Fungal communities decline with urbanization—more in air than in soil

The study addresses how urbanization affects fungal diversity and community composition in both air and soil. Urban biodiversity underpins ecosystem services important for human well-being, yet urban environmental degradation threatens these services and may negatively affect human health via impacts on the microbiome and immune system. Fungi provide critical ecosystem functions (e.g., nitrogen fixation by lichens, nutrient and water uptake via mycorrhizae, plant protection by endophytes, and wood decomposition) and are used as indicators of environmental quality, but comprehensive assessments across urban gradients are scarce due to methodological challenges. The authors aim to quantify differences in fungal species composition, richness, and DNA abundance between urban and natural areas, determine the spatial scale at which communities change from natural to urban environments, and evaluate whether air or soil sampling better captures these changes. They hypothesized that urban areas would harbor poorer fungal diversity due to reduced host resources and higher pollution, and expected stronger dissimilarities in soil than in air given the greater dispersal potential of airborne spores.

Background highlights include: fungi’s diverse ecological roles and use as bioindicators—lichens as air quality indicators, mycorrhizal fungi sensitivity to nutrients and pollutants, and wood-decaying fungi as indicators of forest naturalness. Airborne fungal diversity is particularly high and air is the primary dispersal medium for fungi. Routine air biomonitoring tends to focus on allergy- and disease-causing fungi, leaving broader community composition poorly known. A recent methodological advance—direct spore sampling from air using a cyclone sampler combined with molecular identification—has shown higher detected diversity than substrate-specific DNA sampling, enabling more comprehensive surveys of airborne fungal communities.

Study design: Five Finnish cities (sites) spaced ~100–500 km apart. Within each site, six plots: three urban (settlement areas: noncultivated lawns, roadsides, backyards) and three natural (surrounding forests). Two urban core plots ~1 km apart; two natural core plots ~1 km apart; one urban edge and one natural edge plot ~1 km apart; edge-to-core distance ~10 km. The design yields plot pairs spanning urban-urban, natural-natural, and urban-natural at ~1, 10, 100, and 500 km scales. Sampling period: 16–23 August 2019, synchronized across sites. Sampling: From each plot, three replicate soil and three replicate air samples (total planned 180 samples: 90 air, 90 soil). Soil: remove litter; collect three 2.5 cm diameter cores to 5 cm depth within 1 m square, pool and mix; subsample 2 mL to tube. Air: cyclone sampler collecting particles >1 µm directly to Eppendorf vial, following Global Spore Sampling Project protocol; each air sample accumulated for 24 h. Due to 15 cyclone samplers and 30 plots, air sampling alternated between natural and urban locations but was synchronized across sites. Preprocessing: Soil samples freeze-dried 48 h at 0.57 mbar, −80 °C. Air samples cleaned of large items by adding sterile water, vortexing 10 s, removing visible contaminants with sterile tweezers, then freeze-dried 24 h at 0.57 mbar, −80 °C. Dry samples stored at −20 °C and shipped to University of Guelph for molecular analyses; DNA extraction, PCR, and library prep at Centre for Biodiversity Genomics; Illumina MiSeq at AAC Genomics Facility. DNA extraction and sequencing: Targeted fungal ITS barcode region. Air samples followed Ovaskainen et al. protocol including synthetic DNA spike-in (nine plasmids of synthetic ITS-like sequences) to calibrate raw sequence counts to quantitative DNA estimates (0.1 µl of 0.01 ng/µl spike per 12.5 µl round-one PCR). Soil extractions modified: bead-beating with 3/8" stainless steel bead, ILB+1% PVPP buffer with Proteinase K; homogenization (Genie 2, 5 min), incubations at 56 °C (2 h) and 65 °C (2 h), centrifugation; binding to 1 µm glass fiber membrane with GuSCN buffers; sequential washes; elution in 10 mM Tris-HCl pH 8.0. Some soil DNAs repurified due to humic acids with additional ILB+PVPP and GuSCN binding/washes and elution. Illumina MiSeq amplicon sequencing with primers ITS_S2F, ITS3, ITS4 adding 6 random nucleotides (N6) before Illumina “mis” adapter to increase library complexity for round-one PCR. Bioinformatics: Pairing in Geneious Prime 2019.0.4; discard reads <100 bp; quality trimming (QV20) with BBDuk. Primer trimming with Cutadapt v1.8.1; length filtering (<200 bp) with Sickle v1.33; initial clustering with UCLUST v1.2.22q at 99.5% similarity to reduce computational cost. Spike clusters identified by UBLAST at 95% similarity. For each sample i, total sequences n_i, spike sequences s_i, non-fungal sequences p_i, and fungal sequences f_i; estimate total fungal DNA amount using w_i = f_i/s_i, previously shown to align with qPCR-based absolute DNA estimates. Probabilistic taxonomic placement with PROTAX tailored for fungi to phylum, class, order, family, genus, species with uncertainty estimates; identifications considered reliable at >90% probability. Non-fungal excluded by retaining only sequences reliably classified to a fungal phylum. Constrained clustering: Reliably classified sequences at each taxonomic level served as backbone; hierarchical constrained clustering from class to species. Within each taxon, reliable sequences were clustered to obtain representative sequences. Distributions of within- and among-taxon similarities informed optimal similarity thresholds (equalizing false positives and negatives) per level. Non-reliably classified sequences were mapped (LAST) to representatives; if best similarity exceeded threshold, assigned to that taxon; otherwise de novo clustered with UCLUST at the optimal threshold. Statistical analyses: Excluded six failed samples (<10,000 sequences): five urban soil and one natural soil; remaining 174 samples averaged ~300,000 reads (range 17,000–510,000). OTU abundance computed as product of total fungal DNA amount (from spike-in) and relative OTU abundance among fungal sequences. Because absolute DNA units are not comparable between air and soil, data normalized separately so total DNA amount sums to one within each method. Taxa classification: predominantly air-detectable if at least 10× more abundant in air than soil; predominantly soil-detectable if ≥10× more in soil; natural or urban specialists if ≥10× more abundant in one habitat type; taxa present in fewer than five samples excluded from these classifications. Community analyses: DNA amount data log(x+10) transformed; Euclidean distances computed; ordination via Sammon mapping. Significance of factors assessed with PERMANOVA (adonis, vegan in R): in air- or soil-restricted analyses, factors included habitat type (natural/urban), site, and plot; in combined analyses, also sample type (air/soil) and interactions; OTUs present in fewer than five samples excluded. Richness and DNA amount modeling: Generalized linear mixed models with log-transformed DNA amount (Gaussian LMM) and OTU richness (Poisson GLMM, log link). Models included log sequencing depth as covariate and nested random effects of plot within site. Full models included main effects of sampling method and habitat type (either four-level factor combining natural/urban with core/edge or two-level natural/urban) and their interaction.

• Both aerial and soil fungal communities were markedly poorer in urban than in natural areas.
• Crossing from natural to urban habitats over a 1 km scale led to approximately a fivefold reduction in fungal DNA abundance in both air and soil samples.
• Species richness dropped to nearly half, and fungal DNA abundance to about one fifth, at the 1 km transition from natural to urban areas.
• Contrary to initial expectations, fungal diversity decreased with urbanization more in air than in soil, despite the high dispersal potential of airborne spores.
• A large proportion of fungi detectable in air were specialists of natural habitats, whereas soil fungal communities contained a larger proportion of habitat generalists.
• The pronounced sensitivity of aerial fungal communities to anthropogenic disturbance indicates that air sampling is a reliable and efficient bioindicator of ecosystem health in urban environments.

The study demonstrates that urbanization strongly filters fungal communities, reducing both richness and DNA abundance, and that this effect is spatially abrupt at the natural–urban edge (within ~1 km). The unexpected stronger decline in airborne fungal diversity relative to soil suggests that despite extensive dispersal, many airborne-detected taxa are sourced from natural habitats and are scarce or absent over urban substrates, leading to reduced representation in urban air. In contrast, soil communities include many habitat generalists more tolerant of urban conditions, buffering diversity losses. These findings address the research questions by quantifying compositional, richness, and abundance differences between urban and natural areas, identifying the sharp spatial scale of change at edges, and revealing that air sampling is particularly sensitive to urbanization effects. The results underscore the value of aerial fungal community assessments as indicators of ecosystem health in urban settings and highlight potential pathways through which urban environmental degradation may diminish ecosystem services mediated by fungi.

This study provides quantitative evidence that urbanization substantially diminishes fungal diversity and DNA abundance in both air and soil, with a stronger effect observed in airborne communities. The sharp declines at the urban edge and the prevalence of natural-habitat specialists in air underscore the sensitivity of aerial fungal assemblages to anthropogenic disturbance, supporting air sampling as a powerful biomonitoring tool for urban ecosystem health. Future work could expand temporal coverage to assess seasonality, broaden geographic and habitat scopes beyond Finnish cities, integrate environmental covariates (e.g., pollution, vegetation structure, resource availability), and link community changes to specific ecosystem functions and human health outcomes.

• Geographic scope limited to five Finnish cities; generalizability to other regions and climates may be constrained.
• Short, synchronized sampling window (8 days in August 2019) limits inference on seasonal or interannual variability.
• Six samples failed sequencing (<10,000 reads), potentially reducing power, especially in urban soil.
• Air sampling alternated between natural and urban plots due to a limited number of cyclone samplers, which may introduce temporal allocation effects despite synchronization.
• Absolute DNA amounts are not directly comparable between air and soil; analyses relied on method-specific normalization and spike-in-based estimates.
• Taxonomic assignments, while probabilistic (PROTAX), include uncertainty and depend on completeness of reference databases; taxa present in fewer than five samples were excluded from certain classifications.