Diversity of Cultivable Soil Fungal Taxa Across a Land-Use Gradient in the Andes–Amazon Transition Zone: Insights from Agroecological Systems

Abstract

Land-use change strongly affects soil microbiota, yet the role of agroecological systems in shaping soil fungal communities remains poorly understood in tropical soils. We evaluated the diversity, trophic modes, community composition, and co-occurrence networks of culturable soil fungal taxa across a land-use gradient in the Colombian Andes–Amazon transition zone. Agroecological systems—including improved pasture (IP), cacao and copoazu agroforestry systems (CaAS and CoAS), secondary forest with agroforestry enrichment (SFAE), and a moriche palm swamp ecosystem (MPSE)—were compared with dominant land-uses (degraded pasture, DP and old-growth forest, OF). Fungi were isolated using the soil dilution plate method and identified based on morphological and molecular characteristics, and soil physicochemical properties were measured to evaluate their relationships with fungal community patterns. A total of 420 isolates were assigned to 93 fungal species. Alpha-diversity metrics revealed significantly higher fungal richness in OF and MPSE, and higher Shannon diversity in agroforestry and forest-based systems, whereas DP exhibited the lowest values. Ordination analyses showed clear differences in fungal community composition, with CoAS displaying the most distinct assemblage. Agroecological and forest-based systems favored saprotrophic and symbiotrophic modes. Co-occurrence network analyses indicated that MPSE, OF, and IP supported more complex and modular fungal networks. Soil pH and total phosphorus (TP) were key drivers of fungal community composition, whereas exchangeable calcium, TP, soil organic carbon, and base saturation were associated with network attributes. Overall, our findings highlight the importance of agroecological management for soil fungal diversity and network organization in Amazonian transition landscapes.

1. Introduction

Soils are a non-renewable natural resource and among the most biologically diverse habitats on Earth, providing the foundation for ecosystem productivity, climate regulation, and biogeochemical cycling [1]. However, global land-use change has driven profound alterations in biodiversity and ecosystem functioning in both anthropogenic and natural ecosystems, primarily through modifications in vegetation cover and edaphic properties such as pH, nutrient availability, and soil structure. These changes constitute a major threat to biodiversity, particularly within tropical ecosystems [2,3,4], and directly affect diversity, composition, and the functional potential of soil microbial assemblages, which are widely recognized as indicators of ecosystem health and resilience.

Soil microorganisms play a fundamental role in maintaining soil fertility by mediating nutrient turnover, decomposing organic matter, and driving essential metabolic processes within terrestrial ecosystems [5]. Among them, fungi represent one of the most functionally diverse and ecologically relevant groups, contributing to organic matter degradation, soil aggregation, and the bioremediation of pollutants [6,7]. Beyond their saprotrophic roles, soil fungi regulate plant health through symbiotic and antagonistic interactions—facilitating nutrient acquisition and suppressing pathogens—and thereby contribute to overall ecosystem stability [8,9].

The Amazon basin is one of the world’s most biologically diverse regions and a major global carbon sink, storing nearly 10% of the planet’s terrestrial carbon [10]. Despite the generally low fertility of Amazonian soils, ecosystem productivity is sustained by highly efficient nutrient recycling processes mediated by microorganisms, with fungi playing a pivotal role in litter decomposition and nutrient mobilization [11]. However, the rapid expansion of agriculture, cattle ranching, and other land-use changes in the Andean–Amazon transition zone has altered these microbial-driven processes, potentially leading to biodiversity loss and reduced soil functionality [12,13].

Land-use changes strongly shape soil fungal assemblages, which respond to edaphic factors such as pH, texture, and organic carbon content that regulate their diversity, composition, and network connectivity [2,3,14,15,16]. Although soil fungi play critical roles in nutrient cycling, organic matter decomposition, and plant symbioses, their taxonomic diversity, co-occurrence patterns, and functional organization across contrasting land-use systems remain unevenly explored, particularly in tropical and subtropical soils [17]. Recent studies have documented a high diversity of culturable filamentous soil fungi across agricultural and natural ecosystems, encompassing a wide range of functional guilds and trophic modes, including saprotrophs, endophytes, pathotrophs, and taxa with potential biocontrol activity [18,19,20,21,22]. In the Amazon basin, several works have reported distinct patterns of fungal richness and community assemblages, across forests, agroforestry systems, crops, and regenerating pastures, highlighting the role of environmental heterogeneity in structuring fungal taxa across land-use gradients [12,18,20,23].

Agroecology-based land-use systems offer an integrative framework for reconciling agricultural production with biodiversity conservation in tropical regions. Practices such as diversified tree cover, organic nutrient management, and reduced soil disturbance can enhance soil biological activity, microhabitat heterogeneity, and the provision of multiple ecosystem services [22,24,25]. Understanding how such systems are associated with soil fungal diversity and interaction patterns is therefore essential for informing sustainable land-management strategies under increasing anthropogenic pressure.

Despite growing interest in soil fungal ecology, the taxonomic and functional responses of cultivable soil fungal communities to agroecology-based land-use systems remain poorly documented in Amazonian landscapes. To address this knowledge gap, we evaluated soil fungal diversity, trophic modes, community composition, and co-occurrence network structure using a culture-dependent approach across a land-use gradient in the Andean–Amazon transition zone. This gradient ranges from old-growth forests, representing relatively undisturbed reference ecosystems, to degraded pastures subjected to long-term extensive grazing, and includes intermediate systems managed under agroecological practices.

We hypothesized that (i) agroecological land-use systems support higher culturable soil fungal alpha-diversity than degraded pastures; (ii) the composition and co-occurrence network structure of cultivable soil fungal taxa differ among land-use types, with agroecological systems exhibiting more complex and modular interaction networks; and (iii) variation in culturable soil fungal communities attributes is associated with land-use-driven differences in soil physicochemical properties.

Overall, our study provides new insights into how agroecological systems are associated with patterns of diversity and the organization of isolated soil fungal taxa, contributing to a better understanding of the ecological mechanisms underpinning soil biodiversity and offering relevant implications for biodiversity conservation and sustainable land-use planning in Amazonian transition landscapes.

2. Materials and Methods

2.1. Study Area and Land-Use Description

The research was carried out in the department of Caquetá, in the northwestern Colombian Amazon, within the Andes–Amazon transition zone (Figure 1).

The study area is characterized by pronounced geomorphological variability, encompassing alluvial lowlands, undulating hills, foothill zones, and mountainous terrain [26]. The climate is humid tropical, with a mean annual temperature close to 25 °C and average annual precipitation reaching approximately 3235 mm. Rainfall follows a monomodal pattern, with a distinct wet season extending from March to June and a comparatively drier period occurring between November and February [27]. According to the United States Department of Agriculture (USDA) soil classification system, soils are mainly Inceptisols and Oxisols. These soils are typically fine-textured and exhibit restricted drainage, strong acidity (pH 4.5–5.8), elevated aluminum saturation, and low base saturation. Nutrient availability is generally limited, with low concentrations of organic carbon, phosphorus, potassium, and magnesium, factors that constrain overall soil fertility and aeration capacity [26].

Seven representative land-use types previously defined by Sterling et al. [28] were considered in this study. These comprised five systems integrating agroecological management—secondary forest with agroforestry enrichment (SFAE), copoazu agroforestry system (CoAS), cacao agroforestry system (CaAS), moriche palm swamp ecosystem (MPSE), and improved pasture (IP)—as well as two widely distributed regional land-uses: old-growth forest (OF) and degraded pasture (DP).

SFAE sites consisting of secondary forests were 8–12 years old and primarily composed of Henriettea fascicularis (Sw.) M. Gómez and Miconia elata (Sw.) DC. Enrichment plantings included native palms, windbreak trees, and fruit-bearing species.

CaAS and CoAS sites corresponded to agroforestry systems dominated by cacao (Theobroma cacao L.) and copoazu (T. grandiflorum [Willd. ex Spreng.] Schum.), respectively. Soil management relied exclusively on organic amendments, particularly vermicompost, with no synthetic agrochemical inputs.

MPSE sites were wetlands dominated by Mauritia flexuosa L.f. Restoration efforts involved enrichment planting with native tree species two years before sampling. These wetlands were protected from livestock access and experienced minimal anthropogenic disturbance.

IP sites followed a Voisin rotational grazing system, with grazing periods limited to a maximum of three days per paddock, followed by adequate resting periods. These systems consisted of subdivided pastures dominated by Brachiaria decumbens (Stapf) R.D. Webster, B. brizantha cv. Marandú (A.Rich.) Stapf and B. humidicola (Rendle) Schweick, complemented by mixed forage banks. IP received regular applications of liquid organic fertilizer (pig-slurry digestate) every 35 days, at an approximate dose of 500 L ha⁻¹, representing a more intensive and sustainable livestock management strategy.

OF sites represent relatively undisturbed mature forests with minimal selective logging for household use.

DP sites, originally forested, had undergone more than 15 years of continuous extensive grazing. Grazing periods in these systems typically extended for up to 12 consecutive days, with no rotational management and no application of mineral or organic fertilizers. These areas were characterized by low cattle-carrying capacity, high soil compaction, and the dominance of grasses Paspalum notatum Flüggé and Cynodon nlemfuensis Vanderyst, reflecting long-term degradation under continuous grazing.

2.2. Soil Sampling

Three plots measuring 20 × 30 m were established within each land-use type, resulting in a total of 21 sampling plots, following the general framework described by Sterling et al. [28]. Field sampling was conducted between July and August 2024, with procedural adjustments based on Yang et al. [3]. Within each plot, five 1 × 1 m subplots were positioned diagonally to capture spatial variability. From each subplot, five soil cores were randomly extracted at a depth of 0–20 cm. The cores collected within each subplot were thoroughly mixed to generate a composite sample representative. After collection, composite samples were sieved through a 2 mm mesh to remove coarse fragments and visible plant debris. Each processed sample was subsequently separated into two portions. One fraction was preserved at 5 °C for fungal isolation, quantification, and identification [18], while the second fraction was air-dried at ambient temperature for approximately seven days prior to physicochemical analyses [3,18].

2.3. Soil Physicochemical Analyses

Soil physicochemical properties were quantified using standard analytical procedures. Soil organic carbon stocks (SOC, Mg ha⁻¹) were calculated according to Maia et al. [29] by combining the organic carbon concentration—measured via the Walkley–Black dichromate oxidation method—with bulk density (kg m⁻³) and the thickness of the sampled soil layer (m). Soil pH and electrical conductivity (EC, dS m⁻¹) were assessed in saturation paste extracts following conductometric protocols established by the USDA Salinity Laboratory [30]. Cation exchange capacity (CEC, meq 100 g⁻¹) was determined through titration with 1 M NaOH, and base saturation (BS, %) was derived as the proportion of exchangeable base cations (Ca²⁺, Mg²⁺, K⁺, and Na⁺) relative to total CEC [30]. Exchangeable potassium (E_K, mg kg⁻¹), calcium (E_Ca, mg kg⁻¹), and magnesium (E_Mg, mg kg⁻¹) were extracted using ammonium acetate and subsequently quantified by inductively coupled plasma optical emission spectrometry (ICP–OES) [30]. Total phosphorus (TP, mg kg⁻¹) was analyzed using the Bray II colorimetric method [31], while total nitrogen (TN, %) was quantified through Kjeldahl digestion [32]. Aluminum saturation (Al_S, %) was calculated as the ratio between extractable Al (1 N KCl extraction) and the sum of ammonium acetate–extractable base cations plus extractable Al, expressed as a percentage [30]. Average humidity saturation (AHS, %) was determined gravimetrically from the saturation paste. Soil texture (clay, sand and silt fractions) was obtained using the hydrometer method [33]. Bulk density (BD, g cm⁻³) was estimated from the oven-dry mass of the <2 mm soil fraction divided by the corresponding sample volume, following SSIR-42 guidelines for disturbed samples and correcting for coarse fragments when necessary [30].

2.4. Mycological and Molecular Analyses

2.4.1. Isolation of Culturable Soil Fungi

Culturable soil fungi were obtained using a serial dilution–plating approach following Arévalo-Gardini et al. [18] and Sarker et al. [34] with minor modifications. For each sample, 10 g of soil was homogenized in 90 mL of sterile distilled water (1:10, w/v) and shaken for 30 min. Then, 100 µL aliquots of the 10⁻³, 10⁻⁵, and 10⁻⁷ dilutions were spread onto three replicate Petri dishes per dilution containing potato dextrose agar (PDA; Condalab, Madrid, Spain) and Sabouraud dextrose agar (SDA; Condalab, Madrid, Spain), both supplemented with chloramphenicol (PanReac AppliChem, Barcelona, Spain; 50 mg L⁻¹) to suppress bacterial growth. Plates were incubated at 24–25 °C for seven days [18]. Emerging fungal colonies were transferred to fresh PDA plates to obtain pure isolates, which were subsequently preserved for morphological characterization and molecular identification.

2.4.2. Morphological Identification

Fungal isolates were examined to determine their taxonomic affiliation at the genus level through the evaluation of colony macromorphology and microscopic features, following the taxonomic keys of Crous et al. [35], Domsch et al. [36], and Ulloa and Hanlin [37]. Morphological identification was performed on axenic cultures and independently of colony counting, which was used solely for estimating fungal density.

2.4.3. Molecular Identification

Fungal DNA extraction was carried out following the methods described by Stirling [38], Kim et al. [39], and Huang et al. [40], with minor modifications. Approximately 1 mg of fresh fungal biomass per isolate was disrupted in liquid nitrogen with sterile glass beads and transferred to 2 mL microcentrifuge tubes containing 100 µL of preheated CTAB extraction buffer (0.5 M EDTA, 1 M Tris-HCl pH 8.0, 5 M NaCl, PVP, and β-mercaptoethanol). Following incubation, proteins and other contaminants were removed by chloroform:isoamyl alcohol (24:1) extraction, and samples were centrifuged at 12,000 rpm for 20 min at 4 °C. The aqueous phase was recovered and DNA precipitation was induced with an equal volume of isopropanol at −20 °C overnight. DNA pellets were obtained by centrifugation at 13,000 rpm for 15 min, rinsed with 70% ethanol, air-dried, and resuspended in 50 µL of sterile ultrapure water. The integrity and concentration of extracted DNA were evaluated by 1% agarose gel electrophoresis and spectrophotometric quantification using a NanoDrop One C instrument (Thermo Fisher Scientific, Waltham, MA, USA) prior to amplification [3].

The internal transcribed spacer (ITS) and large subunit (LSU) rDNA regions were targeted for molecular identification. Amplifications were carried out using primer pairs ITS1-f (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2-r (5′-TCCTCCGCCTTATTGATATGC-3′), and LROR-f (5′-ACCCGCTGAACTTAACTTAAGC-3′) and LR6-r (5′-CGCCAGTTCTGCTTACC-3′), respectively [41]. PCR reactions were performed under the following thermal profile: initial denaturation at 94 °C for 2 min; 40 cycles of denaturation at 94 °C for 30 s, annealing at 56.8 °C for 30 s, and extension at 72 °C for 30 s; and a final elongation step at 72 °C for 10 min [42].

The resulting sequences were compared by BLAST (https://blast.ncbi.nlm.nih.gov/) against reference sequences available in NCBI GenBank and MycoBank databases (https://www.mycobank.org/) [19]. Among the 420 fungal isolates obtained in this study, a subset of representative isolates was selected for ITS–LSU molecular characterization based on morphological criteria. Newly generated sequences were deposited in GenBank, and their corresponding accession numbers are provided in Supplementary Table S1.

Multiple-sequence alignments for the ITS and LSU regions were generated separately using CLUSTALW v1.2.4 with default parameters [43] and subsequently concatenated into a single dataset for phylogenetic analyses. Phylogenetic relationships were inferred using concatenated ITS–LSU sequences from the characterized isolates, together with representative reference sequences retrieved from GenBank. The optimal nucleotide substitution model was selected using jModelTest2 under the Bayesian Information Criterion (BIC) framework [44]. Bayesian phylogenetic inference was performed in MrBayes v3.2 under the GTR+Γ+I model with 1,000,000 generations [42].

When ITS–LSU data and morphological traits did not allow confident species-level assignment, isolates were conservatively classified at the genus level based on the combined molecular and morphological evidence. All species names were validated according to the Species Fungorum database (https://www.speciesfungorum.org/).

2.5. Soil Fungal Density, Diversity, and Community Composition

The density of culturable fungi was quantified as colony-forming units per gram of dry soil (cfu g⁻¹), calculated from the dilution yielding countable, non-confluent plates (approximately 30–300 colonies per plate). Colony counts were averaged across replicate Petri dishes, and the corresponding dilution factor was applied following standard microbiological procedures [18]. Relative abundance (%) was calculated as the proportion of colonies assigned to each fungal taxon relative to the total number of colonies counted per sample. Species accumulation curves were generated using the random permutation method with 999 permutations, as implemented with the specaccum function from the vegan 2.5-7 package [45] in R v. 4.3.3 [46].

Alpha-diversity indices, including species richness, the Shannon–Wiener diversity (H′), and the Simpson dominance (D), were computed using the diversity function from the diverse v.0.1. R package [47]. Differences in fungal community composition among land-use types were explored using principal coordinate analysis (PCoA). The ordination was based on Bray–Curtis dissimilarity matrices and performed with the cmdscale function from the vegan v.2.5-7 package in R [45].

2.6. Soil Fungal Trophic Modes

Trophic mode information was initially obtained from the FUNGuild database [48]. To enhance classification reliability and reduce potential misclassification associated with ecological plasticity and incomplete guild-level resolution, assignments were cross-validated using additional curated databases, including FungalTraits [49] and FunFun [50], as well as targeted literature review [2,14,18,51]. To maintain consistency in functional classification, trophic modes were adopted as the primary analytical framework. Species were grouped into three principal trophic categories: pathotroph, saprotroph, and symbiotroph. Taxa assigned to multiple trophic categories were classified under mixed trophic strategies. For ecologically ambiguous taxa (e.g., genera with documented multiple lifestyles), classifications were conservatively defined to avoid overinterpretation.

To facilitate comparisons among trophic modes, relative abundance data were visualized using a row-wise standardized (Z-score) heatmap. This approach emphasizes the relative enrichment or depletion patterns of each trophic mode across land-use types, rather than absolute dominance. The heatmap was generated with the ggplot2 v. 3.3.3 R package [52].

2.7. Construction and Analysis of Soil Fungal Co-Occurrence Networks

To explore potential ecological associations among soil fungal taxa across different land-use types, we constructed co-occurrence networks based on correlation analyses, following the general analytical framework proposed by Zhu et al. [14], Yang et al. [3], and Xiao et al. [2], with minor modifications. Relative abundance data of cultivable soil fungi were used as the input. Taxa with an average relative abundance higher than 0.5% across all plots were retained to reduce noise associated with rare taxa [14]. For each land-use type, subsets of the filtered dataset were used to compute pairwise correlations among fungal taxa using Spearman’s rank correlation, implemented with the corr.test function in the psych v. 2.4.6.26 R package [53]. Correlation matrices were transformed into adjacency matrices by retaining only statistically significant correlations with |r| ≥ 0.6, a threshold commonly applied in soil fungal and microbial network studies to reduce spurious associations and emphasize stronger covariation patterns [3,16].

Network graphs were constructed and analyzed using Gephi (v. 0.10.1) [54]. Networks were treated as undirected and weighted. Visualization was performed using the Fruchterman Reingold layout, with nodes colored according to fungal class and node size scaled by degree centrality. Network-level topological metrics were calculated using the igraph v.2.0.3 R package [55] and cross-validated in Gephi, including the number of nodes and edges, average degree, density, clustering, average path length, positive edges (%), negative edges (%), and modularity (Louvain algorithm). Finally, network complexity was quantified following the approach of Chen et al. [56] using linkage density (number of links per taxon) as an integrative measure of network complexity [14].

2.8. Statistical Analyses

We fitted general linear models (GLMs) to assess the influence of land-use type on soil physicochemical variables, cultivable soil fungal diversity, and network-level co-occurrence metrics using the gls function in the nlme v. 3.1-131.1 R package [57] using the InfoStat v.2020 interface [58]. When overall effects were significant, pairwise comparisons were performed using Fisher’s LSD test (α = 0.05). Model assumptions were verified by assessing normality and homoscedasticity, and the variables were log-transformed when these assumptions were not met. Where heteroscedasticity was detected, a variance structure (varIdent) was incorporated to improve model fit [57]. Differences in fungal community composition across land-use types were examined using permutational multivariate analysis of variance (PERMANOVA) based on Bray–Curtis dissimilarities with 9999 permutations. Analyses were conducted using the adonis2 function from the vegan package. Post hoc pairwise comparisons were implemented with the pairwise.adonis function in the pairwiseAdonis v.0.4.1 package [59].

Associations between fungal community composition and soil physicochemical attributes were further explored using Mantel tests implemented with the mantel_test function from the LinkET v.0.0.7.4 package [60]. Spearman rank correlations were calculated to evaluate relationships between soil variables and fungal community attributes using the correlate function from the same package. Prior to these analyses, multicollinearity in network variables was assessed using hierarchical clustering (varclus, Hmisc v.5.1-3) [61], and highly collinear variables were excluded. To identify the relative contribution of soil properties to community shifts, multiple regression models were fitted using ordination-derived metrics as response variables. Specifically, the second axis of a non-metric multidimensional scaling (nMDS) ordination based on Bray–Curtis dissimilarities (computed with metaMDS, vegan R package) was used as a proxy for fungal community composition. Soil predictors exhibiting strong collinearity were removed prior to model fitting to ensure stability.

Finally, structural equation modeling (SEM) was used to disentangle direct and indirect pathways linking land-use systems, soil physicochemical properties, and fungal community attributes. Candidate models were evaluated using Fisher’s C statistic and Akaike’s Information Criterion (AIC). Prior to SEM construction, we applied a structured variable selection procedure to determine which soil physicochemical properties would be included as predictors. First, Spearman correlations and Mantel tests were used to identify soil variables significantly associated with fungal diversity, community composition, or network metrics. Second, multicollinearity among soil variables was evaluated using hierarchical clustering through the varclus function, and highly collinear predictors were removed to ensure model stability. From the remaining set, only soil properties with demonstrable statistical associations and clear ecological relevance were retained as candidate predictors in the model. SEM analyses were carried out with the psem function in the piecewiseSEM v.2.3.0 R package [62], and model visualization was conducted using the grViz function from the DiagrammeR v.1.0.11 R package [63]. All statistical procedures were conducted in R v.4.3.3 using the RStudio 2025.05.0 interface [64].

3. Results

3.1. Soil Physicochemical Properties

Soil physicochemical properties varied significantly among the evaluated land-use types (DP, IP, CaAS, CoAS, SFAE, MPSE, and OF; p < 0.05) (Figure 2).

BS was the highest in SFAE (48.25%) and CaAS (38.83%), intermediate in MPSE (31.69%) and IP (22.21%), and lowest in DP, CoAS, and OF (all, <14%). Soil pH followed a similar pattern (p < 0.001), with higher values in SFAE (4.79) and CaAS (4.54) and the lowest in OF (4.00). EC differed among land-uses (p < 0.01), reaching the highest value in OF (0.43 dS m⁻¹) and the lowest in MPSE and SFAE (0.19 dS m⁻¹). CEC showed slight variation (p < 0.05), with the highest values in OF and the lowest in DP (5.62 and 2.80 meq 100 g⁻¹, respectively). Meanwhile, AHS did not show significant statistical differences (p > 0.05). SOC differed highly significantly (p < 0.001), with higher contents in OF, SFAE, MPSE, and CaAS (42–53 Mg ha⁻¹) and the lowest in DP (25.34 Mg ha⁻¹). TN was greater in OF and SFAE (0.16–0.18%) and lower in DP (0.10%), whereas BD was significantly higher in DP (1.55 g cm⁻³) and lower in SFAE and OF (<1.25 g cm⁻³). E_K and E_Ca were higher in agroecological systems (CaAS, SFAE, MPSE) than in DP, while E_Mg, TP, and soil texture fractions did not show significant differences (p > 0.05). Al_S also varied significantly (p < 0.01), with the lowest values in CaAS (60.77%) and the highest in OF (92.18%).

3.2. Soil Fungal Diversity and Composition

A total of 420 soil fungal isolates were obtained and assigned to 93 fungal species based on combined morphological assessment and ITS–LSU molecular characterization of 161 representative isolates. Phylogenetic relationships inferred from concatenated ITS–LSU sequences are shown in Supplementary Figure S1.

Species accumulation curves showed asymptotic trends and adequate sampling coverage across all land-use types (Figure 3). Fungal species belonged to Ascomycota (80 species), Basidiomycota (10 species), and Mucoromycota (3 species), representing 47 genera in total (

Table 1. Density (×10⁴ cfu g⁻¹ dry soil) and relative abundance (%) of cultivable soil fungal species and their trophic modes across land-use types in the Andean–Amazon transition zone. Land-use type acronyms are defined in Figure 1.

Species	Trophic Mode	CaAS	CoAS	DP	IP	MPSE	OF	SFAE
Acremonium polychromum	Pathotroph–Saprotroph–Symbiotroph							35 (45.55)
Annulohypoxylon stygium	Saprotroph–Symbiotroph		1.67 (18.6)
Apiotrichum gamsii	Saprotroph						0.33 (0.31)
Arcopilus cupreus	Pathotroph–Saprotroph				35.67 (12.69)
Aspergillus sp. 1	Pathotroph–Saprotroph			0.1 (0.03)
Aspergillus sp. 2	Pathotroph–Saprotroph				35.67 (12.69)
Aspergillus sp. 3	Pathotroph–Saprotroph	2 (7.64)		0.1 (0.03)		13.33 (12.54)		0.67 (0.87)
Aspergillus sp. 4	Pathotroph–Saprotroph	17.17 (65.61)
Aspergillus sp. 5	Pathotroph–Saprotroph					13.33 (12.54)
Aspergillus sp. 6	Pathotroph–Saprotroph	0.17 (0.64)
Circinella simplex	Saprotroph		1.0 (11.1)
Clonostachys farinosa	Pathotroph				23.67 (8.42)
Clonostachys rosea	Pathotroph–Saprotroph–Symbiotroph				24 (8.54)
Clonostachys rossmaniae	Pathotroph				23.67 (8.42)
Corallomycetella elegans	Pathotroph–Saprotroph						0.33 (0.31)
Corallomycetella repens	Pathotroph–Saprotroph					0.9 (0.85)	0.5 (0.47)
Cubamyces menziesii	Saprotroph					0.07 (0.06)	0.33 (0.31)
Curvularia lunata	Pathotroph–Symbiotroph	1.33 (5.1)
Cyanoporus camptogrammus	Saprotroph						40 (37.39)
Daldinia eschscholtzii	Saprotroph					0.5 (0.47)		1.33 (1.73)
Diaporthe sojae	Pathotroph–Symbiotroph		0.67 (7.4)
Earliella scabrosa	Saprotroph		0.67 (7.4)
Fusarium concolor	Pathotroph–Saprotroph–Symbiotroph						0.17 (0.16)
Fusarium polyphialidicum	Pathotroph–Symbiotroph						0.17 (0.16)
Gongronella butleri	Saprotroph	1.33 (5.1)					0.5 (0.47)	0.67 (0.87)
Hypomontagnella monticulosa	Pathotroph–Saprotroph						12 (11.21)	0.33 (0.43)
Hypoxylon investiens	Symbiotroph							16.83 (21.91)
Hypoxylon pulicicidum	Saprotroph						0.67 (0.62)
Lasiodiplodia theobromae	Pathotroph–Symbiotroph						11.67 (10.9)
Metarhizium anisopliae	Pathotroph				1.33 (0.47)
Metarhizium baoshanense	Pathotroph				0.33 (0.12)
Metarhizium lepidiotae	Pathotroph				23.67 (8.42)
Metarhizium robertsii	Pathotroph				0.33 (0.12)			0.67 (0.87)
Montagnula chiangraiensis	Saprotroph	0.33 (1.27)
Montagnula opulenta	Saprotroph	0.33 (1.27)
Nectria augustoi	Pathotroph–Saprotroph–Symbiotroph	0.33 (1.27)	0.33 (3.7)		23.67 (8.42)	0.17 (0.16)
Neomassarina chromolaenae	Saprotroph		0.33 (3.7)
Neomassarina pandanicola	Saprotroph		0.33 (3.7)
Neopestalotiopsis clavispora	Pathotroph						0.33 (0.31)
Paraconiothyrium brasiliense	Saprotroph				23.67 (8.42)
Paraconiothyrium cyclothyrioides	Saprotroph				27.67 (9.86)
Paraconiothyrium zingiberacearum	Saprotroph				0.33 (0.12)
Penicillifer diparietisporus	Pathotroph–Saprotroph			85.13 (23.67)
Penicillifer martinii	Pathotroph–Saprotroph						0.33 (0.31)
Penicillium sp. 1	Saprotroph						0.33 (0.31)
Penicillium sp. 2	Saprotroph			85.13 (23.67)		0.33 (0.31)	7 (6.54)
Penicillium sp. 3	Saprotroph	0.33 (1.27)	0.33 (3.7)
Penicillium sp. 4	Saprotroph					33.57 (31.57)
Penicillium sp. 5	Saprotroph						0.33 (0.31)
Penicillium sp. 6	Saprotroph				36 (12.81)
Penicillium sp. 7	Saprotroph				0.33 (0.12)
Penicillium sp. 8	Saprotroph						24.33 (22.74)
Penicillium sp. 9	Saprotroph	0.17 (0.64)
Phlebia floridensis	Saprotroph					0.17 (0.16)
Polyporus ciliatus	Saprotroph						0.17 (0.16)
Porogramme epimiltina	Saprotroph						0.17 (0.16)
Pseudallescheria angusta	Pathotroph–Saprotroph						0.33 (0.31)	0.67 (0.87)
Pseudodactylaria longidenticulata	Saprotroph						0.33 (0.31)
Rigidoporus microporus	Saprotroph						0.33 (0.31)
Rigidoporus vinctus	Saprotroph						0.5 (0.47)
Roussoella neopustulans	Saprotroph		0.33 (3.7)
Roussoella siamensis	Saprotroph						0.67 (0.62)
Scedosporium apiospermum	Pathotroph						0.33 (0.31)
Scedosporium boydii	Pathotroph	0.5 (1.91)	0.33 (3.7)	0.17 (0.05)		0.23 (0.22)	2 (1.87)	18.67 (24.3)
Scedosporium sphaerospermum	Saprotroph							0.33 (0.43)
Scytalidium synnematicum	Saprotroph						0.33 (0.31)
Sesquicillium essexcoheniae	Saprotroph					0.07 (0.06)	0.17 (0.16)
Sporoschisma juvenile	Saprotroph						0.5 (0.47)
Striaticonidium cinctum	Pathotroph					0.17 (0.16)	0.33 (0.31)
Talaromyces sp. 1	Saprotroph		0.33 (3.7)			0.33 (0.31)
Talaromyces sp. 2	Saprotroph	0.67 (2.55)						1.33 (1.74)
Talaromyces sp. 3	Saprotroph		0.33 (3.7)	85.13 (23.67)
Talaromyces sp. 4	Saprotroph			89.60 (24.92)
Tolypocladium album	Pathotroph–Saprotroph–Symbiotroph				0.33 (0.12)		0.67 (0.62)
Tolypocladium tropicale	Pathotroph–Saprotroph–Symbiotroph					0.33 (0.31)
Trichoderma sp. 1	Pathotroph–Saprotroph–Symbiotroph			1.79 (0.5)
Trichoderma sp. 2	Pathotroph–Saprotroph–Symbiotroph			1.79 (0.5)	0.33 (0.12)
Trichoderma sp. 3	Pathotroph–Saprotroph–Symbiotroph	1 (3.82)				0.17 (0.16)
Trichoderma sp. 4	Pathotroph–Saprotroph–Symbiotroph	0.17 (0.64)
Trichoderma sp. 5	Pathotroph–Saprotroph–Symbiotroph					0.33 (0.31)
Trichoderma sp. 6	Pathotroph–Saprotroph–Symbiotroph					0.67 (0.63)
Trichoderma sp. 7	Pathotroph–Saprotroph–Symbiotroph						0.17 (0.16)
Trichoderma sp. 8	Pathotroph–Saprotroph–Symbiotroph			1.79 (0.5)
Trichoderma sp. 9	Pathotroph–Saprotroph–Symbiotroph				0.33 (0.12)
Trichoderma sp. 10	Pathotroph–Saprotroph–Symbiotroph		1.33 (14.8)
Trichoderma sp. 11	Pathotroph–Saprotroph–Symbiotroph		0.33 (3.7)			41.17 (38.71)		0.33 (0.43)
Trichoderma sp. 12	Pathotroph–Saprotroph–Symbiotroph	0.17 (0.64)				0.33 (0.31)
Truncospora tephropora	Saprotroph						0.67 (0.62)
Umbelopsis angularis	Saprotroph	0.17 (0.64)
Verruconis verruculosa	Saprotroph		0.33 (3.7)
Westerdykella angulata	Saprotroph		0.33 (3.7)	4.43 (1.23)
Westerdykella formosana	Saprotroph			4.43 (1.23)
Xylaria curta	Saprotroph–Symbiotroph					0.17 (0.16)

). Trichoderma (12 species), Penicillium (9 species), Aspergillus (6 species), Metarhizium, and Talaromyces (4 species each) were the most species-rich genera. Species richness varied among land-use systems, with 16 species identified in CaAS, 16 in CoAS, 12 in DP, 18 in IP, 20 in MPSE, 33 in OF, and 12 in SFAE.

At the phylum level, the culturable assemblages were dominated by Ascomycota across all land-use types, accounting for 80–100% of the relative abundance in most systems. Basidiomycota were consistently present but at lower proportions, reaching their highest contribution in OF (~39%), while Mucoromycota represented a minor fraction (<15%), mainly in agroforestry systems. At the class level, Sordariomycetes constituted the most abundant group across the land-use gradient, particularly in CoAS (~54%), SFAE (~85%), and MPSE (~58%). Eurotiomycetes were also prominent, especially in IP (~47%) and CaAS (~58%), whereas Dothideomycetes showed moderate contributions in agroforestry systems (up to 15% in CoAS). Agaricomycetes increased markedly in OF (~39%), reflecting the contribution of Basidiomycota-dominated forest taxa.

The genera Penicillium, Trichoderma, Aspergillus, Talaromyces, and Scedosporium were the most abundant across land-use types. Penicillium was particularly abundant in IP (~31%) and MPSE (~25%), Aspergillus dominated CaAS (~49%), and Scedosporium increased in SFAE (~29%), whereas OF was characterized by a high proportion of less frequent genera (~77%).

The alpha-diversity of cultivable soil fungal taxa differed significantly among land-use types (Figure 4). Species richness varied significantly (p < 0.01), with the highest mean values observed in OF (12.5) and MPSE (9.6), intermediate values in CoAS, CaAS, SFAE, and IP (6.9–7.9), and the lowest richness in DP (5.8) (Figure 4A). The Shannon–Wiener index also differed among land-uses (p < 0.01), reaching its highest value in CoAS (1.9), followed by CaAS (1.5) and MPSE (1.3), whereas DP and IP showed the lowest values (<1.0) (Figure 4B). Simpson’s index showed significant variation among land-use types (p < 0.01), ranging from 0.16 in CoAS to 0.64 in DP (Figure 4C). In addition, fungal density differed markedly among land-use types (p < 0.001), with the highest values observed in DP (33.41 × 10⁴ cfu g⁻¹), followed by IP (3.78 × 10⁴ cfu g⁻¹) and OF (1.43 × 10⁴ cfu g⁻¹), while the lowest density was recorded in CoAS (0.24 × 10⁴ cfu g⁻¹) (Figure 4D).

Cultivable soil fungal community composition differed significantly among land-use types, as evidenced by the PCoA ordination and PERMANOVA results, indicating that approximately 48% of the total variation was explained by land-use differences (Figure 5A). The strongest dissimilarities in fungal community composition were observed among agroecological systems, particularly between CaAS and CoAS, and between CoAS and SFAE, or MPSE (Figure 5B). Moderate but significant dissimilarities were also detected between DP or IP and agroforestry systems.

3.3. Soil Fungal Trophic Modes and Co-Occurrence Network Structure

Fungal trophic modes exhibited distinct relative abundance patterns across land-use types (Table 1; Figure 6). The relative abundance of fungal trophic modes differed among land-use types. Saprotrophs were the dominant category overall (especially in DP and OF), whereas mixed trophic strategies (particularly pathotroph–saprotroph and pathotroph–saprotroph–symbiotroph) showed pronounced variation across systems. Agroecological and forest-based systems (CaAS, CoAS, SFAE, OF, and MPSE) tended to support higher relative proportions of mixed trophic modes, while DP was characterized by a stronger dominance of saprotrophic taxa. Symbiotrophs were consistently low but slightly more represented in SFAE, whereas pathotrophs were more abundant in IP and SFAE.

Co-occurrence networks of cultivable soil fungi taxa exhibited clear structural differences among land-use types (Figure 7A–G). Each network displayed distinct patterns of organization and node connectivity, reflecting variations in the strength of associations and the number of links among fungal taxa. Network complexity and topology differed significantly among land-use systems (

Table 2. Mean (±SE; n = 3) topological metrics of cultivable soil fungal co-occurrence networks across seven land-use types in the Andean–Amazon transition zone.

Network Metrics	DP	IP	CaAS	CoAS	SFAE	MPSE	OF
Total nodes	9.00 ± 0.50 d	12.00 ± 0.59 bc	11.00 ± 0.51 c	12.00 ± 0.53 bc	11.00 ± 0.58 c	14.00 ± 0.52 a	13.00 ± 0.53 ab
Total edges	30.00 ± 0.51 g	63.00 ± 0.58 c	52.00 ± 0.53 e	55.00 ± 0.58 d	46.00 ± 0.55 f	75.00 ± 0.58 a	69.00 ± 0.54 b
Average degree	6.67 ± 0.19 d	10.50 ± 0.30 a	9.46 ± 0.27 b	9.17 ± 0.27 b	8.36 ± 0.24 c	10.71 ± 0.31 a	10.46 ± 0.30 a
Density	0.83 ± 0.02 c	0.96 ± 0.03 a	0.95 ± 0.03 ab	0.83 ± 0.02 c	0.84 ± 0.02 c	0.82 ± 0.02 c	0.87 ± 0.03 bc
Clustering	0.91 ± 0.03 ab	0.96 ± 0.03 a	0.96 ± 0.03 a	0.80 ± 0.02 c	0.84 ± 0.02 bc	0.81 ± 0.02 c	0.89 ± 0.03 ab
Average path length	1.17 ± 0.02 a	1.05 ± 0.02 b	1.06 ± 0.01 b	1.17 ± 0.03 a	1.16 ± 0.02 a	1.18 ± 0.03 a	1.13 ± 0.03 ab
Positive edges (%)	60.00 ± 0.58 a	39.68 ± 1.15 bc	30.77 ± 0.89 d	36.36 ± 1.05 c	39.13 ± 1.13 bc	40.00 ± 0.58 b	38.24 ± 1.10 bc
Negative edges (%)	40.00 ± 0.58 c	60.32 ± 1.74 b	69.23 ± 2.00 a	63.64 ± 1.84 ab	60.87 ± 1.76 b	60.00 ± 0.58 b	61.76 ± 1.78 b
Modularity	0.23 ± 0.01 f	0.29 ± 0.01 e	0.55 ± 0.02 bc	0.61 ± 0.02 a	0.50 ± 0.01 d	0.57 ± 0.02 ab	0.52 ± 0.02 cd
Network complexity	3.35 ± 0.15 d	5.27 ± 0.21 ab	4.75 ± 0.20 c	4.60 ± 0.17 c	4.20 ± 0.17 c	5.37 ± 0.18 a	5.31 ± 0.19 ab

Different lowercase letters within rows indicate significant differences among land-use types (p < 0.05, Fisher’s LSD test). Land-use type acronyms are defined in Figure 1.

). DP showed significantly smaller and less connected networks than all other systems, as indicated by the lowest numbers of nodes and edges and by reduced network complexity metrics. In contrast, MPSE exhibited the highest network complexity, followed by OF and IP. Although these systems consistently ranked higher in terms of nodes and edges, average degree, modularity, and clustering coefficient, differences among MPSE, OF, and IP were not always statistically significant. Cacao- and copoazu-based agroforestry systems (CaAS and CoAS), together with SFAE, displayed intermediate levels of network complexity and connectivity, indicating partial but incomplete reorganization of fungal co-occurrence networks relative to DP.

In addition, agroecological and forested systems exhibited co-occurrence networks of cultivable soil fungal taxa characterized by a significantly higher proportion of negative associations, whereas the DP network showed a significantly higher proportion of positive associations (Table 2). These patterns reflect contrasting modes of network covariation among land-use types, with DP differing significantly from all other systems in the balance between positive and negative edges. As shown in Figure 7, the taxa with the highest centrality values (degree, betweenness, and closeness) were identified as putative keystone taxa within land-use-specific networks. These highly connected taxa were predominantly affiliated with the classes Sordariomycetes, Eurotiomycetes, Agaricomycetes, and Dothideomycetes, although their relative representation varied among land-use types. Across the land-use gradient, these classes consistently included taxa occupying hub positions within co-occurrence networks, suggesting a central structural role in shaping network organization.

3.4. Relationships Between Soil Properties and Fungal Diversity, Trophic Modes, Community Composition, and Co-Occurrence Networks

Mantel tests showed significant correlations (p < 0.05) between cultivable soil fungal community composition and key soil physicochemical properties, particularly SOC, TN, and BD (Figure 8). In addition, variation in fungal community composition along the second NMDS axis was significantly explained by a combination of chemical and physical soil properties (R² = 0.55; p < 0.01) (Supplementary Figure S2). Among the chemical variables, TP, pH, and CEC showed strong positive associations with fungal community composition, whereas sand content was the main physical factor contributing to the observed variation. Overall, chemical properties accounted for 65.75% of the explained variance, while physical properties contributed 34.25%.

Spearman’s correlation analyses further indicated significant associations between soil parameters and fungal diversity metrics and network attributes (Figure 8). SOC, TN, Al_S, and silt content were positively correlated with fungal richness, whereas BD showed negative associations with richness and positive associations with fungal density. E_K was also negatively correlated with fungal density. Furthermore, E_Ca was strongly and negatively associated with network complexity and with the proportion of negative edges, while the proportion of negative edges was also negatively correlated with SOC. In contrast, BS, E_Ca, and TP exhibited positive associations with the proportion of positive edges.

Building on the relationships identified between soil properties and fungal community composition, diversity, and network attributes, Spearman correlation analyses further revealed significant associations between fungal trophic modes and soil physicochemical properties (Supplementary Figure S3). Pathotroph showed a significant negative correlation with EC, while pathotroph–symbiotroph was negatively associated with soil pH. In contrast, saprotroph exhibited significant positive correlations with AHS and silt content. Similarly, saprotroph–symbiotroph was positively correlated with AHS and SOC. Additionally, symbiotroph showed a significant negative correlation with BS. The remaining trophic modes did not show statistically significant correlations with the evaluated soil variables (Supplementary Figure S3).

The SEM indicated that land-use type (AGRO) exerted both direct effects on fungal richness, community composition, and network complexity, as well as indirect effects mediated through soil physicochemical properties, as evidenced by significant multi-step pathways linking land-use, soil variables, and fungal attributes (Figure 9).

Soil chemical properties showed significant direct effects primarily on fungal community composition. The model exhibited a good overall fit (Fisher’s C = 24.11, p > 0.05) and explained a high proportion of the variance in fungal richness (R² = 0.78), community composition (R² = 0.83), and network complexity (R² = 0.89). Significant path coefficients represented independent effects after accounting for the full set of predictors included in the model. In contrast, non-significant paths indicated relationships that were weak, indirect, or context-dependent, and whose effects were not statistically distinguishable from zero once other variables were included. These non-significant paths therefore suggest that certain soil properties or land-use categories do not exert direct control on specific fungal attributes, but may still influence them indirectly through other pathways or in combination with stronger drivers.

Land-use type exerted differentiated effects on fungal richness. DP showed a strong negative direct effect on fungal richness, whereas OF exhibited a significant positive effect. Agroecological systems—including IP, CaAS, CoAS, SFAE, and MPSE—displayed intermediate path coefficients that were not statistically significant, indicating that their effects on fungal richness were weaker or largely mediated through other physiochemical soil properties. Fungal community composition was primarily shaped by soil chemical properties. Soil pH showed a positive direct effect, whereas TP exerted a negative effect. Among land-use types, CoAS exhibited the strongest and statistically significant effect on fungal community composition, whereas the non-significant paths associated with the remaining land-use categories indicate overlapping compositional patterns once soil chemistry was accounted for.

Fungal network complexity exhibited the strongest response to land-use type. DP had a significant negative effect, reflecting simplified and weakly connected networks. In contrast, MPSE and IP showed significant positive effects on network complexity, consistent with denser and more structured co-occurrence patterns. OF also exhibited relatively high complexity; however, its direct path was not statistically significant, suggesting that its influence on network structure may be largely indirect, potentially mediated through fungal diversity and community composition rather than through a direct land-use effect.

4. Discussion

4.1. Agroecological Land-Use Systems Enhance Soil Fungal Alpha-Diversity

Agroecological and forest-based land-use systems were associated with significantly higher soil fungal alpha-diversity than DP. OF and MPSE systems supported the highest richness, highlighting the well-established role of structurally complex, organic-rich habitats in promoting soil fungal diversification [4,51]. Agroforestry systems based on copoazu and cacao (CoAS and CaAS) exhibited intermediate richness and diversity values, consistent with evidence that mixed-canopy systems enhance edaphic microhabitat heterogeneity and resource availability, thereby sustaining more diverse fungal assemblages than simplified production systems [18,65]. In contrast, DP showed the lowest richness and diversity, in agreement with previous studies demonstrating that intensive tropical land-use reduces vegetation cover, alters soil structure, and accelerates soil degradation, all of which constrain fungal diversity [4,65]. In this study, differences in soil fungal alpha-diversity among land-use systems were primarily driven by variation in species richness within dominant fungal lineages rather than by shifts in the presence or absence of major phyla. Ascomycota—mainly represented by Sordariomycetes and Eurotiomycetes—accounted for the highest proportion of species richness and relative abundance across all land-use types, followed by Basidiomycota (Agaricomycetes) and Mucoromycota (Mucoromycetes). Although these phyla are widely distributed and commonly dominant in soils across biogeographic regions, agroecological and forest-based systems supported a greater number of species within these groups, resulting in higher alpha-diversity relative to degraded pastures. Ascomycota include numerous saprotrophic taxa capable of exploiting a wide range of organic substrates, which promotes high within-group diversification under heterogeneous and resource-rich conditions [2,3,18]. Basidiomycota contribute specialized ligninolytic taxa, which play key roles in advanced stages of organic matter decomposition [66,67], whereas Mucoromycota typically comprise fast-growing opportunistic fungi that respond quickly to fluctuating soil conditions [68].

In addition to alpha-diversity patterns, fungal density (cultivable fungal abundance) differed markedly among land-use types, with DP exhibiting substantially higher colony-forming unit counts than agroecological and forested systems. This pattern likely reflects the dominance of fast-growing, opportunistic fungi favored under disturbed conditions, rather than higher ecological integrity [25,69]. DP sites are typically characterized by soil compaction, reduced plant diversity, and simplified organic inputs, which promote r-strategist fungal taxa with high reproductive output and rapid colonization capacity [12,51]. Thus, elevated fungal density in DP coincided with low species richness, indicating numerical dominance by disturbance-tolerant taxa rather than functionally diverse communities [18]. In contrast, agroecological and forested systems supported lower fungal densities but higher taxonomic diversity, consistent with increased niche partitioning, reduced competitive dominance, and more stable community organization [2,3]. These findings underscore that fungal density and fungal diversity respond differently to land-use intensification, and that high fungal density under degraded conditions should be interpreted as a signal of environmental stress rather than ecosystem recovery.

4.2. Land-Use Types Shape Distinct Fungal Community Assemblages

Agroforestry and enriched systems tended to cluster closer to each other in ordination space (Figure 5), reflecting their comparable vegetation complexity and continuous organic matter inputs, factors known to shape fungal niches and promote heterogeneous soil resource environments [51]. In contrast, DP and OF exhibited the most divergent community patterns, likely driven by pronounced differences in disturbance intensity, nutrient availability, and microhabitat heterogeneity. DP systems typically exhibit compacted soils, reduced plant cover, and simplified microbial habitats, conditions that strongly constrain fungal recruitment, persistence and turnover [4,12]. Conversely, OF harbor highly stratified litter layers, stable moisture regimes, and diverse root exudate profiles, all of which favor the establishment of complex and phylogenetically diverse fungal assemblages [3,12,16,51].

The patterns observed in this study reinforce the strong structuring effect of land-use type on cultivable soil fungal community composition in tropical landscapes. Agroecological systems—including IP, CaAS, CoAS, SFAE and MPSE—exhibited the greatest compositional divergence from DP, as reflected by distinct taxonomic assemblages and shifts in fungal composition relative to degraded pastures. These compositional differences indicate that diversified land-use systems support fungal assemblages that are taxonomically differentiated from disturbance-dominated communities [12,16], highlighting the central role of land-use management in shaping soil fungal community identity across tropical agroecological gradients [18,65].

4.3. Agroecological Management Enhances Fungal Network Complexity and Functional Connectivity

Beyond differences in taxonomic composition, analyses of fungal trophic modes and co-occurrence network patterns provide additional insights into how land-use systems influence the structural organization and potential functional attributes of cultivable soil fungal assemblages. The pronounced shifts in trophic mode distribution across land-use types suggest that agroecological management shapes not only fungal diversity but also the functional composition of these communities.

The dominance of saprotrophic taxa, particularly in DP and OF, indicates that decomposer-driven processes remain central across land-use systems, although their relative importance varies with management intensity. The higher relative abundance of saprotrophs in OF likely reflects the sustained availability of organic substrates characteristic of mature forest ecosystems. In contrast, their predominance in DP likely results from simplified vegetation structure and continuous grazing, conditions that may favor generalist decomposers capable of exploiting recurrent organic inputs under disturbance-prone environments [2,20,25,69].

Agroecological and forest-based systems (CaAS, CoAS, SFAE, OF, and MPSE) exhibited higher relative proportions of mixed trophic strategies, particularly taxa combining pathotrophic, saprotrophic, and symbiotrophic attributes. The prevalence of these multi-strategy trophic modes suggests greater ecological complexity, potentially associated with increased plant diversity, structural heterogeneity, and more stable organic matter inputs. Such conditions may favor fungi with broader ecological plasticity, capable of shifting among nutritional strategies depending on resource availability and host interactions [18,22,25,49,51].

Although symbiotrophs were consistently low across systems, their relatively higher representation in SFAE may indicate enhanced plant–fungal associations in enriched secondary forests. Conversely, the higher relative abundance of pathotrophs in IP and SFAE suggests that land-use systems characterized by vegetation restructuring and active management may create ecological conditions that favor parasitic nutritional strategies. Given that pathotrophs include fungi parasitizing a wide range of living hosts—including plants, animals, and microorganisms—this pattern likely reflects shifts in host availability and biotic interactions rather than exclusively plant–pathogen dynamics [25,49,50,51].

Overall, shifts in trophic mode composition across land-use types support the view that agroecological management influences not only fungal diversity but also the functional structure of cultivable soil fungal assemblages. Systems characterized by greater structural and botanical complexity appear to promote a higher proportion of mixed trophic strategies, whereas simplified or intensively managed systems tend to favor dominance by saprotrophic taxa.

Co-occurrence network analyses further revealed that highly connected taxa belonging to Sordariomycetes, Eurotiomycetes, Agaricomycetes, and Dothideomycetes played differentiated structural roles across land-use-specific networks. Taxa affiliated with Sordariomycetes frequently occupied hub positions in pasture and agroforestry systems, consistent with their saprotrophic capacities and contributions to organic matter decomposition under disturbance-driven inputs [70]. In contrast, Agaricomycetes emerged as central taxa primarily in forested systems, aligning with their well-documented roles in lignocellulose degradation and nutrient release in mature forest soils [71]. Eurotiomycetes and Dothideomycetes—groups that include stress-tolerant saprotrophs and plant-associated fungi—also contributed to network organization, reflecting their ecological versatility across contrasting soil conditions [51].

The observation that MPSE, OF, and IP harbored the most complex fungal co-occurrence networks suggests that these systems provide edaphic conditions conducive to high fungal connectivity, functional complementarity, and ecological stability. Network complexity—characterized by increased numbers of nodes, edges, and modules—has been linked to more efficient resource partitioning, enhanced carbon processing, and greater resistance to environmental disturbances [2,3,14]. Moreover, the high modularity observed in MPSE and OF indicates a compartmentalized network architecture, which may buffer ecosystem processes against localized perturbations and enhance system-level resilience [14,69].

In MPSE, elevated network complexity likely reflects a stable hydrological regime, continuous organic inputs, and heterogeneous microhabitats, conditions that promote sustained carbon turnover and the coexistence of multiple fungal trophic modes. Similar patterns have been reported in wetland ecosystems, where complex microbial networks underpin high rates of organic matter decomposition and ecosystem resilience [69]. In contrast, the simplified networks observed in DPs reflect reduced connectivity, diminished functional redundancy, and lower resistance to disturbance, consistent with long-term soil compaction, carbon depletion, and intensive grazing pressure [2,3,14]. Agroforestry systems (CaAS, CoAS, and SFAE) exhibited intermediate levels of network complexity, suggesting that the reintroduction of tree cover and litter inputs enhances microbial connectivity and carbon cycling potential, although longer recovery times may be required to achieve the structural integration observed in natural ecosystems [16,18].

Finally, the balance between positive and negative associations within fungal co-occurrence networks varied consistently along the land-use gradient. Degraded pasture networks were characterized by a high proportion of positive associations, indicating broadly similar environmental responses among taxa, a pattern often linked to simplified assemblages and limited functional differentiation [25,69]. In contrast, agroecological systems (IP, CaAS, CoAS, SFAE, and MPSE) and old-growth forests exhibited networks dominated by negative associations, indicative of stronger niche differentiation and increased competition among taxa for resources [2,16,69]. This shift reflects greater ecological structuring and functional specialization under diversified land-use systems, reinforcing the role of agroecological management in promoting soil fungal networks that are functionally efficient and more resilient to environmental perturbations [22,25].

4.4. Soil Physicochemical Properties Mediate Fungal Diversity, Community Composition, and Network Structure

Our results demonstrate that key soil physicochemical properties exert a strong influence on fungal diversity, taxonomic composition, and co-occurrence network structure, in line with growing evidence that edaphic filters are major determinants of microbial community assembly in terrestrial ecosystems [2,3,14,15]. Moreover, the indirect pathways identified in the SEM indicate that land-use effects on fungal assemblages operate not only through direct management-related impacts, but also indirectly via land-use-driven modifications of soil physicochemical conditions. These conditions act as proximal environmental filters shaping fungal community assembly and network organization across land-use systems [2,16]. Soil fertility indicators such as SOC, TN, and E_Ca, together with physical attributes including BK and silt content, jointly influenced fungal richness, community composition, and co-occurrence structure. Organically enriched and physically well-structured soils tended to support more cohesive and functionally stable fungal networks, likely due to enhanced substrate availability, improved microhabitat conditions, and increased niche differentiation [2,25,69].

Relationships between fungal trophic modes highly connected taxonomic classes, and soil properties further emphasize the role of edaphic filtering in structuring soil fungal assemblages [2,15]. The negative association between pathotrophs and EC, together with the decline of mixed trophic modes such as pathotroph–symbiotroph along soil pH gradients, is consistent with evidence that ionic balance, nutrient availability, and soil chemical conditions can differentially filter fungal life strategies [14,51]. Likewise, the contrasting responses of saprotrophic taxa—showing positive associations with AHS and silt content—suggest that decomposition-related functions are structured along distinct physicochemical gradients of soil heterogeneity. Higher soil moisture and finer particle fractions likely enhance substrate availability, nutrient retention, and microhabitat stability, conditions that favor saprotrophic activity and organic matter turnover [48,49,51]. Moreover, the positive relationship between pathotroph–symbiotroph taxa and SOC and AHS indicates that carbon-rich and moisture-stable environments may support fungi capable of integrating parasitic and mutualistic nutritional modes [51,70,71].

Together, these findings indicate that soil chemical properties exert stronger control over fungal community turnover and network organization than texture-related attributes, underscoring the central role of nutrient availability and chemical balance in shaping fungal ecological strategies [2,14,15]. Consequently, land-use-driven shifts in soil fertility and chemical status emerge as primary environmental gradients structuring fungal assemblages across the landscape [2,16,65]. This highlights the importance of maintaining soil quality for sustaining belowground biodiversity in tropical landscapes and for guiding agroecological management and ecological restoration efforts [72].

These patterns are consistent with the observed variation in fungal richness and network complexity across land-use types. The lowest values recorded in degraded pastures reflect the detrimental effects of soil compaction, nutrient depletion, and vegetation loss on fungal assembly and connectivity [2,3,69]. In contrast, the recovery of fungal richness and network complexity in MPSE, IP, and OF suggests the reestablishment of ecological interactions and nutrient cycling processes driven by increased plant cover, organic inputs, and root exudation [3,16,56].

The positive effect of soil pH on fungal taxonomic composition further supports the role of soil acidity as a key driver of fungal biogeography, as many Ascomycota and Basidiomycota taxa exhibit relatively narrow pH optima that regulate enzyme activity and mycelial growth [51,67,69,73,74]. Conversely, the negative effect of TP suggests reduced fungal turnover under high phosphorus availability, potentially due to competitive exclusion and diminished reliance on saprotrophic or symbiotic nutrient acquisition pathways [14,67].

Finally, the elevated network complexity observed in CoAS indicates that perennial agroforestry systems promote niche diversification and foster complementary associations among soil fungi [18,23]. Similarly, the dense and highly connected networks in MPSE and IP suggest that moderate management intensity combined with continuous organic matter inputs enhances microbial connectivity and ecological resilience [75]. Overall, the gradient observed—from simplified networks in degraded pastures to increasingly complex and interactive networks in agroforestry and forest systems—reflects a progressive restoration of belowground ecological complexity consistent with land-use recovery trajectories in tropical soils [16,18].

4.5. Conservation Implications of Agroecological Systems in Transition Landscapes

Our findings underscore the pivotal role of agroecological land-use systems in maintaining soil fungal biodiversity and ecological functioning during land-use transitions in tropical landscapes. In particular, the high co-occurrence network complexity observed in agroecological systems highlights their potential to enhance ecological stability by fostering structurally robust and functionally diverse fungal networks. Such network configurations are thought to buffer belowground ecosystems against environmental disturbances associated with forest conversion, soil degradation, and extensive cattle ranching [4,12,65]. These results indicate that the promotion of agroecological practices—such as diversified plant cover, continuous organic matter inputs, reduced soil disturbance, and productive restoration strategies—can accelerate the recovery of soil fungal diversity, network organization, and associated ecosystem functions in Amazonian transition landscapes.

From a conservation perspective, our study highlights the value of incorporating soil fungal indicators—encompassing diversity, trophic modes composition, and network attributes—into land-use planning and restoration frameworks. Integrating belowground biodiversity metrics into conservation and agroecological transition strategies can improve the assessment of ecosystem recovery trajectories and support evidence-based decision-making aimed at sustaining long-term soil health and ecological integrity in the Andes–Amazon transition zone.

4.6. Methodological Considerations of the Culture-Based Approach

In this study, culture-based isolation and enumeration represent only a fraction of the total soil mycobiome. As such, we acknowledge that this approach recovers a selective subset of fungal taxa, typically favoring fast-growing saprotrophs while underrepresenting slow-growing oligotrophs, many Basidiomycota, and obligate symbiotic groups. Consequently, the taxonomic and functional patterns reported here should be interpreted as representative of the culturable fraction of the soil fungal community, and caution is warranted when extrapolating these findings to the broader soil mycobiome. Despite these limitations, culture-based metrics remain sensitive indicators of shifts in microbial abundance and community organization under contrasting disturbance and management regimes [18,25]. Moreover, while high-throughput amplicon sequencing captures a broader spectrum of fungal diversity, the culture-dependent approach applied in this study provides direct access to taxonomically verified isolates and enables downstream functional, physiological, and interaction-level assessments. These attributes are particularly valuable for interpreting ecological network structure in agroecological and forest-based systems [69,75]. Increasing evidence indicates that culture-based datasets offer unique insights into fungal physiology, enzymatic capabilities, and biotic interactions—traits that cannot be reliably inferred from sequence data alone [76]. As a result, co-occurrence networks constructed from cultured isolates may capture ecologically meaningful patterns of association and functional complementarity that remain undetectable in metabarcoding-based analyses [76,77]. Together, these considerations underscore the continued relevance of culture-dependent approaches as a complementary framework for advancing fungal ecology, particularly in studies linking land-use management, soil processes, and ecosystem functioning.

5. Conclusions

This study demonstrates that agroecology-based land-use systems strongly influence the diversity, taxonomic composition, and co-occurrence network complexity of cultivable soil fungal taxa across the Andes–Amazon transition zone. Old-growth forests and moriche palm swamp ecosystems supported the highest fungal diversity and, together with improved pastures, exhibited the most complex and structured fungal co-occurrence networks compared with degraded pastures. Agroforestry systems displayed intermediate levels of network complexity, with copoazu-based agroforestry showing the most distinct fungal assemblages among managed systems. Land-use types characterized by greater vegetation heterogeneity and continuous organic matter inputs promoted saprotrophic and symbiotrophic modes, which were associated with denser and more stable fungal network configurations. Soil physicochemical properties—particularly pH and total phosphorus—emerged as key drivers of fungal taxonomic composition.

Overall, our findings highlight that the promotion of agroecological management enhances soil fungal diversity and network stability, reinforcing the ecological sustainability of tropical soils. These results underscore the importance of integrating agroecological practices into land-use planning and conservation strategies to sustain belowground biodiversity and ecosystem functioning in the Andes–Amazon landscape under ongoing land-use change.