pH and Nitrate Drive Bacterial Diversity in Oil Reservoirs at a Localized Geographic Scale

Oil reservoirs are one of the most important deep subsurface biospheres. They are inhabited by diverse microorganisms including bacteria and archaea with diverse metabolic activities. Although recent studies have investigated the microbial communities in oil reservoirs at large geographic scales, it is still not clear how the microbial communities assemble, as the variation in the environment may be confounded with geographic distance. In this work, the microbial communities in oil reservoirs from the same oil field were identified at a localized geographic scale. We found that although the injected water contained diverse exogenous microorganisms, this had little effect on the microbial composition of the produced water. The Neutral Community Model analysis showed that both bacterial and archaeal communities are dispersal limited even at a localized scale. Further analysis showed that both pH and nitrate concentrations drive the assembly of bacterial communities, of which nitrate negatively correlated with bacterial alpha diversity and pH differences positively correlated with the dissimilarity of bacterial communities. In contrast, the physiochemical parameters had little effect on archaeal communities at the localized scale. Our results suggest that the assembly of microbial communities in oil reservoirs is scale- and taxonomy-dependent. Our work provides a comprehensive analysis of microbial communities in oil reservoirs at a localized geographic scale, which improves the understanding of the assembly of the microbial communities in oil reservoirs.


Introduction
Oil reservoirs are typical deep subsurface biospheres, which exhibit extreme environmental conditions, such as high salinity and high temperature [1,2]. Diverse microorganisms have been found in oil reservoirs using culture-dependent and culture-independent approaches, such as sulfate-reducing microorganisms, nitrate-reducing microorganisms, fermenters, acetogens, and methanogens [3][4][5][6]. They affect the quality of crude oil, mediate methanogenesis, and drive the biogeochemical process in the deep subsurface [1,7].
Despite significant efforts to reduce our dependence on fossil fuels for energy, crude oil remains one of the most important resources for industry and energy. Numerous technologies for enhanced oil recovery have been developed, of which microbial enhanced oil recovery (MEOR) is one of the most economic, sustainable, and efficient approaches [8]. Despite its unique advantages, this technique has not been widely applied, as conditions in oil reservoirs such as high temperature, high pressure, anoxic conditions, and high salinity are extremely harsh for microbial survival. Understanding microbial metabolism and the assembly of microbial communities in oil reservoirs is therefore essential. An adequate turbations. They have similar temperatures and depths, but different chemical properties. High-throughput sequencing of 16S rRNA genes was used to analyze both bacterial and archaeal communities. Correlation analysis between physiochemical data and microbial diversity has improved our understanding of how environmental variability shapes microbial communities in oil reservoirs at local geographic scales.

Sample Collection and Geochemical Analysis
All samples in this study were collected in July 2021 from production and injection wells in the Changqing field, a high-temperature, low-permeability oil field in northwest China. This oil field has been exploited over 10 years by long-term water flooding. The shallow groundwater is the major source of the injection water with an in situ temperature of about 20 • C. Three samples of the injection wells (i.e., Z1, Z3, and Z4) and 26 samples of the production wells were collected (Table 1 and Table S1), using the sampling valve on the wellhead. The bottles were filled with samples to exclude air. All bottles were sealed with sterile screw caps and then transported to the laboratory within 48 h below 5 • C after sampling for further analysis.   sured using gravimetric method [28]. The concentrations of cations and anions in the injected and produced waters were analyzed using an ion chromatograph (SHINE CIC-D160) with an SH-AC-3 column (for cation analysis) and an SH-CC-9 column (for anion analysis).

DNA Extraction and 16S rRNA Gene Sequencing
The oil and water phases of the produced water was separated by gravitational precipitation. Then, approximately 500 mL of the water phase of each reservoir production sample was collected for DNA extraction. Microbial cells in the water phase were collected by filtration through 0.22-µm sterilized cellulose ester membrane filters. Genomic DNA was extracted and purified from the collected cells using the FastDNA ® Spin Kit for Soil (MP Biomedicals, Cleveland, OH, USA). The concentration and purity of DNA were determined by NanoDrop2000 UV-vis spectrophotometer (Thermo Scientific, Wilmington, NC, USA), and DNA quality was checked by 1% agarose gel electrophoresis.

16S rRNA Gene Sequence Processing and Statistical Analysis
The raw reads obtained from sequencing were quality-filtered using Trimmomatic [30]. Specifically, reads with quality scores lower than 20, or sequence lengths shorter than 100 bp, or containing N were discarded. The clean reads were analyzed using the Quantitative Insights into Microbial Ecology 2 (QIIME 2) pipeline (v2019.7) [31]. Barcodes and primers were trimmed using the Cutadapt plugin [32]. Sequences were denoised and clustered into amplicon sequence variants (ASVs) using the DADA2 plugin [33]. Taxonomy assignment of the ASVs was performed using the Silva classifier (Silva-138) [34]. Afterwards, samples were rarefied to an even depth of 50,000 sequences per sample for bacteria and 30,000 sequences per sample for archaea, alpha diversity and beta diversity were calculated using the "qiime diversity" function implemented in QIIME2. The unconstrained non-metric multidimensional scaling (NMDS) analysis, Mantel test, Spearman correlation analysis, and hierarchical clustering were performed in R (v3.6.2), with the "vegan" and "pheatmap" packages. The Sloan Neutral Community Model for Prokaryotes was employed in R as described previously [35]. Results were visualized in R (v3.6.2) using the "ggplot2" package.

Physicochemical Characteristics of the Produced Water Samples
Recent biogeography research of microbial communities in oil reservoirs has demonstrated that the assembly of oil reservoir microbial communities is influenced by both physicochemical conditions and geographical locations [14,26]. Because of the dispersal limitation, although the oil reservoir microbial communities showed high heterogeneities at a large geographic level, such as the continental scale, it is not clear how niche selection drives the microbial communities at a localized scale (e.g., within the oil field). In this study, a total of 29 samples were collected from three injection wells and 26 adjacent production wells in the Changqing oil field, a high-temperature, low-permeability oil field in northwest China. The geographic distance between two adjacent wells is less than one kilometer. To characterize the natural habitat of the microbial community in each production well, we measured the geochemical conditions of each sample, which are listed in Table 1. The depth of the sampling wells ranged from 1505 to 1807 m and the temperature ranged from 53 to 62 • C. The most produced water samples had faintly acid pH values ranging from 6.30 to 7.20. High variations were found in the concentrations NO 3 − (0-237.2 mg/L) and SO 4 2− (1.83-767.5 mg/L), which were the potential electron acceptors in the anaerobic hydrocarbon degradation.

Microbial Compositions of the Produced Water Samples
A total of 11,168,180 bacterial 16S rRNA gene sequences and 5,494,843 archaeal 16S rRNA gene sequences were obtained from all samples. A total of 2472 and 555 amplicon sequence variants (ASVs) were obtained for bacterial and archaeal communities, respectively. To investigate whether niche selection drives the microbial communities in the oil reservoirs, we compared the microbial composition of injected water samples and produced water samples. The injected water communities and produced water communities only shared 146 bacterial ASVs and 22 archaeal ASVs, which only accounted for 9.3% and 5.0% of total bacterial and archaeal ASVs in the produced water samples, respectively ( Figure S1). The hierarchical clustering based on the Weighted UniFrac distances of both bacterial and archaeal communities clearly showed that all samples formed two distinct clusters as a function of the sample ( Figure 1). The injected water samples clustered together and all the produced water samples formed a separate cluster. Moreover, the taxonomic analysis showed that the microbial communities in the injected water samples and produced water samples were distinguished by their dominant taxa. Although the bacterial communities of both injected water samples and produced water samples were dominated by Proteobacteria ( Figure S2), they showed distinct bacterial compositions at the genus level. Polaromonas (average 15.3% of the total frequencies), Limnobacter (15.1%), Hydrogenophaga (7.1%), Novosphingobium (8.1%), and Caulobacter (5.9%) were the dominant genera in the injected water. In contrast, Marinobacter (16.4%), Marinobacterium (9.8%), Halomonas (9.4%), and Roseovarius (6.2%) dominated the bacterial communities of the produced water at the genus level. For archaea, the injected water samples were dominated by Crenachaeota and Euryarchaeota, but the produced water samples were dominated by Euryarchaeota and Halobacterota at the phylum level ( Figure 1B). The results indicated that microorganisms in the injected water had a weak effect on the microbial composition in the produced water. Because the dispersal of microorganisms in water-flooded oil reservoirs is primarily mediated by the injected water, this result suggested that niche selection might be the key factor driving microbial communities in oil reservoirs. It is also worth noting that the effect of microorganisms in the injected water on the microbe composition in the produced water may be environment dependent. For example, because the higher the in situ temperature of a reservoir, the fewer microorganisms in the injected water are adapted in the oil reservoir, the effect of microorganisms in the injected water on the microbial composition of the produced water is weaker in the high-temperature reservoirs [16].
water are adapted in the oil reservoir, the effect of microorganisms in the injected water on the microbial composition of the produced water is weaker in the high-temperature reservoirs [16].

Assembly Mechanisms of Oil Reservoir Communities at a Local Geographic Scale
To assess the neutral processes (e.g., migration) to the assembly of oil reservoir microbial communities, we employed the Slogan Neutral Community Model (NCM) [35]. We compared the observed community composition and distribution across the sample at the ASV level with that predicted by the Slogan neutral model (Figure 2). The fit of the neutral model (as indicated by R 2 ) was 0.406 for the bacterial community and 0.182 for the archaeal community. The estimated migration rates were 1 × 10 −4 and 1 × 10 −5 for the bacterial community and archaeal community, respectively, which were extremely low compared with other surface environments [36][37][38]. The results suggested that high dispersal limitation of microorganisms occurred in oil reservoirs even at the localized scale, which was consistent with the fact that although injected water might bring exogenous microorganisms into the deep subsurface reservoirs, they had little effect on the microbial composition ( Figure 1).

Assembly Mechanisms of Oil Reservoir Communities at a Local Geographic Scale
To assess the neutral processes (e.g., migration) to the assembly of oil reservoir microbial communities, we employed the Slogan Neutral Community Model (NCM) [35]. We compared the observed community composition and distribution across the sample at the ASV level with that predicted by the Slogan neutral model (Figure 2). The fit of the neutral model (as indicated by R 2 ) was 0.406 for the bacterial community and 0.182 for the archaeal community. The estimated migration rates were 1 × 10 −4 and 1 × 10 −5 for the bacterial community and archaeal community, respectively, which were extremely low compared with other surface environments [36][37][38]. The results suggested that high dispersal limitation of microorganisms occurred in oil reservoirs even at the localized scale, which was consistent with the fact that although injected water might bring exogenous microorganisms into the deep subsurface reservoirs, they had little effect on the microbial composition ( Figure 1).

The Concentration of Nitrate Determines the Diversity of the Communities
Because the neutral processes (e.g., migration) had little effect on the assembly of oil reservoir microbial communities, we speculated that physiochemical parameters might determine the microbial diversity and community assembly. Correlations of individual physicochemical parameters with alpha diversity showed that the concentration of NO3 − is a major driver of variation in bacterial alpha diversity (Figures 3 and S3). The Shannonwiener index (Spearman's coefficient: rs = −0.43, p = 0.03), faith PD index (Spearman's coefficient: rs = −0.71, p = 5.6 × 10 −5 ), and observed features (Spearman's coefficient: rs = −0.5, p = 0.0087) all decreased as the NO3 − concentrations increased. Note that the evenness of the bacterial communities (i.e., Pielou index) was not correlated with the physiochemical parameters and the results indicated that high concentrations of NO3 − decreased the bacterial richness. Similar results were also found in oil reservoirs with nitrate injection [39] and nitrate-treated pipelines [40]. In contrast, no significant correlation between archaeal alpha diversity and physiochemical parameters was found ( Figure S4).

The Concentration of Nitrate Determines the Diversity of the Communities
Because the neutral processes (e.g., migration) had little effect on the assembly of oil reservoir microbial communities, we speculated that physiochemical parameters might determine the microbial diversity and community assembly. Correlations of individual physicochemical parameters with alpha diversity showed that the concentration of NO 3 − is a major driver of variation in bacterial alpha diversity (Figures 3 and S3). The Shannonwiener index (Spearman's coefficient: r s = −0.43, p = 0.03), faith PD index (Spearman's coefficient: r s = −0.71, p = 5.6 × 10 −5 ), and observed features (Spearman's coefficient: r s = −0.5, p = 0.0087) all decreased as the NO 3 − concentrations increased. Note that the evenness of the bacterial communities (i.e., Pielou index) was not correlated with the physiochemical parameters and the results indicated that high concentrations of NO 3 − decreased the bacterial richness. Similar results were also found in oil reservoirs with nitrate injection [39] and nitrate-treated pipelines [40]. In contrast, no significant correlation between archaeal alpha diversity and physiochemical parameters was found ( Figure S4).

The Concentration of Nitrate Determines the Diversity of the Communities
Because the neutral processes (e.g., migration) had little effect on the assembly of oil reservoir microbial communities, we speculated that physiochemical parameters might determine the microbial diversity and community assembly. Correlations of individual physicochemical parameters with alpha diversity showed that the concentration of NO3 − is a major driver of variation in bacterial alpha diversity (Figures 3 and S3). The Shannonwiener index (Spearman's coefficient: rs = −0.43, p = 0.03), faith PD index (Spearman's coefficient: rs = −0.71, p = 5.6 × 10 −5 ), and observed features (Spearman's coefficient: rs = −0.5, p = 0.0087) all decreased as the NO3 − concentrations increased. Note that the evenness of the bacterial communities (i.e., Pielou index) was not correlated with the physiochemical parameters and the results indicated that high concentrations of NO3 − decreased the bacterial richness. Similar results were also found in oil reservoirs with nitrate injection [39] and nitrate-treated pipelines [40]. In contrast, no significant correlation between archaeal alpha diversity and physiochemical parameters was found ( Figure S4).

pH Drives the Microbial Compositions of the Communities
To reveal how the physiochemical parameters determine the microbial composition in the oil reservoirs, we investigated the correlation between physiochemical parameters and the beta diversity throughout the samples. We found that beta diversity (i.e., weighted UniFrac distance) of bacterial communities in the produced water significantly correlated with pH (Mantel test: r = 0.2876, p = 0.005) ( Table 2). Similar to the results for the correlation between archaeal alpha diversity and physiochemical parameters, the beta diversity of archaeal communities was not correlated with physiochemical parameters either. We fitted the physiochemical parameters to unconstrained non-metric multidimensional scaling (NMDS) ordination ( Figure S5). The result showed that pH significantly correlated with the bacterial compositions in the produced water samples (r = 0.344, p = 0.01) ( Figure 4A). To confirm whether pH drives bacterial communities, we compared beta diversity across samples as a function of pH. We found that nMDS 1 positively correlated with pH values for bacteria (Spearman's coefficient: r s = 0.51, p = 0.0084) ( Figure 5A). Weighted UniFrac distances also showed that the bacterial dissimilarities between samples were significantly correlated with the increase in pH values ( Figure 5B and Figure S6). The results confirmed that pH drove the bacteria compositions of produced water samples at a localized geographic scale.  In contrast, the physicochemical parameters had a weak effect on archaeal compositions ( Figures 4B, S7 and S8). These results suggested that the archaeal communities were homogenized in oil reservoirs at a localized geographic scale. We compared the beta diversity within bacterial communities and archaeal communities, respectively. We found that the weighted UniFrac dissimilarities within bacterial communities were significantly higher than those within archaeal communities ( Figure 5C), suggesting higher homogeneities throughout archaeal communities in oil reservoirs at a localized geographic scale. The homogeneities of archaeal communities were also found at a larger scale [14,26]. Zhao et al. found that the archaeal compositions are conserved in water-flooded oil reservoirs throughout China, which showed more core archaeal taxa [14]. Yun et al. found that although both bacterial and archaeal communities displayed clear distance-decay patterns, the beta diversity within archaeal communities was lower than that of bacterial communities [26]. These findings suggested that the archaeal community is more homogeneous in oil reservoirs at a large scale. Moreover, a stronger correlation between archaeal communities and physiochemical parameters than bacterial communities was also found at the continental scale, which might be explained by the high heterogeneity of bacteria communities at the large scale [14]. In this work, we found that at the local geographic scale the bacteria communities were significantly correlated with pH. The results suggested that the assembly of microbial communities in oil reservoirs is scale-and taxonomy-dependent, i.e., archaeal compositions could be predicted at a large geographic scale, while bacterial compositions could be predicted at the local scale by physiochemical parameters.

Taxonomic Association with Physicochemical Parameters
To determine the influence of physiochemical parameters on the relative abundances of individual microbial taxa, we applied Spearman's correlation analysis between physiochemical parameters and the relative abundances of microbial taxa at the genus level. We found that 34 of the top 50 bacterial genera significantly correlated with at least one physiochemical parameter. Of these genera, 13 bacterial genera including Alcanivorax, Flavobacterium, Rhizobium, and Procabacter negatively correlated with NO3 − and no genus of the top 50 genera showed a positive correlation with NO3 − (Figure 6A). This result might explain why bacterial alpha diversity decreased as nitrate concentration increased. The correlation analysis also showed that seven out of the top 50 bacterial genera including Alkalibacter and Sphaerochaeta were negatively correlated with the pH value ( Figure 6A). In contrast, the physicochemical parameters had a weak effect on archaeal compositions ( Figures 4B, S7 and S8). These results suggested that the archaeal communities were homogenized in oil reservoirs at a localized geographic scale. We compared the beta diversity within bacterial communities and archaeal communities, respectively. We found that the weighted UniFrac dissimilarities within bacterial communities were significantly higher than those within archaeal communities ( Figure 5C), suggesting higher homogeneities throughout archaeal communities in oil reservoirs at a localized geographic scale. The homogeneities of archaeal communities were also found at a larger scale [14,26]. Zhao et al. found that the archaeal compositions are conserved in water-flooded oil reservoirs throughout China, which showed more core archaeal taxa [14]. Yun et al. found that although both bacterial and archaeal communities displayed clear distance-decay patterns, the beta diversity within archaeal communities was lower than that of bacterial communities [26]. These findings suggested that the archaeal community is more homogeneous in oil reservoirs at a large scale. Moreover, a stronger correlation between archaeal communities and physiochemical parameters than bacterial communities was also found at the continental scale, which might be explained by the high heterogeneity of bacteria communities at the large scale [14]. In this work, we found that at the local geographic scale the bacteria communities were significantly correlated with pH. The results suggested that the assembly of microbial communities in oil reservoirs is scale-and taxonomy-dependent, i.e., archaeal compositions could be predicted at a large geographic scale, while bacterial compositions could be predicted at the local scale by physiochemical parameters.

Taxonomic Association with Physicochemical Parameters
To determine the influence of physiochemical parameters on the relative abundances of individual microbial taxa, we applied Spearman's correlation analysis between physiochemical parameters and the relative abundances of microbial taxa at the genus level. We found that 34 of the top 50 bacterial genera significantly correlated with at least one physiochemical parameter. Of these genera, 13 bacterial genera including Alcanivorax, Flavobacterium, Rhizobium, and Procabacter negatively correlated with NO 3 − and no genus of the top 50 genera showed a positive correlation with NO 3 − ( Figure 6A). This result might explain why bacterial alpha diversity decreased as nitrate concentration increased. The correlation analysis also showed that seven out of the top 50 bacterial genera including Alkalibacter and Sphaerochaeta were negatively correlated with the pH value ( Figure 6A). Only one genus (i.e., an unclassified genus in the family of Pseudomonadaceae) was positively correlated with pH. It was notable that the pH value and TDS (including cations and chloride) showed opposite effects on the relative abundances of the bacterial taxa.

Conclusions
In this work, we investigated the microbial communities in different oil production wells from the same oil field. We found that both bacterial and archaeal communities are dispersal limited, even at the localized geographic scale. Compared with the bacterial Compared with bacterial taxa, the correlation between archaeal taxa and physiochemical parameters was weaker ( Figure 6B). Of all archaeal genera, only 16 out of all 46 archaeal genera were significantly correlated with at least one physicochemical parameter. An unclassified genus in the family Methanosarcinaceae was positively correlated with TDS, chloride, and sodium, suggesting that this genus might be resistant to high saline conditions. In contrast, the genus Methanobacterium was less resistant to saline conditions, which decreased when the concentrations of chloride and sodium were increased. These results were consistent with low beta diversity within archaeal communities, which also suggested that archaeal communities in oil reservoirs were stable and homogeneous at a localized geographic scale.

Conclusions
In this work, we investigated the microbial communities in different oil production wells from the same oil field. We found that both bacterial and archaeal communities are dispersal limited, even at the localized geographic scale. Compared with the bacterial communities, the archaeal communities were more similar within different production wells. Further correlation analysis also showed that nitrate concentration negatively correlated with bacterial alpha diversity and pH differences positively correlated with the dissimilarity of bacterial communities. Our results suggest that the assembly of microbial communities in oil reservoirs is scale-and taxonomy-dependent. Our work provides new insights into the assembly of microbial communities in oil reservoirs.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/microorganisms11010151/s1. Figure S1: Distribution of bacterial (A) and archaeal (B) ASVs in the produced water and injected water; Figure S2: Bacterial composition of samples at the phylum level; Figure S3: Relationships between bacterial alpha diversities and physiochemical parameters; Figure S4: Relationships between archaeal alpha diversities and physiochemical parameters; Figure S5: Relationships between bacterial profiles and physiochemical parameters; Figure S6: Relationships between weighted UniFrac distances of bacterial communities and variations in physiochemical parameters; Figure S7: Relationships between archaeal profiles and physiochemical parameters; Figure S8: Relationships between weighted UniFrac distances of archaeal communities and variations in physiochemical parameters; Table S1: Hydrodynamic connection of injection and production wells used in the work.