Response of Soil Organic Carbon and Microbial Metabolic Pathways in Guangxi Karst Regions to Different Vegetation Types

Abstract

This study investigates how different vegetation types influence the molecular structure and abundance of soil organic carbon (SOC), as well as their influence on microbial metabolic pathways and community composition. Soil samples were collected from four different sites: a woodland dominated by Drypetes perreticulata (DP), a woodland dominated by Horsfieldia hainanensis (HM), a Zea mays L. field (ZL), and a citrus reticulata orchard (CB). The molecular structure of soil organic carbon (SOC) was characterised using Fourier Transform Infrared (FTIR) spectroscopy, identifying aromatic carbon (ArC), polysaccharide carbon (PSC), alkyl carbon (AlkC), amine carbon (AmC), ether carbon (EtC), and olefin carbon (OleC). Our results indicated significant variations across vegetation types: DG exhibited a significantly higher ArC content, while maize fields showed lower PSC levels. To analyse the relationships between different samples, we employed principal component analysis (PCA), which revealed distinct organic carbon structures across vegetation types, with the forests (DG and HM) significantly differing from agricultural sites (ZL and CB). Additionally, the 16S V3_V4 region of soil bacteria was sequenced using high-throughput sequencing. We employed PICRUSt2 to predict microbial metabolic pathways, revealing consistent core metabolic functions across samples but significant variations in secondary metabolism, with HM samples exhibiting the most distinctive metabolic profiles. Redundancy analysis (RDA) further demonstrated that microbial metabolic pathway variation explained 55.66% of organic carbon structure variance. Key microbial taxa exhibited significant associations with specific carbon source types and functional pathways. These findings highlight the pivotal mechanisms by which different vegetation types regulate soil organic carbon structure and composition by driving changes in microbial metabolic traits and community assembly. This study provides a mechanistic basis for understanding the coupling between vegetation, microorganisms, and carbon cycling, offering significant guidance for optimising vegetation restoration strategies, enhancing soil carbon sequestration capacity, and advancing carbon management practices based on microbial regulation.

1. Introduction

Soil organic carbon (SOC) plays a pivotal role in sustaining soil fertility and carbon storage functions [1]. Vegetation type has been identified as a key driver of soil organic carbon sequestration, primarily by altering the quantity and quality of organic inputs through litterfall chemical composition and root exudation patterns [2]. However, quantitative research remains scarce on how vegetation-induced changes in microbial community composition and metabolic functions subsequently regulate the molecular structure of soil organic carbon. This knowledge gap is particularly pronounced in karst ecosystems, where a high calcium content and shallow soil profiles create a unique environment for organic–mineral interactions [3]. Preliminary studies by Chen et al. and Wei et al. indicate substantial variations in soil organic carbon stability across different land-use types, yet the underlying microbial regulatory mechanisms require further elucidation [4,5]. Consequently, understanding how specific microbial groups and functional pathways influence soil organic carbon molecular composition under diverse vegetation types is particularly crucial for predicting outcomes in karst regions with their distinctive ecosystems.

Changes in different vegetation types (such as forests, grasslands, and scrublands) can directly influence the material composition of plant litter inputs [6] shaping unique soil microbial community structures and functional pathways [7], and regulating the accumulation of derived organic compounds [8,9]. For instance, broad-leaved trees have been demonstrated to enhance the stability of soil SOC aggregates compared with coniferous species, primarily linked to differences in leaf decomposition and litterfall volume [10]. Typically, stronger litter decomposition correlates with higher soil carbon sequestration rates [11]. Compared with agricultural land, forests can enhance slow-cycling carbon and sequester it in deeper soils by extending root systems into these layers [12]. Conversely, frequent tillage and monoculture cropping in agricultural fields typically lead to declining SOC levels [13]. Deng et al. found that the process of restoring vegetation to farmland significantly impacts the recovery of soil carbon pools [14]. This vegetation restoration can promote soil organic carbon accumulation through biomass input, thereby contributing to the stability of soil aggregates [15,16]. Beyond vegetation effects, both natural and anthropogenic disturbances significantly alter the quantity and quality of soil organic carbon. Noppol Arunrat et al. [17] contend that fires not only reduce total soil organic carbon but also modify its molecular composition, frequently leading to the formation of highly aromatic and persistent pyrolytic carbon. Understanding the effects of vegetation restoration types and environmental factors on SOC sequestration enables more accurate predictions of carbon dynamics [18]. Consequently, in ecologically fragile karst regions, the influence of vegetation types on organic carbon structure is paramount. Further research is urgently needed to elucidate the molecular structural alterations in organic carbon induced by different vegetation types.

Soil organic carbon (SOC) pools constitute fundamental drivers of ecosystem function, playing a pivotal role in maintaining soil fertility, enhancing water retention capacity, improving soil aggregate stability, and mitigating atmospheric carbon dioxide concentrations [19,20,21]. Soil microorganisms, as the primary mediators of soil organic matter (SOM) transformation, directly regulate its composition and stability through metabolic activities [22,23]. These microbial communities exhibit rapid responsiveness to environmental conditions—including nutrient availability, moisture, temperature, and the quantity and quality of organic matter—during nutrient acquisition and energy cycling processes [24]. Consequently, soil organic matter can be regarded as being formed through the partial decomposition of plant aboveground and belowground biomass by microorganisms [25]. Different vegetation types influence microbial community structure [26] and the abundance of functional genes associated with soil carbon cycling (e.g., genes related to carbon degradation, carbon fixation, and methane metabolism) [27,28,29] by altering the quality and quantity of litterfall and root exudates, as well as differences in soil properties. This abundance is closely linked to soil chemical properties and enzyme activity, thereby leading to variations in carbon and nitrogen cycling [30,31]. However, the precise mechanisms by which vegetation-shaped microbial communities modulate soil organic carbon molecular structure through functional gene regulation remain unclear. Our recent study indicated that woodland soil exhibits a higher richness of microbial diversity compared with farmland soils. There were significant differences between forest and farmland soil’s microbial community composition [32]. However, how microbial communities regulated by different vegetation types modulate soil organic carbon structure formation through changes in associated functional genes remains unknown.

This study aims to systematically investigate how vegetation types influence the formation and transformation of soil organic carbon molecular structures by regulating microbial functional pathways. We propose the following core research hypotheses: (1) vegetation types drive the structural composition diversity of soil organic carbon by altering litter composition; (2) vegetation-induced changes in soil microbial community structure further shape organic carbon compositional diversity through differences in their metabolic functional genes. To test these hypotheses, this study integrates Fourier Transform Infrared Spectroscopy, 16S rRNA high-throughput sequencing, and PICRUSt2 functional prediction. Three specific research objectives are established: first, to quantify the differences in key functional group compositions of soil organic carbon under different vegetation types; second, to reveal how microbial community structure and functional pathways respond to vegetation type; third, to establish links between microbial metabolic functions and organic carbon molecular structures, thereby elucidating the coupling mechanisms within the vegetation–microbe–carbon cycle.

2. Materials and Methods

2.1. Study Area

This study was conducted in Nonggang National Nature Reserve, Guangxi, China (longitude 106°42′28″–107°04′54″ E, latitude 22°13′56″–22°39′09″ N) (Figure S1). The altitude of the sample area ranged from 215 to 370 m, and the slope ranged from 15 to 45°. The sample area is located south of the Tropic of Cancer, with strong solar radiation and a high temperature, which belongs to the tropical monsoon climate. The average annual temperature was 22 °C, with an absolute maximum temperature of 39 °C and an absolute minimum temperature of −3 °C. The average annual rainfall was 1150–1550 mm, and the annual evaporation was 1344–1748 mm. The area is characterised by a typical karst landscape. The region exhibits a distinct pattern of alternating wet and dry seasons, with an annual frost-free period of approximately 350 days and a frost period lasting 13 days. The soil of the sample site is limestone soil. Detailed site information is found in a previous study [32]. Four representative vegetation types were selected for the experiment, including Horsfieldia hainanensis Merr. (HM) and Drypetes perreticulata Gagnep. (DG) as the main species. Both tree species have extremely limited distributions, occurring only in Guangxi and a few other regions, making them characteristic tree species of karst landscapes. Additionally, two farmland plots were selected adjacent to the natural forest, planted with Zea mays L. (ZL) and citrus (CB). Both of these areas were previously natural forest and were cut down for clearing. The maize arable land has been under continuous cultivation for 20 years, while the citrus groves have been under continuous cultivation for 10 years, and both areas are regularly fertilised during routine management.

2.2. Sample Plot Design

This study was conducted in October 2022 within the Guangxi Nonggang Nature Reserve, selecting four vegetation types—HM, DG, ZL, and CB—as research subjects. Within each vegetation type, four 10 m × 10 m plots were established, spaced 25 metres apart. Soil samples were collected from the 0–10 cm layer using an 8 cm diameter soil sampler following the removal of surface litter via the “S”-shaped sampling method. Four replicate samples were taken from each plot, pooled to form a composite sample. A total of 16 soil samples were collected across the four vegetation types. Plots representing different vegetation types did not overlap spatially but were situated within the same protected area, exhibiting a degree of habitat association.

2.3. Soil Sample Processing

Stones and plant residues (such as roots and dead branches and leaves) were removed from the soil, after which the soil sample was sieved through a 2-millimetre mesh to eliminate coarse particles. One part was stored in a refrigerator at −80 °C for extraction of total soil DNA, and the other placed in a well-ventilated, cool, dry room to air-dry for determining soil chemical properties.

2.4. Determination of Soil Organic Carbon Structure

The molecular structure of soil organic carbon (SOC) was characterised using Fourier Transform Infrared Spectroscopy (FTIR), identifying aromatic carbon (ArC), polysaccharide carbon (PSC), alkyl carbon (AlkC), amine carbon (AmC), ether carbon (EtC), and olefin carbon (OleC). Soil organic carbon structure was determined using Fourier Transform Infrared Spectroscopy (FTIR) according to Ellerbrock and Gerke using the Kbr press method [33]. A total of 5 mg of soil (<0.5 mm) was mixed with 200 mg of potassium bromide in an onyx mortar and pressed into tablets by applying a pressure of 25 KPa. The pressed, smooth, and undamaged samples were analysed using a Fourier infrared spectrometer. FTIR spectra were acquired at a resolution of 4 cm⁻¹ and using 16 scans in the wave number range of 4000–400 cm to collect absorbance spectra of organic matter. Upon obtaining the baseline-corrected and smoothed FTIR spectrum, characteristic peaks are first identified to determine the absorption peaks representing different carbon groups. Peak areas are then integrated. Finally, the relative percentage of the functional group is obtained by normalising the intensity of each individual peak by dividing it by the total intensity of all selected characteristic peaks.

Relative content (%) = (Intensity of a specific peak/Total intensity of all selected peaks) × 100%

2.5. DNA Extraction and Analysis

Soil DNA was extracted from eight soil samples using the OMEGA Soil DNA Kit (M5635-02) (Omega Bio-Tek, Norcross, GA, USA), with its quantity and quality assessed via the NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The V3–V4 region of the bacterial 16S rRNA gene was amplified using primers 338F (5′-ACTCCTACGGGAGGCAGCA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) [34]. The fungal ITS region was amplified using primers ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2 (5′-GCTGCGTTCTTCATCGATGC-3′) [35].

The PCR reaction system for both bacteria and fungi comprised 25 µL, including the following: 5 µL 5× reaction buffer, 5 µL 5× GC buffer, 2 µL dNTP (2.5 mM), 1 µL upstream primer (10 µM), 1 µL downstream primer (10 µM), 2 µL DNA template, 8.75 µL ddH₂O, and 0.25 µL Q5 DNA polymerase. The PCR programme was as follows: 98 °C pre-denaturation for 2 min; 25 cycles comprising 98 °C denaturation for 15 s, 55 °C annealing for 30 s, and 72 °C extension for 30 s; followed by a final extension at 72 °C for 5 min, with samples stored at 4 °C. The PCR amplicons were purified using Agencourt AMPure Beads (Beckman Coulter, Indianapolis, IN, USA) and detected by the PicoGreen dsDNA Detection Kit (Invitrogen, Carlsbad, IN, USA). PCR products were sequenced using the Illumina NovaSeq 6000 sequencing platform at Shanghai Personal Biotechnology Co., Ltd, Shanghai, China

2.6. Data Processing

The 16S rRNA sequencing data were processed using the QIIME 2 bioinformatics workflow (version 2019.4). First, the cutadapt plugin (version 2.3) was employed to remove primers, followed by demuxing the raw sequence data using the demux plugin [36]. Subsequently, the qiime dada2 denoise-paired tool was invoked to execute DADA2 for quality control, denoising, assembly, and chimera removal [37]. The specific parameter settings were as follows: --p-trunc-len-f 240, --p-trunc-len-r 200, --p-trim-left-f 10, --p-trim-left-r 10. Following denoising and individual sequence analysis, feature sequences were merged with the ASV table, and ASVs with a total sequence count of only 1 across all samples were excluded (Table S1). The feature classifier plugin was employed: the Naive Bayes classifier and SILVA Release 132 database were used to assign taxonomic information to ASVs [38].

2.7. Statistical Analysis of Data

The baseline correction and normalisation of Fourier Transform Infrared Spectra (FTIR) were performed using Omnic software (Thermo Electron Scientific, USA version 7.3) [39]. Automatic baseline correction was first applied to eliminate scattering effects, followed by vector normalisation to minimise spectral intensity fluctuations. Processed spectral data spanned the 1800–800 cm⁻¹ wavelength range. Principal component analysis (PCA) was performed using the multivariate analysis package within OriginPro software (2018 edition) to extract loadings and scores, thereby distinguishing differences in soil organic carbon structure influenced by varying vegetation types. The percentage contribution of each characteristic organic carbon functional group was obtained by integrating the peak areas of each functional group and calculating the proportion of each group’s peak area relative to the total functional group absorption peak area.

To investigate the functional potential of microorganisms, PICRUSt2 was employed to predict metabolic pathways from 16S rRNA gene sequences [40]. The specific workflow is as follows: First, the feature sequences obtained from sequencing were aligned against the 16S rRNA gene sequences in the reference genome to construct a phylogenetic tree; the hidden state prediction model within the Castor algorithm was utilised to infer the functional composition of the feature sequences based on the gene family copy numbers of the reference sequences. Next, gene family copy numbers were calculated for each sample by integrating feature sequence abundance across samples, with hierarchical output preserving sequence species information. Finally, predicted gene families were mapped to the KEGG functional database, and MinPath was employed to infer metabolic pathway presence and abundance.

The percentage contribution of each functional group was determined by integrating the peak areas of characteristic organic carbon functional groups and calculating their proportion relative to the total functional group absorption peak area. The statistical evaluation of differences in relative functional group abundances across vegetation types was conducted using a one-way analysis of variance (ANOVA) in SPSS 26.0 (IBM Armonk, NY, USA). RDA was performed using CANOCO 5, with Duncan’s multiple range test (ANOVA, Duncan) determining primary correlations between organic carbon functional groups and bacterial community metabolic pathways. We used the ggplot2 package (version 3.3.6) within the R environment (version 4.2.0) to generate multi-group differential volcano plots, serving to visualise the abundance-based analysis of microbial metabolic pathways across different vegetation types. Employing Python (version 3.6.6) to construct interactive Sankey’s diagrams, thereby revealing interactions among microbial metabolic pathways. Heatmaps were generated using ComplexHeatmap 2.12.0 within R version 4.2.0 to display trends in metabolic pathway abundance across samples.

3. Results

3.1. Molecular Structures of Different Vegetation Types

The soil from each treatment was analysed using FTIR spectroscopy, and the results are shown in Figure 1. It can be seen that the main characteristic peaks and their attributions are as follows: 914 cm⁻¹ originates from the symmetric C-O-C stretching of cyclic ether [41], 1030 cm⁻¹ and 1380 cm⁻¹ are attributed to the C-O of polysaccharides and the C-O stretch of polysaccharides [42,43], and the peak observed at 1260 cm⁻¹ in the C-O-C stretching vibration band may reflect the presence of ether bond structures in lignin [44]. The peak at 1340 cm⁻¹ is attributed to the C-N telescopic vibration of amines and ammonium salts [45]. The peak at 1430 cm⁻¹ can be attributed to C-H asymmetric deformations [46], 1630 cm⁻¹ to aromatic compounds C=C and aromatic C=H stretching vibrations [47], and 1640 cm⁻¹ to polysaccharides’ adsorbed O-H and conjugated C=O chemical bonds [48].

As can be seen from Figure 1, the FTIR spectral features of soil organic carbon structure under the influence of different vegetation types exhibit relative similarity, indicating that the carbon skeleton structure remains fundamentally consistent across varying vegetation influences. The spectral features of DG samples are more pronounced, suggesting a relatively higher inorganic content and greater carbon composition heterogeneity. Some characteristic peaks exhibit varying degrees of absorption intensity, reflecting the influence of differing numbers of organic carbon structural units and functional groups under various vegetation types. This indicates that varying terrain slopes exert a significant impact on the quantity of organic carbon structural units and functional groups.

To further analyse the spectral differences among vegetation types, principal component analysis (PCA) was employed. The principal component score plot (Figure 2A) indicates that the first two principal components collectively explain 92.0% of the total variance (PC1 = 78.6%, PC2 = 13.4%). The samples exhibit a distinct separation: DG clusters significantly apart from HM, ZL, and CB along the PC1 axis, whilst ZL and CB form a tight cluster, reflecting their highly similar organic carbon compositions. The loadings plot (Figure 2B) reveals the specific wavenumbers driving the separation. The principal wavenumbers contributing to PC1 include 916 cm⁻¹ (aromatic compounds C-H out-of-plane chemotaxis), 1020 cm⁻¹ (lignin C-H chemotaxis), and 1030 cm⁻¹ (C-O stretching vibration). Within PC2, the dominant loadings were 875 cm⁻¹ (C-H out-of-plane chemical bond), 990 cm⁻¹ (-HC=CH- out-of-plane deformation), and 1050 cm⁻¹ (aromatic C-H in-plane deformation or C-O stretching vibration). These results indicate that the compositional differences among the HM, DG, ZL, and CB vegetation types are primarily attributable to variations in aromatic compounds and oxygen-containing functional groups.

The semi-quantitative analysis of soil infrared spectra showed that aromatic compounds, olefinic compounds, and polysaccharides were the main organic carbon components, with polysaccharides having the highest relative content, and olefinic and carboxylic compounds having a lower relative content. Alkyl compounds were relatively elevated in the DG, and ammonium and ether compounds were also present. Polysaccharides were the highest in active components, while aromatic compounds were the most stable. While the absorption peak at 1630 cm⁻¹ can indicate the degree of organic carbon decomposition, the higher the absorption peak at this location indicates a higher percentage of stable organic carbon compounds. Furthermore, 1630 cm⁻¹ absorption peaks appeared in DG, and the content of aromatic compounds was the highest the polysaccharide content was lowest, and soil stability was optimal. The highest content of polysaccharides in HM resulted in the most active soil (

Table 1. Relative content of organic carbon fractions under the influence of different vegetation types (%).

	ArC	PSC	AlkC	AmC	EtC	OleC
CB	6.10 ± 0.13 b	132.04 ± 2.36 a	-	-	-	16.48 ± 0.37 d
ZL	6.24 ± 0.31 b	92.18 ± 12.29 b	-	-	-	14.61 ± 0.58 c
DG	105.79 ± 20.52 a	152.48 ± 14.26 a	38.82 ± 4.45 b	3.76 ± 3.76 b	3.52 ± 3.53 b	8.82 ± 0.83 b
HM	4.83 ± 0.15 b	147.58 ± 3.61 a	416.87 ± 2.30 a	503.94 ± 1.73 a	578.18 ± 1.62 a	11.12 ± 0.35 a

ArC: aromatic carbon; PSC: polysaccharide carbon; AlkC: alkyl carbon; AmC: amine carbon; EtC: ether carbon; OleC: olefin carbon; CB: citrus; ZL: Zea mays L.; HM: Horsfieldia hainanensis Merr; DG: Drypetes perreticulata Gagnep. The data in the table are mean ± standard error; Different lowercase letters indicate significant differences between different treatments. (p < 0.05).

).

3.2. Responses of Functional Pathways in Soil Microbial Communities of Different Vegetation Types to Molecular Structure

This study employed PICRUSt, based on KEGG comparative annotations, to clarify the differences in soil bacterial functional diversity across different vegetation types. The results of the relative abundance analysis of primary metabolic pathways in soil bacteria across the study area revealed that the primary metabolic functions of soil bacteria in different vegetation types were similar, comprising seven categories. The relative abundances, from highest to lowest, were biosynthesis (65.88%), generation of precursor metabolite and energy (15.72%), degradation/utilisation/assimilation (14.44%), metabolic clusters (2.53%), glycan pathways (0.78%), macromolecule modification (0.55%), and detoxification (0.10%). Secondary functional prediction revealed that soils from different vegetation types harbour 60 secondary metabolic functions, 12 of which belong to the biosynthesis function at the primary level. The cluster analysis of the abundance and intergroup differences in the 60 secondary metabolic functions across different vegetation types yielded the results shown in Figure 3. As can be seen from the figure, the functional pathways of HM differ most significantly from the other three categories, while the functional pathways of CB and DG are relatively similar. Furthermore, the distances between CB and ZL were significantly closer in the cluster analysis.

The metabolic pathway differential abundance analysis revealed significant differences in microbial functional pathways among the four vegetation types. As shown in the volcano plot (Figure 4), the number of significantly altered metabolic pathways (|logFC| > 1, p < 0.05) exhibited marked differences in pairwise comparisons. The comparison between HM and ZL yielded the highest number of differentially abundant pathways (n = 27), whereas the comparison between ZL and CB produced the lowest (n = 11). Notably, two functional pathways demonstrated significant differences across all four vegetation types: PWY-7354 (Aclacinomycin biosynthesis) and PWY-7046 (4-coumarate degradation (anaerobic). Furthermore, 13 pathways exhibited significant changes in three out of the four comparisons, whilst 17 pathways showed significant differences in two comparisons.

3.3. Differential Functional Pathway Interpretation and Species Composition Analysis

The RDA indicates that variation in microbial functional pathways is the primary driver of organic carbon structural changes, explaining 55.66% of the total variance. In Figure 5, the direction of each arrow represents the gradient of increasing values for the corresponding variable (e.g., metabolic pathway, organic carbon fraction, or microbial taxonomic unit); the length of the arrow is proportional to its importance in explaining variation; while the angle between arrows reflects correlation—acute angles denote strong positive correlations, obtuse angles indicate negative correlations, and angles approaching right angles suggest weak or absent correlations. In Figure 5, the arrow representing the PWY-7354 pathway is closely adjacent to the aromatic carbon arrow, indicating a strong positive correlation between the two. However, the small angle between this arrow and the polysaccharide carbon also suggests a moderate association with this carbon component. The DHGLUCONATE-PYR-CAT-PWY metabolic pathway exhibits strong correlations with both fatty carbon and ether carbon, whereas PWY-6749 is primarily associated with ether carbon. Chlorophyll-SYN and PWY-5499 demonstrate a strong directional alignment with alkenyl carbon, suggesting a potential key role in alkenyl carbon metabolism.

By associating differentially abundant metabolic pathways with their corresponding microbial taxa, key functional communities can be identified. The first two RDA principal component axes (RDA1 = 55.85%; RDA2 = 31.84%) collectively explained 87.69% of constrained variance. The differentially abundant taxon Phormidium_IAM_M-710 within the CHLOROPHYLL-SYN pathway exhibited a significant positive correlation with aromatic carbon, suggesting a potential specialisation in utilising aromatic compounds as a carbon source. Differentially abundant species, Fimbrimonadaceae and Xanthobacteraceae, within the PWY-6479 pathway both showed strong correlations with polysaccharides, indicating their potential significance in polysaccharide degradation. Micromonospora associated with PWY-7354 exhibited a significant positive correlation with aliphatic carbon, highlighting its potential role in aliphatic carbon metabolism. Conversely, Inquilinus associated with PWY-5499, whilst positively correlated with aliphatic carbon, exhibited negative correlations with both aromatic compounds and polysaccharides, indicating its specialisation towards a metabolic niche centred on aliphatic carbon utilisation. Within the DHGLUCONATE-PYR-CAT-PWY metabolic pathway, no significant correlations were observed between microbial taxa and carbon fractions.

This integrated analysis reveals how key microbial taxa shape soil organic carbon composition and transformation processes through their specific metabolic capabilities.

4. Discussion

This study systematically investigates how different vegetation types alter the molecular structure of soil organic carbon—specifically through modifying the carbon skeleton, functional groups, and molecular weight—thereby regulating microbial metabolic pathways and shaping community composition.

4.1. Vegetation Effects on Organic Carbon Structure

The composition and stability of soil organic carbon (SOC) structure are significantly regulated by vegetation type, primarily through variations in litter physicochemical properties and root-mediated processes [49,50]. Within subtropical forests, differences in tree species composition simultaneously modulate both the quantity and quality of organic matter input, thereby shaping the diversity and stability of soil organic carbon’s structural composition [51,52]. Our findings unequivocally demonstrate that vegetation type exerts a selective influence on soil organic carbon composition, with marked variations in the relative abundance of key carbon fractions.

It is noteworthy that the aromatic carbon (ArC) content in the DG plot was significantly higher than that in the HM, ZL, and CB plots. This pattern can reasonably be attributed to the rich content of lignin and tannins in evergreen broad-leaved forest litter; these polyphenolic compounds directly contribute to a stable aromatic carbon pool during decomposition [2]. Within these soils, the persistence of aromatic carbon further accumulates with increasing depth, with a marked rise in the proportion of deep soil aromatic carbon. This may be attributable to the enhanced formation of lignin derivatives and organo-mineral complexes in deeper soil horizons. Conversely, the markedly lower polysaccharide carbon (PSC) content in the ZL plot aligns with typical management practices in maize cropping systems, which prioritise grain yield over biomass retention. Although crop residues such as straw contain relatively high polysaccharide levels (17–24%), these labile polysaccharides undergo rapid mineralisation in intensively managed soils due to a low lignin content, frequent tillage, and heightened microbial activity [53]. Moreover, water-soluble sugars present in maize, such as glucose and arabinose, exhibit high bioavailability, stimulating microbial activation and accelerating carbon loss through synergistic metabolism [54,55].

AlC, EtC, and AmC occur exclusively in evergreen broad-leaved trees, with an HM content significantly higher than in DG, further corroborating the role of forest-specific plant inputs. These compounds likely originate from cuticular waxes, cork tissues, and microbial residues accumulating in less-disturbed forest ecosystems [56]. In contrast, the observed increase in olefinic carbon (OleC) content in agricultural soils likely stems from the incomplete microbial degradation of lipids and the subsequent accumulation of unsaturated aliphatic intermediates. Antibiotic-producing bacteria such as streptomycetes possess extensive gene clusters for secondary metabolite synthesis, including antibiotics derived from fatty acids or those interfering with fatty acid metabolism. Frequent tillage disturbances disrupt soil aggregate structure, exposing previously protected lipids to microbial action and significantly accelerating this process [57].

4.2. Carbon Structure-Driven Selection of Functional Pathways

Soil bacterial communities maintain highly conserved core metabolic functions that exhibit consistency across all vegetation types, with biosynthetic pathways (65.88%) and energy metabolism (15.72%) constituting the dominant functional modules. This functional conservatism may reflect fundamental microbial life requirements, as primary metabolic processes such as carbon fixation and energy generation constitute core cellular functions that remain relatively stable across diverse environments [58,59]. In contrast, secondary metabolic processes (60 pathways) exhibit marked vegetation-dependent variation, revealing how plant-specific biochemical inputs drive microbial metabolic specialisation [60].

The observed metabolic patterns can be mechanistically attributed to vegetation-induced alterations in soil physicochemical properties and resource acquisition strategies. Forest ecosystems, characterised by a high organic matter content and low disturbance, prompt microbes to increase investment in energy conversion and carbohydrate metabolism [61,62]. Specifically, forests’ robust carbon storage capacity correlates positively with heightened microbial activity in carbon processing pathways, including glucose metabolism and unsaturated fatty acid synthesis [63]. These functional enhancements likely represent adaptive responses to complex organic inputs from forest litter, which provides diverse carbon substrates for specialised metabolic pathways [64].

Notably, the transition from agricultural to forest ecosystems is accompanied by direct alterations in soil properties—decreased pH and increased organic carbon, nitrogen, and water content—establishing distinct environmental mechanisms that reshape microbial functional composition [65]. Within mature forest systems (HM), adaptation to low-light and humus-rich environments promotes the enrichment of lignocellulose-degrading actinomycetes [66]. We hypothesise that these microbes play a pivotal role in decomposing complex plant polymers by activating carbohydrate metabolism (0.78%) and macromolecular modification pathways (0.55%). These taxa may promote soil carbon stabilisation by converting plant-derived compounds into microbial by-products that adhere to mineral surfaces. Conversely, frequently disturbed farmland soils (ZL/CB) are dominated by r-strategy bacteria, which prioritise amino acid and fatty acid metabolic pathways. We propose that this metabolic preference reflects rapid nutrient cycling dynamics within agricultural systems: tillage and straw incorporation generate readily degradable organic compounds, favouring fast-growing microbes with streamlined metabolic capabilities. This functional configuration indicates reduced metabolic complexity in carbon processing pathways under agricultural management, potentially diminishing stability.

These findings reveal that core metabolic functions remain stable across different vegetation types, whereas secondary metabolism exhibits significant plasticity in response to vegetation-mediated variations in soil properties and resource acquisition. This functional diversification highlights the pivotal role of plant–soil interactions in shaping the metabolic potential of soil bacterial communities, offering crucial insights for ecosystem-scale carbon cycling and sequestration.

4.3. Pathway-Mediated Shifts in Microbial Community Composition

Within evergreen broad-leaved forests (HM/DG), litter rich in lignin and lipids leads to the accumulation of recalcitrant carbon forms, particularly aromatic carbon and ether-bonded carbon. The complex chemical structures of these compounds may select microbial communities capable of producing oxidising enzymes (such as peroxidases), which are crucial for breaking down aromatic rings and ether bonds within lignin derivatives. This enzymatic strategy constitutes a key mechanism for microbial adaptation to resource environments dominated by plant secondary metabolites. Conversely, in agroecosystems (ZL/CB), increased inputs of crop residues rich in polysaccharides and aliphatic compounds elevate the proportion of polysaccharide carbon and alkyl carbon. These readily degradable carbon sources promote rapid metabolic pathways such as glycolysis, supporting faster microbial growth rates and accelerated carbon turnover. The starkly contrasting carbon structures between forest and agricultural systems thus drive the development of fundamentally divergent microbial metabolic strategies.

Within the core functional groups of evergreen broad-leaved forests, *Phormidium_IAM_M-710* exhibits a significant positive correlation with aromatic carbon (r > 0.7) and is enriched in the chlorophyll synthesis pathway. We hypothesise that this species may possess a specialised capacity for degrading lignin derivatives, potentially converting phenolic breakdown products into carbon sources via aromatic ring-cleaving enzymes. This metabolic specialisation confers a competitive advantage in environments rich in complex carbon sources. Similarly, Inquilinus-associated bacteria exhibit the preferential utilisation of aliphatic carbon while showing negative correlations with aromatic and polysaccharide carbon. This pattern suggests a focus on plant lipid metabolism (e.g., olefins), thereby avoiding competition for more readily available carbon sources. Such metabolic differentiation may enhance overall functional efficiency within communities in litter-rich environments. In agricultural fields (ZL/CB), the key core bacterial groups, Fimbrimonadaceae and Xanthobacteraceae, are enriched in PWY-6479 and strongly positively correlated with polysaccharide carbon (r > 0.8). We hypothesise that these taxa are primary decomposers of cellulose and hemicellulose. Their dominant position in agricultural systems stems both from abundant cellulose inputs via crop residues and from frequent disturbance favouring rapidly growing microbial strategies. Micromonospora was enriched in the PWY-7354 strain and associated with alkyl carbon metabolism, potentially playing a key role in hydrocarbon degradation via alkyl monooxygenase-mediated pathways. This functional capacity is particularly significant in agricultural soils, where plant root exudates and decaying residues contribute substantial amounts of aliphatic compounds to the soil carbon pool.

Collectively, these findings reveal how vegetation-specific litter chemistry selects for distinct microbial functional groups possessing specialised enzymatic activities, thereby establishing predictable patterns of carbon composition and metabolic potential throughout ecosystems.

5. Conclusions

This study integrates soil organic carbon molecular structural characteristics with microbial functional prediction to elucidate the compositional mechanisms of soil organic carbon structure under different vegetation types in karst regions. The key findings include the following: (1) Forest vegetation (particularly evergreen broad-leaved forests) promotes the formation of stable carbon components such as aromatic carbon by introducing refractory constituents rich in lignin and lipids via litter input. (2) Agricultural systems (maize fields and citrus orchards) exhibit reduced soil organic carbon pool stability due to the rapid conversion of readily degradable polysaccharides in crop residues. (3) Differences in microbial secondary metabolic pathways drive the differentiation of organic carbon structures. Specific bacterial groups (e.g., *Phormidium IAM M-710*, Micromonospora) regulate carbon transformation and sequestration pathways through functional coupling with particular carbon fractions. Therefore, in ecological restoration and carbon management practices within karst regions, priority should be given to selecting native forest tree species (such as evergreen broad-leaved species) that promote the formation of stable carbon fractions. This approach enhances soil carbon sequestration capacity by modulating microbial communities. Future research should delve into the molecular mechanisms of key microbial groups in carbon transformation. By validating the long-term impacts of diverse vegetation management strategies on soil carbon sinks, this work will provide more precise theoretical foundations for ecosystem carbon cycle management and vegetation restoration in karst regions.