Distinct Ocular Surface Microbiome in Keratoconus Patients Correlate With Local Immune Dysregulation

Abstract

Purpose: Keratoconus (KC) is characterized by irregular astigmatism along with corneal stromal weakness and is associated with altered immune status. Tissue resident microbiomes are known to influence the immune status in other organs, but such a nexus has not been described in ocular conditions. Therefore, we examined the ocular surface microbiome of patients with KC and correlated it to the immune cell and tear molecular factor profiles.

Methods: Sixty-two patients with KC and 21 healthy controls underwent corneal topography analysis and eye examination followed by a collection of Schirmer's strip, ocular surface wash, and ocular surface swabs. Microbiomes were analyzed by extracting DNA from the swabs followed by 16S rRNA gene V3-V4 amplicon sequencing and analyzed using QIIME. Fifty-two molecular factors from Schirmer's strip tear extracts and 11 immune cells from ocular wash were measured using multiplex ELISA and flow cytometry. Alpha diversity, linear discriminant analysis effect size (LEfSe), relative abundance and receiver operating characteristic – area under the curve (ROC-AUC) analysis were performed. Unsupervised clustering at the genus level with clinical parameters, soluble factors, and immune cells was performed.

Results: Fifty-two phyla/class, 132 order, 283 family, and 718 genera were identified in our cohort. Alpha diversity indices were comparable between patients with KC and the healthy controls. Dominant phyla across groups were Actinobacteria, Proteobacteria, Firmicutes, and Bacteroidetes. Alphaproteobacteria increased in KC eyes whereas Actinobacteria, Firmicutes_Bacilli reduced compared with the healthy controls. We found a significant positive correlation of Microbacterium, Cutibacterium, and Brevundimonas genera abundance with keratometry and corneal thickness. Levels of IL-21, IL-9, Fractalkine, and VEGF positively correlated with Tetrasphaera (P < 0.05). β2-microglobulin and CD66b^high cells correlated with Bacteroides (P < 0.05). CD45⁺ cells correlated with Escherichia_Shigella (P < 0.02).

Conclusions: We discovered a unique microbiome signature of KC which correlated to disease grades and secreted molecular factors and immune cells. Therefore, the altered microbiome on the ocular surface may drive immune dysregulation in KC and provide scope for potential interventions in the future.

Keratoconus (KC) is characterized by asymmetric, progressive corneal ectasia, and thinning of the stroma leading to vision impairment worldwide.^1,2 Higher incidence and prevalence rate³ of KC ranges from 1.5 to 25 cases per 100,000 individuals per year and 0.2 and 4790 per 100,000 persons, respectively, without any gender bias.^4,5 Despite the implementation of numerous management strategies, the etiology of KC continues to be an unresolved enigma.⁶ KC is mostly prevalent in the working age group and adds a disproportionate economic burden to the society.⁷ Numerous studies^8–10 have reported the dysregulation of genes, enzymes, cytokines, extra cellular matrix proteins, and role of environment in KC pathogenesis.¹¹ 

Humans are supra-organisms, composed of both human and microbial cells, which maintain normal physiology and any imbalance can predispose to disease.¹² It exhibits different microbial diversity across the biological niches.¹³ There is an emerging paradigm in ocular surface health and disease which can be studied by microbiota sequencing.¹⁴ A recent study has indicated similarities and differences in corneal microbiome abundances in patients with KC and healthy subject's corneal epithelium using 16s ribosomal RNA gene analysis.¹⁵ Treatment naïve patients with KC ocular swabs identified Pelomonas and Ralstonia genera compared with healthy controls.¹⁶ Prior studies have shown that KC severity correlates with inflammatory factors, cytokines in the corneal epithelium and in the tear film.^17,18 However, none of these studies have correlated the ocular microbiome with disease presentation and their immune status in KC eyes. In an animal model of infection, ocular surface resident bacteria have been shown to protect the ocular surface via local signaling through IL-17 and mucosal γɗ-T cells.¹⁹ Our group has demonstrated a distinct ocular immune cell profile in KC pathogenesis.^9,20 Therefore, we hypothesized that there may exist a link between the altered immune status and the microbiome that can contribute to KC pathology. However, given the inherent diversity within human subjects, it was important to perform this study in coordinated microbiome, immune factor, and immune cell containing samples collected from the same eyes within a defined time. Our present study aims for an integrated approach in building the knowledgebase by understanding the microbial and immunological contributors to ocular surface health and disease for refining management strategies in KC. 

Methods

Study Design and Participants

This cross-sectional, observational study is approved by institutional ethics committee (ECR/187/Inst/Kar/2013/RR-16) of Narayana Nethralaya, Bangalore, India, and in harmony with the tenets of the Declaration of Helsinki. All study participants were voluntary and samples were collected after written informed consent was received. The study includes patients with KC (total n = 62 of which grade 1, n = 16; grade 2, n = 27; and grades 3+4, n = 19) and healthy controls (n = 21) visiting Narayana Nethralaya from the year 2017 to 2021. The mean age of the healthy controls was 27.1 ± 0.4 years (female controls = 13 and male controls = 8) and patients with KC were 22.7 ± 0.9 years (female patients = 23 and male patients = 39). Topography measurements and pachymetry of greater than 400 microns were included. Patients with KC on treatment, subjects with a history of or ongoing concurrent ocular comorbidities and systemic inflammatory conditions, contact lens use, ocular medications use, and history of ocular surgery in the last 3 months were excluded from the study. 

Ocular Swab Collection and DNA Isolation

A sterile, DNA and RNase-free swab (PW003; HiMedia Laboratories Pvt Ltd, Mumbai, India) was used for collecting 83 ocular swabs from 62 eyes of 62 patients with KC and 21 eyes of 21 healthy controls by the same clinician (author R.S.). The lower conjunctival fornix was swept first from the nasal to temporal side in a single stroke, followed by eversion and collection of samples from the upper lid in a similar fashion. The procedure was performed without the use of any anesthetic agent. The collected swabs were immediately placed in a sterile tube containing lysis buffer, sealed, and stored at −80°C. DNA isolation using QIAamp BiOstic Bacteremia DNA kit (Qiagen, USA), as per manufacturer’s protocol, was carried out. In brief, the microcentrifuge tube containing the swab in lysis buffer is centrifuged at 13,000 x g for 2 minutes to collect the (transparent) pellet. The kit utilizes an IRT solution for removing all the inhibitors from the medium. This lysate is transferred into a 2 mL PowerBead Tube Garnet, which purifies nucleic acids using chaotropic buffers and ethanol wash buffers. The eluted DNA is quantified using Qubit DNA HS Assay (Invitrogen, Cat #Q32854). 

Library Preparation and Sequencing

All DNA samples were amplified and sequenced by Illumina HiSeq2500 system using 16S rDNA universal 16S forward (5′AGAGTTTGATCCTGGCTCAG-3′) and 16S reverse (5′GGTTACCTTGTTACGACTT-3′) primers. V3V4 amplicon was generated from the 16S product by a nested PCR strategy using V3V4 forward (5′CCTACGGGNGGCWGCAG-3′) and V3V4 reverse (5′GACTACHVGGGTATCTAATCC-3′) primer. The V3V4 amplified products were cleaned using Sera-Mag beads (1X, GE Healthcare, Cat# 29343057) and processed for library preparation. DNA sequencing library was prepared using NEBNext Ultra DNA Library Prep Kit for Illumina (NEB, Cat# E7370L). All the prepared libraries were assessed for fragment distribution by 5300 Fragment Analyzer using dsDNA 910 Reagent Kit (35–1500 bp, DNF-910-K1000). The 2 × 250 base paired end reads per sample are generated using Illumina MiSeq version 2. 

Bioinformatics Analysis

The analysis was performed using in-house pipeline in QIIME environment. All the reads generated were assessed for quality control and trimmed below Q30. The 16s rDNA were determined using presence of conserved region (CR) in the targeted amplicons (V3, V4, or V3-V4). The sequenced reads are first overlapped and stitched to form long reads greater than 400 bp using the Fast Length Adjustment of SHort reads (FLASH) program. Dereplication and chimeric sequences were removed using the USEARCH and the UCHIME algorithms. 

Tear and Ocular Wash Sample Collection

Sterile Schirmer's strips (5 × 35 mm²; Contacare Opthalmics and Diagnostics, India) were used for tear collection from a subset of controls (n = 12) and patients with KC (n = 49) and processed as described earlier.²¹ Briefly, the tear fluid was extracted by chopping Schirmer's strip in 1.5 mL microcentrifuge tubes and agitating them in 300 µL of 1 × phosphate buffer saline (PBS) at 4°C for 1.5 hours. Tear fluid was eluted by centrifugation and stored at −80°C. Ocular surface wash was collected from the subjects for immune cell profiling using flow cytometry as described earlier.⁹ Briefly, 2 mL of 0.9% sterile saline solution was used for gently irrigating the ocular surface from nasal to temporal side. The run-off washed saline solution was collected by placing a sterile tube at the lateral canthus of the irrigated eye. Then, 0.05% paraformaldehyde was added to fix the cells and they were stored at 4°C. 

Flow Cytometry Analysis of Tears and Ocular Wash

We estimated the levels of secreted cytokines, chemokines, and various soluble factors in the tears using a flow cytometer (BD FACS Calibur; BD Biosciences) based cytometric bead array (BD CBA Human Soluble Protein Flex Set System; BD Biosciences) and LEGENDplex (Biolegend Inc.), following the manufacturer's instructions.^9,20 Briefly, soluble analytes were captured using capture antibody beads of defined size and distinct fluorescence acquired by a flow cytometer and analyzed with the assistance of BD FACSDiva software (BD Biosciences, USA). Immune cells profile from the ocular wash of healthy controls and patients with KC were analyzed for the proportions of various immune subsets using fluorochrome conjugated antibodies specific for immune cells, as performed earlier.²⁰ Fixed ocular wash samples were spun at 2000 revolutions per minute (rpm) for 5 minutes at 4°C to collect the cell pellet. A cocktail of fluorochrome conjugated antibodies was incubated with these cell pellets for 45 minutes at room temperature followed by washing and resuspension in 300 µL of 1 × PBS solution (pH 7.4). These fluorochrome conjugated antibodies were procured and data was acquired using flow cytometer and analyzed using FCS Express 6 (De Novo Software, USA). Single stained control compensation and manual gating strategy were used for immune cell subset identification. A schema for the methodology used in the study is shown in Figure 1. 

Figure 1.

 

Schematic representation of our study cohort. (A) Microbiomes from ocular swabs from healthy controls and patients with KC were assessed. Matched (B) ocular wash and (C) tears were isolated for flow cytometric based analysis.

Statistical Analysis

The sample diversity within the major groups (the control and the patients with KC) was calculated at the genus level using the R (3.7) library Vegan (version 2.4-2) and Fisher's alpha parameter for alpha diversity index calculation. These exploratory analyses were performed with custom R scripts. The Galaxy module for Linear discriminant analysis Effect Size (LEfSe) was used to generate cladogram with hierarchy based on the phylum, class, family, and genus. The relative differences among classes were calculated by Linear Discriminant Analysis. Then the low-expressed Operational Taxonomic Unit (OTU) reads with identification < 50% of the samples in the major groups (the control, KC1, KC2, and KC3 & 4 groups) were dropped. After filtering out the low expressed microbes, the common and unique microbial expressions were deduced. Differential expression analysis was performed. receiver operating characteristics and area under the curve (ROC-AUC) were calculated for patients with KC at the genus level and for significant soluble factors and immune cells. Fold change of abundance was calculated for the patients with KC as compared with the healthy controls. The Mann-Whitney U Test was performed using GraphPad Prism (version 8). The P value < 0.05 were considered statistically significant. Volcano plot was constructed using the absolute abundance of genus observed in patients with KC compared with the healthy controls. The Seaborn version 0.11 package was used for visual representation. Spearman rank correlation was performed among the differently abundant microbes, immune cells, and cytokines. Linkage module from Scipy version 1.10.1 library was used to make the different clusters based on the correlation values. Further, the distant matrix was plotted as heatmap. 

Results

Clinical Characteristics

Increased Belin-Ambrósio enhanced ectasia display total deviation (BAD-D) index and significantly higher keratometry values (K1, K2, Kmax, and Kmean) indicate corneal steepening in patients with KC compared with the healthy control. The mean central corneal thickness (CCT) of 464.79 ± 48.63 microns and thinnest corneal thickness (TCT) of 447 ± 45.10 microns were measured in the patients with KC. However, control corneal thickness measured at the thinnest and central points were ≥ 530 ± 23 microns (Table 1). 

Table 1.

 

Clinical Characteristics of KC Cohort

 

Altered Ocular Surface Microbiota in KC

OTU identification and taxonomy classification revealed 52 phyla/class, 132 orders, 283 families, and 718 genera. Venn-diagram shows their distribution in two groups (Fig. 2A). There was no statistical difference in alpha diversity metrics based on biodiversity and richness indices between the patients with KC and the healthy controls (Figs. 2B, 2C). This specifies that the diversity within a niche is not different. Interestingly, KC grade 1 exhibited significant (P = 0.02) variation in biodiversity compared with the healthy controls and other patients with KC groups (Supplementary Fig. S1). A cladogram displayed the differences in ocular surface microbiota between the patients with KC and the healthy controls (Fig. 2D). Linear discriminant analysis (LDA) effect size using LEfSE analysis revealed significant bacterial differences in ocular surface microbiota between the KC (positive score) and control groups (negative score). We observed higher abundances of f_Intrasporangiaceae; g_Tetrasphaera, f_Microbacteriaceae; g_Microbacterium (Actinobacteria phylum), and f_Family-XII; g_Exiguobacterium (Firmicutes) in the KC group. The abundance of f_Family-XI;g_Gemella, g_Granulicatella (Firmicutes), f_Porphyromonadaceae; g_Porphyromonas, f_Flavobacteriaceae, f_Sphingobacteriaceae, o_Sphingobacteriales (Bacteroidetes), f_Geodermatophilaceae (Actinobacteria) were lower in the KC group than in the control group (Fig. 2E). In addition, less abundant OTUs were filtered by identification cutoff of < 50% in patients with KC and the healthy controls. This resulted in narrowing the OTU identification to 12 phyla, 29 orders, 54 families, and 76 genera abundantly observed in the cohort. Venn diagram illustrates the relationship of common 10 core phyla_class and unique Patescibacteria; c_Saccharimonadia to the control group and Firmicutes; c_Erysipelotrichia to the KC group (Fig. 3A). A significant reduction in the percentage of Firmicutes; c_Bacilli abundance was observed in the patients with KC compared with the healthy controls (Fig. 3B). The heat map represents abundance of these 12 phyla individually for (n = 21) the controls and (n = 62) the patients with KC (Fig. 3C). Further, at order level, the Venn diagram has 26 common taxa between patients with KC and the healthy controls. Bacteroidetes; o_Sphingobacteriales, Patescibacteria; o_Saccharimonadales unique to the controls, and Firmicutes; o_Erysipelotrichales unique to KC. Similarly, 41 families were found common between both the groups and 10 families unique to the control group along with 3 families unique to patients with KC were identified (Figs. 3D, 3E). Similar Venn diagrams and heat maps for KC grades reveal distinct abundance patterns. At the phyla level, grade 1 KC exhibits a notable inverse abundance pattern when compared to the control group, namely, Actinobacteria, Coriobacteria, Bacilli, Clostridia, Negativicutes, and Alphaproteobacteria. Additionally, at the genus level, the unique abundance of four microbiota were observed in the controls, namely, Sphingopyxis, Sphingobacterium, Ochrobactrum, and Gemella. We also observed unique genera in KC grade 1 (n = 17), KC grade 2 (n = 2), and KC grades 3 + 4 (n = 3; Supplementary Table S1). The mean ± SEM abundance of 76 genera observed in the patients with KC and the healthy controls are listed with the respective P values (Mann Whitney U test; Supplementary Table S2). Taxa uniquely observed across all the grades of KC with their fold change with control and respective P values are listed in Supplementary Table S3. Similarly, the taxa abundance was observed uniquely to higher grade (KC 3 + 4) and lower grade KC1 in Supplementary Tables S4 and S5. These observations suggest that dysbiosis may contribute toward disease pathogenesis. In addition, the alterations in microbiota abundance varied with disease severity. These indicate that ocular microbiome dysbiosis can be associated with KC (see Supplementary Figs. S1, S2). 

Figure 2.

 

Descriptive status of ocular microbiome. (A) Identified OTUs across phyla, order, class, and genus in patients with KC and healthy controls. Alpha diversity measures are estimated by (B) biodiversity and (C) richness comparison between healthy controls (n = 21) and patients with KC (n = 62). Violin plot indicates the data distribution observed in the present study cohort. (D) Cladogram representation of differences between patients with KC (green) and healthy controls (red). Nodes represent taxa ranging from phylum_class, order, family, and genus levels from the inner to the outer circle. The size of the node represents the taxa abundance. (E) Linear discriminant analysis (LDA) score (log 10) plot derived from LEfSE analysis showing the biomarker taxa with LDA score > 3.

Figure 3.

 

Ocular microbiota composition in patients with KC and healthy control subjects. Venn diagram comparing the operational taxonomical unit (after stringent filter of < 50%) at (A) phylum, (D) order, and (E) family level between patients with KC (green) and healthy controls (red). (B) Bar graph showing the percentage of abundant phylum in patients with KC and healthy controls. Mann Whitney U test was performed for statistical analysis (*P = 0.05). (C) Heat map represents the individuals contributing the absolute abundance at phylum level in the healthy control subjects (n = 21) and patients with KC (n = 62).

Unique Genus Level Diversity in Patients With KC

Overall, the relative abundances for the 718 identified genera in KC were analyzed by volcano plot. Of these, 25 genera were found to be differentially abundant. Sphingomonas (Proteobacteria) and Atopobium (Actinobacteria) genus were significantly reduced in patients with KC (P = 0.04 and P = 0.02, respectively, Mann Whitney) compared with the control subjects. Uniquely, genus Roseburia and Dialister of Firmicutes phyla were significantly abundant in patients with KC (P = 0.04 and P = 0.03, respectively; Fig. 4A). Following a rigorous cutoff of < 50% of OTU reads within the groups; 52 genera were commonly present in both groups and 8 unique to patients with KC and 16 unique to the healthy controls. KC had unique genera under Actinobacteria phyla namely Kytococcus, Ornithinicoccus, and Collinsella and under Firmicutes phyla namely Roseburia, Exiguobacterium, Dialister, Megamonas, and Megasphaera. The healthy controls had a total of 16 unique genera of which Actinotalea, Pseudarthrobacter under Actinobacteria phyla; Porphyromonas, Prevotella, Sphingobacterium under Bacteroidetes phyla; Gemella, Granulicatell, Leuconostoc, Finegoldia under Firmicutes phyla; Fusobacterium (Fusobacteria); Methylobacterium, Aureimonas, Ochrobactrum, Shinella, Sphingopyxis, and Luteimonas under Proteobacteria phyla (Fig. 4B). The notable and significantly prevalent 17 genera present in either the KC or the control group were identified for further refinement. Interestingly, Exiguobacterium and Collinsella unique to KC and Porphyromonas, Gemella, Fusobacterium, Sphingobacterium, and Actinotalea genera unique to control were observed in alluvial (Sankey) plot. In addition, a significantly (P ≤ 0.05) higher abundance of Tetrasphaera, Microbacterium, Bacteroides, Brevundimonas, and Pantoea genera were observed in patients with KC. Similarly, Cutibacterium, Staphylococcus, Sphingomonas, Escherichia_Shigella, and Pseudomonas genera exhibited significantly reduced abundance in patients with KC as compared with the healthy controls (Fig. 4C). Four of the significantly altered genera, namely Microbacterium, Pantoea, Pseudomonas, and Sphingomonas achieved an AUC value of 0.723 (95% confidence interval [CI] = 0.609 to 0.820, P < 0.001); 0.706 (95% CI = 0.590 to 0.806, P = 0.001); 0.713 (95% CI = 0.596 to 0.811, P = 0.003), and 0.711 (95% CI = 0.595 to 0.810, P = 0.002), respectively, indicating the reliable microbial markers for KC (Figs. 4D–G). 

Figure 4.

 

Validation and independent diagnosis of microbial markers for KC at the genus level. (A) Volcano plot showing the degree of differential abundant genus in patients with KC compared with the healthy controls (x-axis, log₂ fold change, and y-axis, minus log₁₀ of P value). The dashed vertical and horizontal lines at fold change ± 1.0 and minus log 1.3 corresponds to P value of 0.05. The red dots denote the higher abundant genus in patients with KC and the green dots are the significantly reduced taxa in patients with KC. (B) Venn diagram (after stringent filter of < 50%) at the genus level. (C) Alluvial plot illustrating the absolute abundance of genera observed uniquely, significantly more or less in the healthy controls and the patients with KC. Area under the curve receiver operator characteristics (AUC-ROC ≥ 0.7) achieved for genera (D) Microbacterium, (E) Sphingomonas, (F) Pseudomonas, (G) Pantoea with the significant P value and 95% confidence interval (CI) demonstrating the significantly altered microbial genera in patients with KC.

Immune Profile of Patients With KC From Tears and Ocular Wash

A distinct profile of soluble factors was detected in KC tears. KC, being an inflammatory disorder, showed significantly higher levels of interleukins IL-6, IL-9, IL-21, fractalkine, soluble Fas Ligand (sFasL), growth factors like erythropoietin (EPO), vascular endothelial growth factor (VEGF) and matrix metallopeptidase 2 (MMP2), neutrophil gelatinase-associated lipocalin (NGAL) enzymes. However, significantly reduced levels of few cytokines IL-2, IL-18, interferon alpha (IFNα), and soluble factors, such as sTNFRI and sIL-1R2, were also observed (Fig. 5A, Table 2). Among these, few soluble factors achieved an AUC value between 0.8 and 0.9 with significant P values. Cytokines, such as IL-6, has an AUC value of 0.825 (95% CI = 0.697 to 0.915, P < 0.001); IL-21 has an AUC value of 0.805 (95% CI = 0.675 to 0.901, P < 0.001), and IL-2 achieved an AUC value of 0.826 (95% CI = 0.698 to 0.915, P < 0.001) between patients with KC and the healthy controls. MMP2 has an AUC value of 0.882 (95% CI = 0.765 to 0.954, P < 0.001); β2-microglobulin with an AUC value of 0.805 (95% CI = 0.675 to 0.901, P < 0.001), and, interestingly, EPO with highest AUC value of 0.934 (95% CI = 0.832 to 0.984, P < 0.001) were observed in the KC cohort (Figs. 5B–G). Inversely, the proportion of natural killer (NK) cells, CD66b^high (activated neutrophils), and ratio of activated to inactive neutrophils (CD66b^High/CD66b^Low) increased significantly in the KC group with respect to the healthy control group (Fig. 5H, Table 3). The proportion of CD45⁺ cells, a pan leukocyte marker, was significantly reduced in the patients with KC compared with the healthy control subjects and achieved an AUC value of 0.843 (95% CI = 0.700 to 0.936, P < 0.001; Fig. 5I). ROC with significant P < 0.001 values were achieved for CD66b^high cells having 0.786 AUC (95% CI = 0.634 to 0.896) and ratio of CD66b^High/CD66b^Low cells having 0.849 AUC (95% CI = 0.705 to 0.941; Figs. 5J, 5K). These data collectively indicate the potential role of altered soluble factors and immune cells in patients with KC. 

Figure 5.

 

Soluble factor and immune cells profile of KC. Volcano plot of (A) soluble factors and (H) immune cells subset in patients with KC versus healthy controls. X-axis denotes fold change (log2) and Y-axis denotes P value (−log10). Cutoff is set as ≥ 2-fold change. Red dots represent significantly upregulated levels and the blue dots represent significantly reduced levels in patients with KC. Area under the curve receiver operator characteristics (AUC-ROC) analysis of patients with KC with the signification soluble factors (B) IL-6, (C) IL-21, (D) IL-2, (E) MMP-2, (F) β2-microglobulin (G) EPO and immune cells, such as (I) CD45+ (J) Ratio of CD66b_High/CD66b_Low and (K) CD66b_High cells.

Table 2.

 

Ocular Soluble Factors From Tears of KC and Control Subjects

 

Table 3.

 

Ocular Surface Immune Subset Proportions in Control and KC Study Cohort

 

The Relationship Among Ocular Surface Microbiota, Clinical Parameters, and Immune Profile in Patients With KC

We have performed correlations amongst the 17 significant disease-associated genera. Collinsella, being uniquely abundant in KC showed strong positive correlation with Bacteroides (r = 0.89, P < 0.001), moderate positive correlation with Tetrasphaera (r = 0.46, P = 0.01), and a moderate negative correlation with Cutibacterium (r = −0.54, P = 0.003). Exiguobacterium genus unique to KC showed weak positive correlation with Tetrasphaera (r = 0.31, P = 0.02) and Sphingomonas (r = 0.31, P = 0.02; Fig. 6A). Further, unsupervised hierarchical clustering demonstrated that the association (six distinct clusters) among the significantly dysregulated immune cells, soluble factors, genera, and clinical parameters (Fig. 6B). The abundance of Tetrasphaera genus has a weak positive correlation with Fractalkine (r = 0.32, P = 0.01), VEGF-A (r = 0.37, P = 0.003), IL-21 (r = 0.30, P = 0.02), and IL-9 (r = 0.26, P = 0.04). Microbacterium genus are significantly abundant in KC, and weakly yet positively correlated with the clinical parameters, K1, K2, Kmean, Kmax, and BAD-D (r = 0.28, 0.33, 0.32, 0.36, and 0.37, P = 0.012, 0.002, 0.004, 0.001, and 0.001, respectively). An increase in Microbacterium abundance increased with keratometry values is indicative of KC progression. NK T cells and IL-6 also showed a weak positive correlation with Microbacterium genus. Bacteroides genus showed a weak positive correlation with β2-microglobulin (r = 0.26, P = 0.04), and a moderate positive correlation with CD66b^High cells (r = 0.43, P = 0.002) and MMP-2, EPO, NGAL, and ratio of CD66b^High/low cells clustered together (dark blue). Notably, the genera with lesser abundance (Sphingomonas, Staphylococcus, Cutibacterium, and Pseudomonas) in patients with KC clustered together with IL-2, sIL-1R2, and corneal thickness. Of which Cutibacterium showed a positive (weak) correlation with corneal thickness parameters CCT (r = 0.24, P = 0.02), TCT (r = 0.23, P = 0.03), IL-2, and sIL-1R2. Brevundimonas genus of Proteobacteria phylum also exhibited a weak positive correlation with CCT (r = 0.22, P = 0.05) and TCT (r = 0.24, P = 0.02), IL-2, and sIL-1R2. This cluster shows significant negative correlation with preceding three clusters suggesting the presence of parameters that are distinctive to KC. A positive negligible correlation between Collinsella, Exiguobacterium, and IL-18 (r = 0.07 and r = 0.1, respectively) was also identified. Porphyromonas, Fusobacterium, Gemella, Actinotalea, and Sphingobacterium genera were abundantly observed in the healthy controls compared to the patients with KC showed differential weak to moderate correlation status with CD45⁺ cells (r = −0.28, r = −0.3, r = 0.41, r = 0.33, and r = 0.58, respectively). The abundance of Pantoea showed negative weak correlation with CD45⁺ cells (r = −0.2). and Escherichia_Shigella had weak positive correlation with CD45⁺ cells (r = 0.32, P = 0.02). The analysis revealed the association of inter-relationship among unique microbiota, immune factor, and clinical indices, albeit this was only observed in a few genera, such as Tetrasphaera (positively correlated with IL-21; and negatively correlated with CCT and TCT), Microbacterium (positively correlated with MMP2, EPO, K1, K2, Kmean, Kmax, and BAD-D; and negatively correlated with CCT and TCT), Bacteroides (positively correlated with β2-microglobulin; and negatively correlated with CCT), and β2-microglobulin showed positive correlation with K1 and Kmean, as shown in Supplementary Table S6. Collectively, a distinct pattern was observed among the microbiota and with immune cells, soluble factors, and characteristic clinical features of KC. The summary of the results is represented in Figure 7. 

Figure 6.

 

The relationship among ocular surface microbiota, clinical parameters, and immune profile for KC. Heatmap showing the (A) Spearman correlation between the 17 shortlisted genera and (B) unsupervised clustering of the genus with other parameters reveals 6 distinct clusters. The heatmap represents the positive and negative correlation between parameters in each cluster. The magnitude of the boxes in the plot is proportional to the R-value and the color scale bar shows the interpretation of correlation coefficient based on the standard guidelines. The red color represents the positive correlation and the blue color represents the negative correlation. ***P values = 0.001; **P values = 0.01; *P values = 0.05.

Figure 7.

 

Schematic summary of the altered ocular microbiome with changes in the soluble factors and the immune cells leading to clinical pathology in KC. Ocular microbiome with KC disease presentation and its immune status. Microbes higher in abundance (Brevundimonas, Microbacterium) correlates with keratometry indices of KC. Lesser abundance of microbes (Cutibacterium, Pseudomonas, Sphingomonas, Staphylococcus) correlates with corneal thinning. Ocular surface microbiome influences immune cells activation and inflammatory milieu in tear film.

Discussion

Inflammatory status and microbiome influence each other within the same environment.²² Whereas it has been proposed that ocular surface commensal microbiome might regulate host metabolism, aid in immune system development, and help against pathogen invasion at the ocular surface,^23,24 experimental data in human disease is scarce. An imbalance in the microbiota may lead to localized or systemic inflammation or vice versa leading to ocular pathology.²⁵ The precise contributions from the local microbiota toward mechanisms responsible for triggering harmful immune responses in different chronic inflammatory eye diseases are not yet fully understood.^22,26 Thus, we analyzed ocular surface microbiome, tear, and ocular wash profile in KC patients. Our data uncovers a unique microbiome signature of KC that also correlated with disease severity suggesting a causal relationship with the pathology. We did not observe any significant alpha diversity differences in the healthy controls and the patients with KC, except for the initial grade of KC1 indicating the change in microbial diversity impacts the disease initiation within a particular environment. Lower alpha diversity parameters in the patients’ with KC epithelia compared with the healthy controls with no significant differences were reported.¹⁵ Most abundant phylum observed across ocular microbiome are Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria.^14,27 Overall, at the phyla/class level, we found lesser abundance of Actinobacteria, Firmicutes_Bacilli and a higher abundance of Proteobacteria (alpha and gamma), Firmicutes_Clostridia in patients with KC compared to the healthy controls. Our cohort identified Patescibacteria; c_Saccharimonadia phyla was more commonly found in the healthy controls and Firmicutes; c_Erysipelotrichia in patients with KC. 

An imbalance in gut microbiota may lead to systemic inflammation potentially involving translocation of microbes through epithelial barrier. The gut-eye axis holds significant importance in ophthalmology, given that gut microbiota appears to affect immune responses in remote locations, including the eyes.²⁸ Numerous ocular disorders, including Sjögren’s syndrome-related dry eye, infectious keratitis, uveitis, myopia, and glaucoma have been linked to irregularities in the gut microbiome.^29–31 Patients’ with multiple sclerosis gut microbiome reveals the abundance differences primarily in Firmicutes especially the Bacilli and Clostridia classes and Bacteroidetes and Actinobacteria compared with healthy controls.³² Collagen-induced arthritis (CIA) rats showed higher abundance of Erysipelotrichi and Alphaproteobacteria class in gut microbiota.³³ Presence of Patescibacteria in gut microbiota of patients with Alzheimer's disease was noted.³⁴ Presence of these major phylum in our data from KC eyes is supported by results from previous studies.^35–38 LEfSE score has identified Tetrasphaera, Microbacterium, and Exiguobacterium genera as markers in KC disorder. Dominant bacterial level of Tetrasphaera with LDA > 4 has been reported in tumor tissue samples.³⁹ Thus, our data suggest that at least certain altered microbial genera have similarities with alterations observed in other tissues of the body. Hence, it is plausible that the microbiota driven immune dysregulation is also occurring locally at the cornea. 

Numerous studies have reported the dysregulation of corneal extra cellular matrix (ECM) and remodeling making the cornea biomechanically instable.^17,40–42 At the genus level, Collinsella (Actinobacteria) abundance was observed as unique in our KC cohort. Collinsella was shown to increase the gut permeability by reducing the expression of tight junction protein (ZO-1) in human intestinal epithelial (CACO-2) cells via IL-17 and associated cytokines in the context of rheumatoid arthritis.⁴³ Reports suggest Collinsella has been associated with rheumatoid arthritis, type 2 diabetes (T2D), cholesterol metabolism, and glucose metabolism in pregnancy,⁴⁴ linking it with chronic inflammatory status. We observed that keratometry parameters (K1, K2, Kmean, Kmax, and BAD-D) had a significant positive correlation with higher abundance of Microbacterium (Actinobacteria) and with NKT⁺ cells and IL-6, indicating a unique KC specific immune-microbiome interaction axis. Higher abundance of Microbacterium has been associated with inflammatory diseases, such as endophthalmitis,⁴⁵ interstitial pulmonary inflammation,⁴⁶ and esophageal microbiome of pediatric esophageal eosinophilia (EE), a chronic inflammatory disorder.⁴⁷ More than 50 Microbacterium spp. were encountered in human clinical specimens⁴⁸ suggestive of their role in exacerbating chronic inflammation across different tissue types. Our data show significantly lesser abundance of Pseudomonas, Sphingomonas, and Staphylococcus in patients with KC, which positively correlated with corneal thickness parameters, such as CCT, TCT, and IL-2. Pseudomonas, Sphingomonas, and Staphylococcus were reported to be the most abundant genera observed within the ocular surface microbiota.⁴⁹ Presence of Sphingomonas has been related with soft tissue and bone infections.⁵⁰ A higher abundance of Corynebacterium, Pseudomonas, and Staphylococcus were reported in human infectious ocular conditions like keratitis, blepharitis, and scleritis.⁵¹ Pseudomonas and Staphylococcus are reported to release several bacterial enzymes, such as alkaline protease, elastase, and phospholipase C which act on extra cellular matrix and affect immune responses.⁵² This is an interesting observation because these genera are usually associated with ECM degradative and inflammatory activities. Thus, their reduced abundance in KC eyes is indicative of the unique nature of sterile, chronic inflammation like status of the ocular surface in patients with KC, but further studies are needed to determine the underlying mechanisms. We observed higher abundance of Brevundimonas (Proteobacteria), that positively correlated with corneal thickness parameters, such as CCT and TCT and soluble IL-1R2 and IL-2 levels. Corynebacterium, Finegoldia, and Brevundimonas are reported to be core genera in the conjunctival microbiome.⁵³ Zeng et al. have reported enriched Brevundimonas in the gastric fluid of patients of Chinese origin with gastric cancer.⁵⁴ Brevundimonas was identified as a key genus in patients with non-neoplastic large intestine mucosa compared to matched neoplastic tissue of patients with colorectal cancers.⁵⁵ Thus, in KC the elevation of this genera may be associated with immune and ECM-related factors modulating the corneal matrix remodeling process underlying ectasia. 

The abundance of Pantoea, Corynebacterium, and Staphylococcus observed in the KC eyes may have a relation with tear film quality because they have been reported in meibomian gland dysfunction and dry eye disease.⁵⁶ Higher abundance of Bacteroides (Bacteroidetes) significantly correlated with β2-microglobulin in our patients with KC. An autocrine function of β2-microglobulins has been shown to increase collagenase synthesis in rabbit synovial cells,⁵⁷ thereby directly linking this genus with ectatic processes in the cornea. Advanced glycation end product (AGE) modified β2-microglobulin stimulates cytokines, TNFα, and IL-1β synthesis and enhances collagenase leading to collagen degradation and connective tissue breakdown in cultured human synovial cells.⁵⁸ A subsequent study showed AGE modified β2-microglobulin modulates collagen synthesis in human fibroblast cells.⁵⁹ Collagenases are expressed by Bacteroides and has broader substrate (collagen) specificity with an ability to digest triple helix of collagen fibrils.⁵² An infectious ocular condition, orbital cellulitis, has Bacteroides and Prevotella as one of the causative organisms.³⁸ Therefore, Bacteroidetes may be directly linked to enzyme mediate destabilization of the corneal extracellular matrix via β2-microglobulin in patients with KC. A significantly reduced abundance of Escherichia-Shigella was observed in our patients with KC when compared with the healthy controls and negatively correlated with CD66b^high and EPO levels. In case of bacterial keratitis, both the Escherichia-Shigella and Bacteroides were found to be significantly more when compared with the healthy subjects.²⁷ Taken together, our data argue that underlying interactions of the immune mediators with resident microbiome on the ocular surface may be of pathologic consequence in KC. It is also important to note that genetic, dietary, and environmental aspects can dynamically affect the composition of the local microbiome as well as the immune milieu in any subject. Further studies on direct interactions between specific genera and molecular factors are warranted to determine their roles in causality and severity. However, the study has limitations such as cohort size and selective measurement of molecular factors and immune cells. Future studies on larger cohorts and large-scale assays, such as proteomics and metabolomics, from patients will expand the understanding beyond this study. 

In conclusion, a specific, altered immune-microbiome axis exists in KC which may be of diagnostic or prognostic use. Importantly, there may be a role for probiotics or specific antibiotics in KC alongside immunomodulators to prevent progression of the disease. 

Acknowledgments

The authors thank the late Bhujang Shetty, Founder and Chairman, Narayana Nethralaya, Bangalore, India, for resource support, Vrushali Deshpande, GROW Research Laboratory, Narayana Nethralaya Foundation, Bangalore, India, for assistance with bioinformatics and statistics. 

Supported by Narayana Netralaya Foundation, Bangalore, India. 

Disclosure: N.R. Kumar, None; P. Khamar, None; R. Kannan, None; A. Padmanabhan, None; R. Shetty, None; S. D'Souza, None; T. Vaidya, None; S. Sethu, None; A. Ghosh, None