Apples are an important temperate tree fruits worldwide1,2. Apple surfaces teem with a wide variety of microorganisms, mainly fungi, that are closely associated with fruit development, post-harvest preservation and quality control3,4,5. Specifically, certain fungi may be recognized as endophytes and bio-control agents, providing a competitive advantage over biotic and abiotic stresses6. Phytopathogenic fungi are able to cause plant disease or lead to fruit post-harvest deterioration, significantly affecting fruit yield, quality and marketing value7,8. In addition, fungi can produce toxic secondary metabolites, namely, mycotoxins9. The accumulation of mycotoxins such as patulin, alternariol toxins, ochratoxins and aflatoxins in fruits and the derived products can seriously affect quality and cause consumer concerns9,10. Despite the improvement of technologies for disease control and post-harvest preservation, the complexities of fungal communities and their diversity on apple surfaces, as well as their potential effect on quality, are just beginning to be revealed.

Previous work has characterized fungi in fruit tissues of tens to a hundred species using traditional culture-based techniques11,12,13. Culture-based techniques play important roles in research related to the diagnosis of fungal infections, the confirmation of pathogenicity, and clarification of certain invasive biological mechanisms14,15. However, only a small proportion (<5%) of microbes are cultivable16, which prevents researchers from understanding the overall fungal community. In recent years, next-generation high-throughput DNA sequencing techniques with improved sequencing capability have been highlighted in microbial community analyses17,18,19,20. Fungal ribosomal DNA (rDNA) internal transcribed spacer sequences (ITS) are regarded as the credible regions for fungal identification at the species level21. Next-generation sequencing platforms (including SOLiD, Illumina and 454 sequencing) coupled with a powerful database and user-friendly software are beneficial for the analysis of a complex microbial community at deeper, more comprehensive level17,22. To date, next-generation high-throughput DNA sequencing techniques have been successfully applied in microbial community analyses of various environmental and organic samples, such as soil23, air18, water24, gut20,25, leaves and roots26. Over the past decade, several studies of fruit microbial communities have been conducted using next-generation sequencing methods27,28,29. However, only a few studies have reported the fungal diversity of apple fruit. Glenn et al. first studied the microbial diversity of apple leaves and fruit using a sequencing technique, but the samples were subject to cryopreservation over an extended period30. Abdelfattah et al. recently reported the fungal communities of organically and conventionally grown apples at the consumer point of purchase using ITS1 sequencing31. However, fungi on apple surfaces are still not well known. Furthermore, the effects of environmental factors associated with rural and peri-urban planting on the apple-surface fungi must be further explored.

We expected the fungal community on apple surfaces to be determined by various factors. Fruit species28, location32, orchard production strategies19, and organic/conventional agricultural practices32 have been investigated and demonstrated to be important factors that affect microbial community and diversity. Based on FAO statistics ( for 2014, China is the largest producer of apples and devotes the largest harvest area to apples worldwide, accounting for 48.4% (4.09 × 107/8.46 × 107 tonnes) and 45.0% (2.27 × 106/5.05 × 106 hectare) respectively. In China, many apple trees are planted on hilly land because of ancient planting habits and land-use policy33. The Bohai Bay area and Northwest Loess Plateau region— largely hilly, rural areas— are mainly Chinese apple-producing areas (accounting for approximately 80% of the total acreage)34. In recent decades, many peri-urban areas have become sightseeing and fruit-picking parks, as part of efforts to explore tourism resources35,36. Many apple orchards have recently been planted near cities to attract citizens to fruit-picking activities37. Driven by economic interests, the shift in apple fruit cultivation areas from rural to peri-urban has recently become imperative. However, a variety of urban pollutants such as organic toxins, harmful gases, heavy metals, and pathogenic microorganisms are contaminating water, soil, and air near the city38,39. Several studies have demonstrated that heavy metals and organic toxins in peri-urban areas can ultimately contaminate agricultural products grown in these regions40,41. However, no reference exists to compare the microbial differences among agricultural produce grown in peri-urban and traditional planting areas. The microbial security of peri-urban planted produce requires further attention.

The objective of the present study was to evaluate the fungal community and diversity on the surface of apples at the point of harvest using fungal ITS1 sequencing. In our study, ‘Fuji’ apple samples were grown in two different geographical locations (Fig. 1)42. One orchard is far from the city (in a rural area), and the other is near the city. The samples were treated with the microbial community collection and analysed immediately after harvest. We focused on the following questions: (i) What are the fungal community and diversity on apple surfaces at the point of harvest? (ii) How do the fungal communities on the apple surfaces vary according to different factors of the rural/peri-urban environment? This study highlights the overall fungi communities on apple surfaces and compares the fungal differences of apples grown under various environmental conditions to provide a basis for further evaluation.

Figure 1
figure 1

Locations of two orchards (rural orchard and peri-urban/suburb orchard) in Xingcheng, Liaoning, China, sampled in this study. Shen Youming created this figure. The simplified Chinese map was generated by modifying the map described by Zhang et al.42 by using Adobe PhotoShop (version CS5).


ITS-based fungal community sequencing statistics

A total of 538,433 ITS reads were obtained from eight samples, with an average of 67,305 reads per sample. After quality control, we obtained 532,273 high-quality sequences (98.9% of total reads), with an average of 66,543 per sample. The length of the high-quality reads was mostly distributed between 200 and 300 bp. After filtering the rare OTUs (≤0.001% of the total sequences, ≤4 hits in this study), a total of 530,764 sequences were clustered in 421 identified OTUs taxonomies. The fungal OTUs’ taxonomic compositions and abundances are summarized in Table S1. The clustered OTUs of peri-urban samples (ranging from 204 to 220 OTUs and averaged 213 OTUs) were lower than those of rural samples (ranging from 215 to 263 OTUs and averaged 234 OTUs) (one-tail t-test, p = 0.05). The shared OTUs in samples were determined via a Venn diagram (Figure S1). A total of 294 OTUs overlapped between rural and peri-urban samples, occupying 69.8% of all OTUs (Figure S1a). A total of 55 OTUs were shared by all the samples (Figure S1b).

Fungal diversity on the surface of apple fruits

Rarefaction curves, rank abundance curves, and alpha diversity indexes were applied in fungal alpha diversity analyses. The rarefaction curves and rank abundance curves all had a steep slope at the beginning, followed by long, flat tails, indicating that an increase in the amount of sequencing data would not yield more new OTUs (Figure S2). The rarefaction curves tended towards saturation, illustrating that the current sequencing depth was sufficient for fungal diversity investigation (Figure S2a). As shown in Figure S2b and Table S1, the majority of OTUs were detected with low abundance. The ACE and Simpson indexes of samples were calculated, and the results are summarized in Figure S3. The indexes of ACE were higher in rural samples than peri-urban samples, with a significantly difference in one-tail t-test (p = 0.042), indicating rural samples had an increase in fungal diversity.

Overall characteristics of fungal community composition

A total of 360 fungal species belonging to four phyla, 17 classes, 50 orders, 79 families, and 111 genera were detected on the apple-surface samples. Ascomycota was the most abundant phylum, accounting for 90.2% of the total sequences. Basidiomycota was detected with relative lower abundance, accounting for 8.4% of the total sequences. We also detected the minor phyla Chytridiomycota and Rozellomycota. The distributions and percentages of predominant fungi at different classification levels (relative abundance ≥ 0.5%, including two phyla, 8 classes, 10 orders, 11 families, 12 genera, and 16 species) of rural and peri-urban samples are shown in Fig. 2. The relationship between fungal evolution and the fungal abundance of the two group samples were fully displayed by MEGA 5, and the classification taxa are shown in Figure S4. Additionally, a hierarchical tree of the overall fungal taxa composition at each taxonomic level was constructed by GraPhlAn, and the results are shown in Figure S5. The 20 most abundant taxa were marked by letters (A-T) and patterned using colour nodes (Figure S5). Moreover, the 50 most dominant fungal genera, coupled with cluster analysis are shown in the heat map (Fig. 3). Accordingly, certain genera were simultaneously enriched in samples; for example, Nectria, Stilbella and Exobasidium were enriched in sample suburb 4, and Hannaella, Ramichloridium and Sphaerulina were enriched in sample rural 2. Certain genera such as Acremonium and Tilletiopsis were enriched in samples from the peri-urban orchard, whereas the genera Periconia, Coniothyrium and Phoma were enriched in rural samples. The heat map also shows that samples from the same orchards tended to cluster together. Table S1 shows that the 5 most predominant genera on the apple surface included Acremonium (21.4%), Aureobasidium (17.8%), Cryptococcus (2.1%), Tilletiopsis (1.5%) and Phoma (1.0%), together accounting for 43.8% of the total sequences. As shown in Table S1, the majority of genera were detected with relatively low abundance.

Figure 2
figure 2

Distributions of predominant fungi (relative abundance ≥0.5%) at different taxonomic levels (phylum, class, order, family, and genus) of samples from two orchards. The significances were tested by one-tail Student’s t-test and marked as follow: p < 0.05 *p < 0.01 **p < 0.001 ***p ≥ 0.05 without a mark.

Figure 3
figure 3

Heat map showing relative abundance of the 50 most dominant fungal genera in eight samples. Blue represents a fungus of relatively low abundance, and red represents a fungus of relatively high abundance. Cluster analyses of samples (vertical) and classification units (horizontal) were performed according to similarity.

Association networks among fungal genera

Fungal co-occurrence networks among the 50 most abundant fungal genera are shown in Fig. 4. A total of 26 interaction-pairs related to 22 genera were obtained (rho > 0.8 and p < 0.01), mainly including Acremonium, Aspergillus, Aureobasidium, Entyloma, Penicillium, Phoma, Stephanonetria, and Tilletiopsis. There were 14 pairs of cooperative relationships and 12 pairs of competitive relationships. Among them, Cryptococcus, Paraphaeosphaeria, Ramichloridium, Septobasidium and Stephanonectria exhibited a higher degree of linkage (≥4 linkages) with other genera. As shown in Fig. 4, the genera Cryptococcus, Dioszegia, Hannaella, and Ramichloridium were generally observed to cooperate with other genera, whereas the genera Acremonium, Paraphaeosphaeria, Septobasidium, and Stephanonectria were observed to compete with other genera.

Figure 4
figure 4

Network diagram of the 50 most dominant genera showing cooperative and competitive associations. Red represents cooperative association, and green represents competitive association.

Environment-related changes of fungal communities

The beta diversity of fungal communities among samples was determined by unweighted UniFrac distances principal coordinates analysis (UniFrac PCoA) (Fig. 5). The UniFrac PCoA showed that samples from the two orchards had significant differences in fungal communities. These significant intergroup differences were further tested using several types of statistical analyses, including Adonis permutational multivariate analysis (Adonis/PERMANOVA), analysis of similarities (ANOSIM), and random forest analysis. The Adonis/PERMANOVA was scored with p = 0.013 (p < 0.05, in significant level), and the ANOSIM’s R statistical analysis was scored with R = 0.925, p = 0.01 (p < 0.05, in significant level). The random forest analysis indicated no estimated error or class error, further demonstrating the statistical significance of the fungal differences between samples. Cluster analysis of the 50 most dominant fungal genera indicated that samples from the same orchard trended to cluster together (Fig. 3). The significantly differently abundant taxa between samples from rural and peri-urban orchards were obtained using linear discriminant (LDA) effect size analysis (LEfSe), and the results are shown in Fig. 6. A cladogram presented 18 significantly different taxa between two groups. Fourteen taxonomies were remarkably enriched in rural samples including the two orders Pleosporales and Septobasidiales; the 5 families Helotiaceae, Montagnulaceae, Leptosphaeriaceae, Phaeosphaeriaceae, Septobasidiaceae; the class Pucciniomycetes; and the 6 genera Septobasidium, Articulospora, Sclerostagonospora, Coniothyrium, Ramichloridium, and Paraphaeosphaeria. The other four taxonomies were enriched in peri-urban samples including the 2 families Bionectriaceae and Microstromataceae, and the 2 genera Stephanonectria and Microstroma.

Figure 5
figure 5

Beta diversity analysis was performed by principal coordinate analysis (PCoA) of the unweighted UniFrac distance among samples.

Figure 6
figure 6

Comparison of fungal variations between rural and peri-urban samples using LEfSe. Differences in taxa are represented in red (peri-urban) and green (rural) colour. (a) Taxonomic cladogram presenting significant differences between groups. (b) Histogram of LDA scores (logarithmic LDA score > 2.0 and p < 0.05) for differentially abundant features between groups.

Compositional variations of fungal genera between rural and peri-urban samples

The differential fungal compositions at the genus level were analysed using Student’s t-test, and the results are listed in Table 1. A total of 15 genera statistically differed in abundance between rural and peri-urban samples. Ten genera were enriched in rural samples, namely Articulospora, Coniothyrium, Cryptococcus, Dioszegia, Entyloma, Paraphaeosphaeria, Periconia, Phoma, Sclerostagonospora and Septobasidium. The five genera Acremonium, Aspergillus, Microstroma, Penicillium, and Tilletiopsis showed a significantly large distribution in peri-urban samples.

Table 1 Differential fungal compositions between groups at the genus level. Rural RA/Peri-urban RA represents the average relative abundance of rural/peri-urban samples at the genus level. The p-values were generated by one-tail Student’s t-test (p < 0.05 indicates significance).


The diversity of fungi on fresh apple surfaces is closely associated with fruit development, preservation and quality control3,4,5. A comprehensive and thorough investigation of fungal communities on apple surfaces is essential to improving fruit cultivation and disease prevention practices. Therefore, we evaluated the fungal community and their diversity on apple surfaces and identified the potential effects of peri-urban environments on fruit fungal signatures, which helps determine the complexity of fruit fungal communities, identify potential risks and improve fruit quality.

We investigated the overall fungal community using the Illumina sequencing technique. Compared with previous culture-based techniques11,12,13, the ITS1 sequencing technology is efficient and less labour-intensive and provides deeper insight into fungal community and diversity. The obtained OTU rarefaction curves (Figure S2a) tended towards saturation, indicating that the current sequencing depths were sufficient to detect the majority of fungi43. BLAST searching with 97% identification classified the fungi at the species level, in agreement with many previous studies44,45. We obtained a vast number of low-abundance OTUs (Table S1), which contributed the majority of fungal diversity. A total of 111 fungal genera belonging to 4 phyla were identified, indicating the remarkable fungal diversity on the apple surfaces. The present results pertaining to fungal profiles on apple surfaces are similar to those reported by Abdelfattah et al. regarding apple peel, who indicated comparable high-quality sequences ranging from 65,419 to 129,369 and identified similar fungi in 4 phyla and 24 predominant genera31. Previous reports on fruit fungal studies have mainly been conducted using culture-based techniques and have identified several fungi, such as Alternaria, Aspergillus, Cladosporium and Penicillium11,12,13. The most serious drawback of culture-based techniques is that they only quantify proportions of fungi belonging to specific cultivable taxonomic groups, which are mainly determined by the culture media used28. Significant differences in fungal profile have been reported in several studies of fruits29,46 and grains47,48, using similar DNA sequencing methods. Abdelfattah et al. reported on fungal communities on strawberry, obtaining 218,164 high-quality sequences, and identifying fungi mainly belonging to the genera Botrytis and Cladosporium29. Xing et al. reported on fungal communities on peanut kernels, obtaining an average of 36,718 reads (belonging 196 OTUs) and identifying fungi belonging to 37 genera48. As discussed in the literature, such fungal differences might be related to plant species28, environmental factors27,48 and farming systems49. Moreover, differences in plant pattern50, sample size51, sampling time point27, and extraction method can potentially affect fungal diversity.

Ascomycota was the predominantly detected phylum, which was present in all samples, coinciding with the results of Abdelfattah et al.31. Ascomycota is the largest phylum of fungi kingdom, which contains approximately 64,000 fungal species52. Basidiomycota was detected with relative lower abundance. Genera of Tilletiopsis, Cryptococcus and Gyrophanopsis were mainly detected in Basidiomycota. Tilletiopsis (mainly enriched in peri-urban samples) might cause white haze disease in apple53. Cryptococcus (mainly enriched in rural samples) in particular contains several potential pathogenic species (such as Cryptococcus gattii and Cryptococcus neoformans), which are agents of human Cryptococcosis54. The 50 most predominant genera shown in the heat map may play important roles in maintaining fruit fungal stability. As shown in Fig. 3, the presence of potential pathogenic fungi may be related to a variety of plant diseases or fruit post-harvest rots; these fungi include Acremonium55, Aspergillus10, Entyloma56, Exobasidium57, Microstroma58, Mycosphaerella59, Nectria55, Penicillium10, Phoma60, Podosphaera60, Ramichloridium61, Sarocladium62, Stemphylium63 and Tilletiopsis53. In particular, Acremonium (mainly enriched in peri-urban samples) might be associated with black spot disease in apple fruit55,64; Penicillium and Aspergillus (mainly enriched in peri-urban samples) might be associated with apple post-harvest deterioration and rot10; Mycosphaerella causes apple plant diseases59; Nectria might cause apple tree canker65; Phoma (mainly enriched in rural samples) causes apple rot66; and Podosphaera causes apple mildew60. The potential endophytes living in plant tissues might play essential roles in disease control, particularly in Aureobasidium67, Paraphaeosphaeria68, Stephanonectria69, and Talaromyces70,71. Aureobasidium (enriched mainly in rural samples), covered with a slimy mass of spores, has been used in the biological control of plant and post-harvest diseases67,72. Potential entomopathogenic fungi genera, including Cordyceps73 and Simplicillium74, are parasitic toward insects and other arthropods and are regarded as beneficial for pest control. Potential human foodborne pathogenic fungi include Aspergillus75, Candida76, Cryptococcus54, Penicillium75, Phaeoacremonium77, Rhodotorula78, and Wallemia79. Compared with apple fungi profiles reported in previous studies31,46, we observed more detailed fungal communities and classifications. However, the relationships between the obtained fungal communities and those described by Abdelfattah et al.31 were difficult to define because many factors influence fungi formation or sample isolation. To understand the full community, its diversity and the corresponding ecological interactions, fungi on apple surfaces require further study.

Inter-genus interactions are extremely important in shaping fungal dynamics. The network analysis results shown in Fig. 4 clearly demonstrate potential intra-genus relationships. The entangled interactions were difficult to identify because one genus can directly or indirectly influence another genus. Despite the complicated networks of many fungal genera, we compared our results with those of several other related reports. We observed that Stephanonectria was competitive with both Entyloma and Phoma. Stephanonectria was reported as an endophytic fungi69,80, whereas Entyloma56 and Phoma60 were reported as plant pathogens. From this point of view, it is reasonable to propose that Stephanonectria is a beneficial symbiont and might potentially protect apple fruit from diseases caused by the pathogens Entyloma and Phoma. The genus Paraphaeosphaeria was observed to be competitive with Acremonium. Paraphaeosphaeria was reported to be an endophytic fungi and yielded antifungal metabolites68,81. Acremonium was reported to cause black spot disease in apple fruit64. This result indicates that Paraphaeosphaeria might be used as a bio-control agent for controlling Acremonium-caused black spot disease. The genus Aureobasidium was used in the biological control of plant and post-harvest diseases67,72. We observed that Aureobasidium was competitive with Aspergillus, consistent with previous reports indicating competition between Aureobasidium and Aspergillus82,83. These results generally indicate that protective fungi guard fruit against pathogenic fungi. In fact, the inter-genus interactions obtained in our study require further investigation and verification. The present network analysis provides a basis for further fungal interactions analysis and fruit disease bio-control.

Environmental-related changes of fungal communities were tested by comparative analyses. Using Student’s t-test, we found that OTU numbers and ACE indexes of rural samples were higher than those of peri-urban samples. These results indicate that rural samples share a more diverse fungal community than peri-urban samples do. Similarly, several studies reported that organic-labelled fruits and vegetables have a greater richness of microbial OTUs than conventionally labelled produce28,32. From a bacterial point of view, a more diverse community might represent a more stable, healthier ecosystem28,84. The obtained results indicate that rural areas might provide a healthier environment for apple growing. Therefore, the rural orchard environments examined in this study are beneficial for apple fungal diversity.

Beta diversity revealed significantly different fungal communities between rural and peri-urban samples. A total of 294 OTUs (69.8% of 421 OTUs) overlapped between samples from the two different orchards (Figure S1a), indicating the existence of considerable amounts of distinguished OTUs (approximately 30%). The results of the beta diversity analysis (Fig. 5) and the heat map of the 50 most dominant fungal genera (Fig. 3) tend to cluster samples from identical orchards. Additionally, these results are generally consistent with the results of the Adonis/PERMANOVA, ANOSIM and random forests analyses. LEfSe results showed significantly different classification taxa between rural and peri-urban samples (Fig. 6). Overall, these results indicate a significant difference in fungal communities on the apple surfaces from two types of peri-urban and rural environments. Previous studies have indicated that fungal structural differences might be related to factors such as fruit species28, farming practices19, and climatic or environmental conditions32. In the present study, different levels of environment pollution in peri-urban and countryside areas were regarded as one of the most distinguishing conditions, over farming practices, climatic conditions and others. We observed that the fungal differences identified on apple surfaces from peri-urban and rural orchards might be due to different levels of pollution in their respective environments. Further analysis of fungal communities on produce and their diversity in several environments should be tested to confirm this theoretical mechanism.

Different fungal compositions at the rural and peri-urban genus level (Table 1) indicate that environmental factors significantly affect fungal compositions. A total of 10 genera were enriched in rural samples, whereas a total of 5 genera were enriched in peri-urban samples. Notably, the genera enriched in rural samples were reported as protective fungi, mainly Coniothyrium85, Paraphaeosphaeria68,81, Periconia86 and Sclerostagonospora87. The genus Coniothyrium might have the ability to control fruit disease. For example, the Coniothyrium minitans species has been widely reported as a bio-control agent against the plant-pathogenic fungus Sclerotinia sclerotiorum88. Genera of Paraphaeosphaeria68, Periconia86 and Sclerostagonospora87 have been reported to be endophytic fungi. Entyloma56 and Phoma60 (enriched rural samples) have been reported to show pathogenic features. Articulospora, Dioszegia and Septobasidium have rarely been studied. In contrast, the fungal genera enriched in peri-urban samples have all been reported to exhibit pathogenic features and are associated with fruit corruption. Specifically, Acremonium might be associated with black spot disease in apple fruit55,64. Aspergillus and Penicillium are the major microbes causing fruit deterioration and the accumulation of mycotoxins (mainly patulin, ochratoxins and aflatoxins)10. Tilletiopsis might cause white haze disease in apple53. Microstroma has been reported to cause walnut downy leaf spot58. As a result, we observed significantly different fungi genera enriched in peri-urban and rural samples. Certain fungal genera enriched in rural samples have been reported to exhibit protective functions. Conversely, the genera enriched in peri-urban samples have mainly been reported to exhibit pathogenic features and are associated with fruit corruption. Therefore, the rural environment provides certain advantages in maintaining healthier fungal communities on apple surfaces. Apple are grown in peri-urban environments carry potential microbiological risks.

In conclusion, fungal ITS1 region sequencing techniques were successfully applied to the study of fungal community and diversity on apple surfaces. We observed high fungal diversity on the surfaces of apples, which enhances our understanding of fungal communities on apples. Furthermore, several potentially protective and pathogenic genera were identified. Co-occurrence relationships indicate that protective fungi guard plants against pathogenic fungi. The diversity analyses reveal the presence of significant differences in fungal community and diversity between samples from rural and peri-urban orchards. The rural samples maintained more protective fungal genera on apple surfaces, whereas the peri-urban planted apples carried potential microbiological risks. We observed that the fungal differences identified on apple surfaces from peri-urban and rural orchards were due to different levels of pollution in their respective environments. However, further studies should be valuable in confirming this theoretical mechanism. This study will help to reveal the complexity of fungal communities on apple fruit and shed light on ways to improve fruit disease prevention and quality control.


Orchard characteristics and sample treatment

This study was conducted in two ‘Fuji’ apple orchards: the first, located near the city comprises approximately 5 acres of flat land at the Hot Springs Experimental Base of the Institute of Pomology, Chinese Academy of Agricultural Sciences (E: 120.72, N: 40.61); the second, located in a rural area, spans approximately 10 acres on the sunny side of a hilly, gentle slope, approximately 40 km away from the city (E: 120.14, N: 40.65) (Fig. 1)42. A total of 24 trees selected in our study were evenly distributed between the two orchards (not on the border). The orchards were supervised under conventional management; the pesticides λ-cyhalothrin, pyridaben and carbendazim were utilized within the recommended concentrations for insect and disease control. Four replicates of apple samples were collected from each orchard on October 10th, 2016.

Sample preparation

Eight samples (4 replicates × 2 orchards) of apple-surface microbial DNA were collected after harvest. The microbial samples were gathered by wiping or swabbing each apple with a pre-moistened cotton swab. Swab samples were collected and stored at −40 °C for less than two weeks before microbial DNA extraction. Microbial genomic DNA from swab was extracted using the MoBio Power Water® DNA Isolation kit (MoBio Laboratories, Inc., Carlsbad, CA, USA) according to the manufacturer’s instructions. The concentration and molecular size of DNA extraction were measured using an NC 2000 spectrophotometer (Thermo Scientific Fisher, Waltham, MA, USA) and 1.0% agarose gel electrophoresis, respectively. The spectrophotometric A260:A280 ratios were above 1.8, and DNA concentrations were above 80 ng/µL, indicating that the DNA extractions were sufficient for subsequent analysis. Finally, the DNA extractions were stored at −80 °C until further use.

DNA amplification and sequencing

The fungal ITS1 was PCR-amplified according to processes described previously89. The primer pairs were ITS5F: 5′-GGAAGTAAAAGTCGTAACAAGG-3′ and ITS1R: 5′-GCTGCGTTCTTCATCGATGC-3′. The PCR incubation system contained 5 U of DNA polymerase (Pyrobest TaKaRa, Japan), 15 pmol of both primers, 2.5 mM of dNTP mixture, 10 μL of 10x Buffer II, and 40 ng of template DNA for a total volume of 25 μL. Next, a PCR series was performed using an ABI 9600 instrument under the following conditions: first denaturation at 94 °C for 4 min; 25 cycles of denaturation at 98 °C for 45 s, annealing at 60 °C for 45 s, extension at 72 °C for 45 s; and final elongation at 72 °C for 8 min. The amplicons were purified with Agencourt AMPure Beads (Beckman Coulter, Indianapolis, IN) by following the recommended procedures. The obtained ITS amplicons were sent to Shanghai Personal Biotechnology Co., Ltd. (Shanghai, China), and sequenced on an Illumina MiSeq paired-end sequencing platform.

Data quality control and analysis

Data quality control and analysis were mainly conducted using the software package QIIME ( The raw data were first filtered by removing sequences of base quality ≤ Q20, without primers, and with terminals mismatched with both primers (by FLASH v1.2.7, To obtain high-quality sequences, we discarded low-quality sequences, including sequences shorter than 150 bp, sequences with any ambiguous bases, sequences containing more than 6 mononucleotide repeats, sequences with a Phred score below 25, and chimeric sequences (by QIIME). These high-quality sequences were clustered into operational taxonomic units (OTUs) by the QIIME-uclust OTU-picking workflow at 97% sequence identity44,92. OUT taxa were assigned by BLAST searching against the UNITE database using the best hits (Release 5.0, The OTU taxonomy was simplified by removing rare OTUs (≤0.001% of the total sequences) to reduce the complexity of the subsequent analysis93. Qualitative and quantitative information about the OTUs is listed in Table S1 and was used for subsequent analysis.

Based on the OTU results, rarefaction curves and rank abundance curves43 were plotted to explain both the richness and evenness of the fungi community. The alpha diversity indexes, including ACE94 and Simpson95 indexes, were calculated to investigate the sequencing depth and fungal diversity. Student’s t-test and ANOVA were used to identify significant differences in alpha indexes and differential fungal compositions at the genus level of different groups (p < 0.05). Venn diagrams were generated by the mothur software package to identify shared and unique OTUs among groups92. The relationship between fungal evolution and taxonomy abundance was displayed using MEGA 5 (; a hierarchical tree was constructed using GraPhlAn96. Beta diversity analysis was conducted to investigate the similarity of fungal communities among samples using UniFrac PCoA97. The differences in multi-group UniFrac distances for pairwise comparisons between groups were determined using Student’s t-test and the Monte Carlo permutation test with 1,000 permutations, and visualized using boxplots (by QIIME)98. Adonis permutational multivariate analysis (Adonis/PERMANOVA)99, analysis of similarities (ANOSIM)100, and random forest analysis101 were performed to evaluate the fungal similarity among samples. LEfSe was used to detect differentially abundant taxa using the online Galaxy workflow framework (logarithmic LDA score > 2.0 and p < 0.05) ( The Spearman rank correlation coefficients of the 50 most abundant genera were calculated to observe the potential cooperative and competitive network among the genera using the mothur software package and visualized by Cytoscape (rho > 0.8 and p < 0.01) (

Data availability

All data generated or analysed in this study are included in this published article (and its Supplementary Information files).