Multi-Scale Environmental Gradients Govern Microbial Succession and Structure Functional Gene Divergence in Element Cycling Along a Desert Lakeshore

Abstract

As a critical aquatic–terrestrial ecological transition zone, the lake littoral zone exhibits steep biogeochemical gradients and plays a vital role in regulating submerged microbial communities and their functions. This study aims to reveal how multi-scale environmental gradients influence microbial succession processes along desert lake littoral zones, as well as the distribution patterns of functional genes involved in carbon (C), nitrogen (N), and sulfur (S) cycling. The results demonstrated that microbial alpha-diversity in the vadose zone exhibited significant individual variability horizontally, while showing pronounced inter-group differences vertically. Horizontally, a distinct functional succession was observed from the shore to the water’s edge, with microbial potential shifting progressively from aerobic oxidative types toward anaerobic reductive types. Vertically, the root-intensive layer fostered more complex co-occurrence networks through enhanced interspecific interactions, suggesting higher functional resilience compared to other layers. Further analysis identified soil moisture as the primary environmental filter driving microbial composition, explaining 27.7% of the variation. Structural equation modeling (SEM) further elucidated that pH and Total Organic Carbon (TOC) were the key regulators of carbon fixation and sulfur oxidation genes, while Total Nitrogen (TN) dominated the distribution patterns of nitrogen cycling genes. These findings deepen the mechanistic understanding of microbial-mediated element cycling in desert lakeshore zones and provide a theoretical basis and data support for maintaining the functions of these fragile ecosystems.

1. Introduction

The lakeshore vadose zone, a transitional area between land and water, features diverse environmental gradients and high habitat heterogeneity [1,2]. Microbes in this zone are key drivers of organic matter dynamics and the biogeochemical cycles of carbon, nitrogen, and sulfur, processes essential for ecosystem balance [3,4,5]. Unlike other settings, the physicochemical properties of lakeshore vadose zones in desert lake littoral zones are significantly influenced by lake water, groundwater, and precipitation. In this environment, frequent fluctuations in groundwater levels directly regulate soil moisture conditions [6], profoundly affecting vegetation composition and microbial metabolic potential [7,8]. Therefore, studying the microbial composition and function of the subaqueous zone in lake littoral zones contributes to a deeper understanding of the mechanisms underlying microbial interactions with carbon, nitrogen, and sulfur cycles in this zone.

Microbial communities are not homogeneously distributed but exhibit significant spatial heterogeneity, with pronounced vertical stratification being a dominant spatial pattern across diverse ecosystems [9,10]. This stratification arises from steep vertical gradients in pH, moisture availability, substrate, and nutrient content [11,12], whose ecological influence often surpasses that of horizontal variation. The intensified evaporation–concentration cycles inherent to sandy environments exacerbate salt sequestration and hydro-fluctuations within the capillary fringe [13], reinforcing vertical stratification as the primary regulator of biogeochemical processes [14]. Nevertheless, despite our current understanding of how vegetation responds to groundwater fluctuations [15], a significant knowledge gap persists regarding the influence of horizontal landscape and vertical profile gradients on the spatial configuration of microbial functional genes.

Against this backdrop, our research focuses on Buzhai Naoer Lake—a typical endorheic lake system situated within the Maowusu sandland, functioning as a closed biogeochemical laboratory. Covering approximately 1 square kilometer, the lake experiences negligible external surface runoff, with its hydrological balance entirely dependent on the equilibrium between local precipitation input and evaporation discharge, constituting a truly closed ecosystem. Our objectives are to: (1) reveal the diversity and compositional dynamics of microbial communities and functional genes along both vertical and horizontal dimensions; (2) identify key functional gene groups associated with C, N, and S cycling and their potential ecological roles; (3) analyze the co-occurrence network structures and stability differences of microbial communities across various habitats; and (4) elucidate the key environmental drivers controlling microbial community assembly and functional gene distribution.

2. Materials and Methods

2.1. Study Area

The study area is situated on the southern edge of the Maowusu sandland in the northern Ordos Plateau (1300–1430 km), characterized by a typical arid to semi-arid climate. This region experiences low and unevenly distributed precipitation (200–400 mm annually, primarily from July to September) and high potential evaporation (1900–2500 mm per year). The mean annual temperature is 6.95 °C. Seasonal soil freezing occurs from November through early May. From early May to late September, the biogeochemical cycles of carbon, nitrogen, and sulfur become particularly active, driven by rising temperatures and plant growth.

Based on lithological features and site conditions, the area was classified into three distinct eco-hydrological zones: a natural vegetation ecosystem, a lakeshore ecosystem, and a farmland ecosystem (Figure 1). These zones exhibit pronounced hydrological gradients. The natural vegetation ecosystem shows low soil moisture (3–25%), a deeper groundwater depth (1.2–1.5 m), and lower total dissolved solids (TDS: 0.4–0.6 g/L). In contrast, the lakeshore ecosystem is defined by high moisture (30–45%), a very shallow groundwater depth (0.5–0.8 m), and moderate TDS (0.6–0.8 g/L). The farmland ecosystem has a moisture of (12–32%), an intermediate groundwater depth (0.8–1.2 m), and the highest TDS levels (0.8–1.0 g/L) (Table S1).

Figure 1. Location of the Study Area.

2.2. Soil Sample Collection and Testing

Soil samples were collected from the lakeshore zone in June 2023. Based on comprehensive topographic conditions, site conditions, and lithological structural characteristics, six horizontal soil monitoring points (S1–S6) were selected. Each monitoring point was then vertically subdivided into four layers—the surface layer, root-intensive layer, deep soil layer, and capillary fringe layer—according to regional water movement patterns (groundwater recharge to lakes), groundwater depth (shallow water table), and soil water infiltration characteristics (predominantly vertical infiltration). Consequently, each monitoring point features four sampling points at different depths.

Soil samples collected at each sampling point were divided into three portions for processing: one fresh soil sample was used for on-site determination of soil moisture content, pH, and salinity, as well as laboratory determination of ammonium nitrogen (NH₃-N) content, nitrate nitrogen (NO₃-N) content, and the activities of β-glucosidase (β-GC), nitrite reductase (NiR), and peroxidase (POD). One soil sample (air-dried) was used to determine total carbon (TC), total organic carbon (TOC), and total nitrogen (TN) contents. A fresh soil sample was stored frozen at −80 °C for subsequent microbial DNA extraction and sequencing analysis.

The laboratory analytical workflow is as follows: Soil moisture content, pH, and salinity were measured on-site using a portable soil indicator tester. The parameters requiring laboratory analysis are as follows: NH₃-N and NO₃-N concentrations are determined using a continuous flow analyzer (San plus, SKALAR, Breda, The Netherlands). TC and TOC contents were determined using the direct measurement method with a TOC analyzer (TOC-L, Shimadzu, Kyoto, Japan). TN content was determined using the sulfuric acid-accelerator digestion Kjeldahl method; The activities of soil enzymes, including β-glucosidase (β-GC), nitrite reductase (NiR), and peroxidase (POD), were determined following the standardized protocols described by Burns (2013) [16]. These methods are widely recognized in Chinese soil science and align with the principles used in the Chinese Ecosystem Research Network (CERN). Specifically, NiR activity was measured using the colorimetric method based on NO₂⁻−N depletion, and β-GC was quantified using p-nitrophenyl-β-D-glucopyranoside as the substrate.

2.3. DNA Extraction and 16S rRNA Amplicon Sequencing

Total genomic DNA was extracted from approximately 0.5 g of soil using the E.Z.N.A.^® Soil DNA Kit (Omega Bio-tek, Inc., Norcross, GA, USA) in a sterile laminar flow hood with filter-tipped pipettes to minimize contamination. The quality and integrity of the extracted DNA were verified by 1% agarose gel electrophoresis. The hypervariable V3–V4 region of the bacterial 16S rRNA gene was amplified by polymerase chain reaction (PCR) using barcoded specific primers and the TaKaRa Premix Taq^® Version 2.0 (TaKaRa Biotechnology Co., Dalian, China). The PCR products were then purified automatically using the Agencourt AMPure XP nucleic acid purification kit (Beckman Coulter, Inc., Brea, CA, USA). Subsequently, the purified amplicons were used to construct sequencing libraries, which were paired-end sequenced (PE250) on an Illumina MiSeq/Novaseq 6000 platform (Illumina, Inc., San Diego, CA, USA).

2.4. High-Throughput qPCR for Functional Gene Quantification

Following DNA extraction (as described in Section 2.3), DNA purity was assessed, and the extracts intended for the gene chip were normalized to a concentration of 20 ng·μL⁻¹ using sterile deionized water and stored at −20 °C for subsequent analysis. Absolute abundances of key functional genes involved in carbon (C), nitrogen (N), and sulfur (S) cycling were quantified using the Wafergen SmartChip Real-Time PCR System (WaferGen Bio-systems, Inc., Fremont, CA, USA) in combination with a high-throughput qPCR microfluidic chip (QME chip). The custom QMEC chip was pre-configured with 63 primer sets targeting key metabolic processes: carbon cycling (including fixation, degradation, methanogenesis, and methane oxidation), nitrogen cycling (including nitrification, organic nitrogen mineralization, nitrogen fixation, and denitrification), and sulfur cycling (including oxidation and reduction).

The qPCR amplification protocol consisted of the following steps: (i) initial enzyme activation at 95 °C for 10 min; (ii) 35 cycles of denaturation at 98 °C for 10 s, annealing at 54 °C for 30 s, and extension at 72 °C for 45 s; and (iii) a final extension at 72 °C for 10 min before cooling to 4 °C for storage. All high-throughput qPCR runs included three technical replicates per sample and negative controls. Only genes that exhibited positive amplification in all three technical replicates were considered for downstream data analysis, and the specificity of the PCR products was confirmed by 2% agarose gel electrophoresis [17]. The raw cycle threshold (Ct) values obtained from the SmartChip system were subjected to quality control and normalization procedures to obtain relative quantities. The absolute abundance of each target gene (copies per gram of dry soil) was then calculated by normalizing its relative quantity to the absolute abundance of the 16S rRNA gene, which was determined separately using a Roche LightCycler system. The absolute quantification of the 16S rRNA gene (in copies per μL) was used as an internal standard to convert the relative gene abundances into absolute copy numbers, using the following formula:

$$\mathrm{Absolute} \mathrm{abundance} \mathrm{of} \mathrm{Target} \mathrm{Gene} = {\displaystyle \frac{\mathrm{b} \times \mathrm{c} \times \mathrm{m}}{\mathrm{a} \times \mathrm{d} \times \mathrm{n}}}$$

where a: Relative copy number of 16S rRNA gene; b: Relative copy number of the target functional gene; c: Absolute quantification of 16S rRNA gene (copies/μL); d: DNA concentration loaded onto the chip (ng/μL); m: Total mass of extracted DNA (μg); n: Mass of the original soil sample (g).

2.5. Data Analysis

Statistical analyses were performed using SPSS 25.0 and R (v3.6.0). One-way ANOVA followed by Tukey’s HSD test was used to identify significant differences (p < 0.05) among spatial zones. Redundancy analysis (RDA) was conducted via Canoco 5 to elucidate relationships between microbial communities and environmental filters. Furthermore, structural equation modeling (SEM) was employed using the lavaan (v0.6-5) and semPlot (v1.1.2) packages in R to evaluate the direct and indirect effects of environmental gradients on element cycling potential.

3. Results

3.1. Distribution Characteristics of Soil Physicochemical Properties

This study revealed significant gradient variations in the physicochemical properties of lakeshore soils both horizontally and vertically through analysis of indicators such as pH, salinity, moisture content, TN, TP, and TOC. The pH ranges from 7.21 to 9.29, indicating an alkaline environment. Salinity ranged from 440 to 1027 mg/kg, while NH₃-N and NO₃-N concentrations varied between 0.91 and 2.54 mg/kg and 1.05 and 11.20 mg/kg, respectively. The contents of TC, TOC, and TN ranged from 1.58 to 12.52 g/kg, from 1.07 to 4.56 g/kg, and from 0.31 to 0.70 g/kg, respectively, Alkaline phosphatase content ranged from 0.032 to 0.064 mg/(g·d), while β-glucosidase content ranged from 0.16 to 10.95 μg/(g·h). Vertically, from the surface zone to the fluctuated zone, both pH and salinity decrease with increasing soil depth, with salinity accumulating most in the surface zone. TC, TOC, and TN contents, as well as soil alkaline phosphatase and β-glucosidase activity, decrease with increasing soil depth. Horizontally, the pH levels on the south bank of the lake are generally higher than those on the north bank of the lake. Soil salinity shows an increasing trend from the lake shores toward the lake center, with particularly pronounced signs of salinization at the lakeshore ecosystem (S3 and S4) (Figure 1). NH₄-N concentrations were lower at points S1 and S6, which were farthest from the lake shore. The highest NO₃-N concentration was observed at corn field S6. Alkaline phosphatase and β-glucosidase activities at the lakeshore ecosystem (S3 and S4) were significantly lower than at other monitoring sites. The spatial variation trends of TP content, nitrite reductase activity, and peroxidase activity were not significant (Figure 2).

Figure 2. Spatial Distribution Characteristics of Soil Physicochemical Properties. TN: Total Nitrogen; TP: Total Phosphorus; TC: Total Carbon; TOC: Total Organic Carbon; POD: Peroxidase; β-GC: β-Glucosidase; AKP: Alkaline Phosphatase; NiR: Nitrite Reductase. B1–B4 represent surface layer, root-intensive layer, deep soil layer, and capillary fringe layer.

3.2. Soil Bacterial Communities and Their Co-Occurrence Networks

3.2.1. Composition and Variation of Soil Bacterial Communities

This study found that dominant bacterial communities at the phylum level were consistent across both vertical and horizontal scales through analysis of soil bacterial community composition. The dominant bacterial communities at the phylum level were primarily Gammaproteobacteria (13%). Alphaproteobacteria (10%), Blastocatellia_Subgroup_4 (7%), Deltaproteobacteria (8%), Acidimicrobia (7%), Subgroup_6 (7%), and Anaerolineae (7%) collectively accounted for 59% of the total relative abundance (Figure 3). Significantly enriched bacterial groups at S3 included the Chloroflexi phylum and δ-Proteobacteria class. At S4, the bacterial community showed significant enrichment of the Blasocatellia (Subgroup_4) phylum (Figure S1). These specific microbial communities reflect the selection of bacterial populations by the ecological environment of different sites (such as nutrients and microhabitats). Through LEfSe multi-level species differential analysis, it was found that the bacterial phylum Bacilli was significantly enriched in the natural vegetation ecosystem S2.

Figure 3. Changes in dominant genera and alpha diversity along the lake shoreline in vertical and horizontal directions: (a) Horizontal dominant genera; (b) Vertical dominant genera; (c) Horizontal alpha diversity; (d) Lateral alpha diversity. (* p < 0.05; ** p < 0.01).

3.2.2. Soil Bacterial Symbiotic Networks

To understand differences in soil microbial network interactions among distinct microbial communities, symbiotic networks between bacterial species were constructed using OTUs under different microbial communities (Figure 4). Based on calculations of bacterial network topology, the surface zone comprises 976 nodes and 8770 edges, the root-intensive layer contains 1220 nodes and 25,277 edges, deep soil layer includes 937 nodes and 12,059 edges, and the fluctuated zone consists of 689 nodes and 7245 edges (Table S2). The number of nodes and edges in the root-intensive layer was significantly higher than in other soil zones, indicating that the soil environment in the root-intensive layer is conducive to forming more complex symbiotic networks, with potentially stronger functional interactions among bacteria.

Figure 4. Co-occurrence networks of bacterial communities at different depths in the vadose zone: (a) Surface layer; (b) Root-intensive layer; (c) Deep soil layer; (d) Capillary fringe layer; (e) Natural vegetation ecosystem; (f) Lakeshore ecosystem; (g) Farmland ecosystem; Red lines indicate positive correlations, green lines indicate negative correlations, and nodes represent operational taxonomic units (OTUs). Different modules are represented by distinct colors.

3.3. Key Functional Genes Involved in Carbon, Nitrogen, and Sulfur Cycles

To clarify the potential functions of soil bacteria in the study area, HeatMap analysis revealed significant differences in microbial functions at both vertical and horizontal scales (Figure 5). At the horizontal scale, potential functional groups in natural vegetation and farmland ecosystems predominantly exhibited oxidative traits, while lakeshore ecosystem showed predominantly reductive traits. Within natural vegetation ecosystem, functional groups such as methanol oxidation, aromatic compound degradation, aerobic nitrite oxidation, and nitrification exhibited higher relative abundances. In farmland ecosystem, potential functional groups such as xylanolysis, nitrogen fixation, aerobic chemoheterotrophy, and chemoheterotrophy exhibited higher relative abundances. In lakeshore ecosystem, dominant functional groups primarily concentrate on reductive functions within nitrogen and sulfur cycles, dominated by sulfate respiration, iron respiration, nitrate respiration, nitrogen respiration, respiration of sulfur compounds, and fermentation (Figure 5a). Vertically, from the surface zone to the fluctuated zone, the potential functional community shifts from oxidative to reductive. In the surface layer, the dominant functions include aerobic respiration with oxidative capacity, urea hydrolysis, phototrophic nutrition, and methanotrophic nutrition. In the root-intensive layer, the potential functional group is dominated by intracellular parasitism, xylan hydrolysis, urea hydrolysis, and nitrate reduction. In the deep soil layer, the potential functional group is dominated by sulfide oxidation, sulfur compound oxidation, iron respiration, fermentation, aerobic ammonium oxidation, and nitrification. In the Capillary fringe layer. the potential functional community is dominated by nitrate respiration, nitrogen respiration, sulfate respiration, and nitrate reduction (Figure 5b).

Figure 5. Functional gene prediction under different partitioning schemes. (a) Horizontal direction; (b) Vertical direction.

Based on the predicted microbial functional genes, the primary elements in the study area are carbon, nitrogen, and sulfur, with no other pollutants or heavy metals detected.

Therefore, this study employed gene chip technology to amplify functional genes primarily related to carbon, nitrogen, and sulfur, identifying 24 functional genes associated with carbon, nitrogen, and sulfur cycling (Figure S2).

3.3.1. C Cycling

Measurement and statistical analysis of soil carbon cycling genes identified four principal functional categories: carbon fixation, carbon degradation, methanogenesis, and methane oxidation (Figure 6a). Genes associated with carbon fixation demonstrated the highest relative abundance (79.9%), substantially surpassing those for carbon degradation (19.9%), methanogenesis (3.1%), and methane oxidation (0.9%). Spatial analysis revealed distinct distribution patterns: Horizontally, carbon fixation genes were predominantly enriched at sites S1 and S6. The potential for carbon degradation decreased in the order: S5 > S4 > S3 > S6 > S2 > S1. Methanogenesis genes were primarily localized to the lakeshore sites (S3 and S4), whereas methane oxidation genes were most prevalent at the upland sites S1 and S2, following the sequence: S1 > S2 > S5 > S6 > S3 > S4. Vertically, the abundances of carbon fixation and carbon degradation genes exhibited a slight but consistent decrease with increasing soil depth. In contrast, the genetic potential for methane cycling (methanogenesis and methane oxidation) displayed a non-linear trajectory, initially decreasing and then increasing with depth. Notably, the highest gene abundance for methanogenesis was recorded in the fluctuated zone (Figure 6b). The dominant methane-cycling process varied by zone: the surface layer was characterized by methanogenesis, the deep soil layer and capillary fringe layer were dominated by methane oxidation, while the root-intensive layer maintained a relatively balanced ratio between methanogenesis and methane oxidation genes.

Figure 6. Quantitative profiles of the metabolic potential for carbon (C) and nitrogen (N) cycling. Panels (a,b) depict the abundance of key functional genes related to C cycling, while panels (c,d) illustrate the abundance of key functional genes associated with N cycling.

3.3.2. N Cycling

Through detection and statistical analysis of genes involved in the nitrogen cycle within the vadose zone, it was determined that seven major gene categories participate in nitrogen cycling: ammonification, nitrogen fixation, mineralization, aerobic ammonium oxidation, nitrification, anaerobic ammonium oxidation, and denitrification (Figure 6c). Among these, ammonification and denitrification genes showed the highest potential (65%), followed by denitrification gene abundance (18%). The study area primarily exhibited anaerobic ammonium oxidation, denitrification, nitrogen fixation, nitrification, and nitrate reduction. Horizontally, nitrogen fixation primarily occurs at S1, S3, and S5; nitrification mainly takes place at S1, S2, S5, and S6; nitrate reduction occurs predominantly at S1, S5, and S6; anaerobic ammonium oxidation occurs mainly at S1 and S6; and denitrification primarily occurs at S3 and S4. Vertically, the potential for organic nitrogen mineralization, aerobic ammonium oxidation, nitrification, and denitrification decreases with increasing depth, following the pattern: surface layer > root-intensive layer > deep soil layer > capillary fringe layer. In the fluctuated zone, denitrification dominated the potential nitrogen cycling function, accounting for 48.04%, followed by ammonification at 35.36% (Figure 6d).

3.3.3. S Cycling

Through detection and statistical analysis of nitrogen cycle genes in the vadose zone, it is evident that two main categories of genes participate in the sulfur cycle: sulfur oxidation and sulfur reduction (Figure 7a). Among these, the sulfur oxidation functional genes include dsrA, dsrB, and apsA; the sulfur reduction functional genes include soxY and yedZ. Horizontally, sulfur reduction functional genes exhibited higher abundance at S3 and S4, while sulfur oxidation functional genes showed the highest abundance at S5. Vertically, the trend of sulfur oxidation activity was: surface layer > root-intensive layer > capillary fringe layer > deep soil layer, with the surface zone accounting for as much as 88.84% of sulfur oxidation functional genes. Sulfur reduction activity decreased in the following order: deep soil layer > capillary fringe layer > root-intensive layer > surface zone. Among these, functional genes in the root-intensive layer, deep soil layer, and capillary fringe layer primarily focused on sulfur reduction (Figure 7b).

Figure 7. Quantitative profiles of the metabolic potential for sulfur (S) cycling. Panel (a) shows the relative abundance of key S-cycling genes, while panel (b) depicts their spatial distribution across the soil profile.

4. Discussion

4.1. Factors Influencing Soil Bacterial Community Composition and Co-Occurrence Networks

Redundancy analysis (RDA) revealed that the first axis (RDA1) explained 53.86% of the total variation, and the second axis (RDA2) explained an additional 22.89%. Collectively, these two axes accounted for 76.75% of the variability in microbial community composition (Figure 8). Several environmental factors were significantly correlated with microbial community structure, including soil moisture, depth, total nitrogen (TN), nitrite reductase (NiR) activity, total carbon (TC), salinity, and β-glucosidase (β-GC) activity. Among these, soil moisture was the predominant factor, independently explaining 27.7% of the community variation, followed by soil depth (7.3%). Notably, the activities of soil enzymes (such as NiR and β-GC), total phosphorus (TP), and TN generally exhibited positive effects on the abundance of most bacterial taxa. In contrast, factors including salinity, total organic carbon (TOC), nitrate nitrogen (NO₃⁻-N), soil moisture, and pH showed significant negative effects on the microbial communities.

Figure 8. RDA results. TN: Total Nitrogen; TP: Total Phosphorus; TC: Total Carbon; TOC: Total Organic Carbon; POD: Peroxidase; β-GC: β-Glucosidase; AKP: Alkaline Phosphatase; NiR: Nitrite Reductase.

The dominant bacterial phyla identified in this study, such as Gammaproteobacteria and Alphaproteobacteria, are commonly found in soils of arid regions, including grasslands, steppes, and pastures [18]. Our observations resonate with broader surveys of arid ecosystems, where Proteobacteria frequently predominate owing to their inherent metabolic plasticity. Nevertheless, unlike typical desert edaphic environments where Actinobacteria often serve as co-dominant taxa, the heightened relative abundance of Proteobacteria within the Maowusu sandland lakeshore may be ascribed to the pronounced hydraulic permeability of the sandy matrix. This physical characteristic streamlines the flux of both nutrients and moisture, effectively distinguishing these porous lakeshore interfaces from more consolidated desert soils [19]. This study revealed that the α-diversity of vadose zone microorganisms exhibited significant individual variability horizontally and notable inter-group differences vertically. Horizontally, this pattern may be attributed to variations in soil texture, moisture, and temperature under different vegetation types, which alter local microenvironments and subsequently shape microbial community diversity and composition through local adaptation. For instance, the significantly lower microbial diversity at site S2 corresponds well with its lower TP and TOC contents. Vertically, the strong physicochemical gradient formed with increasing soil depth significantly influences microbial community structure and diversity [20,21]. The α-diversity followed the order: root-intensive layer > surface layer > deep soil layer > capillary fringe layer (Figure 3). This vertical pattern is not only closely associated with the decreasing nutrient content along the soil profile but may also be linked to rhizodeposits [22,23]. As demonstrated by Bi et al. (2021) in their study on the metabolism of five typical desert-adapted plants in the Mu Us Desert, plant-mediated metabolism serves as a primary factor influencing microbial α- and β-diversity [24]. The organic compounds in rhizodeposits—such as sugars, amino acids, and vitamins—provide abundant nutritional resources that stimulate microbial growth and reproduction.

We detected that the root-intensive layer supported a co-occurrence network with significantly greater node number, edge number, and average connectivity compared to the fluctuated zone (Figure 4 and Table S2). This pattern can be attributed to two primary mechanisms: First, the microbial community in the typically undergoes long-term evolutionary adaptation, fostering stable ecological relationships. In contrast, the microbial community in the fluctuated zone must constantly adapt to changing environmental conditions, which likely suppresses the formation of a complex cooperative network. Previous research has demonstrated that fluctuations in moisture and temperature can negatively impact microbial growth [25], a finding consistent with our results identifying soil moisture as a key driver of microbial composition (Figure 8). Second, network connectivity appears intrinsically linked to microbial diversity. We observed that the pattern of microbial co-occurrence network complexity paralleled the trend in microbial diversity. Furthermore, existing studies confirm a significant positive correlation between microbial diversity and network connectivity (links) [17,26]. Greater connectivity generally indicates a community with more diverse ecological functions and enhanced stability [26].

4.2. Drivers of the Potential of Soil C, N, and S Cycling Functional Genes

Path analysis of the carbon cycle via structural equation modeling (SEM) revealed that soil depth indirectly influenced carbon cycling by affecting soil enzyme activities and microbial community composition through physicochemical properties (pH, p = −0.122) and nutrient contents (TOC, p = −0.436; TN, p = −0.268). Soil depth influenced β-glucosidase activity through two pathways: (1) a direct negative effect (p = −0.615), and (2) an indirect negative effect mediated by its influence on pH, salinity, TN, TOC, peroxidase activity, and nitrite reductase activity. Furthermore, soil depth indirectly regulated the potential of both carbon fixation and carbon degradation via its effects on pH, salinity, and TOC, exerting an overall negative influence. Among these intermediary factors, pH exhibited a significant positive effect (p < 0.05) on the carbon fixation potential, whereas TOC showed a significant negative effect (p < 0.05; Figure 9a).

Figure 9. Structural Equation Models. (a) Carbon Cycle Structural Equation Model; (b) Nitrogen Cycle Structural Equation Model; (c) Sulfur Cycle Structural Equation Model. * p < 0.05; ** p < 0.01; *** p < 0.001.

For the nitrogen cycle, the SEM adequately represented three key nitrogen transformation processes: nitrogen fixation, mineralization, and aerobic ammonia oxidation. Soil depth was found to indirectly affect the potential for nitrogen mineralization and aerobic ammonia oxidation primarily by regulating the soil TN content. Conversely, the potential for nitrogen fixation was indirectly influenced by soil depth through a more complex pathway involving TOC, TN, peroxidase, pH, aerobic ammonia oxidation, and nitrite reductase (Figure 9b).

Regarding the sulfur cycle, the model incorporated sulfur oxidation and sulfur reduction. Sulfur oxidation potential was affected directly by soil depth and indirectly through the mediation of pH and TOC. Sulfur reduction potential was primarily controlled by soil depth and pH. Specifically, soil depth exerted indirect control over sulfur reduction by influencing pH (Figure 9c).

The cycling of carbon (C), nitrogen (N), and sulfur (S) in the vadose zone involves multiple critical biogeochemical processes [27], which are governed and driven by relevant functional genes. This study detected a total of 10 C-cycling, 10 N-cycling, and 4 S-cycling functional genes. Horizontally, the gene abundances for methanogenesis, denitrification, and sulfur reduction were higher at the lakeside monitoring sites S3 and S4. In contrast, genes associated with carbon fixation, carbon degradation, nitrification, and sulfur oxidation were predominantly more abundant at sites farther from the lake. These findings align with the predicted microbial metabolic potential a pattern potentially linked to the higher nutrient levels near the lake shoreline. Previous studies have indicated that elevated concentrations of soil nitrogen (e.g., TN, NH₄⁺-N) and carbon (e.g., TC, TOC) can increase the abundance of key functional genes, such as amoA (AOB) for nitrification and nirK/nosZ for denitrification [28,29]. Consistently, our results also showed significant correlations between TN, TOC, and the genes involved in ammonification and denitrification (Figure S3).

Interestingly, analysis of both the predicted microbial functional groups and the quantified C-, N-, and S-cycling genes revealed a clear vertical stratification. The surface zone was dominated by oxidative functional types and genes primarily involved in carbon fixation, ammonification, nitrification, and sulfur oxidation. In stark contrast, the fluctuated zone was characterized by reductive functional potentials, such as nitrate respiration, nitrogen respiration, sulfate respiration, and nitrate reduction, along with a high abundance of genes for methanogenesis, denitrification, and sulfur reduction (Figure 4b). This observed shift from oxidative processes at the surface to reductive processes at depth mirrors findings reported by Zhang et al. (2023) [30]. This stratification is primarily driven by environmental gradients: the surface zone provides ample oxygen, suitable temperatures, and sufficient nutrients, favoring the activity of microbes and genes responsible for methane oxidation, nitrification, and sulfur oxidation. Organic matter, a key substrate for methanogenesis, influences the abundance and activity of methanogens depending on its decomposability and availability. With increasing soil depth, oxygen and nutrient contents decline while moisture increases. Particularly in the fluctuated zone, oxygen and nutrients are minimal, and moisture approaches saturation. As soil moisture conditions can alter resource availability and thereby influence soil microbial biomass and composition—a finding supported by our study identifying soil moisture as a key driver of microbial composition (Figure 8)—the microbial community in this deep zone becomes predominantly dominated by reductive metabolisms.

This study indicates that pH and TOC are key factors influencing carbon fixation genes, but other factors beyond pH and TOC also drive carbon fixation genes. Furthermore, in addition to soil physicochemical properties, climatic variations such as precipitation and temperature significantly impact carbon fixation potential [31,32], an aspect not addressed in our research.

This study demonstrates that soil depth can indirectly influence soil nitrogen mineralization potential and aerobic ammonium oxidation potential by regulating soil total nitrogen (TN) content. This is likely because the soil surface zone, influenced by the accumulation of animal manure and plant litter, exhibits higher TN content. This provides ample mineralization substrates, exerting a positive effect on nitrogen mineralization genes, leading to increased NH₄⁺ concentrations in the soil. This, in turn, enhances the expression of ammonium oxidation-related genes (e.g., hao and nxr genes), thereby enhancing the potential for aerobic ammonium oxidation [33]. With increasing depth, TN content progressively decreases, correspondingly reducing substrates for nitrogen mineralization and aerobic ammonium oxidation, which in turn inhibits related reactions. This study also revealed that soil depth indirectly influences nitrogen fixation potential by regulating multiple factors including TOC, TN, peroxidase activity, pH, and nitrite reductase activity. This is primarily because soil pH significantly affects the activity of nitrogen-fixing microorganisms [34], with extreme acidic or alkaline environments potentially inhibiting their growth and nitrogen fixation processes. TOC and TN serve as primary energy sources for nitrogen-fixing microorganisms. By providing sufficient energy, they can enhance microbial activity and thereby increase soil nitrogen fixation potential [35]. Regarding functional gene differentiation, we validated the regulatory effects of factors such as TN and TOC on genes like amoA and nirK/nosZ in this study against findings from Ding et al. (2022) and Ke and Lu (2012) [29,36]. The results indicate that the driving mechanisms revealed in this study largely align with patterns observed in other arid systems globally. However, within the specific context of desert lakeshores, water content (explaining 27.7% of variance) plays a more critical and decisive role.

This study indicates that pH and TOC significantly influence sulfur-oxidizing bacteria in the vadose zone, consistent with previous research findings [37,38]. Zhen et al. (2019) investigated sulfur-oxidizing bacteria in Tibetan hot spring sediments by constructing a soxB gene clone library, revealing that these bacteria exhibit broad pH adaptability [39]. Elevated soil pH accelerates soil sulfur oxidation and organic sulfur mineralization rates, directly benefiting plant uptake of generated sulfate. Beyond pH, TOC exerts a significant positive effect on sulfur-oxidizing genes, likely because high TOC typically correlates with improved soil structure and enhanced water-holding capacity. Such environments optimize soil aeration and moisture conditions, promoting sulfur-oxidizing bacterial growth and metabolism. Furthermore, this study revealed that sulfur-oxidizing functional genes exert a significant positive effect on β-glucosidase activity, suggesting sulfur oxidation may enhance carbon degradation.

4.3. Research Limitations

Despite the small SEM sample size (n = 24) in this study, the results remain credible, consistent with the methodological consensus on small-sample SEM: Hoyle et al. (2015) indicate that SEM with n < 50 can yield reliable results when the model is parsimonious and each latent variable has ≥3 reliable indicators [40]. All latent variables in this study correspond to 34 highly loaded indicators (factor loadings > 0.65), and the number of parameters is only half the sample size, meeting this criterion. Furthermore, the empirical correction statistics proposed by Yuan et al. (2017) for small samples (n < 2p) effectively control Type I error rates [41]. This study employed robust ML estimation and chi-square value correction, achieving acceptable model fit indices (CFI = 0.96, RMSEA = 0.068), further validating the reliability of the results. Simulation studies by Rosseel et al. (2024) also confirm that parameter estimation accuracy using fit-based estimation methods (e.g., bounded ML, robust ML) in small samples can meet research requirements, consistent with the methodological choices and performance of this study [42].

5. Conclusions

This study conducted a multi-scale investigation of the lakeshore zone at Buzhainaoer Lake in the Mu Us Desert, elucidating the spatial distribution patterns of microbial communities and their functional genes under the combined effects of horizontal landscape gradients and vertical physicochemical gradients. It revealed how soil moisture and nutrient gradients drive the succession of microbial functions through filtration processes. Although based on sampling at specific spatio-temporal points, this study reveals the connection between micro-rhizosphere interactions and macro-environmental gradients. These findings provide data support for assessing the potential ecological sensitivity of this fragile region under future environmental changes.