Distinct Assembly Mechanisms for Prokaryotic and Microeukaryotic Communities in the Water of Qinghai Lake

0. INTRODUCTION

Saline lakes are distributed globally, and their surface area accounts for 45% of total area of inland water bodies (Jellison et al., 2008; Williams, 1996). Microorganisms are one important component of saline lake ecosystems (Oren, 2008) and contribute significantly to the biogeochemical cycles of carbon, nitrogen and sulfur elements (Wang W D et al., 2019; Paul Antony et al., 2013; Kuznetsov, 1975) and biomass production (Cole et al., 1988; Hammer, 1981). It is widely recognized that microbial community composition is usually associated with its ecological functions (Bertolet et al., 2022; Reed and Martiny, 2013; Waldrop et al., 2000). Therefore, it is of great importance to study the key factors influencing microbial community compositions in order to understand biogeochemical functions in saline lakes.

Microbial assembly is the formation mechanism that determines the diversity, function, composition, and biogeographical distribution of microbial communities (Liu et al., 2020; Martiny et al., 2006). A large number of studies have demonstrated that the microbial assembly was jointly governed by deterministic (niche-related processes) and stochastic processes (neutral-related processes) (Dini-Andreote et al., 2015; Zhou et al., 2014; Wang et al., 2013; Stegen et al., 2012), and their relative importance varies with spatial and temporal scales and environments. Deterministic processes emphasize environmental filtering (e.g., salinity, pH, temperature) and species interactions (e.g., competition and cooperation) caused by habitat preferences and microbial fitness, while stochastic processes emphasize dispersal limitation and random processes such as immigration, speciation, emigration, and extinction (Chase and Myers, 2011; Hubbell, 2011). Prokaryotes and microeukaryotes are important components of lake microbial communities (Liu et al., 2011), and key populations for studying the assembly mechanisms for lake microbial communities (Ma et al., 2018; Zhao et al., 2017). Previous studies demonstrate that deterministic and stochastic processes jointly drive the assembly of prokaryotic microbial communities in lakes (Yang et al., 2021; Monchamp et al., 2019; Wang et al., 2013; Xiong et al., 2012). However, the relative importance of deterministic and stochastic processes in the shaping of microeukaryotic microbial communities in saline lake water has not been fully understood (Liu et al., 2020; Logares et al., 2018).

Furthermore, the relative importance of deterministic and stochastic processes varied greatly on the assembly of prokaryotic and microeukaryotic microbial communities in different niches (Gad et al., 2020; Liu et al., 2020; Yang et al., 2020; Logares et al., 2018). Microeukaryotes have larger cell sizes, more complex cell structures and more functional enzymes than prokaryotes, which may lead to their different niches (Fabian et al., 2017; Danger et al., 2016; Massana and Logares, 2013). Therefore, it is reasonable to hypothesize that the relative influence of deterministic and stochastic processes on prokaryotic communities on the shaping of microbial communities may be distinct from that of microeukaryotic communities. However, so far such information remains limited in saline lake water.

The Qinghai-Tibetan Plateau (QTP) hosts thousands of lakes with surface area larger than 1 km², of which approximately 70% are saline lakes (Li et al., 2022; Ma et al., 2011; Zheng, 1997). In recent years, due to the impact of climate change and human activities, the water environmental factors (e.g., salinity) of many QTP lakes have changed (Zhang et al., 2020). For example, the water level of Qinghai Lake (QHL) is rising by more than three meters, and the lake area has increased by about 3% since 2005 in the context of current climate warming (Ge et al., 2021; Tang et al., 2018; Cui and Li, 2016). As the largest (area 4 432 km²) saline (salinity 14.13 g·L^-1) lake on the QTP, QHL plays a key role in maintaining ecosystem function and regulating regional climate change (Li et al., 2022; Jiang et al., 2008; Dong et al., 2006). Therefore, QHL is considered as an ideal and representative site to examine lake ecosystem's response to climate change. Microorganisms are the basic component of the lacustrine ecosystem and are very sensitive to environmental changes. They have become a model group for studying community assembly, biodiversity maintenance, and environmental assessment (Liu et al., 2020; Yang et al., 2019; Wu et al., 2006). However, there is limited knowledge about how environmental changes (e.g., variation of depth, salinity and nutrient availability) affect the distribution and assembly mechanisms of prokaryotic and microeukaryotic communities in saline lake water. In this study, we systematically investigated the basic physicochemical parameters and the assembly mechanisms of prokaryotic and microeukaryotic community compositions in Qinghai Lake at different water depths. The objectives of this study were to: (1) quantify the relative contribution of deterministic and stochastic processes in the shaping of prokaryotic and microeukaryotic communities, and (2) determine the key environmental factors shaping the prokaryotic and microeukaryotic community compositions in the studied saline lake water.

1. MATERIALS AND METHODS

1.1 Sample Collection and Lake Water Characterization

Sampling cruise was conducted in Qinghai Lake in June 2018. A total of 29 lake water samples were collected from eleven sites (Q1–Q11) at different water depths in Qinghai Lake (each site was divided into 3 depths, with only top samples for sites Q10 and Q11) (Table 1, Fig. S1). According to the depth of each site, the water depths of 0.1, 6–14, and 11–27 m from the horizontal plane were defined as top, middle, and bottom water columns, respectively. At each sampling site, water depth was measured using a length-marker rope; the GPS location was measured using a portable GPS unit (eTrex H, Garmin, US); triplicate water samples were collected with a Schindler sampler at different water depths. The resulting triplicate water samples from each depth were mixed into one unique representative analytical sample. All water samples were subsequently divided into two subsamples for geochemical analysis (2 L) and microbial community (0.5 L) analyses.

Table  1.  Depth profiles of main physicochemical parameters of the studied saline lake water on the day of sampling

Sample	Depth (m)	Salinity (g L-1)	pH	Turbidity (NTU)	DOC (mg L-1)	DIC (mM)	Ammonium (mg L-1)	Nitrite (mg L-1)	Nitrate (mg L-1)	TN (mg L-1)	TP (mg L-1)	Chlorophyll_a (ug/L)	Ferrous iron (mg L-1)	Reduced sulfur (mg L-1)
Q1_T	0.1	14.47	9.48	1.04	38.7	27.8	0.528	0.001	4.301	6.429	0.055	0.40	0.979	0.004
Q1_M	14	14.45	9.50	0.82	42.3	28.1	0.586	0.001	4.348	7.284	0.063	0.44	0.508	0.006
Q1_B	27	13.26	9.52	0.95	41.6	29.2	0.461	0.001	5.206	7.262	0.027	0.64	0.058	0.002
Q2_T	0.1	11.91	9.53	0.95	47.8	30.2	0.553	0.001	0.807	7.198	0.008	0.34	0.104	0.003
Q2_M	12	12.95	9.50	0.70	40.4	30.6	0.624	0.001	1.426	6.369	0.047	0.39	0.108	0.001
Q2_B	25	12.52	9.53	0.63	37.7	30.4	0.511	0.002	1.546	6.017	0.003	0.48	0.137	0.002
Q3_T	0.1	12.84	9.45	0.71	41.3	31.5	0.394	0.001	4.525	7.203	0.021	0.49	0.088	0.014
Q3_M	10	14.17	9.47	0.65	47.2	28.9	0.430	0.001	3.945	7.453	0.017	0.55	0.103	0.008
Q3_B	20	12.57	9.46	1.05	37.5	30.6	0.492	0.001	5.315	6.846	0.053	0.42	0.133	0.012
Q4_T	0.1	13.73	9.54	0.68	39.0	30.6	0.433	0.001	4.329	7.796	0.077	0.43	0.105	0.006
Q4_M	11	13.70	9.50	1.37	33.4	26.6	0.432	0.001	4.088	6.520	0.011	0.57	0.122	0.014
Q4_B	22	13.24	9.52	2.55	37.6	28.6	0.440	0.001	6.224	7.351	0.078	0.55	0.101	0.016
Q5_T	0.1	13.13	9.54	1.60	35.8	26.8	0.463	0.001	3.727	7.383	0.003	0.48	0.148	0.007
Q5_M	12	13.83	9.51	0.98	36.2	26.6	0.470	0.001	3.004	7.780	0.340	0.45	0.078	0.006
Q5_B	24	14.35	9.52	1.19	33.4	26.2	0.543	0.001	5.088	6.870	0.074	0.46	0.094	0.006
Q6_T	0.1	14.12	9.50	2.38	35.3	27.8	0.274	0.001	2.907	9.322	0.039	0.29	0.020	0.006
Q6_M	11	12.84	9.50	0.75	34.7	28.1	0.254	0.001	3.624	8.448	0.329	0.75	0.126	0.006
Q6_B	23	13.46	9.50	1.69	33.9	27.0	0.360	0.001	3.080	8.314	0.526	0.58	0.015	0.005
Q7_T	0.1	13.51	9.50	1.38	32.3	27.4	0.216	0.001	5.597	9.331	0.026	0.66	0.012	0.006
Q7_M	10	13.73	9.50	1.08	33.1	26.4	0.271	0.001	2.405	8.988	0.015	0.54	0.001	0.006
Q7_B	20	13.42	9.50	1.44	35.6	27.7	0.214	0.001	3.312	8.970	0.001	0.41	0.010	0.008
Q8_T	0.1	13.51	9.51	0.97	34.3	27.1	0.344	0.001	4.705	8.686	0.055	0.52	0.007	0.007
Q8_M	6	13.56	9.51	1.04	35.6	26.9	0.260	0.001	3.436	9.546	0.053	0.39	0.007	0.006
Q8_B	11	13.78	9.51	0.78	38.1	28.2	0.225	0.001	3.533	9.329	0.058	0.28	0.010	0.004
Q9_T	0.1	13.83	9.50	0.91	35.5	27.6	0.368	0.001	1.993	9.182	0.070	0.39	0.011	0.005
Q9_M	10	13.95	9.50	0.75	36.7	25.3	0.236	0.001	2.349	12.360	0.052	0.48	0.001	0.002
Q9_B	21	13.79	9.50	0.79	37.7	29.0	0.308	0.001	2.932	10.180	0.051	0.45	0.616	0.006
Q10_T	0.1	13.82	9.49	1.86	38.9	32.5	0.252	0.001	2.288	8.855	0.053	0.69	0.022	0.006
Q11_T	0.1	13.09	9.48	1.50	39.9	32.1	0.736	0.001	3.738	9.145	0.054	0.50	0.028	0.006
NTU, DOC, DIC, TN and TP indicate nephelometric turbidity unit, dissolved organic carbon, dissolved inorganic carbon, dissolved nitrogen, and total phosphorus, respectively. T. Top; M. middle; B. bottom.

 | Show Table

DownLoad: CSV

Environmental variables were measured according to the methods used in our previous study (Yang et al., 2021). Briefly, in the field, the salinity and pH of the sampling water were immediately measured with portable water multiparameter device (SANXIN, Shanghai, China). Water turbidity was measured by using a turbidimeter (HANNA instruments/HI98703). Field colorimetric Hach kits were applied to measure the concentrations of ferrous iron, reduced sulfur, ammonium, nitrite, and nitrate. All geochemical samples were stored in ice in the field and during transportation to the laboratory. On arrival in the laboratory, the samples were analyzed immediately. In the laboratory, concentrations of dissolved organic carbon (DOC) were determined on a multi N/C 2100S analyzer (Analytik Jena, Germany). Dissolved inorganic carbon (DIC) concentrations were measured by means of the potentiometric acid titration method (Bradshaw et al., 1981). The content of total phosphorus (TP) and total nitrogen (TN) were analyzed using a published colorimetric method (Neal et al., 2000; Willis et al., 1996). Chlorophyll-a was extracted in 90% acetone overnight and analyzed by using a fluorospectrophotometer (Shimadzu Corp., Japan) (Yang et al., 2018). For aquatic microbial community analyses, about 0.5 L of lake water samples were filtered through 0.22 μm pore-size polycarbonate filters (Whatman, UK), and then the resulting biomass-containing membrane filters were aseptically collected into 1.5 mL sterilized tubes. All biomass-containing membrane filters were stored in dry ice in the field and then were transferred to a -80 ºC freezer in the laboratory until further processing.

1.2 DNA Extraction and Illumina Sequencing

Genomic DNA was extracted from the biomass-containing filters by using the Fast DNA SPIN Kit (MP Biomedical, USA) following manufacturer's instructions. The 16S and 18S rRNA genes were amplified using universal primer sets of 515F/806R and EK-82F/EK1520R for prokaryotic and microeukaryotic communities, respectively, following previously published procedures (Tamaki et al., 2011; López-García et al., 2001). Briefly, unique 12 bp barcodes were attached between the sequencing adapter and the reverse primes (806R and EK 1520R) to allow multiplexing of the samples. Triplicate PCR products were purified using a DNA Gel Extraction Kit (Axygen, USA) and then normalized in DNA amounts. Qualified samples were sent for sequencing on an Illumina Hiseq 2500 platform (paired‐ends sequencing of 2 × 250 bp) (Caporaso et al., 2012).

1.3 Processing of 16S/18S Sequence Data

Raw 16S rRNA and 18S rRNA gene sequences were firstly processed using the software QIIME2 (version 2019.7) (Bolyen et al., 2019) following the recommended tutorials. Major steps were: the forward and reverse reads were merged and assigned to samples based on barcodes. Sequence quality control (i.e., chimeras checking, sequence filtering, dereplication, and denoising) and feature table generation were achieved using the DADA2 method in QIIME2. In the resulting feature table, the effective sequences were grouped into amplicon sequence variants (ASVs) at a 100% sequence identity level (Callahan et al., 2017). Representative ASVs were then taxonomically determined at a confidence threshold of 0.8 against the database of the Ribosomal Database Program (RDP) v.11.5 and UNITE V.7.1 (http://unite.ut.ee) for 16S rRNA and 18S rRNA genes, respectively. ASVs belonging to the chloroplast, chlorophyte, mitochondrial and those unassigned to prokaryotes and microeukaryotes were filtered from the feature tables. Singletons in each sample and the ASVs with total reads less than 10 in all 29 samples were removed from the feature tables. The obtained feature tables were finally rarified to equal sequence numbers (n = 13 189 and 4 503 for 16S and 18S rRNA genes, respectively) before downstream analyses.

1.4 Statistical Analyses

Statistical analyses of ecological data were conducted using various packages in the R program (version 4.1.1) unless otherwise indicated. Spearman's rank correlations were run to assess correlation for each pair of environmental variables using the "Hmisc" package. Alpha diversity indices including richness (i.e., observed ASVs), Simpson, Shannon, and Equitability indices were calculated with the use of R package "vegan". Linear regression was employed to assess the correlation between the Shannon index of prokaryotic/microeukaryotic communities and the measured environmental variables by using the "vegan" and "ggplot2" packages. Furthermore, Spearman correlations analysis was performed to discern significant (P < 0.05) correlations between the relative abundance of prokaryotic/microeukaryotic phyla/classes and measured environmental variables with the use of the R package "Hmisc".

A redundancy analysis (RDA) was selected to analyze the relationships between whole prokaryotic/microeukaryotic community compositions and environmental variables in the studied water samples. Before RDA, all the measured environmental variables were normalized to values ranging between 1 and 100 in the "vegan" package as described previously (Larsen et al., 2012). Because the detrended correspondence analyses (DCA) indicated that the longest gradient lengths were between 3 and 4 for prokaryotic and microeukaryotic community data, suggesting RDA is suitable for the analysis of this study. In order to avoid collinearity, a forward selection procedure was conducted to choose significant (P < 0.05) environmental variables through the 'ordiR2step' function in the "vegan" package (Oksanen et al., 2020). Only significant (P < 0.05) environmental variables were shown in the RDA ordination.

A null modeling-based statistical framework was used to quantify the relative contributions of deterministic and stochastic processes in structuring the prokaryotic and microeukaryotic communities in the studied lake water. The detailed classification and calculating procedure were described previously (Stegen et al., 2013). Briefly, the deterministic processes contain variable selection and homogeneous selection, whereas stochastic processes contain dispersal limitation, homogenizing dispersal, and non-dominant processes (Liu et al., 2020; Dini-Andreote et al., 2015; Stegen et al., 2015, 2013). The detailed quantification microbial community assembly procedure has been described in our and many previous studies (Jiao et al., 2020; Ning et al., 2020; Yang et al., 2020; Zhou et al., 2014; Stegen et al., 2013).

1.5 Data Availability

All the raw sequences obtained from this study have been deposited at the National Omics Data Encyclopedia (NODE) under the project OEP003218: the 16S rRNA and 18S rRNA gene sequences were under accession numbers of OES137767-OES137795 and OES137738-OES137766, respectively.

2. RESULTS

2.1 Physicochemical Properties of the Studied Saline Lake Water Samples

The basic physicochemical parameters of the studied water samples were summarized in Table 1. Briefly, salinity of the water samples ranged 11.91–14.47 g/L with pH 9.45–9.54; turbidity varied from 0.63 to 2.55 nephelometric turbidity unit (NTU); DOC concentrations were 32.3–47.8 mg/L; DIC ranged 25.3–32.5 mM; the concentrations of ammonium, nitrite, nitrate, TN and TP were 0.21–0.74, 0.001, 0.8–6.2, 6.02–12.36 and 0.001–0.53 mg/L, respectively; the content of chlorophyll-a ranged 0.28–0.75 μg/L; the concentrations of ferrous iron and reduced sulfur were 0–27.5 and 0.001–0.016 mg/L, respectively (Table 1). Spearman correlation analyses showed that some physicochemical variables were correlated with each other (Fig. S2). For example, DOC was significantly (P < 0.05) correlated with DIC; ammonium was significantly (P < 0.05) correlated with TN and ferrous iron; nitrate was significantly (P < 0.05) correlated with reduced sulfur; and TN was significantly (P < 0.05) correlated with ferrous iron (Fig. S2).

2.2 Microbial Diversity and Community Compositions

A total of 831 585 and 283 506 high-quality sequences were obtained for the prokaryotic 16S rRNA and microeukaryotic 18S rRNA genes in this study, respectively. Alpha diversity indices of prokaryotic and microeukaryotic community were summarized in Table S1 and Table S2. The prokaryotic richness ranged from 127 to 282 and Shannon Wiener index was 3.23–4.39 among the studied lake water (Table S1), while the observed microeukaryotic ASVs were 53–112 and Shannon_Wiener index ranged 2.08–3.52 (Table S2). Linear regression analysis indicated Shannon index of prokaryotic community was significantly negatively correlated with pH (R² = 0.23, P < 0.01) (Fig. 1a), and that of microeukaryotic community was positively correlated with turbidity (R² = 0.16, P < 0.05) (Fig. 1b) and chlorophyll-a (R² = 0.24, P < 0.01) (Fig. 1c), respectively. However, the diversity of prokaryotic (Fig. S3a) and microeukaryotic (Fig. S3b) communities did not show significant correlations (P > 0.05) with depths.

Figure  1.  Linear regression analysis showing correlations between the prokaryotic/microeukaryotic Shannon index and environmental variables in the studied Qinghai Lake water samples. (a) Prokaryotic Shannon index and pH; (b) microeukaryotic Shannon index and turbidity; (c) microeukaryotic Shannon index and chlorophyll-a. Red shade indicates 95% confidence.

DownLoad: Full-Size Img PowerPoint

Taxonomic assignment indicated that the prokaryotic 16S rRNA gene sequences obtained from in the studied lake water were mainly affiliated with Acidimicrobiia, Actinobacteria, Alphaproteobacteria, Bacteroidia, Gammaproteobacteria, Oxyphotobacteria, Planctomycetacia, Rhodothermia and Verrucomicrobia (Fig. 2a). The obtained microeukaryotic 18S rRNA gene sequences were mainly affiliated with Cercozoa, Chlorophyta, Ciliophora, Cryptophyta, Dinoflagellata, Fungi, Metazoa, Ochrophyta and Telonemia (Fig. 2b). Across all samples, five most abundant phyla of prokaryotes were Alphaproteobacteria (average 23.7%), Verrucomicrobiae (21.7%), Bacteroidia (18.8%), Gammaproteobacteria (13.0%) and Oxyphotobacteria (10.2%) (Fig. 2a). In the studied lake water, microeukaryotic communities were dominated by Chlorophyta (42.1%), Metazoa (29.3%) and Cercozoa (13.9%) (Fig. 2b).

Figure  2.  Composition variations of prokaryotic (a) and microeukaryotic (b) communities (at the phylum/class level) among all the samples at different water depths in the Qinghai Lake. T. Top; M. middle; B. bottom.

DownLoad: Full-Size Img PowerPoint

Furthermore, Spearman correlation analysis showed that the relative abundances of prokaryotic and microeukaryotic phyla/classes were significantly (P < 0.05) correlated with multiple environmental variables (i.e., depth, pH, turbidity, DOC, DIC, ammonium, TN, chlorophyll-a, ferrous iron and reduced sulfur) (Fig. 3). Interestingly, most of these prokaryotic and microeukaryotic phyla/classes were significantly correlated with nutrients contents, such as TN, ammonium, DIC, DOC and ferrous iron. For example, the relative abundances of Actinobacteria, Gammaproteobacteria, Rhodothermia, Cryptophyta, Dinoflagellata, Fungi and Ochrophyta were positively (P < 0.05) correlated with TN, whereas the relative abundances of Acidimicrobiia, Oxyphotobacteria, Verrucomicrobiae, Chlorophyta and Ciliophora were negatively (P < 0.05) correlated with the content of TN; the relative abundances of Actinobacteria and Gammaproteobacteria were negatively (P < 0.05) correlated with ammonium, DOC and DIC (Fig. 3).

Figure  3.  Spearman correlations between the prokaryotic/microeukaryotic communities and environmental variables in the studied saline lake water samples. ***P < 0.001; **P < 0.01; and *P < 0.05; DOC. dissolved organic carbon; DIC. dissolved inorganic carbon; TN. total nitrogen.

DownLoad: Full-Size Img PowerPoint

2.3 Influence of Environmental Factors on Whole Prokaryotic and Microeukaryotic Community Compositions

The RDA analysis indicated that environmental factors exhibited significant (P < 0.05) influence on the whole prokaryotic and microeukaryotic community compositions (Fig. 4). For example, DIC (R² = 0.14, P < 0.05), TN (R² = 0.06, P < 0.05) and pH (R² = 0.11, P < 0.05) significantly affected the whole prokaryotic community composition (Fig. 4a), whereas ammonium (R² = 0.35, P < 0.05), depth (R² = 0.31, P < 0.05), TN (R² = 0.15, P < 0.01), chlorophyll-a (R² = 0.38, P < 0.05), DIC (R² = 0.23, P < 0.01) and turbidity (R² = 0.27, P < 0.05) showed significant influence on the distribution of microeukaryotic community in the studied saline lake water (Fig. 4b).

Figure  4.  A redundancy analysis (RDA) showing the correlations between prokaryotic (a) microeukaryotic (b) communities and environmental variables in the studied saline lake water, respectively. Values on RDA axis indicate the percentages of total variation explained by each axis. DIC. Dissolved inorganic carbon; TN. total nitrogen.

DownLoad: Full-Size Img PowerPoint

2.4 Ecological Processes Underlying the Assembly of Prokaryotic and Microeukaryotic Communities

Null model analysis revealed the relative contribution of deterministic and stochastic processes in shaping the community compositions of prokaryotes and microeukaryotes (Fig. 5). The prokaryotic community assembly in the studied saline lake water was predominately governed by non-dominant processes, homogeneous selection, variable selection, and dispersal limitation, which explained on average 60.4%, 15.8%, 13.6% and 10.2% of the variations of prokaryotic community compositions, respectively (Fig. 5a). However, the microeukaryotic community assembly in the studied lake water was mainly regulated by non-dominant processes, variable selection, and homogenizing dispersal, which accounted for 68.9%, 23.6% and 6.3% explanation on average for the variations in microeukaryotic community compositions, respectively (Fig. 5b).

Figure  5.  Donut charts showing the relative contributions of deterministic and stochastic processes in the of structuring the prokaryotic (a) and microeukaryotic (b) communities in the studied saline lake water, respectively.

DownLoad: Full-Size Img PowerPoint

3. DISCUSSION

3.1 Relative Contributions of Deterministic and Stochastic Processes in Shaping of the Prokaryotic and Microeukaryotic Community Structures

Stochastic processes (e.g., non-dominant processes and dispersal limitations) appeared to dominate (> 70%) the assembly processes of prokaryotic and microeukaryotic communities, while deterministic processes (e.g., variable selection) played a secondary role (Fig. 5). This dominance of stochastic processes was consistent with many other microbial studies in lakes (Roguet et al., 2015), rivers (Tang et al., 2020), ponds (Chase, 2010), deserts (Caruso et al., 2011) and soils (Hussain et al., 2021). For example, stochastic processes were found to be dominant in governing the assembly of the bacterial communities in wetland sediments of the Hulun Lake Basin (Ma et al., 2022) and microeukaryotic communities in surface water samples of a subtropical river across wet and dry seasons (Chen et al., 2019). In contrast, deterministic processes (environmental filtering) have also been found to play a prominent role in shaping community assembly (Hanashiro et al., 2022; Yang et al., 2021; Ren et al., 2015). For example, the elevated salinity and nutrients had comprehensive effects on the aquatic bacterial community in the aquatic ecosystem, and promoted the importance of deterministic processes in bacterial community assembly in semi-arid Inner Mongolia Plateau (Tang et al., 2021). Such inconsistency can be ascribed to the difference in geochemistry, hydrology and microbiota of the studied ecosystems (Wang B C et al., 2019). Generally, stochastic processes emphasize the influence of dispersal processes or ecological drift on the variation of microbial community structure at different spatial and temporal scales (Zhou and Ning, 2017; Volkov et al., 2003). In this study, water samples were collected from different depths on a small spatial scale in saline Qinghai Lake. So it is reasonable to observe the prominent role of stochastic processes in community assembly in the investigated saline lake water.

Among the stochastic processes, non-dominant processes (e.g., drift) dominated the assembly of prokaryotic and microeukaryotic communities. This finding was in accordance with one previous study, in which ecological drift showed a critical role in shaping the structure of microbial communities (Ramoneda et al., 2020; Zhou et al., 2014, 2013). It is well recognized that communities are more susceptible to drift when environmental selection is weak (Zhou and Ning, 2017; Chase and Myers, 2011). Therefore, it is reasonable that non-dominant processes had strong influence on the assembly of prokaryotic and eukaryotic communities in the studied lake water. Except for the non-dominant processes, distinct stochastic processes affected the prokaryotic and microeukaryotic community compositions, with the former being governed by dispersal limitation (10.2%), while the latter being mainly regulated by homogenizing dispersal (6.3%) (Fig. 5). Dispersal limitation has been widely accepted as one major driver for shaping microeukaryotic community assembly. While dispersal of microeukaryotes is limited since their cell size is usually larger than that of prokaryotes (Zhang et al., 2021; Yang et al., 2020; Powell et al., 2015). In addition, previous studies showed that the influence of homogenizing dispersal on the distribution of microorganisms is mainly caused by their high dispersal rates (Stegen et al., 2013). Therefore, homogenizing dispersal may possess stronger influence on the distribution of prokaryotes than that of microeukaryotes. However, our results in saline lake water showed an opposite scenario. For instance, dispersal limitation plays a more important role in prokaryotes (10.2%) than in microeukaryotes (0.7%), and dispersal rate (homogenizing dispersal) of microeukaryotes (6.3%) is higher than that of prokaryotes (0) (Fig. 5). Indeed, the dispersion of microeukaryotes is still controversial. Some studies reported that microeukaryotes (e.g., protozoa, diatoms, and other microalgae) were more dispersive than prokaryotes (Finlay, 2002), while other studies support the 'size-dispersal' hypothesis, i.e., the distribution of smaller ones (prokaryotes) is less affected by dispersal limitation than that of larger organisms (microeukaryotes) (Liu et al., 2020; Farjalla et al., 2012). So it is not surprising to observe that the homogenizing dispersal may possess stronger influence on microeukaryotes than that on prokaryotes in this study.

Additionally, our results demonstrated that deterministic (environmental filtering) also plays a vital role in shaping the assemblies of prokaryotic and microeukaryotic communities. For example, homogeneous selection (13.6%) and variable selection (15.8%) jointly explained the variation of the prokaryotic communities, and variable selection (23.6%) was the main deterministic processes that accounted for the structure of microeukaryotic communities (Fig. 5). Moreover, prokaryotes and microeukaryotes were significantly (P < 0.05) correlated with various environmental variables (e.g. depth, pH, turbidity, DOC, DIC, ammonium, TN, chlorophyll_a and reduced sulfur), which was evidenced by Spearman correlations and RDA analyses (Figs. 3, 4). This was consistent with previous studies, in which environmental filtering (deterministic processes) dominated in the shaping of the prokaryotic and microeukaryotic communities (Gu et al., 2021; Sadeghi et al., 2021; Xue et al., 2018). It is well known that environmental factors, such as salinity, pH, temperature, depth and nutrient availability shape the microbial community by enhancing ecological selection at the regional and local scales (Yang et al., 2016b; Roguet et al., 2015). Therefore, it is not surprising to observe that the deterministic processes also play important roles in driving variations of prokaryotic and microeukaryotic community compositions in the investigated lake water.

3.2 Environmental Influence on the Prokaryotic and Microeukaryotic Community Structures

Physiochemical variables are one type of factors for variable selection. It is notable that nutrients (e.g., TN, DIC, DOC, ammonium) were the most important variables for environmental filtering of prokaryotic and microeukaryotic communities in saline lake water (Figs. 3, 4). This finding was consistent with previous studies. For example, TN was shown to significantly affect the prokaryotic and fungal community compositions in QTP saline lake sediments (Yang et al., 2020); Carbon (C) and nitrogen (N) jointly shaped benthic microbial communities in Qinghai Lake sediments (Zhang et al., 2022). In addition, nutrients were the dominant environmental filtering force for controlling microbial community compositions in different ecosystems (Frade et al., 2020; Zhou et al., 2017; Koorem et al., 2014; Ma et al., 2004). This can be ascribed to the fact that nutrients (e.g., C sources and N fraction) substantially enhanced the growth and basic metabolic capacity of microorganisms (Liu et al., 2019; Demoling et al., 2008). Although salinity has long been recognized as the most important variable in prokaryotic and microeukaryotic communities (Liu et al., 2020; Yang et al., 2016a; Liu et al., 2013), nutrients can offset salinity constraints to some extent (Yue et al., 2021). Therefore, it is reasonable to observe the strong effects of nutrients on prokaryotic and microeukaryotic communities in the investigated saline lake water.

Furthermore, other measured environmental factors (e.g., pH, chlorophyll-a) can also affect the distribution patterns of prokaryotic and microeukaryotic communities in Qinghai Lake (Fig. 4). For instance, it is not surprising to observe that pH was one of the environmental factors influencing the distribution of prokaryotic community in this study. This finding is consistent with previous studies (Ren et al., 2015; Xiong et al., 2012). In addition, chlorophyll-a content was found to be significantly correlated with microeukaryotic community composition. One previous study indicated that chlorophyll-a content was related to the abundance of photosynthetic algae (Liu et al., 2020). Therefore, the significant influence of chlorophyll-a content on microeukaryotic community composition was expected. In summary, environmental drivers (e.g., nutrients, pH and chlorophyll-a) play important roles in shaping the prokaryotic and microeukaryotic community structures in saline lake water, and their contribution to microbial composition variations cannot be ignored.

4. CONCLUSIONS

Our results demonstrated that stochastic processes played overwhelming roles (> 70%) in driving the assembly of prokaryotic and microeukaryotic communities, with non-dominant processes contributing the most. In addition to non-dominant processes, prokaryotic and microeukaryotic microbial communities were shaped by distinct assembly processes: prokaryotes were mainly governed by homogeneous selection (15.8%), variable selection (13.6%) and dispersal limitation (10.2%), whereas microeukaryotes were strongly driven by variable selection (23.6%) and homogenizing dispersal (6.3%). Nutrient contents (e.g., TN, DIC, DOC and ammonium) were identified as important driving forces of the observed differences in prokaryotic and microeukaryotic diversity and community compositions. Overall, this study provides a better understanding of ecological patterns and assembly mechanisms of prokaryotic and microeukaryotic communities in the water of saline Qing Lake. Further studies are needed to disentangle the temporal influences on microbial assembly mechanisms in saline lake water.