst Caves with High CO₂

0. INTRODUCTION

The increase in the atmospheric CO₂ concentration has an impact on the carbon cycle of terrestrial ecosystems (Paerl et al., 2016; Willis and MacDonald, 2011; Ebersberger et al., 2003). As the main life form of terrestrial ecosystem, microbes are involved in element cycling of ecosystems, which produces feedbacks to environment and climate. Therefore, studying microbial community compositions and functions is of utmost importance to understanding of the biogeochemical cycling of carbon in ecosystems (Fuhrman, 2009; Rothschild and Mancinelli, 2001). Previous studies have showed the influence of CO₂ concentration on microorganisms in terrestrial ecosystems. For example, elevated CO₂ concentration will change the structure and functions composition of soil microbial community (Yang et al., 2019). CO₂ concentration enhancement can lead to decrease of Gram-positive bacteria in grass soil (Fu et al., 2019) and to decrease of the content of phospholipid-derived fatty acids in bacteria, fungi, and Gram-negative bacteria in forest soil (Deng et al., 2016). However, the atmospheric CO₂ concentration mainly has an indirect effect on soil bacteria via plant root system in grassland and woodland (Wang et al., 2017). Caves are one of the most important terrestrial ecosystems, in which CO₂ can directly affect microbial growth and ecological functions. Therefore, the study of microbial diversity and function and their response to environmental factors (particularly, different CO₂ concentrations) in caves can provide a data basis for assessing the impact of CO₂ concentration on terrestrial ecosystems and for improving our understanding of interactions between microorganisms and environment.

The karst area accounts for 15% of the global ice-free land area (Palmer, 2017). Karst caves are formed mainly by geological processes such as dissolution, physical weathering and volcanic activity. Karst caves are dark, lacking organic matter from photosynthesis, and geographically relatively isolated. Therefore, they are considered as terrestrial oligotrophic environments under extreme conditions (Chen et al., 2016; Deininger et al., 2016; Wong and Breecker, 2015; Zhao et al., 2012). However, caves host high diverse microbes (Pedersen, 2000). In the meanwhile, the cave entrances mostly directly connected to the outside and is more easily accessible than other subsurface environment (Barton et al., 2007), thus karst caves are an ideal sites to study microbial diversity and function of the subsurface biosphere (Barton et al., 2004).

In the past, a lot of researches have been performed on cave microbial community. For example, approximately 350 Actinomycetes species were retrieved from rock walls in the Altamira Cave, northern Spain (Groth et al., 1999). Fifty-one different bacterial and archaeal species were discovered in the Wind Cave, the United States (Chelius et al., 2009). Cave microbial communities showed a great diversity and many ecological functions, among which bacteria were mainly involved in biogeochemical cycle in cave environments. For example, bacterial genes were involved in carbon degradation, carbon fixation, methane metabolism, nitrification, nitrate reduction and ammonia assimilation in five caves in Mizoram State, northeastern India (de Mandal et al., 2017). The functional genes involved in sulfur oxidation and CO₂ fixation could be detected in Acidithiobacillus thiooxidans, the dominant species in the bacterial biofilm of the sulfide-rich Frasassi Cave, Italy, indicating bacterial involvement in carbon cycle in this cave (Bauermeister et al., 2012). The metabolic functions of the bacterial community in drip water of the Buddhist Monk Cave in Hubei, China, were mainly affected by temperature and pH, and carbon-fixing bacteria increased with increasing CO₂ concentration in microcosm experiments (Yun et al., 2018). Therefore, the CO₂ concentration is an important factor affecting the bacterial community and function in caves. However, the effect of high CO₂ concentration on bacterial community remains elusive. It is even more difficult to compare the impact of CO₂ concentration with other environmental factors on the bacterial community, which directly limits understanding of the carbon cycle in caves.

The concentration of CO₂ in karst caves is high and can reach several thousand ppm due to dripping and degassing (Baldini et al., 2008). Therefore, karst caves are a natural laboratory for studying microbial response to elevated atmospheric CO₂ concentrations. The CO₂ concentration in Xueyu Cave (XYC) in southwestern China is as high as 8 751 ppm. Previous studies showed that variations of CO₂ concentration inside the XYC were derived from groundwater degassing. Rainfall is an important factor that controls the partial pressure of CO₂ in the overlying soil, and gas exchange because temperature changes is an important reason for sudden changes in the partial pressure of atmospheric CO₂ in the cave (Lü et al., 2019). Lü et al. (2019) found that the microbial diversity in the XYC exhibited temporal and spatial variations and bacteria, fungi and Actinobacteria show different changes, and that natural conditions (such as precipitation) had an important effect on the microorganisms in the groundwater of the cave. However, it is unclear how CO₂ affects the microbial community. This also limits our understanding of element cycle in karst caves with various CO₂ concentrations. Therefore, further research is urgently needed in this field. In order to fill the above knowledge gaps, XYC and Shuiming Cave (SMC, an undeveloped cave near the Xueyu Cave, was a simple, stable and relatively closed cave system with CO₂ concentration of 3 303 ppm.) were selected as research objects in the present study. The purposes of this study were to explore microbial community structure and its relationship to CO₂ concentration and other related environmental factors, and to predict microbial functions at different locations (cave rock walls, sediments and groundwater) and their response to high CO₂.

1. METHODS, TECHNIQUES, STUDIED MATERIAL, AND AREA DESCRIPTION

1.1 Site Description

The XYC (29º47′00″N, 107º47′13″E) is located in downstream of the Longhe River, a tributary on the South Bank of the Yangtze River, Fengdu County, Chongqing (Fig. 1). The temperature is 16–18 ºC, and the humidity is always more than 95% inside the XYC. The XYC area is dominated by subtropical humid monsoon climate and is affected by the southwestern Indian Monsoon and southeastern Asian Monsoon. Precipitation is mainly concentrated from April to October every year, and the average annual precipitation is 1 072 mm. The vegetation mainly consists of evergreen broad-leaved forests and thickets. The XYC is 1 643.97 m long, and can be divided into three layers: upper, middle, and bottom. Underground rivers exist all year round, with an accessible length of about 1.2 km. The entrance of the XYC is 55.5 m higher than the level of the Long River, and the elevation is 233 m (Chen et al., 2020). The CO₂ concentration of the XYC (8 751 ppm) is similar to that in the atmosphere of the primitive Earth.

Figure  1.  A geographic map showing the locations of the studied caves in Chongqing Municipality, SW China.

DownLoad: Full-Size Img PowerPoint

The SMC (29º47′00″N, 107º47′00″E) is not undeveloped. CO₂ concentration is 3 303 ppm inside the SMC. The SMC entrance is ten meters higher than the Longhe River water surface. Its underground river water flow is greater than that in the XYC, and there is a spring accumulation platform at the entrance. Near the entrance is a larger cavern, where groundwater converges into a pool and eventually flows into the Longhe River. The microbial ecosystem is simple and less disturbed by human activities.

1.2 Field Measurement and Samples Collection

The XYC (29º47′00″N, 107º47′13″E) is located in downstream of the Longhe River, a tributary on the south bank of the Yangtze River, Fengdu.

Sampling cruise was performed in October 2018. The water temperature (T), dissolved oxygen (DO), and pH value were measured using the ODEON multi-parameter water quality monitor produced by Ponsel, France. The accuracy was 0.1 ℃, 0.01 mg·L^-1, and 0.01 pH unit, respectively. For CO₂ and CH₄ concentration measurements, gas samples at the entrance and in the interior of the XYC and SMC were collected into gas bags. For measuring the contents of anions and cations, water samples were collected into 60-mL polyethylene plastic bottles and were stored in a dark box at 4 ℃. Inside the two caves, sediments and rock wall surface soils were collected into sterile bags for geochemistry measurement and microbial analyses, and they were marked as SMCSediment, SMCWall, XYCSediment, and XYCWall respectively. Water samples inside the caves ("IWater") or at the entrance of the caves ("OWater") were collected with the use of a glass suction filter fixed with one filter membrane (pore size 0.22 μm), and they were marked as SMCIWater, SMCOWater, XYCIWater and XYCOWater, respectively. The resulting biomass-contained filters were collected into sterilized centrifuge tube. Both sediments/soils and biomass-contained filters for microbial analyses were stored on dry ice in the field and during transportation. The samples were transported to the laboratory and stored at -80 ℃ until analyzed.

1.3 Geochemical Analysis

In laboratory, anions (HCO₃^- and NO₃^-) were determined by an ion chromatograph (861 Advanced Compact IC Metrohm, Switzerland), and cations (Ca²⁺ and Mg²⁺) were measured using an ICP-OES spectrometer (IRIS Intrepid II XSP, Thermo Fisher Scientific, USA). The accuracy of both methods was 0.01 mg·L^-1. Total organic carbon (TOC) was measured by the Multi-N/C3100 carbon and nitrogen analyzer (Jena Company, German). CO₂ (ppm) and CH₄ (ppm) contents were measured using gas chromatography (GC) Agilent 8890. Sulfuric acid accelerator digestion and Kjeldahl method were used to measure the soil total nitrogen (TN) content (Bremner, 1960). Nitric acid-hydrogen peroxide-hydrofluoric acid digestion was used to measure the content of total phosphorus (TP) and total calcium in soil by ICP-AES (Tian and Shao, 2013).

1.4 DNA Extraction

Genomic DNA was extracted using the Fast DNA SPIN kit (MP Biomedical, Solon, OH, USA). 16S rRNA gene was amplified from the extracted DNA using a universal primer set of 515F (5'-GTGYCAGCMGCCGCGGTA-3') / 806R (5'-GGACTACVSGGGTATCTAAT-3'). The DNA gel extraction kit from Axygen (Union City, CA, USA) was used to purify the successful PCR products. Sequencing was performed in an Illumina MiSeq platform at Guangdong Magigene Biotechnology Co. Ltd., Guangzhou, China.

1.5 Bioinformatics and Statistical Analysis

Data analysis followed the UPARSE process (Edgar, 2013). FLASH was used to stitch and trim paired samples (Magoč and Salzberg, 2011). The forward and reverse primer sequences were removed, and QIIME v1.8.0 was then used for mass filtering (Caporaso et al., 2012). Sequence reads were removed when theycontained (1) three consecutive low-quality (Phred quality scores < 30) bases, (2) ambiguous bases, and (3) less than 75% of continuous read length. Uchime module and de novo method in USEARCH1 were used for chimera inspection (Edgar et al., 2011). Qualified reads were clustered using USEARCH. All eligible fragments were truncated to 235 nt, and all singletons and fragments shorter than 235 nt were removed. The UPARSE-OTU algorithm (Edgar, 2013) was used to classify the operating classification unit (OTU) according to 97% cutoff, and then selected the representative OTU sequences to compare in silvav132 database. Alpha diversity (i.e., Simpson, Shannon, Equitability, and Observed OTU) was calculated through QIIME v1.8.0.

SPSS 25 software (SPSS Inc., Chicago, IL, USA) was used to analyze microbial difference among samples (Louca et al., 2016). The number of OTUs exceeding 0.5% of the total OTUs was defined as the dominant OTUs, and a heat map of the dominant OTUs was drawn (Luo, 2009). In the meanwhile, Pearson correlation was performed to analyze the correlation between geochemical parameters and dominant OTUs. The weighted UniFrac distance algorithm of Bray-curtis and PCoA were used to analyze the similarity of OTUs among samples (Team, 2009). In order to predict the potential microbial ecological function in the cave, the obtained OTUs were compared with the FAPROTAX 1.1 database (Anderson and Willis, 2003).

1.6 Carbon Utilization Test on Selected Bacterial Isolates

Biolog ECO plates (Biolog, America) containing 31 different carbon sources were used to test the carbon source utilization of the selected strains retrieved from the wall and sediment samples (Chen et al., 2020). For each of the selected strains, 150 μL of bacterial culture was inoculated per well of the ECO plate, followed by incubation at 28 ºC. During cultivation, readings were taken at a wavelength of 590 nm at regular intervals (24 h) using a microplate reader. The average of the optical density (OD) values of 3 replicates of each substrate minus the OD value of the negative control was taken and used for data analysis. The average color change rate (AWCD) was considered as an indicator for evaluation of the carbon source utilization capacity of each tested strain (Jiang et al., 2006). The R language was used to draw a heat map based on the color change of the carbon source to show the carbon source utilization preferences and the capabilities of the tested strains.

1.7 Availability of Data

The sequences retrieved in this study were deposited in the National Center for Biotechnology Information under the BioProject accession numbers SRR12328005-SRR12328012 and GenBank accession numbers of MT378306-MT3783010.

2. RESULT

2.1 Geochemical Characteristics of the Studied Caves

The geochemical characteristics of the samples were shown in Table S1. The HCO₃^- concentrations of inside XYC, inside SMC, XYC entrance, and SMC entrance were in a descending order. The CO₂ concentration was the highest of inside XYC and the lowest at the XYC entrance. The CH₄ content in the SMC was significantly lower than that in the XYC. The TOC content of the solid sample was lower and the total P content and total Ca content were higher in the XYC than that in the other cave. The XYC water samples had higher pH, Ca²⁺, NO₃^- and HCO₃^-, but lower Mg²⁺ content than the SMC water samples. Almost no difference in temperature was observed between the XYC and SMC. And, there was less difference of DO in the underground river water.

2.2 Microbial Community Structures in the Studied Caves

The microbial distribution differed at the phylum level among samples (Fig. 2). The dominant phyla in the studied water samples inside the two caves and at the entrance of the two caves were Proteobacteria, Actinobacteria and Bacteroidetes (Fig. 2a). The proportions of the three dominant bacteria in water samples were 96.41% (relative frequency, hereinafter unless stated otherwise) inside the SMC, 96.76% at the entrance of SMC, 95.14% inside the XYC and 93.44% at the entrance of XYC. Bacteroidetes accounted for a higher proportion and Actinobacteria accounted for a lower proportion in the studied water samples at the entrance of XYC than that from the three other sampling sites. In the sediment samples, the dominant phyla of SMC were Proteobacteria (23.24%), Chloroflexi (16.90%) and Actinobacteria (11.23%); and the dominant phyla of XYC were Proteobacteria (33.34%), Actinobacteria (15.37%) and Bacteroidetes (13.16%). In the studied rock wall samples (Fig. 2b), the dominant phyla of SMC were Thaumarchaeota (29.24%), Proteobacteria (22.85%), Actinobacteria (8.80 %), Acidobacteria (6.70%) and Rokubacteria (6.15%); and the dominant phyla of XYC were Proteobacteria (24.77%), Actinobacteria (20.96%), Bacteroidetes (13.16%), Thaumarchaeota (7.05%) and Cyanobacteria (10.74%).

Figure  2.  Comparison of the average quantitative contribution of the sequences affiliated with different microbial phyla from the water body (a) and the cave wall (b). In each sample, the microbial phyla with the relative frequency less than 1% were included as others.

DownLoad: Full-Size Img PowerPoint

The heatmap showed the distribution of dominant OTUs among samples (Fig. S1). Among the water samples, the water sample inside SMC was dominated by OTUs affiliated with Acinetobacter, Aeromonas, Micrococcaceae, and Micrococcaceae (2 396, 2 444, 996, 2 652, and 1 054, respectively) (Fig. S1a); the water sample at theentrance of SMC was dominated by OTUs affiliated with Glutamicibacter, Micrococcaceae, Enterobacteriaceae, Vogesella, and Micrococcaceae (336, 2 067, 3 167, 809, and 996, respectively); the water sample inside XYC was dominated by OTUs affiliated with Massilia, Acinetobacter lwoffii, and Comamonas (647, 2 942, and 2 897, respectively); the water sample at the entrance of XYC was dominated by OTUs affiliated with Acinetobacter, Flavobacterium, and Thermomonas (705, 3 109, and 2 742, respectively) (Fig. S1a). The sediment inside SMC was dominated by OTUs affiliated with Nitrosomonadaceae and Flavobacterium (1 000 and 20, respectively) (Fig. S1b); the sediment inside XYC was dominated by OTUs affiliated with Acinetobacter, Flavobacterium, Nitrospira, and Gaiella (955, 20, 1 248, and 2 146, respectively). The SMC rock wall sample was dominated by OTUs affiliated with Nitrososphaeraceae, Micrococcaceae, and Acinetobacter (1 640, 2 652, 198 and 705 respectively); the XYC rock wall sample was dominated by OTUs affiliated with Micrococcaceae and Psychroflexus (2 652 and 1 087, respectively) (Fig. S1b).

The microbial alpha diversity varied among the studied samples (Table 1). The observed OTUs, Shannon, Simpson, and Equitability ranged 450–976, 3.191–5.445, 0.861–0.989, and 0.519–0.817, respectively (Table 1). The microbial diversity of the water samples was higher at the entrance of the two caves than that inside the caves. However, the microbial diversity of the water samples inside XYC was higher than that inside SMC. The microbial diversity of the sediment samples did not show significant difference between the two caves, while the microbial diversity of the rock wall samples was more diverse in XYC than that in SMC.

Table  1.  Alpha diversity of microbial communities in each of the studied samples

Sample ID	Observed_OTUs	Shannon	Simpson	Equitability
SMCIWater	470	3.191	0.861	0.519
SMCOWater	450	3.599	0.920	0.589
XYCIWater	570	3.648	0.914	0.575
XYCOWater	602	3.769	0.869	0.589
SMCSediment	787	5.445	0.989	0.817
XYCSediment	976	5.443	0.980	0.791
SMCWall	787	5.378	0.978	0.806
XYCWall	721	5.144	0.975	0.782

 | Show Table

DownLoad: CSV

PCoA was plotted to show the effect of HCO₃^- concentration on microbial communities of water samples and of CO₂ on microbial communities of rock wall samples (Fig. 3). The water samples inside the two caves and at the entrance of the SMC were grouped together, indicating that their microbial community structures were similar. The water samples at the entrance were separated from the sediments and the rock wall samples, and microbial difference between the sediment and rock wall samples of one cave was smaller than that between caves, suggesting that the HCO₃^-/CO₂ concentration had greater influence on the microbial communities in the studied cave sediment and rock wall samples.

Figure  3.  PCoA plots showing microbial difference among the studied samples ("IWater" represents the water samples from inside the studied caves. "OWater" stands for water samples at the entrance of the studied cave. P1 and P2 represent the interpretation degree of x-axis and y-axis, respectively).

DownLoad: Full-Size Img PowerPoint

2.3 The Influence of Geochemical Variables on Cave Bacterial Community

Most dominant OTUs in the studied water and sediment samples were significantly correlated with T, DO, Mg²⁺, NO₃^-, pH, HCO₃^-, and Ca²⁺ (Fig. 4a). Most dominant OTUs of the rock wall samples were significantly correlated with TOC, N, CO₂, P and Ca (Fig. 4b). In the water and sediment samples, OTUs belonging to Micrococcaceae and Aeromonas (2 652 and 2 444, respectively) were significantly correlated with NO₃^- and Mg²⁺. OTUs belonging to Micrococcaceae and Enterobacteriaceae (2 067 and 3 167, respectively) were significantly correlated with DO. OTU955 belonging to Acinetobacter was significantly correlated with pH, temperature and HCO₃^-. OTUs belonging to Flavobacterium, Nitrospira, and Gaiella (20, 1 248, and 2 146, respectively) were significantly correlated with pH, temperature, DO, HCO₃^- and Ca²⁺. OTU647 affiliated with Massilia was significantly correlated with NO₃^-, pH, T, DO, Ca²⁺ and Mg²⁺. OTU1559 affiliated with Pseudomonas was significantly correlated with pH and T. OTU996 affiliated with Micrococcaceae was significantly correlated with NO₃^-, T, DO, HCO₃^-, Ca²⁺ and Mg²⁺. OTU336 affiliated with Glutamicibacter was significantly correlated with DO, HCO₃^- and Ca²⁺. OTU809 affiliated with Vogesella was significantly correlated with NO₃^-, DO and Mg²⁺. In the rock wall samples, OTUs belonging to Psychroflexus, Thermosynechococcus, Hydrogenobacter, Beijerinckiaceae, Chloroplast, Microbacteriaceae, Ralstonia, Synechococcus and Sulfurimonas (1 087, 2 437, 648, 1 544, 1 573, 1 175, 763 305, and 2 817, respectively) were significantly positively correlated with CO₂ concentration, P and Ca, and significantly negatively correlated with TOC and N (Fig. 4b). While an opposite trend was observed for the other OTUs except those affiliated with Micrococcaceae, Acinetobacter, Paeniglutamicibacter, Pirellulaceae Mitochondria, Woesearchaeia, Micrococcaceae and Woesearchaeia (2 652, 2 396, 1 054, 533, 1 367, 2 965, 2 067 and 629, respectively).

Figure  4.  Heatmap showing the relationship between the physicochemical parameters and the most abundant OTUs (> 0.5%) in water body samples (a) and rock wall samples (b).

DownLoad: Full-Size Img PowerPoint

2.4 The Potential Microbial Ecological Functions Identified by Prediction of the Studied Cave Samples

There were different microbial ecological functions in the studied samples of different types. Among the studied samples, a total of twenty microbial functions were identified as dominant (relative abundance > 1%), and they were aerobic ammonia oxidation, aerobic nitrite oxidation, nitrification, induction of sulfur compounds, knallgas bacteria, dark hydrogen oxidation, dark sulfide oxidation, dark thiosulfate oxidation, dark oxidation of sulfur compounds, fermentation, aerobic chemoheterotrophy, animal parasites or symbionts, aromatic compound degradation, nitrate respiration, nitrate reduction, chloroplasts, cyanobacteria, oxygenic photoautotrophy, photoautotrophy, phototrophy and chemoheterotrophy. In the water and sediment samples shared six common microbial metabolic functions: aerobic chemoheterotrophy, animal parasites or symbionts, aromatic compound degradation, nitrate reduction, and chemoheterotrophy. The sediment samples shared aerobic nitrite oxidation and nitrification. Aerobic ammonia oxidation was found in the rock wall sample of SMC. These abovementioned 21 microbial functions (except aerobic nitrite oxidation) were found in the rock wall of XYC.

Among them, there were six different microbial functions with evident differences in various points, thus methanogenesis, methylotrophy, nitrification, aromatic compound degradation, photoautotrophy, and chemoheterotrophy were selected from the studied samples (Fig. 5). It was found that methanogenesis and methylotrophy functions were exclusively co-existing in the sediments of SMC. The nitrification function was present in the rock wall samples of SMC and XYC with low abundance. Aromatic compound degradation function was present in all the studied samples regardless of sample types, with being more abundant in the water samples than that in the sediment samples. Chemoheterotrophy function was present in all the studied samples, with being more abundant in the water samples of the two caves and in the sediment samples of XYC than the other samples. Photoautotrophy function was more abundant in the rock wall of XYC and less abundant in the sediment of SMC (about 0.1%–0.15%) than that in the other samples.

Figure  5.  Relative abundance of six microbial metabolic functions at different sampling sites (×10⁵).

DownLoad: Full-Size Img PowerPoint

These six different microbial functions had significant correlations with the measured geochemical parameters (Figs. S2 and S3). Specifically, the methanogenesis and methylotrophy functions were positively correlated with TOC and N, and was negatively correlated with CO₂, P and Ca. Nitrification function was positively correlated with HCO₃^-, pH, Ca²⁺, TOC and N, and was negatively correlated with T, DO, CO₂, P, and Ca. The function of aromatic compound degradation was positively correlated with TOC, N, and was negatively correlated with CO₂, P and Ca. Photoautotrophy was positively correlated with CO₂, P, and Ca, and was negatively correlated with TOC, and N. Chemoheterotrophy was positively correlated with NO₃^- and was negatively correlated with Mg²⁺.

2.5 Carbon Source Utilization of Test Bacterial Strains from the Studied Caves

In one of our previous studies, a total of 244 bacterial strains were obtained from the same batch of studied samples (Chen et al., 2020). In this study, five strains (representing different species present in different types of samples and in samples of different caves, as shown in Table 2) were selected for BIOLOG test for their carbon source utilization. These bacterial strains showed different carbon source utilization preferences (Fig. S4). Among them, Gordonia sp. preferred esters and alcohol carbon sources. Bacillus amyloliquefaciens preferred amino acids, amines and carbohydrate carbon sources, and Acinetobacter junii preferred carbohydrates, esters and amino acids. Pseudomonas sp. prefers amino acids and ester carbon sources. Lysobacter sp. preferred ester and acid carbon sources. Overall, strain XYC_W_90 from the cave samples with higher CO₂/HCO₃^- concentration exhibited lower AWCD and lower number of utilized carbon sources (S) than the other strains (e.g., SMC_S_097, SMC_S_234, SMC_W_131, SMC_S_312) (Fig. 6). Strain SMC_W_131 retrieved from the rock wall samples of SMC had the highest AWCD and S values.

Table  2.  The isolates selected for the Biolog carbon utilization test

ID of Isolates	Isolation source	The closest reference strains (GenBank Accession no.)	Identity
SMC_S_097	More abundant in SMC and less abundant in XYC	Acinetobacter junii (MK053916)	100%
SMC_S_234	More abundant in SMC and less abundant in XYC	Pseudomonas sp. (HE614849)	99%
SMC_W_131	Present in the sediment (SMC and XYC) and rock wall (SMC)	Lysobacter sp. (JX415443)	99%
SMC_S_312	Present in the sediment (SMC and XYC)	Gordonia sp. (MH101280)	100%
XYC_W_90	Present in the rock wall (SMC and XYC)	Bacillus amyloliquefaciens (MK336717)	100%

 | Show Table

DownLoad: CSV

Figure  6.  Time-course variations of Biolog test results of the selected representative strains retrieved from the studied samples. (a) and (b) panels were for AWCD and S values, respectively.

DownLoad: Full-Size Img PowerPoint

3. DISCUSSION

3.1 Microbial Diversity in the Studied Caves

Microbial community in the water samples at the entrance was more diverse than the other samples inside the SMC and XYC (Table 1), suggesting that high CO₂ may depress microbial diversity in caves. This finding was consistent with previous studies that CO₂ decreases bacterial diversity in the western Pacific Ocean (Xie et al., 2016). Microbial community in the XYC water samples (pH 8.75) was more diverse than that in the SMC (pH 7.63) (Table S1), indicating that high pH may enhance microbial diversity in caves. This finding was also consistent with previous studies on microbial diversity in cave drip water of Monk Cave (Yun et al., 2018). Therefore, it is reasonable to speculate that pH may be one main factor affecting the microbial diversity of the water samples from the two caves. The microbial diversity in the XYC rock wall samples was lower than that in SMC, suggesting CO₂ concentration may be important in shaping microbial community in rock wall samples in caves considering that the CO₂ concentration in XYC was about three times that of XYC (8 751 ppm vs 3 303 ppm). This finding was in agreement with previous results in soil and seawater (Zhao et al., 2014). Taken together, high CO₂ concentrations can inhibit microbial diversity in the rock wall of caves.

In addition, the community composition in the sediment and rock wall samples was quite different between the two caves. The proportion of Thaumarchaeota was higher than that of Proteobacteria in the SMC rock wall samples, while the trend was opposite in the XYC. Previous studies had shown that the soil microbial community structure changes significantly with increasing CO₂ concentration. For example, the abundance of Bacteroidetes and Proteobacteria (Könneke et al., 2005) increased with increasing CO₂ concentration. We speculated that the higher proportion of Proteobacteria than that of Thaumarchaeota in XYC was caused by the high concentration of CO₂/HCO₃^-. Accordingly, a high concentration of CO₂/HCO₃^- can affect the microbial community structure of karst cave walls and sediments.

3.2 Influence of Environmental Factors on Microbial Communities in the Studied Caves

The dominant OTUs in the cave samples showed different correlations with environmental factors. For example, the above mentioned Thaumarchaeota was dominant in the rock wall samples of the two caves with a high concentration of CO₂. Previous studies showed that Thaumarchaeota was independent of photosynthesis but only relying on ammonia oxidation in the salt water of the cave to generate energy, and thus thrived in dark caves (Pearson et al., 2001). Thaumarchaeota could obtain energy supplement by fixing CO₂ (Zhang and He, 2012; Wuchter et al., 2003). Therefore it is reasonable to observe the high abundance of Thaumarchaeota in the rock wall samples.

In addition, high CO₂ can inhibit the growth of most bacteria. For example, in the cave wall samples, 23.9% of the dominant OTUs were positively correlated with the CO₂ concentration, and 76% of the dominant OTUs were negatively correlated with CO₂. Microorganisms that tolerate high CO₂ or even prefer CO₂ were often mixotrophic such as Rhodoferax and Chloroflexi. The former had chlorophyll-a and carotene, and the latter contained green pigments, but their photosynthesis did not produce oxygen (Garrity et al., 2001; Hiraishi et al., 1991). The autotrophic growth of mixotrophic bacteria relied on nutrients in caves. For example, Hyphomicrobium, the closest relative of OTU 557, can exhibit aerobic growth by using one-carbon compounds such as methanol and methylamine as the only carbon source and energy source, and anaerobic growth with the use of nitrate (Zhang et al., 2012). Hyphomicrobium can use CO₂ instead of methane (Jeong and Kim, 2019). The presence of Hyphomicrobium-related sequences suggested that cave bacteria were involved in carbon cycle. So it is reasonable to observe more Hyphomicrobium-related sequences inside the cave with higher CO₂.

It is notable that high CO₂/HCO₃^- concentration inhibited microbial diversity and abundance in the studied caves. The reasons may be explained by the effect of CO₂ on bacterial metabolism (Chen et al., 2020). Moreover, high CO₂ reduced pH in the environment, resulting in affecting microbial growth (Pierce and Sjögersten, 2009). Besides that, high CO₂ affected the enzyme activity of soil (Zhang et al., 2016). Therefore, it is reasonable to observe the effect of high CO₂/HCO₃^- concentration on the cave microbial community.

3.3 Microbial Ecological Function Potential and Utilization of Carbon Source of Characteristic Bacteria

Potential microbial metabolic functions differed among the studied samples of different types. Among the identified microbial functions, degradation of aromatic compounds and chemoheterotrophy were found in each of the studied samples. However, the abundance of the aromatic compound degradation was the lowest in the XYC rock wall. This may be due to the fact that most bacteria mediating aromatic compound degradation are capable of growing on organic matter instead of inorganic carbon (He et al., 2016). XYC had the lowest TOC and the highest CO₂ concentration (Table S1), which suppresses the abundance of bacteria involving aromatic compound degradation. It is notable that bacteria with chemoheterotrophy potential were abnormally higher in XYC than that in SMC, considering that this type of bacteria require organic matter for growth and that the TOC content of XYC was low. This could be explained by the presence of a large quantity of bacteria with photoautotrophy potential in XYC, which may be caused by the presence of light in XYC. Autotrophic bacteria could produce enough organic matter for the growth of bacteria of chemoheterotrophy potential. In contrast, no visible light was present in the SMC, so that the sediment samples contained much less (about 0.1%–0.15%) bacteria with photoautotrophy function. For example, there was weak-band biokarstification near the entrances of dark cave in Guangxi, Guizhou and other places in China (Wang et al., 1998); eleven typical bryophyte communities were discovered among three karst caves in the Rhone-Alps region of France, and diverse bryophytes (8 families 12 genera 15 species) were widely distributed in caves in Europe and worldwide; in addition, two kinds of fluorescent bryophytes (Cyathodium cavernarum and C. smaragdium) were discovered in one karst cave of Guangxi, China, and green light was found to be emitted in dark caves from these two variants (Zhang et al., 2004). C. cavernarum was found at the entrance of the cave, while C. smaragdium grew 0.5 to 22 m away from the entrance inside the cave. Therefore, we assume that the presence of bacteria with photoautotrophy function in the SMC sediment may be caused by fluorescence and/or other invisible light.

Furthermore, it is remarkable that sequences related to Sphingomonas sp., aerobic anoxygenic phototrophic bacteria (AAPB), were in the two caves. AAPB are a type of heterotrophs that can supplement organic carbon deficiency through CO₂ fixation. So more organic carbon might have been detained because of AAPB, which extends the period of karst carbon cycle (Li et al., 2017) Therefore, AAPB-like bacteria might play an important role influencing element cycle in the studied caves. In addition, the high abundance of Cyanobacteria in XYC is reasonable because a proper amount of CO₂ can promote cyanobacterial growth and photosynthetic rate (Gao and Yu, 2000).

In addition, the strains from XYC had a lower carbon source utilization rate than different carbon source utilization preference from SMC (Figs. 6 and S4). Previous studies had found that microorganisms with different phylogenetic categories showed different preferences for carbon substrates (Goldfarb et al., 2011; Vanfossen et al., 2009; Kramer and Gleixner, 2008). For example, addition of sucrose can promote the growth of Actinobacteria, β-Proteobacteria (Burkholderiales and Rhodocyclales) and some other Proteobacteria. Addition of glycine can stimulate the growth of certain γ-Proteobacteria (Enterobacteriales and Alteromonadalaes) (Goldfarb et al., 2011). Different types of microorganisms in one ecosystem prefer different carbon sources, which alleviates nutrition competition pressure among species and maintains microbial genetic, functional diversity, and stability of one ecosystem (McCann, 2000). Therefore, different carbon utilization preferences may affect microbial functional diversity and stability of one ecosystem. Microorganisms tended to use different carbon sources (amino acids vs. sugar) under different CO₂ concentrations (350 ± 50 ppm vs. 750 ± 50 ppm) (Liu et al., 2015). It can be speculated that high CO₂ may affect microbial growth in caves through changing their preference for different carbon sources, and thus affect the entire cave ecosystem.

3.4 Characteristics of the Two Studied Caves with Respect to CH₄ Gas

It is notable that microbes with methanogenesis potential were dominant in the sediment of SMC and were significantly negatively correlated with CO₂ (Fig. S2), and bacteria with methanotrophy function were more abundant in SMC than that in XYC (Fig. S5). In the meanwhile, microbes with syntrophic methanogenesis potential, i.e., a bacterium that symbiotically coexists with methanogens (Dubbs and Whalen, 2010), existed in XYC and SMC, which was also negatively correlated with the CO₂ concentration (Figs. S3 and S6). These results indicated that high CO₂ can inhibit the growth of methanogenesis and syntrophic methanogenesis. The CH₄ gas concentration at the entrances of the two caves was higher than that inside caves, indicating that XYC may be a sink of atmospheric methane, which is consistent with previous studies (McDonough et al., 2016). Interestingly, it was surprising to find that although microbes with methanogenesis potential in XYC was in relatively low abundance, the methane content in XYC was significantly higher than that in SMC (Table S1). Such high concentration of CH₄ could be of biological origin due to the following reason: air in soil and caves could be involved in methanogenesis (McDonough et al., 2016), and the use of CO₂ to produce methane is the most common metabolic pathway known in methanogens (Liu and Whitman, 2008); the high concentration of CO₂ (i.e., substrate for methanogenesis) in XYC might provide basis for high methane through methanogenesis. In addition, high concentration of CH₄ was ever found to co-exist with high CO₂ in caves (Hutchens et al., 2004). However, other sources could not be excluded for the methane in the studied caves due to the lack of stable carbon isotopic characteristic of methane and microbial methane production rate test, which awaits further investigation.

4. CONCLUSION

Karst caves with different CO₂ concentrations had different microbial community compositions and potential functions. Microbial community responded differently to various environmental factors in the studied caves with high CO₂. The HCO₃^-/CO₂ concentration was an important factor affecting the microbial community composition of the two caves. The high CO₂ may affect the microbial growth by changing the pH of the cave environment, affecting the microbial enzyme systems, and by changing microbial preference for different types of carbon sources.