Introduction

Human mucosal surfaces are normally colonized by a diverse community of microorganisms. In recent years, the mucosal microbiome has been extensively studied in tissues including the oral mucosa, vaginal mucosa, and nasopharynx1. The ocular surface is mucosal, but unlike other mucosa, has a relatively paucibacterial resident microbiome consisting primarily of low numbers of coagulase-negative Staphylococcus, Cutibacteria (formerly Propionibacteria), and Corynebacteria2,3,4,5,6,7. The bacterial load of the conjunctiva is approximately 1% that of oral mucosa or skin2,3. Multiple mechanisms have been hypothesized to account for the paucity of bacteria on the ocular surface, including the mechanical action of eyelid blinking, the presence of lysozyme and lactoferrin in tears, and the abundance of IgA in tears8,9. Recent research has also demonstrated a resident conjunctival virome including torque teno virus (TTV), a small single stranded DNA anellovirus3,10,11 and Merkel cell polyoma virus (MCPyV), a small double stranded DNA polyomavirus3. Both TTV and MCPyV have also recently been described in intraocular fluids from post-operative endophthalmitis cases10,12.

Previous studies have suggested that the ocular microbiome is disrupted and re-organized following enucleation13,14. The anophthalmic socket may be altered by absence of normal ocular tissues such as the cornea; by presence of a prosthesis; by medication history preceding and following enucleation; and by the underlying condition necessitating enucleation or evisceration. Prior studies have used standard bacterial culture techniques to characterize changes in ocular surface flora of the anophthalmic socket and found increased microbial density and a modest shift toward more pathogenic organisms. The purpose of the present study was to apply molecular DNA detection techniques, which have the advantages of being able to detect unculturable or poorly culturable bacteria as well as fungi, parasites and DNA viruses, to compare the anophthalmic conjunctiva with intact fellow eyes.

Methods

Study subjects

Institutional approval for this study was obtained from the University of Washington Institutional Review Board. This study followed the tenets of Helsinki as well as ARVO guidelines for research. This study was a registered clinical trial at ClinicalTrials.gov (NCT02306668). Written informed consent was obtained from all subjects.

20 subjects with unilateral anophthalmia were recruited from the oculofacial plastic and reconstructive surgery service of the University of Washington Medicine Eye Institute. Inclusion criteria were history of enucleation or anophthalmia with prosthesis and scleral shell placement in one eye, and ability to provide consent. Exclusion criterion was evidence of active infection at time of sampling; and any evidence of acute prosthetic malfunction such as erosion or extrusion. All subjects wore ocular prostheses (scleral shell) at the time of this study, which was removed aseptically immediately before sampling. After instillation of sterile, topical proparacaine, DNA swabs (Isohelix, SK-2S) were used to obtain samples from the lower conjunctival fornices of the anophthalmic and fellow-eye conjunctiva for molecular analysis. Genomic DNA was isolated from the conjunctiva swabs using the DNeasy Blood & Tissue Kit using manufacturer’s suggested protocol (Qiagen, Inc., Venlo, Netherlands). DNA was quantified using the Qubit® dsDNA HS Assay Kit (ThermoFisher Scientific, Waltham, MA).

Molecular analysis

To calculate bacterial load in each sample, pan-bacterial PCR was performed using 16S ribosomal DNA universal primers as well as primers for human actin to normalize for total DNA recovery (Integrated DNA Technologies, San Diego, CA) as described previously3. Quantitative PCR was used to confirm presence of TTV as described previously11. Primer sequences are given in Table 1. All quantitative PCR assays were performed on the Applied Biosystems 7500 Fast Real-Time PCR system (Foster City, CA) for 30 cycles and quantified using standard curves of cloned positive control template. The final PCR mix contained 0.8 µL each of forward and reverse primers (final concentration of each, 0.4 mM), 10  µL of the FastStart SYBR Green Master Mix without ROX (Roche, Basel, Switzerland), and 2 µL of unamplified genomic DNA. Master mix was prepared under a laminar hood to minimize contaminations. The final reaction volume was 20 µL. For standard curve, a plasmid complementary DNA of the cloned region of interest (e.g., the target sequence for 16S, MCPyV, TTV, or actin) was serially diluted tenfold to obtain copy numbers ranging from 101 to 108 copy/mL. Quantitative PCR routinely was able to detect 10 copies/mL of each control complementary DNA. The PCR reaction consisted of initial holding stage at 50 °C for 2 min and at 95 °C for 10 min, followed by a cycling stage at 95 °C for 15 s and 60 °C for 1 min. 30 cycles of amplification were performed. Copy number of experimental samples was calculated by interpolation of delta CT number against the standard curve derived from the cloned product. Double-distilled H2O was used as negative control in all experiments, and was tested at time of PCR reaction. Water/air controls were not obtained at the time of swabbing.

Table 1 PCR primer sequences.
Full size table

Biome representational in silico karyotyping (BRiSK) is a whole genome representational deep sequencing technique that allows for an unbiased characterization of the DNA-based metagenomic constituents in biopsy samples15. In brief, a type IIS restriction endonuclease was used to isolate 33 bp fragments associated with a specific 6-mer recognition sequence in the DNA and subjected to deep DNA sequencing. Genomic DNA was subjected to whole genome amplification using phi29 polymerase and digested with BsaXI. Illumina-compatible barcoded paired-end adapters were ligated to digested DNA, followed by isolation of sequences with asymmetric adapters using streptavidin column and PCR-amplified using adapter sequences. The resulting product was gel-purified and subjected to high throughput sequencing on the Illumina MiSeq sequencer (Illumina, San Diego, CA). BRiSK data were analyzed for presence of bacteria, fungi and viruses as described previously11. DNA tags were cross-referenced to a comprehensive database for taxonomic classification.

Statistical analysis

Statistical analyses were performed using SPSS for Windows software version 26 (IBM Corporation, Armonk, North Castle, NY, USA). Descriptive analyses are reported as means and standard deviation for continuous variables and as percentage for qualitative variables. All continuous and paired statistical analyses were completed by the related-samples Wilcoxon signed rank test, while the categorical data were analyzed by Fisher’s Exact Test. A 2-tailed probability of 0.05 or less was considered statistically significant.

Results

Demographic data for subjects in this study are shown in Table 2. Among the 20 participants (7 female, 13 male) the average age was 59.5 ± 15 years (mean ± standard deviation). The most common cause of anophthalmia was trauma (40%) but also included previous corneal damage (20%), birth defects (microphthalmos 20%) and retinoblastoma (5%). Mean time since anophthalmic surgery was 26.1 ± 18 years with current prosthetic mean age of 6.7 ± 9 years. The majority of patients (70%) did not endorse symptoms in their anophthalmic socket (Table 3). For those who did, discharge and erythema of the conjunctiva were most common (both 10%). 75% of patients were not using any ocular treatments in their anophthalmic socket. Topical medication use for the remainder included erythromycin (10%), artificial tears (5%), and blephamide (5%). One subject was taking systemic methotrexate (5%). 75% of patients did not have a significant ocular history in their contralateral eye. No subject had evidence of active infection at the time of sampling.

Table 2 Patient characteristics.
Full size table
Table 3 Clinical history.
Full size table

We first sought to determine the overall bacterial load present on anophthalmic and intact conjunctiva using quantitative 16S universal bacterial PCR. Actin was used as positive control for DNA extraction. As shown in Fig. 1, all samples showed robust amplification for actin indicating sufficient sample acquisition. Double distilled water was used as negative control in all experiments and yielded no amplification product. Quantitative PCR for actin revealed no difference in recovery between anophthalmic and control conjunctiva (Table 3). Qualitative 16S PCR demonstrated presence of bacteria in 16/20 anophthalmic samples and 15/20 contralateral samples (Fig. 1). Quantitative PCR was performed for 16S rDNA. By this more sensitive technique, all samples showed products, generally ranging between 100 and 1000 copies/ng starting DNA. Normalization of 16S quantitative PCR for actin revealed average bacterial loads of ~ 0.2 16S copies per actin copy (Table 4). As bacteria have an average of ~ 4 16S copies per cell16, this suggests an average bacterial load of ~ 1 bacterium per 10 human cells in both groups (on the same order of magnitude as reported in previous studies of healthy conjunctiva3). No significant difference in normalized bacterial load was observed in anophthalmic compared with control conjunctiva (0.229 vs. 0.225, p = 0.765 by related-samples Wilcoxon signed rank test).

Figure 1
figure 1

Quantitative (top) and qualitative PCR for 16S ribosomal DNA from anophthalmic (left) and contralateral (right) conjunctiva. Lower images are from ethidium-bromide stained gel electrophoresis. H20 = double distilled water negative control.

Full size image
Table 4 Quantitative PCR in anophthalmic and contralateral conjunctiva.
Full size table

16S metagenomic sequencing is prone to substantial artifact when applied to samples with low bacterial loads, including amplification of environmental DNA and erroneously high estimates of diversity17. BRiSK is a representational deep DNA sequencing technique in which ~ 1% of all DNA in a sample is sequenced, and the presence of non-human DNA is catalogued15. This technique is capable of detecting any DNA-based life-form, including bacteria, fungi, parasites, and DNA viruses, and has been used to characterize the healthy human microbiome at the genus level3,18. Because of potential biases in amplification, it is not a quantitative technique when applied to samples with low bacterial loads, but may be considered similar to culture, which is generally reported as present vs. absent. Presence of bacteria using BRiSK in conjunctival samples from the anophthalmic socket vs. the healthy contralateral socket is shown in Table 5. In both the healthy and anophthalmic sockets, the most common bacterial genera detected were Corynebacterium (9/20 anophthalmic vs 8/20 contralateral), Staphylococcus (8/20 anophthalmic vs 10/20 contralateral), and Cutibacterium (formerly Propionibacteria, 8/20 anophthalmic vs 4/20 anophthalmic). None of these differences were statistically significant by Fisher exact test. Of note, two anophthalmic sockets had detectable Clostridium sequences while none was seen in contralateral control eyes. Clostridium sequences have been detected occasionally in previous studies as a component of normal conjunctiva3,19.

Table 5 BRiSK Detection of Bacteria and Viruses in Anophthalmic vs. Contralateral Socket.
Full size table

Four DNA viruses – TTV, MCPyV, SEN virus (a small DNA virus closely related to TTV20), and papilloma-family viruses – were found by BRiSK in at least some samples (Table 5). All were found more commonly in the anophthalmic socket than in the contralateral socket. More sensitive quantitative PCR was used to determine the number of copies of MCPyV and of TTV from the anophthalmic socket compared to the healthy contralateral conjunctiva (Table 3 and Fig. 2). 19 of 20 samples demonstrated higher quantitative load of MCPyV in the anophthalmic conjunctiva than in the contralateral conjunctiva (p < 0.0001 by Fisher exact test). Quantitative PCR demonstrated an average of 891 MCPyV copies/ng DNA in the anophthalmic conjunctiva compared to 193 MCPyV copies/ng DNA on the healthy contralateral conjunctiva (Fig. 2, p < 0.001 by related-samples Wilcoxon signed rank test). TTV copy number was higher in the anophthalmic conjunctiva than the contralateral conjunctiva in a majority of cases (13/20), and TTV copies/ng were higher in anophthalmic vs. contralateral conjunctiva (41 copies/ng vs 8 copies/ng), but this result was not statistically significant on paired significance testing (p = 0.232).

Figure 2
figure 2

A. Quantitative and qualitative PCR for Merkel cell polyoma virus (MCPyV) in control conjunctiva. B. Quantitative and qualitative PCR for MCPyV in conjunctiva of anopththalmic eyes of same subjects. H20 = double distilled water negative control. C. Comparison of quantitative MCPyV viral load normalized to actin (see Fig. 1) from anophthalmic and contralateral eyes.

Full size image

Discussion

Characterizing the normal microbiome of the ocular surface, specifically the bulbar and palpebral conjunctivae, is important for understanding the etiology of ocular infections (such as corneal ulcers and post-operative endophthalmitis) and ocular inflammatory conditions (such as keratitis)4. While traditional conjunctival culture techniques have consistently identified several common ocular resident bacterial species, 16S rDNA PCR deep sequencing has revealed the presence of additional bacteria and, surprisingly, resident viruses5,7. The sensitivity of BRiSK to detect viruses was demonstrated by Muthappan et al.15 who used paraffin-embedded microscope slides of nasopharyngeal carcinoma to detect Epstein Barr Virus (HHV-4), the causative agent of this malignancy. In a more recent study, the conjunctival microbiome of healthy individuals was characterized using BRiSK analysis, with results consistent with the hypothesis that there is both a bacterial and a viral community in normal human conjunctiva3. The bacterial community is described as paucibacterial, with less than one bacterial genome recovered per 10 human cells. This is approximately two orders of magnitude lower bacterial load than is seen in buccal mucosa or skin. A subsequent study using qPCR of 16S DNA suggested even lower bacterial load in the eyes of contact lens wearers, on the order of one bacterium per 500 human cells18. The most prevalent bacterial genera identified in conjunctiva in these studies were Corynebacterium, Cutibacterium (Propionibacterium), Staphylococcus and Streptococcus. TTV was also identified by BRiSK on the ocular surface3,18, and was confirmed to be present by PCR in approximately half of ocular surface samples. Doan et al. also detected MCPyV and papilloma viruses on healthy conjunctiva in small subsets of patients3.

The goal of the present study was application of this type of molecular microbiome analysis to the conjunctiva of the anophthalmic socket. The anophthalmic socket is categorically different from that of normal ocular adnexa. In most cases it has undergone trauma (enucleation or evisceration, and often surgeries such as ruptured globe repair which predated the removal of the globe), has been exposed to topical antibiotics, and no longer has normal corneal tissue present. In the current study, we find these changes have minimal long-term impact on the bacterial components of the conjunctival surface, with no significant changes seen either in the load of bacteria present (as measured by quantitative 16S PCR) nor their general type. These results differ somewhat with the findings of Toribio et al.13,14, who found greater microbial density and modestly increased prevalence of gram-negative bacteria and coagulase-positive Staphylococcus on the anophthalmic conjunctival surface using standard culture techniques. This discrepancy likely has several causes. The overall bacterial load found by BRiSK is largely made up of Cutibacteria (previously named Propionibacteria21) and Corynebacteria species. These species are recovered poorly in culture and so are likely underestimated, but are fully represented by BRiSK and quantitative 16S PCR. Additionally, BRiSK is unable to distinguish bacteria at the species level, and the technique cannot reliably distinguish coagulase-positive from coagulase-negative Staphylococcus. Our results were consistent with Toribio et al. in finding increased prevalence of Streptococcus in anophthalmic sockets; however, in the present study this did not reach statistical significance as it did in their larger study (which looked at over 180 anophthalmic surfaces). Overall, however, our results suggest that the conjunctival surface in the anophthalmic socket does not have grossly altered bacterial flora in either quantity or general constituency, but may have more subtle changes in the bacterial microbiome.

The current study did demonstrate significantly higher recovery of several viruses from the anophthalmic conjunctiva than the contralateral eye. MCPyV, TTV, SEN virus, and papilloma viruses were all identified by BRiSK more frequently from anophthalmic conjunctiva than from contralateral tissue. Quantitative PCR demonstrated substantially higher recovery of MCPyV DNA from the anophthalmic conjunctiva, with ~ fourfold higher average viral copy number. While higher recovery of TTV was also noted, this result was not statistically significant in our relatively small study.

MCPyV belongs to the Polyomaviridae family of viruses which are icosahedral double-stranded DNA viruses. MCPyV DNA is normally present in many locations in humans including the GI tract22, lymphoid tissues23, urine24, the respiratory tract25, and saliva26. In these locations, the virus does not appear to induce malignancy with significant frequency, suggesting that it is part of a normal virome. MCPyV has been shown to be a causative agent in Merkel cell carcinoma (MCC)27,28,29, a rare but often aggressive neuroendocrine tumor of the skin. Consistent with a viral etiology for cancer, MCC is significantly more common in immunosuppressed individuals such as those with HIV30. Higher loads of MCPyV have been observed in normal skin of patients with MCC than in matched controls, suggesting the hypothesis that higher viral loads predispose to disease31. MCC affects the eyelids in approximately 2.5% of cases32, and typically affects patients in their 6th decade or older33. This tumor has a poor prognosis with approximately 40% mortality. As the eyelids represent only ~ 0.03% of the surface area of human skin, MCC is over-represented in eyelid tumors by about 100-fold. It is conceivable that this may be related to the presence of MCPyV on the ocular surface. While our review of the literature did not yield any cases of MCC in lids of previously enucleated eyes, given the rarity of this tumor (fewer than 50 cases of eyelid MCC annually in the US32) and the relative rarity of enucleation and evisceration (rate of ~ 2.5/100,000 in one survey in the United Kingdom34 suggesting prevalence of anophthalmia of less than 1/1,000) we would calculate expectation of fewer than one case of MCC in the lid of an enucleated individual every 20 years. The fact that enucleation appears to dysregulate MCPyV on the ocular surface suggests that other factors may also affect levels of this virus, and could constitute risk factors accounting for the high relative frequency of this tumor on the eyelids.

The current study has several limitations. This study had a relatively small sample size which, as noted previously, may have limited power to measure subtle changes in the bacterial microbiome of the anophthalmic surface. Time of day and pre-collection routine (such as face washing) were not standardized. Additionally, we did not account for systemic co-morbidities in our subjects. Systemic diseases such as diabetes mellitus could conceivably have an effect on ocular surface flora (although we used fellow eyes as control to attempt to normalize for this). The use of BRiSK and confirmatory PCR was entirely DNA-based. As such, we did not look for the presence or levels of RNA viruses on the ocular surface such as hepatitis C virus. By using DNA for our assays, we additionally cannot distinguish between live and dead bacteria in the present study. Additionally, as with any amplification-based technique, the possibility exists for contamination of reagents generating false-positive results (although this risk appears lower for BRiSK than for 16S metagenomic approaches3). Finally, our normalization assumes that the expression of actin per human cell remains constant following enucleation; it is conceivable that scarring could change the level of actin expression which could bias these results. However, the inability to detect MCPyV by PCR in the majority of contralateral eyes suggests the observed viral load difference between fellow eyes is substantial.

In conclusion, the current study demonstrates the diverse ocular surface microbiome and virome of the anophthalmic socket. In comparison to the healthy ocular surface, MCPyV is a more frequent component of the anophthalmic virome. The factors contributing to higher recovery of MCPyV in anophthalmic sockets remain to be identified, but these results suggest the normal eye contributes to regulation of ocular surface flora, particularly its virome. The effects of these viruses on the normal and abnormal conjunctiva remains to be determined.