The adult microbiome of healthy and otitis patients: Definition of the core healthy and diseased ear microbiomes

Otitis media (OM) and externa (OE) are painful, recurrent ear conditions. As most otitis publications focus on the bacterial content of childhood ears, there remains a dearth of information regarding the adult ear microbiome including both bacteria and fungi. This study compares the outer ear microbiome of healthy adults to adults affected by OE and OM using both intergenic-transcribed-spacer (ITS) and 16S-rDNA sequencing. The adult ear core microbiome consists of the prokaryote Cutibacterium acnes and the eukaryotic Malassezia arunalokei, M. globosa, and M. restricta. The healthy ear mycobiome is dominated by Malassezia and can be divided into two groups, one dominated by M. arunalokei, the other by M. restricta. Microbiome diversity and biomass varied significantly between healthy and diseased ears, and analyses reveal the presence of a potential mutualistic, protective effect of Malassezia species and C. acnes. The healthy ear core microbiome includes the bacteria Staphylococcus capitis and S. capitis/caprae, while the diseased ear core is composed of known bacterial and fungal pathogens including Aspergillus sp., Candida sp., Pseudomonas aeruginosa, S. aureus, and Corynebacterium jeikeium. The data presented highlight the need for early detection of the cause of otitis to direct more appropriate, efficient treatments. This will improve patient outcomes and promote improved antimicrobial stewardship.


Introduction
Acute and chronic ear infections are of considerable importance, among the most common reasons for physician visits and resulting antibiotic treatment [1][2][3]. Otitis was previously considered more prevalent in children under the age of 15 [4], but a recent study found that about half of physician visits for otitis complaints were from adults [2]. The most common ear infections include otitis media (OM) and otitis externa (OE). OM is characterized by middle ear inflammation with subtympanic fluid accumulation, caused or promoted by growth of researchers. When access to raw data files is needed, the interested researcher should submit a request to the corresponding author (T Dawson) or Dr. Yap Ching Chew, COO of Pangea Laboratory LLC. at ychew@pangealab.com. Pangea will grant access to raw data files once a Confidentiality Disclosure Agreement has been signed between the requesting party and Pangea Laboratory LLC. Microbiome abundance tables for both the bacterial and fungal profile for each sample included in this study is provided in the supplemental material.

Sample collection
Samples were collected using the collection kit provided with the PrecisionBIOME NGS Microbial test (Pangea Laboratory, Tustin, CA, USA). This kit contains sterile oropharyngeal and nasopharyngeal swabs and collection tubes pre-filled with a nucleic acid-stabilizing solution called DNA/RNA Shield TM (Zymo Research Corp., Irvine, CA). This solution can preserve the microbial DNA present in the specimen at room temperature for up to 30 days (S1 Fig). Both left and right ears from all healthy participants were swabbed using a sterile nasopharyngeal swab provided with the kit. Swabbing was performed in a standard way by rotating the swab at least five full turns inside the ear canal until swab was saturated. Right after swabbing, the swab was placed inside the collection tube and the rest of the swab stem was discarded at the breakpoint. Tubes were tightly capped and sent for 16S and ITS analysis through the PrecisionBIOME TM NGS Microbial test. A total of 92 healthy ear specimens were collected and analyzed. Specimens from adult subjects with different ear infections including OM and/or OE were collected using a similar procedure as described above. A total of seventy otitis patients (70) with a median age of 64 years and suffering from either OM or OE (Table 1) were recruited for microbial analysis and pathogen identification using the PrecisionBIOME TM NGS Microbial

PLOS ONE
The adult human ear microbiome: Healthy and otitis test. For sample collection, an otoscope and microscope were used to identify the site of infection. For all OM cases, sampling was performed through the outer ear canal into the middle ear. Middle ear fluids were collected by aspiration (tympanocentesis) and placed directly inside the specimen collection tube. In cases of ruptured ear drums, fluids were collected with a flexible nasopharyngeal swab using an auditory speculum. For OE cases, the outer ear canal skin including ear exudates were collected using a sterile nasopharyngeal swab included in the collection kit. Otitis media was the most common type of ear infections analyzed in this study (N = 48/70).

Analysis of microbiota
Microbiota profiling was determined using the PrecisionBIOME™ bioinformatics analysis pipeline. Uclust was used to perform taxonomic classifications using a PrecisionBIOME™ custom proprietary database. Phylotypes were computed as percent proportions based on the total number of sequences in each sample. Relative abundances of bacteria compared to fungi were determined assuming an equivalency of one 16S rDNA copy to one fungal ITS copy. While it is unlikely that the number of 16S or ITS copies change in any given, this is an assumption, which may be considered a limitation of the analysis presented here. Absolute microbial quantification was achieved using a real-time PCR approach using primers targeting the V1-V3 and ITS regions for bacterial and fungal quantification, respectively. Species level resolution of this sequencing approach was previously confirmed by shotgun sequencing [19].

Statistical analyses
Unless otherwise stated, results were expressed as average values with standard deviation. Measurements of α-diversity and evenness were calculated using the Shannon index and the number of observed species. β-diversity was calculated using Bray-Curtis distance at the species taxonomic level for both bacteria and fungi. Linear discriminant analysis (LDA) and effect size (LEfSe) were used to identify taxa that were significantly enriched in each group using the default settings (QIIME version 1.9.1, P value > 0.05 was considered significant) [20]. Analyses of variance and false discovery rate control to correct for type I errors were performed on the species-level relative abundance data of this analysis. Species with a P value < 0.05 were considered significant. A presence-absence data matrix of species by site was generated by assuming species with abundance greater than 1% as presence and less than 1% as absent. The "cooccur" function from R ("cooccur" package in R version 3.5.2, R Core Team, 2013) was used to generate pairwise classification of species having positive, negative, and random associations. The core microbiome was determined based on taxa detected with � 5% relative abundance and in � 50% of all samples. GraphPad Prism (version 8) software was used for data visualization.

Results
To identify possible causative agents of adult ear infections, we compared the bacterial and fungal microbial profiles of three study groups: healthy, asymptomatic adults, otitis media (OM) patients, and otitis externa (OE) patients. NGS analysis detected both bacterial and fungal organisms, with a total of 877 different bacterial and 277 fungal species detected (S1 and S2 Tables). To measure similarities/differences in the microbial composition a β-diversity analysis showed a clear bacterial composition separation of healthy vs infected groups, with OE and OM specimens clustering. No clustering was seen for the fungal component (S2 Fig). As a further control for internal consistency, we compared the left and right samples from healthy ears and found no significant differences (S3 Fig).
Alpha diversity, which measures the diversity within a given sample (as opposed to beta diversity that measures diversity between samples) showed no significant differences for the bacteriome of the study groups (
As was seen in the bacterial analyses, species found in healthy ears were also present in otitis cases, though with lower abundances and frequencies. At the species level, the most abundant Ascomycota fungal species detected in the OE group was Aspergillus sp.  (Fig 1E). In addition to Aspergillus sp. and A red letter (P) highlights those species known to be otitis pathogens. To ease comparison, bacterial and fungal species are consistently colored between graphs. Further information about the relative abundances of all detected species per individual can be found in S1 Table ( 1F). In addition to Aspergillus sp., C. parapsilosis, and C. metapsilosis, three other otopathogens were detected among the 15 most abundant species (Alternaria sp. (1.97%, 4/48), C. albicans (1.62%, 2/48), and Aspergillus flavus (1.51%, 1/48)). Consistent with the OE group, single outliers in the data significantly drive the group-average species abundances.
There was a significant increase in the number of observed species between the healthy and the OM group (Fig 2D). Shannon diversity showed a significant increase in the number of detected species in both OE and OM compared to the healthy group ( Fig 2E). Interestingly, OM and OE groups had a significantly lower fungal biomass as measured by ITS copy numbers compared to the healthy group ( Fig 2F). Therefore, a healthy mycobiome was characterized by a lower fungal diversity and a higher fungal biomass compared to clinically affected samples. This is the opposite to what was found for the bacterial biomass.

Ear microbiome is distinct between healthy and otitis ears
To identify members of the microbiome that could differentiate the study groups, a LEfSe analysis was conducted at all taxonomic levels. The analysis indicated the phyla Actinobacteria

PLOS ONE
The adult human ear microbiome: Healthy and otitis and Firmicutes as significantly enriched in the healthy group, and Proteobacteria in the OE group (Fig 3A). At the species level, 36 significantly distinct bacterial species were identified. For brevity and readability of the graph, only those species that had an abundance of at least 5% in any group are shown (Fig 3B). After applying the 5% filter, four species were significantly enriched in the healthy group, including two commensals, i.e., C. acnes and S. auriscalis, and two otopathogens, A. otitis and S. capitis/caprae. In the OE group, three species were identified, including the two otopathogens S. aureus and P. aeruginosa. OM group was significantly enriched with five previously known otopathogens species, i.e., C. jeikeium, C. striatum, S. marcescens, Streptococcus agalactiae, and Achromobacter xylosoxidans.

PLOS ONE
The adult human ear microbiome: Healthy and otitis The fungal analysis showed that the phyla Basidiomycota were significantly enriched in the healthy group and Ascomycota in the OE group (Fig 3C). Five species were significantly distinct between groups (using � 5% filter). Specifically, M. arunalokei, Malassezia spp., and M. restricta were significantly enriched in the healthy group (Fig 3D). Two potentially pathogenic species were significantly enriched in the OE group, Aspergillus sp. and C. albicans. The OM group had a significant enrichment of the otopathogen C. parapsilosis.

Ear core microbiome
A core microbiome analysis is important for "understanding the stable, consistent components across complex microbial assemblages" and is defined as "the suite of members shared among microbial consortia from similar habitats" [21]. Four species were shared among all groups, C. acnes, M. arunalokei, M. globosa, and M. restricta (Fig 4). Staphylococcus epidermidis was shared between OE and OM. Part of the OE core were Aspergillus sp., P. aeruginosa, and C. pseudogenitalium/tuberculostearicum, and the healthy group had S. capitis and S. capitis/caprae as core members. A taxon from the order Malasseziales was shared only between normal and OM samples.

Microbial composition clustering
Co-occurrence analysis, which highlights interspecies dependencies based on patient health context (Fig 5), showed that interspecies dependencies were health context and species specific. In the healthy group, S. capitis and S. capitis/caprae showed a positive co-occurrence with each other, and negative interactions with M. arunalokei and A. otitis, which was only seen in that group (Fig 5A). The OE group showed two positive interactions, one between C. acnes and S. capitis/caprae, and one between S. aureus and C. simulans (Fig 5B). A negative cooccurrence between M. restricta and the otopathogen Aspergillus sp. was detected, hinting at a protective effect of Malassezia species. Two pathogens from different kingdoms, Aspergillus sp. and P. aeruginosa, had a negative interaction. Interestingly, the OM group was characterized by a positive co-occurrence between S. marcescens and C. acnes. C. acnes further had a negative

PLOS ONE
The adult human ear microbiome: Healthy and otitis interaction with S. aureus. The OM samples were characterized by positive co-occurrences between ear commensals i.e., Malassezia species, and C. acnes and S. epidermidis (Fig 5C). The otopathogen S. aureus had a negative co-occurrence with the commensal C. acnes.

PLOS ONE
The adult human ear microbiome: Healthy and otitis

NGS reveals pathogen dysbiosis by bacterial, fungal, or joint players in otitis patients
Several individual samples were characterized by a very low species diversity, i.e., a single species represented > 90% of either the bacteriome (n = 29) or mycobiome (n = 35). Based on the previous observation that the microbial profile is highly individualized for OE and OM patients (Fig 1B, 1C, 1E and 1F), the bacterial and fungal profiles for each OM and OE patient were analyzed to identify potential pathogens that may be driving the infection in each case. A total of 18% (4/22) of OE patients showed a single bacterial species representing �80% of the total bacteriome, and 32% (7/22) of patients had a single fungal species at �80% of the total mycobiome. Dual bacterial and fungal pathogen dysbiosis was seen in 9% (2/22) of patients ( Table 2). For OM patients, overrepresentation of a single bacterial pathogen was more common with 35% (17/48) of patients, while 15% (7/48) showed a single fungal pathogen overrepresentation (Table 3). There was no fungal otopathogen overrepresentation (�90%) seen in the healthy group and only one sample showed a bacterial pathogen at 92% (A. otitis) (Fig 1A  and 1D). Overall, 38% (11/92) of healthy group samples had a single bacterium, C. acnes, represent �90% of the bacteriome. A possible assumption when such a dysbiosis is detected, is that there may be an overgrowth, i.e., increase of microbial biomass, in the sample. A correlation analysis between the number of observed species and the biomass, as estimated by 16S and ITS copy numbers, however, showed no significant correlation (Bacteria: r = -0.042, P = 0.595; Fungi: r = 0.110, P = 0.162, S4 Fig).
Polymicrobial infections are a concern for OE and OM, which may be either multiple bacterial or fungal pathogens or a combination. OE and OM samples were analyzed for the presence of multiple pathogens in each sample with a cutoff of �10% relative abundance in the bacteriome or mycobiome. All OE patients had at least one bacterial otopathogen present at �10% of the bacteriome, 36% (8/22) had two (�10% each), one patient had three (Fig 6A,  sample A), and one patient showed four otopathogens present at �10% (Fig 6A, sample B). Between the two patients that showed three and four pathogens respectively, there was no

PLOS ONE
The adult human ear microbiome: Healthy and otitis overlap in the pathogens detected. The mycobiome in the OE group showed that 45% (10/22) patients had one and 9% (2/22) patient had two fungal pathogens (Fig 6A).
In the OM group, 40% (19/48) of patients had one bacterial pathogen present at �10%, 21% (10/48) had two and for one patient three otopathogens were detected (Fig 6B). In the mycobiome, 40% (19/48) of patients had one and six individuals had two fungal otopathogens present at �10% (13%). About 35% (17/48) of patients had both a bacterial and a fungal pathogen present at �10% (Fig 6B). As seen for OE and OM polymicrobial analysis, the combination of bacterial and/or fungal pathogens is highly individualized in this data set.

Discussion
Acute and chronic adult ear infections are common, but their etiology and the role of the eukaryotic mycobiome remain inadequately investigated. We investigated microbial community profiles in healthy adult and OM and OE patient ears, and all harbored both bacteria and

PLOS ONE
The adult human ear microbiome: Healthy and otitis fungi. In the healthy ear, the most abundant bacterial species were C. acnes, S. auriscularis, and S. captis/caprae, consistent with previous investigations [22]. The healthy mycobiome was characterized by high numbers of Malassezia with two distinct groups harboring either M. arunalokei or M. restricta. While M. restricta is well characterized and generally regarded as a commensal on normal skin with no reports of it being a human otopathogen, it can be associated with skin diseases like dandruff [23]. M. arunalokei has only recently been identified [24] and remains minimally investigated [25]. Further investigation into the role of these species in healthy ears is warranted. In previous NGS studies M. arunalokei would have appeared as "Malassezia spp." or in unidentified sequence "dark matter". Compared to OE and OM, the commensals C. acnes and S. capitis/caprae were significantly enriched in the healthy group, which confirms previous reports, as well as M. arunalokei and M. restricta. The otopathogens S. auricularis [10] and A. otitidis [11] were also significantly enriched in this group, which was unexpected and highlights the need for continual refinement of research databases and careful consideration of unidentified species. An unexpected frequency of previously postulated otopathogens i.e., C. otitidis or S. caprae, in the healthy ear was detected, however their frequencies and abundances were increased in OE and OM groups. These findings are most likely due to the increased sensitivity of the 16S pipeline [19] versus earlier culture-based methods and the potential lack of testing on clinically healthy ears. This highlights the need for continual refinement of research databases and careful consideration of unidentified species.
In OE the most abundant species were P. aeruginosa, C. acnes, and S. aureus with eight otopathogens among the top 15 most abundant; and in OM by P. aeruginosa, S. aureus, and C. jeikeium, with ten otopathogens among the top 15. The commonality of P. aeruginosa and S. aureus as dominant pathogens in the bacterial analyses of both OM and OE supports previous findings and validates the NGS study results. Previous studies had identified Turicella otitidis, Alloiococcus otitidis, and Staphylococcus auricularis as primary otopathogens, which were present in OM cases at lower abundances [10]. Interestingly, this study only reported OM cases and did not compare results with a healthy age-matched cohort. From our results, these organisms are part of the normal ear flora. This shows that molecular testing has the potential to provide more detailed information about the bacterial etiology of AOM.
The mycobiome was surprisingly distinct between groups. In both OE and OM, Malassezia spp. remained highly abundant mycobiome members, but newly appearing non-Malassezia species were increased in abundance. In OE the two most abundant non-Malassezia species were Aspergillus sp., followed by C. albicans, and Aspergillus spp. in the OM group (17/48 samples). Aspergillus sp. [26] and C. albicans [27], potential pathogens, were significantly enriched in OE, and OM was enriched for the previously characterized otopathogen C. parapsilosis [28,29]. As highlighted by the dysbiosis and overgrowth analysis of individual OE and OM samples, the microbial composition was highly personalized. Therefore, averaging relative abundance across all OM samples may dilute the contribution of kingdom-specific pathogens. This issue may be resolved by analyzing a larger sample size, which may confirm the findings presented here that the microbiome of OE and OM patients is highly individualized. The high degree of inter-subject variances provides a challenge to building microbiome models/profiles associated with each disease state.
High level beta-diversity analysis indicated sample clustering for bacteria but none for fungi. However, the mycobiome alpha-diversity was significantly altered: the number of detected fungal species was significantly higher in the OM group compared to healthy samples, and the Shannon diversity was significantly higher in both otitis groups compared to healthy samples. Thus, healthy ears were characterized by a lower fungal diversity, dominated by two Malassezia species, i.e., M. restricta and M. arunalokei, and higher fungal biomass compared to infected ears. Comparison of subjects with limited microbial diversity (dysbiosis) showed no correlation to healthy or diseased ears, indicating disease is unlikely directly linked to bacterial overgrowth but may indicate a mutualistic fungal/host relationship, where loss of a protective fungal mutualist correlates with the increase in bacterial biomass. Biomass was estimated as bacterial and fungal cell numbers using 16S rDNA and ITS analysis by enumerating copy numbers [30]. While complicated by diversity in fungal ITS copy number, the ITS copy number may also serve as a proxy for fungal cell number [19]. These enumerations showed a significant increase in bacteria and a decrease in fungi for both OE and OM compared to the healthy group. It is important to note that while the 16S bacterial copy numbers were higher than the fungal ITS copy numbers for all groups, it must be considered that due to their significantly larger size (9-12μm versus 0.8-1μm diameter) and the cubic relationship between diameter and mass, each fungal genome represents greater than 1000x more active biomass. Further longitudinal studies and development of more accurate estimates of fungal ITS copy number per cell may help untangle the inter-kingdom relationships [31].
Interkingdom interactions are complex and difficult to decipher via presence/absence analyses, but co-occurrence analysis highlights potential interspecies interaction. These interactions were species and health state dependent. In healthy ears S. capitis and S. capitis/caprae showed a positive co-occurrence, but interestingly negative interactions with the healthy earassociated M. arunalokei. The OE group showed positive interactions between C. acnes and S. capitis/caprae, and between S. aureus and C. simulans. C. simulans is known to cause skin and soft tissue infections [32] but has not previously been identified as an ear pathogen. Based on the data presented here, C. simulans could potentially represent a new biomarker for OE infections. The negative co-occurrence between M. restricta and the known otopathogen Aspergillus was detected, potentially indicating mutualistic benefit of Malassezia. Malassezia have the potential to be commensal (as they are found on all human skin [25]), pathogenic (as in seborrheic dermatitis [33] and Crohn's disease [34]), or mutualistic (as in some cases of atopic dermatitis) [35]. In this study Malassezia, particularly M. arunalokei and M. restricta, appear to most likely be mutualists. Inter-kingdom antagonistic effects were seen between two pathogens from different kingdoms, Aspergillus sp. and P. aeruginosa, but in one case, OM, there was a positive co-occurrence between S. marcescens and C. acnes. It is possible this represents strain specificity in C. acnes, as strains can be either pathogenic or commensal [31]. In the case of OM there were positive interactions between the commensals Malassezia species and C. acnes and S. epidermidis, while the otopathogen S. aureus had a negative co-occurrence with the commensal C. acnes. Specific implications of this finding for diagnostic and treatment purposes remain to be investigated.
The core microbiome represents the stable, consistent components of a microbial community. The shared human ear core microbiome consisted of prokaryote C. acnes and eukaryotic M. arunalokei, M. globosa, and M. restricta, which confirms the potential mutualistic effect of Malassezia species as discussed above.
This detailed ear microbiome NGS analyses revealed new, more complex relationships between microbial community members. Inclusion of the fungal kingdom disclosed significant changes in mycobiome diversity and biomass between healthy and diseased ears, and reveals the presence of a potential mutualistic, protective effect of Malassezia species.

Conclusions
The human ear microbiome remains inadequately investigated and further investigation will be important in definition of ear health and treatment of ear disease. Using NGS as a testing tool could improve treatment outcome, guide the selection of appropriate therapy, and limit inappropriate antibacterial treatments and thus improve global antimicrobial stewardship. This study highlighted the importance of pathogen classification on a per case basis.  Table. Relative abundance table for bacterial profile per sample.  (TXT) S2 Table. Relative abundance table for fungal profile per sample.  (TXT)