Spatial and temporal dynamics of bacterioplankton community composition in a subtropical dammed karst river of southwestern China

Abstract River damming influences the hydro‐physicochemical variations in karst water; however, such disruption in bacterioplankton communities has seldom been studied. Here, three sampling sites (city‐river section, reservoir area, and outflow area) of the Ca2+–Mg2+–HCO 3 −–SO 4 2− water type in the dammed Liu River were selected to investigate the bacterioplankton community composition as identified by high‐throughput 16S rRNA gene sequencing. In the dammed Liu River, thermal regimes have been altered, which has resulted in considerable spatial‐temporal differences in total dissolved solids (TDSs), oxidation‐reduction potential (Eh), dissolved oxygen (DO), and pH and in a different microenvironment for bacterioplankton. Among the dominant bacterioplankton phyla, Proteobacteria, Actinobacteria, Bacteroidetes, and Cyanobacteria account for 38.99%–87.24%, 3.75%–36.55%, 4.77%–38.90%, and 0%–14.44% of the total reads (mean relative frequency), respectively. Bacterioplankton communities are dominated by Brevundimonas, Novosphingobium, Zymomonas, the Actinobacteria hgcIclade, the CL500‐29 marine group, Sediminibacterium, Flavobacterium, Pseudarcicella, Cloacibacterium, and Prochlorococcus. Their abundances covary with spatial‐temporal variations in hydro‐physicochemical factors, as also demonstrated by beta diversity analyses. In addition, temperature plays a pivotal role in maintaining bacterioplankton biodiversity and hydro‐physicochemical variations. This result also highlights the concept that ecological niches for aquatic bacteria in dammed karst rivers do not accidentally occur but are the result of a suite of environmental forces. In addition, bacterioplankton can alter the aquatic carbon/nitrogen cycle and contribute to karst river metabolism.


| INTRODUC TI ON
Karst rivers contain the surface networks of water resources for domestic, industrial, and agricultural use and represent an exclusive habitat for microbes that perform critical functions in biogeochemical cycles under the influence of carbonate rock dissolution (Han & Liu, 2004). Karst rivers are commonly regulated by damming, yet the influence of these dams on changes in hydrological series of water discharge is negative or positive (Miao, Ni, Borthwick, & Yang, 2011). Although the diversity and dynamics of microbes in karst springs (Farnleitner et al., 2005;Ohad et al., 2015;Savio et al., 2018), unsaturated and saturated karst aquifers (Cooper et al., 2016;Gray & Engel, 2013;Johnson et al., 2011;Lin et al., 2012;Menning et al., 2018), and water pools (Shabarova et al., 2014) as well as in groundwater-surface water exchange systems (Li, Song, et al., 2017) have been discussed in the literature, much less attention has been paid to the structure of bacterioplankton communities in dammed karst rivers. In addition, previous studies on bacterioplankton communities in the canyon-shaped and meso-eutrophic Rimov Reservoir (Simek et al., 2008), the dammed Ebro River (Ruiz-González, Proia, Ferrera, Gasol, & Sabater, 2013), and the rivers controlled by the Three Gorges Dam Li, Lu, et al., 2017;Yan et al., 2015) did not include the seasonal variation or depth dynamics in bacterioplankton.
Thus, a major challenge in understanding bacterioplankton ecological function is to determine the role of physicochemical properties in dammed karst rivers or the ecological factors that shape bacterioplankton biodiversity and species coexistence (Ávila, Staehr, Barbosa, Chartone-Souza, & Nascimento, 2017). Despite the controlling factors (such as trophic interactions, evolutionary perspective, spatial heterogeneity, and temporal heterogeneity) of prokaryotic diversity summarized by Torsvik, Øvreås, and Thingstad (2002), the basic principles governing their distribution and abundance in aquatic environments are just beginning to be explored. For instance, Fisher, Klug, Lauster, Newton, and Triplett (2000) highlighted that nutrition (inorganic nitrogen and phosphorus as well as carbon in the form of glucose) and trophic interactions determined bacterioplankton diversity in an oligotrophic lake in northern Wisconsin. Ruiz-González et al. (2013) used surface water samples from the dammed Ebro River and noted that damming caused a pronounced decline in Betaproteobacteria, Gammaproteobacteria, and Bacteroidetes from upstream to downstream sites, whereas Alphaproteobacteria and Actinobacteria significantly increased after reservoirs were constructed. Ávila et al. (2017) asserted that thermal stratification and oxygen depletion dictated the bacterioplankton diversity in two tropical shallow lakes in the Brazilian Atlantic Forest. Ren et al. (2017) found that spring bacterioplankton community composition shifted significantly under enhanced warming and nutrient-enriched conditions.
Although the above studies provide an exceptional opportunity to gain insight into the controlling factors of bacterioplankton community composition and structure in inland aquatic ecosystems, factors related to bacterioplankton diversity and communities in the city-river section, reservoir area, and outflow area of dammed karst rivers are still unknown.
The presence of dams is problematic for many aquatic ecosystems (Bednarek, 2001). Consequently, altered thermal regimes in dammed rivers have been observed at a spatial scale (Kelly, Smokorowski, & Power, 2017;Weber et al., 2017). In addition, temperature stratification usually occurs in dammed rivers because of the change to a more reservoir-like habitat (Bednarek, 2001). It should be noted that temperature can influence the hydrochemistry and recycling of nutrients, etc. (Bednarek, 2001;Li, Sun, Han, Liu, & Yu, 2008). Here, we hypothesized that water temperature is the key factor controlling bacterioplankton community composition in dammed karst rivers.
As a result, three sampling sites (city-river section, reservoir area, and outflow area) were selected to investigate bacterioplankton communities, water hydro-physicochemical properties and their relationship in the dammed Liu River (Figure 1). Consequently, how the bacterioplankton communities changes in relation to hydrophysicochemical parameters was determined via high-throughput 16S rRNA gene sequencing.

| Study area
The Liu River (24°N-27°N, 107°E-111°E) is a tributary within the Pearl River system in Guangxi, China, and was formed by the confluence of the Rong and Long Rivers in Fengshan. The Liu River passes through Liuzhou city (https ://en.wikip edia.org/wiki/Liu_River ) as well as a sand/shale stone area and limestone area, as indicated in Figure 1. According to water quality monitoring data in China (http://123.127.175.45:8082/), the water environmental quality of the Liu River belongs to class II or III, suggesting that the river can be used as a water resource for domestic use. In addition, under the influence of the East Asian monsoon and South Asian monsoon, 71% of the annual precipitation (1004 mm) occurs between April and August. The mean temperature from December to March is 12.6°C (dry-cold season), the mean temperature from April to August is 25.5°C (rainy hot season), and the mean temperature from April to August is 22.4°C (dry hot season). In addition, the water flow of the Liu River is controlled by many dams, including a constructed rubber dam in the city of Liuzhou and the Honghua dam (between sampling sites B and C) at the hydroelectric station (Figure 1), resulting in slow water flow and higher nutrient concentrations. From upstream to downstream in dammed Liu River, the sampling site before the rubber dam is named A (city-river section), the sampling site before the Honghua dam is named B (reservoir area), and the sampling site after the Honghua dam is named C (outflow area).

| Sampling procedure and hydrological monitoring
A total of 23 water samples for the analysis of water hydro-physicochemistry and bacterioplankton community structure were collected in March, June, and September 2016 using a standard water sampler Acc. to Ruttner 2 L (HYDRO-BIOS, Germany) at three sites in the Liu River (Figure 1). To assess the damming influence on the spatial-temporal dynamics of the bacterioplankton community composition and hydro-physicochemistry, water samples were taken at three different depths (0, 5, and 10 m). However, due to water level changes, the samples in the reservoir area and outflow area lacked a layer at 5 and 10 m. Samples were named according to time (M, March; J, June; and S, September), sampling site (A, B, and C), and specific depth (0, 5, and 10 m), in that order (e.g., MA0).
F I G U R E 1 Map showing localization of the dammed Liu River in Liuzhou, Guangxi, P. R. China (a). Timing and depth of sampling locations in the Liu River (b). Sites (A, B, and C) illustrate the sampling locations in the Liu River. The blue triangles indicate the depth of the water samples Water samples (approximately 3 L) were prefiltered using 3 μm filter membranes, and then filtered through 0.22 μm pore-size filter membranes (Merck Millipore, Germany) in situ for bacterioplankton samples. After that, the filter membranes were stored at −80°C until further processing.
Water temperature, pH, electrical conductivity (EC), DO, turbidity, chlorophyll-α (Chlα), dissolved organic nitrogen, TDSs, and Eh were obtained in situ using a multiprobe sensor (YSI, USA  Figure 2a and b). Samples for total nitrogen (TN), total carbon (TC), dissolved organic carbon (DOC) and dissolved organic nitrogen were collected according to Li, Song, et al. ( 2017) and analyzed using a multi N/C R 3100 total organic carbon (TOC analyzer) (Analytik Jena AG, Germany).
Particulate organic carbon (POC) is a broad term that encompasses suspended organic matter such as phytoplankton; consequently, fluvial δ 13 C POC values are a reflection of the relative contributions from freshwater phytoplankton (−25‰ to −30‰) and particulate terrestrial organic matter (−25‰ to −33‰) (Lamb, Wilson, & Leng, 2006).  USA); the error of analysis was better than 0.2 ‰ (1σ). The results were expressed in δ 13 C relative to the Pee Dee Belemnite (PDB) standard, as shown in Figure 3d. The hydro-physicochemical characteristics of the water samples are summarized in Table 1.

| Bioinformatics analysis and statistical analyses
The achieved 16S sequence data from 23 water samples were processed using the QIIME 1.7.0 software (Kuczynski et al., 2012;Li et al., 2018). Low-quality sequences with lengths below 150 bp and an average quality score below 30 were excluded. In addi-

| Hydro-physicochemical characteristics of dammed Liu River
The spatial-temporal hydro-physicochemical characteristics of the dammed Liu River are listed in Table 1  ]≈0.95) (Gao et al., 2009). Although water column thermal stratification is not evident, the spatial-temporal dynamics of water temperatures are clear. In dammed rivers, water temperature usually increases from upstream to downstream, resulting in changed thermal capacities (Hanna, Saito, Bartholow, & Sandelin, 1999). In addition, for strongly seasonal rivers with varying water temperatures, 26% of the variation in water temperature is attributed indirectly to low flow changes, and the remaining fraction is attributed directly to changed atmospheric energy input (van Vliet et al., 2013). Consequently, the temperature can be clustered into three groups as seen in Figure   indicate that bacterioplankton production is relating to water temperature differences (Figure 2d), which confirms the findings that water temperature and oxygen have strong positive correlations with bacterioplankton (Araújo & Godinho, 2008).

| Spatial-temporal variations in bacterioplankton community composition
Of the reads, 96.35% were assigned to 10 major phyla, including Proteobacteria, Actinobacteria, Bacteroidetes, Cyanobacteria, Verrucomicrobia, Planctomycetes, Firmicutes, Acidobacteria, Chloroflexi, and Armatimonadetes, as illustrated in Figure 4 (Archaea are not included). Among these bacterioplankton phyla, Proteobacteria, Actinobacteria, Bacteroidetes, and Cyanobacteria account for 38.99%-87.24%, 3.75%-36.55%, 4.77%-38.90%, and 0%-14.44% of the total reads (mean relative frequency), respectively. These phyla have also been found in different relative proportions in other freshwater ecosystems worldwide (Ávila et al., 2017;Newton et al., 2011). Notably, if a "core" assemblage as the subset of OTUs is present in all samples, then this assemblage can be defined as a core OTU (Engel, 2010 River, as shown in Figure A2. They have also been found to be the numerically dominant group in other freshwater ecosystems (Newton et al., 2011). This result is in accordance with the classification of Betaproteobacteria as r-strategists (Novello et al., 2017), that is, taxa able to grow rapidly under conditions of high resource availability. In addition, the abundance of Proteobacteria increases with depth in March and June (except for JC0, JC5, and JC10); however, the abundance of Proteobacteria decreases with depth in September. Betaproteobacteria have a similar variation pattern, though they are often the most abundant bacteria inhabiting the upper waters of lakes (Newton et al., 2011).
Previous studies demonstrated that Proteobacteria are involved in a variety of biogeochemical processes in aquatic ecosystems (Damashek & Francis, 2018;Xu et al., 2014;Zhang et al., 2015).
Alphaproteobacteria are at the hub of the global nitrogen cycle (Newton et al., 2011). Indeed, the genus Brevundimonas (OTU 1) has a small angle with dissolved organic nitrogen and nitrate, as indicated in Figure 3a, suggesting that they are nitrogen-fixing bacteria (Liu, Peng, & Li, 2012  Acid obacteria Chloroflexi Armatimonadetes Others Unassigned clade belonging to the phylum Actinobacteria are known to have a competitive advantage over others in lakes that are characterized by low DOC and low temperature (Glöckner et al., 2000).
Cyanobacteria are the largest and most widely distributed group of photosynthetic prokaryotes, found in ecosystems ranging from marine and freshwater to terrene (Stanier & Bazine, 1977).
Interestingly, Cyanobacteria account for only a small fraction and are hardly detected in MA10 and MB10. In addition, the Cyanobacteria abundance increases with depth (except for MC and SB), and their abundances are higher in March and September than in June.
This observation was supported by the fact that river damming leads to the disappearance of cyanobacterial blooms (Domingues, Barbosa, & Galvão, 2014) and that cyanobacterial growth is usually enhanced by high water residence times during the dry, cold season and dry, hot season with low freshwater flows (Domingues, Barbosa, & Galvao, 2005). Under these conditions, Cyanobacteria can grow abundantly and form extensive blooms, as confirmed by Betaproteobacteria (Pearson's r = 0.864, p = 0.00, n = 23, twotailed). Some Cyanobacteria are able to produce potent toxins and have drastic impacts on the ecosystem and surrounding communities (Steffen et al., 2012). Consequently, cyanobacterial blooms will disrupt aquatic food webs and act as a driver of hypoxia, especially changing the sensitivity of Proteobacteria to grazing pressure (Eiler, Olsson, & Bertilsson, 2006). As exposed, common freshwater lake genera belonging to Cyanobacteria include Microcystis, Anabaena, Aphanizomenon, Oscillatoria, Planktothrix, Synechococcus, and Cyanothece (Newton et al., 2011); however, in our study, the top Cyanobacteria-related OTUs 23 and 49 are classified into Prochlorococcus. Prochlorococcus (OTUs 23 and 49) are positively correlated with Chlα, as indicated by their small angles in Figure 3a, as previously reported by Domingues et al. (2014). The results suggest that they might contribute significantly to global primary productivity through oxygenic photosynthesis (Boekema et al., 2001;Newton et al., 2011;Stanier & Bazine, 1977). In addition, Prochlorococcus has small angles with dissolved organic nitrogen and nitrate, suggesting that they can play a key role in nutrient cycling in freshwater (Stanier & Bazine, 1977). Cyanobacteria-related OTU 23 is positively correlated with iron (Figure 3a), which is supported by the iron limitation of Prochlorococcus sp. (Mann & Chisholm, 2000).

| Spatial-temporal variations of bacterioplankton community diversity
To investigate the effects of spatial (sampling site and depth) and temporal (season) changes on bacterioplankton communities, we examined alpha and beta diversity (Figure 4 and Appendix Table A1).
According to the numbers of observed and estimated OTUs as well as Shannon and Simpson diversity in the Liu River, alpha diversity shows highly spatial-temporal variations, however, alpha diversity Indeed, Ávila et al. (2017) found that DO showed a significantly negative correlation with the Shannon index in two tropical shallow lakes in the Brazilian Atlantic Forest, as indicated by regression analysis; however, the regression results about DO and the Shannon index are not significant in our study (R 2 = 0.368, p = 0.084).
Moreover, higher alpha diversity is usually found at the surface water in the JB and JC samples, whereas the JA sample has higher alpha diversity at a depth of 5 m. Although alpha diversity measures have no significant difference with depths (Appendix Table A1), the high alpha diversity values of the JA5 sample may be attributed to a less stressful environment due to higher nutrient availability and isolation from external disturbances, such as UV radiation, wind, and waves at this layer (Ávila et al., 2017). In contrast, the alpha diversity values are lower in the surface and bottom layers of the MB0 and SA10 samples, suggesting that anthropogenic activity (e.g., shipping activity, fishing or swimming) can decrease bacterioplankton diversity in the surface layer, and the input of sand/mud restricts supplemental energy generation by light harvesting for bacterioplankton (Gómez-Consarnau et al., 2007), as confirmed in Table 1. Compared with site C, site A and B under the impact of a long water-retention time is quite stable with slow rates of water flow, high water transparency and high nutrient levels, which in turn enhance the difference of bacterioplankton (Yang et al., 2018), as seen in Appendix Table A1. Interestingly, in our study, the alpha diversity with minimal spatial-temporal variations in other C samples may be attributed to the influence of water discharge of the Honghua dam resulting in a normalized bacterioplankton community; however, the hydro-physicochemical characteristics of site C are different. Moreover, because site C and B are directly connected along the Honghua dam, the JC0 sample also has high alpha diversity values.
On the basis of the fact that bacterioplankton community members turn over quickly in response to changing environmental conditions and beta diversity is the variation in species composition among sites in a geographic area (Legendre, Borcard, & Peres-Neto, 2005), we used the unweighted UniFrac and Bray-Curtis distances of beta diversity, independent of changes in alpha diversity, to compare the range of bacterioplankton diversity in spatial-temporal variations ( Figure 5). Highly similar communities (three clusters) are observed at the same sampling time, suggesting that a mixed seasonal environment can facilitate bacterial coexistence (Huang, Dong, Jiang, Wang, & Yang, 2016), as confirmed by Appendix  (Brandão, Staehr, & Bezerra-Neto, 2016). In addition, unweighted UniFrac and Bray-Curtis analyses revealed an enhanced dissimilarity between communities, suggesting that stratification determines the phylogenetic diversity in each community layer, as previously reported by Ávila et al. (2017) in two tropical shallow lakes in the Brazilian Atlantic Forest. Overall, our results suggest that spatial-temporal variations in bacterioplankton community structure are shaped by hydro-physicochemical variability relating to water temperature differences.

| Relationship of bacterioplankton communities with hydro-physicochemical properties
To explore the key drivers shaping bacterioplankton communities in the dammed Liu River, we provided comprehensive results using a variety of statistical methods. The partial Mantel test (permutations = 999) shows the significant effects of temperature on the bacterioplankton community (p < 0.01) when pH and nutrition factors were controlled (Table 2). pH is significantly correlated with bacterioplankton communities (r = 0.161, p = 0.041) when the nutrition factor is controlled.
DO is also significantly correlated with bacterioplankton communities (p < 0.01) when temperature, pH, and nutrition factors are controlled.
In addition, the PLS-PM is represented here with a goodness-offit (GoF) value of 0.501 to integrate the complex interrelationships among environmental factors and bacterioplankton communities ( Figure 6). According to the PLS-PM, temperature, and nutrition exert direct positive effects on bacterioplankton composition and alpha diversity, and pH exerts direct negative effects on bacterioplankton composition and alpha diversity; however, DO exerts a direct negative effect on bacterioplankton composition and a direct positive effect on alpha diversity. Notably, temperature exerts significant positive or negative effects on pH, DO, and nutrition, which in turn cast the influences on bacterioplankton composition and alpha diversity ( Figure 6).
pH is a major environmental determinant shaping the patterns of bacterioplankton biodiversity and bacterioplankton community structures (Yun et al., 2016); however, we have very limited information about the patterns and processes by which overall bacterioplankton communities assemble across wide pH gradients in karst waters (Ren et al., 2015). DO exerts a direct negative effect on bacterioplankton composition, thus contributing to the shape of community structures of anoxygenic and oxygenic phototrophic bacteria in the dammed Liu River (Taipale, Jones, & Tiirola, 2009). It should be noted that pH, DO, and nutrition are affected by water temperature; consequently temperature plays pivotal roles in maintaining aquatic bacterial biodiversity patterns and bacterioplankton community composition, as previously reported by Wang, Pan, Soininen, Heino, and Shen (2016).

| CON CLUS ION
In the dammed Liu River, thermal regimes have been altered, which has resulted in considerable spatial-temporal differences in TDS, Eh, and Prochlorococcus, which covary with spatial-temporal variations of hydro-physicochemical factors. In addition, these groups played a key role in the carbon/nitrogen cycle and contributed to karst

CO N FLI C T O F I NTE R E S T S
None declared.

AUTH O R CO NTR I B UTI O N S
SY, RXH, and QL conceived and designed the experiment. SY, RXH, AS, and QL performed the experiment. AS, ZJJ, YML, YDH, QL, XHW, WEGM, and JHC analyzed the data. YDH and QL led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

E TH I C S S TATEM ENT
None required.

Raw sequence reads have been deposited to NCBI Sequence Read
Archive under the accession number SRP126836. Values are the mean of analytical replicates for each sample ± standard deviations. Statistical pairwise multiple comparisons of data homogeneity were carried out by the Tukey test: means with the same letter in the same column are not significantly different at P < 0.05.