This study was approved by the ethics committee of Renji Hospital, School of Medicine, Shanghai Jiaotong University, Shanghai, China (protocol 2017001). All individuals provided informed consent. Healthy young adults aged between 18 and 25 years old living in the same school in Shanghai, China were recruited in January 2018 and February 2023, respectively. General information including age, sex, medical history, and mask wearing status of the previous 6 months was collected by questionnaires. Each participant in 2023 has claimed to be a long-term mask user and has been wearing masks since January 2020, the early stages of the COVID-19 pandemic. Young adults recruited in 2018, before the COVID-19 pandemic, were regarded as mask-free controls. A total of 82 prolonged mask wearers and 172 mask-free controls fulfilling the following criteria were enrolled: (1) no evidence of opportunistic respiratory tract infections or respiratory diseases; (2) no history of antibiotic intake within the previous 30 days; (3) no history of diabetes, malnutrition, immunodeficiencies, or cancers; and (4) no history of smoking. Specimens were collected by professionals using a sterilized cotton swab (COPAN LQ Stuart Transport Swab; COPAN Italia). The swabs were submerged into 1 mL of sterile saline preservation solution immediately for further use. More details on the sampling process are provided in Data Sheet 1 .

Nasal swabs were fully vortexed in 1 mL of sterile saline for 2 min. Samples (100 μL) from each swab were plated on 5% sheep blood agar and incubated at 37°C for 24 h. Eight random colonies were isolated and identified by MALDI-TOF-MS (Bruker Daltonics, Bremen, Germany) in each sample. All identified bacterial isolates are listed in Data Sheet 2 .

DNA was extracted as follows. Nasal swabs were fully vortexed in 1 mL of sterile saline for 2 min. Bacteria pellet was collected by centrifugation for 10 min at 12,000 rpm. The pellets were resuspended in 180 μL of Buffer ATL (QIAamp DNA Mini Kit, QIAGEN 51306) with 5 μL of lysozyme (50 mg/mL, Sigma L6876) and 5 μL of lysostaphin (1 mg/mL, Sigma L4402) for 30-min incubation at 37°C. Then, DNA was extracted according to the manufacturer’s instructions. Hypervariable V3 and V4 regions of the bacterial 16S rRNA gene were amplified using the primer pair 341F (5′-CCTACGGGNBGCASCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′). The amplicons were then sequenced on the Illumina platform with 30,000 reads volume by the same sequencing company.

The data derived from sequencing were processed using QIIME for further analyses (12). Only reads with high-quality bases (Phred quality score≥25) exceeding 90% were retained for analysis. Chimeras were detected and filtrated by USEARCH 8.0 implemented in QIIME. Filtrated sequences were clustered into operational taxonomic units (OTUs) at a similarity level of 99% using the Uclust algorithm implemented in QIIME. A total of 986 OTUs were obtained and the average number of sequences per sample was 24,708. Taxonomic annotation was performed using the SILVA v138.1 database (13). Low-abundance OTUs fewer than 2 counts in at least 10% of samples were filtered. Taxonomic α-diversity was estimated using Shannon and Simpson indices, and compared using the Mann–Whitney U test. β-diversity between groups was measured using principal coordinate analysis (PCoA) based on Bray–Curtis dissimilarity and compared using the permutational multivariate analysis of variance (PERMANOVA) method.

Microbiome taxonomic abundance data were analyzed using the Mann–Whitney U test or the Kruskal–Wallis H test to identify differentially abundant features. MaAsLin2 was used to adjust covariates including age and sex (14). KEGG pathway abundance was predicted using PICRUSt2 (15). The normalized abundance value of the predicted pathways for each sample was visualized in a heatmap.

All statistical analyses were performed in R. Unpaired, two-tailed t-tests were used for the comparison of continuous variables. Chi-square tests or Fisher’s exact tests were used for the comparison of categorical variables. Wilcoxon rank-sum tests or Kruskal–Wallis H tests were used for the comparison of metadata and taxa relative abundance. Associations between mask wearing frequency or duration and microbiome taxonomic abundance data were analyzed with Spearman’s non-parametric correlation tests. The Benjamini–Hochberg procedure [false discovery rate (FDR)] was used to correct for multiple testing of taxon and metadata differences, with significance defined as FDR < 0.05.

A total of 82 prolonged mask wearers (mean age: 19.05 ± 0.68 years) and 172 mask-free controls (mean age: 18.89 ± 0.90 years) were enrolled. Average concentration of PM2.5 and PM10 in Shanghai in the past 6 months before sampling did not show significant difference in the two groups ( Data Sheet 4 ). Moreover, there were no significant differences in age or sex between the masked and non-masked groups ( Table 1 ). Every participant in 2023 self-reported as a prolonged mask user in the past 3 years. The average weekly duration of wearing masks for each masked participant was 16.06 ± 10.89 h in the previous 6 months; 46.34% (38/82) of these mask users wore masks for at least 16 h per week. With the relaxation of COVID-19 response by the Chinese government, a COVID-19 infection peak occurred in January 2023, just before our sampling time. The masked group was composed of 58.54% (48/82) COVID-19 convalescents and 41.46% (34/82) uninfected controls. Average interval between first COVID-19 infection time and sampling time was 44.50 ± 12.34 days; 97.56% (80/82) of participants in the masked group received at least two doses of COVID-19 inactive vaccine ( Data Sheet 2 ).

To verify whether extended mask wearing changes the nasal microbial characteristics of the general population, we determined the nasal microbial profile of 82 prolonged mask wearers and 172 mask-free controls by 16S rRNA sequencing. The α-diversity was significantly lower in the masked than in the non-masked group (Shannon or Simpson diversity index, p < 0.0001) ( Figure 1A ), and there was a significant difference in β-diversity (PCoA of Bray–Curtis distance, R ^2 = 0.44, p < 0.01) ( Figure 1B ).

Analysis of relative abundance at both phylum and genus level revealed significant changes in the abundance of almost all major taxonomic groups, indicating remarkable changes in the microbiome composition after long-term use of face masks ( Figures 1C, D ). The top four predominant phyla—Proteobacteria, Actinobacteriota, Firmicutes, and Bacteroidetes—composed over 99% of nasal microbiota in both groups. We observed significantly increased abundance of the Proteobacteria and Actinobacteriota phylum (FDR < 0.01), as well as decreased abundance of the Firmicutes and Bacteroidetes phylum (FDR < 0.0001) in the prolonged mask wearing group ( Figure 1C ). At the genus level, there were a total of 108 differential genera. Compared to the mask-free group, the abundance of some major genera including Staphylococcus, Pseudoalteromonas, Corynebacterium, Vibrio, and Cutibacterium was significantly increased (FDR < 0.0001) in the nares of prolonged mask wearers. Interestingly, the abundance of most genera such as Bacteroides, Faecalibacterium, Agathobacter, Bifidobacterium, and Klebsiella was significantly decreased (FDR < 0.0001) in the prolonged mask wearing group ( Figure 1D ). Furthermore, nasal microbiomes did not differ significantly by sex in the two groups ( Figures 1E, F ). Sex makes a contribution to the variation of nasal microbiota in both the non-masked and the masked population ( Figures 1G, H ). However, we did not observe significant differences of nasal microbiota between male and female individuals in the two sex-matched groups ( Figure 1F ).

The microbiome composition was further compared by surveying the mask wearing status of the previous 6 months. We observed that duration of wearing masks influenced the diversity of nasal microbiota. Compared to wearing masks for fewer than 16 h per week, wearing masks for at least 16 h per week showed decreased α- diversity (Shannon diversity index, p < 0.05) ( Figure 2A ), as well as significant difference in β-diversity (PCoA of Bray–Curtis distance, R ^2 = 0.03, p < 0.05) ( Figure 2B ). Redundancy analysis was used to illustrate the relationships between duration of wearing masks per week and other host factors, including age, sex, and COVID-19 infection status. The first two axes accounted for 65.37 of the variability. Sex (R ^2 = 0.56, p < 0.001) and duration of wearing masks per week (R ^2 = 0.44, p < 0.001) predominantly contributed to the variations of nasal microbiota profiles of the masked population ( Figure 2C ). After adjusting for sex covariate, the abundance of a total of 15 genera was negatively correlated with the weekly duration of mask wearing including Alloprevotella, Fusicatenibacter, and Colidextribacter, while Aeromonas was positively correlated with the duration of mask wearing (p < 0.05) ( Figure 2D ).

With the purpose of verifying whether COVID-19 infection history affected the nasal microbiomes, the taxonomic abundance and composition of nasal microbiomes were further compared in uninfected and recovered young adults. There was no significant difference between microbial α-diversity or β-diversity ( Figures 3A, B ), and the microbial communities were similar in the two groups ( Figure 3C ). The results suggested COVID-19 convalescents displaying few changes in the nasal microbiota compared to uninfected individuals, indicating that COVID-19 infection history did not affect the composition of nasal microbiota significantly. After sex correction, no significant changes in nasal microbiota were observed either.

By far, we observed significant differences in the composition of nasal microbiota by prolonged mask wearers. To further investigate how nasal microbiota was altered by wearing masks at the species level, we carried out culture-based analysis. A total of 43 bacterial species were successfully cultured and identified ( Data Sheet 2 ). S. epidermidis, S. aureus, Corynebacterium accolens, Corynebacterium propinquum, and S. capitis were the major culturable species in the nasal cavity. Among them, the nasal colonization rate of S. aureus increased from 18.71% (32/171) to 31.71% (26/82) (p < 0.05), and C. accolens increased from 20.47% (35/171) to 42.68% (35/82) (p < 0.001) after extendedly wearing masks ( Data Sheet 3 ), which were also confirmed by analysis of the relative abundance of all culturable bacteria ( Figures 4A, B ). We also observed the decrease of S. epidermidis and Corynebacterium pseudodiphtheriticum in the nares of the mask wearing population (p < 0.05) ( Figures 4A, B ; Data Sheet 3 ). Moreover, six species of Streptococcus genera (S. oralis, S. infantarius, S. mitis, S. pneumoniae, S. pyogenes, and S. salivarius) were isolated in the prolonged mask wearing group, and none in the non-masked group ( Data Sheet 3 ). Taken together, culture-based analysis supported the alteration of nasal microbiota we observed via 16S rRNA sequencing.

The metabolic process was associated with the composition of nasal microbiota (16). To access the potential functional changes in the microbiota after prolonged mask wearing, we employed PICRUSt2 for functional predictions. A total of 393 MetaCyc pathways and 157 KEGG pathways (level 3) were identified in all the samples. Prolonged mask wearing has extremely changed metabolization of nasal microbiomes; 65.61% (103/157) of KEGG pathways changed in abundance significantly (FDR < 0.0001), involving biosynthesis and metabolism of secondary metabolites, signal transduction, membrane transportation, and infectious diseases. The heatmap showed the 15 most differentially abundant KEGG pathways, displaying increased abundance of biosynthesis of unsaturated fatty acids and the D-Alanine metabolism pathway, as well as decreased abundance of the sphingolipid metabolism and linoleic acid metabolism pathway in the prolonged mask wearing group ( Figure 5 ).