A Tear-Based Approach for Rapid Identification of Bacterial Pathogens in Corneal Ulcers Using Nanopore Sequencing

Abstract

Purpose: This prospective observational study assesses the efficacy of using portable next-generation sequencing directly on tear samples to identify bacterial pathogens in corneal ulcers.

Methods: Tear samples were collected from ulcerated and contralateral eyes using Schirmer strips. Corneal scrapings and cultures were performed as medically indicated. The 16S rRNA gene was amplified from tear samples using polymerase chain reaction (PCR), and Nanopore sequencing was used for bacterial species identification and taxonomic classification.

Results: Bacterial DNA was identified in 8 of 10 samples using the tear-based sequencing method. Nanopore sequencing accurately identified the causative bacteria in all four samples that exhibited bacterial growth on culture and detected bacterial pathogens in two of the four ulcers that did not show bacterial growth on culture. In two cases where cultures could not be obtained due to the ulcer's small size, tear sequencing successfully identified bacterial species. Among the nine contralateral tear samples collected, Nanopore sequencing identified commensal bacteria in four samples.

Conclusions: PCR amplification of 16S rRNA directly from tears followed by Nanopore sequencing is an effective, noninvasive method to identify bacterial pathogens in corneal ulcers, offering noninferior results to traditional culture methods.

Translational Relevance: By eliminating the need for corneal scrapings and nucleic acid extraction, this tear-based method improves the timing and accuracy of bacterial pathogen diagnosis in corneal ulcers, allowing for prompt detection of causative organisms and enabling earlier targeted antimicrobial therapy, thereby improving patient outcomes.

Introduction

Corneal ulcers represent a significant ophthalmologic emergency, often leading to irreversible corneal damage and visual impairment. Globally, microbial keratitis, which is the primary cause of corneal ulcers, is estimated to contribute to approximately 1.5 to 2 million cases of unilateral blindness annually.¹ In the United States, bacterial infections are most frequently responsible for these conditions.² Traditionally, the diagnosis of corneal ulcers has relied on corneal scraping to collect samples for culturing and identifying the causative organism. This method, while considered the gold standard, is invasive, is time-consuming, and exhibits variable sensitivity (38% to 66%),³ largely depending on the size of the ulcer and the type of organism. Culture results, which can take several days to weeks, often delay the initiation of appropriate antimicrobial therapies, thereby extending patient discomfort and increasing the risk of complications and vision loss. 

In contrast, next-generation sequencing technologies, such as Nanopore sequencing, offer a compelling alternative for microbial identification. This portable sequencing technology enables real-time basecalling, which significantly accelerates the pathogen identification process. Prior studies have validated the efficacy of Nanopore sequencing in various clinical scenarios, demonstrating its capability not only to identify microbial agents but also to predict antibiotic resistance.^4–8 However, traditional approaches still require DNA extraction—and, in the case of corneal ulcers, invasive corneal scrapings—which necessitates specialized equipment and skills. 

Tears possess a robust antimicrobial defense system, which is crucial for protecting the ocular surface from infections. Factors such as immunoglobulin A, which neutralizes and prevents pathogen adhesion; lysozyme, an enzyme that breaks down bacterial cell walls; lactoferrin, which possesses direct bactericidal effects; and enzymes like secretory phospholipase A2 and β-lysin, which disrupt bacterial membranes, all contribute to this protective mechanism.^9–11 These antimicrobial agents not only protect the eye but also facilitate the lysis of microbial cells, leading to the release of microbial DNA directly into the tears.^9,12 In addition, endonucleases such as tear lipocalin further degrade DNA, leading to microbial DNA fragments in tears.¹³ 

Under this premise, we hypothesized that it would be possible to detect microbial DNA directly from tear samples obtained from eyes with corneal ulcers, thus eliminating the need for corneal scrapings and DNA extraction. This would provide a noninvasive, efficient method for diagnosing ocular surface infections. Here, we conducted a prospective study on individuals presenting with bacterial corneal ulcers to compare the diagnostic accuracy of the tear-based sequencing method to the traditional culture method, using the nonulcerated contralateral eyes as controls. 

Methods

Study Design and Patient Enrollment

This is a prospective observational cohort study conducted in accordance with the principles of the Declaration of Helsinki and approved by the Institutional Review Board of Yale University under Protocol #2000033805. Informed consent was obtained from all patients. Data on demographics and medical history were collected from the patients’ medical records at enrollment. Statistical analysis was performed using IBM SPSS Statistics software version 29 (IBM Corp, Armonk, NY, USA). 

Participants included in the study were adults aged 18 years and older who had a clinical diagnosis of microbial keratitis confirmed by an ophthalmologist. Eligible individuals presented with symptoms consistent with corneal ulcers, such as eye redness, pain, and visual impairment, and were able to provide informed consent. Exclusion criteria included individuals with a history of corneal transplantation within the prior 6 months. We also excluded those with concern for viral or fungal keratitis based on preliminary clinical assessments, culture results, or prior medical records. We screened and approached 21 patients for potential inclusion in the study. Of those, six patients did not consent to provide tears, and one lacked capacity to consent. Among the 14 patients whose tear samples were collected, 4 were excluded. One was excluded due to fungal growth on culture, one due to history of recurrent herpes simplex virus (HSV) keratitis, one due to corneal transplantation within the prior 6 months, and one because the tear strip broke during centrifugation, rendering the sample unusable. 

Tear Collection

Tear samples were collected from the inferior fornix of both the ulcerated eye and the contralateral eye using sterile Schirmer strips (HUB Pharmaceuticals, Scottsdale, AZ, USA), avoiding contact with the eyelashes. The strip was placed in each eye until tear liquid reached approximately the 15-mm mark on the strip, which took between approximately 20 seconds and 2 minutes, depending on the eye's tear production. The strip was subsequently placed in a microcentrifuge tube. Tears were collected following centrifugation for 2 minutes at 5000 rpm, resulting in approximately 5 µL of tear sample, and they were stored at –80°C. 

Corneal Scraping

Corneal scraping was performed as part of the standard of care for patients with symptoms indicative of microbial keratitis, in accordance with established clinical guidelines.¹⁴ The decision to perform corneal scraping was made at the discretion of the treating physician based on the severity of the ulcer, size, location, and clinical presentation of the infection. The procedure was carried out using a slit lamp for magnification. The patient's eye was anesthetized with topical proparacaine eye drops to ensure comfort during the procedure. Using a sterile Beaver blade, corneal scraping was performed. To preserve microbial viability, the collected samples were immediately inoculated onto the appropriate culture media or transport media by using the blade to make C-shaped streaks on the media. A small portion of the sample from the blade was also placed into thioglycolate broth. The samples were promptly transported to the laboratory for prompt culturing and analysis. 

Nanopore 16S Ribosomal RNA Amplification and Sequencing

Amplification and sequencing of the 16S ribosomal RNA (rRNA) gene, which is universal across bacterial species, were performed using the 16S Barcoding Kit 1-24 (SQK-16S024; Oxford Nanopore Technologies, Oxford, UK). The 16S rRNA gene was amplified using the LongAmp Hot Start Taq 2X Master Mix (New England Biolabs, Ipswich, MA, USA). One µL of the tear sample was added to the master mix and barcoded primers in a reaction mix volume of 50 µL. Polymerase chain reaction (PCR) was performed at the following settings: an initial denaturation step at 95°C for 1 minute, followed by 50 cycles at 95°C for 20 seconds, 55°C for 30 seconds, and 65°C for 2 minutes, and a final extension step at 65°C for 5 minutes. The rest of the library preparation was performed using the recommended protocol for ONT kit SQK-16S024 for compatibility with the Flongle flow cell. PCR products were purified using AMPure XP Beads (Beckman Coulter, Brea, CA, USA) following the Nanopore protocol. Sequencing runs were performed using the R9.4.1 Flongle flow cell (FLO-FLG001; Oxford Nanopore Technologies) on a MinION Mk1B Nanopore sequencer (Oxford Nanopore Technologies, Oxford, UK). Sequencing was allowed to proceed until a plateau of reads was achieved, which generally took between 4 and 12 hours. 

Analysis and Bacterial Identification

Basecalling was performed using the built-in Guppy basecaller on MinKnow, which translates the raw signal data from the Nanopore sequencer into nucleotide sequences. For species identification, the cloud-based EPI2ME FASTQ16S pipeline (v2023.04.21) provided by Oxford Nanopore Technologies was utilized. The pipeline classifies the results from each sequencing run and compares these sequences against individual entries within the reference NCBI 16S rRNA gene BLAST database. A minimum qscore of 7 was set, which indicates the quality threshold for basecalling accuracy; a higher qscore represents a higher confidence in the accuracy of the nucleotide base calls. In cases where multiple reads were identified, the determination of the causative organism was based on the highest read count. 

Results

Demographics of Study Participants and Clinical Presentation

Ten participants were included after meeting the study criteria. The mean age was 50.2 years. There were six males and four females. The ulcer was present in the left eye of five participants and the right eye of five participants. Patient demographics are summarized in Table 1, along with recorded size of the infiltrate at the time of diagnosis. All tear samples were collected on the day of corneal ulcer diagnosis, except for participant 1, whose tear sample was collected 6 days after initial diagnosis in the setting of a long hospital course. All tear samples were collected between April 2023 and March 2024. 

Table 1.

 

Demographics and Clinical Characteristics of the Study Participants

 

Symptoms at presentation included blurred vision, pain, conjunctival injection, foreign body sensation, discharge, tearing, and sensitivity to light. Time of symptom onset before presentation ranged from 1 to 14 days with a mean of 4.7 days. Risk factors for the development of corneal ulcer included use of soft contact lenses (n = 6), history of long hospitalization with associated bacteremia (n = 1), blepharitis (n = 1), following superficial keratectomy 7 days before diagnosis (n = 1), and loose suture in a patient with a penetrating keratoplasty (n = 1). Best-corrected or pinhole visual acuity at presentation ranged from 0 to 2.40 logarithm of the minimum angle of resolution (logMAR), with a mean (SD) of 1.07 (0.88) logMAR. 

On examination, nine patients had a single corneal infiltrate while one patient had two infiltrates of comparable size. Corneal ulcers were centrally located in 50% of the cases, with the remainder being peripheral. The average size of infiltrates (by estimating <1 mm in participant 6 as 1 mm and accounting for two 1-mm infiltrates in participant 10) measured 1.73 mm horizontally and 1.81 mm vertically. No ulcers extended to the sclera or limbus. Two patients developed hypopyon. 

Eight participants underwent corneal scraping for culture, while two had ulcers considered too small for scraping. A total of 19 tear samples were collected since one patient did not have a sufficient tear sample drawn from the contralateral eye. 

Summary of Culture Results

Of the eight participants who underwent corneal scraping and culture, four exhibited bacterial growth on culture. The bacteria identified included coagulase-negative Staphylococcus in broth (participant 1), Staphylococcus aureus (participant 2), and Pseudomonas aeruginosa (participants 3 and 4), all identified 1 day after diagnosis. No other bacterial species were detected on culture. Results from culture are listed in Table 2. The remaining 4 participants showed no bacterial growth on culture media, including blood agar, chocolate agar, Sabouraud agar, and thioglycolate broth, which were monitored for growth for 5 days, except for Sabouraud agar, which was held for approximately 1 month. 

Table 2.

 

Summary of Culture Results and Nanopore Sequencing Results

 

Nanopore Sequencing Results

For the four participants whose ulcer scrapings exhibited growth on culture, Nanopore sequencing identified the predominant bacterial pathogens in each tear sample as follows: Staphylococcus saccharolyticus, Staphylococcus roterodami, P. aeruginosa, and again P. aeruginosa, respectively. S. saccharolyticus is coagulase negative,¹⁵ and S. roterodami is within the S. aureus complex.¹⁶ 

Among the four participants whose ulcers did not exhibit growth on culture, Nanopore sequencing detected no bacteria in two cases while identifying S. saccharolyticus and Cutibacterium acnes in the remaining 2. 

For the two participants with ulcers too small for culture, Nanopore sequencing identified S. saccharolyticus in one and Kocuria rhizophila in the other. Results from Nanopore sequencing compared to results from culture are listed in Table 2. Detailed results featuring taxonomic classification for each case, including all bacterial species with a minimum abundance cutoff of 3% along with a synopsis of the clinical presentation, are provided in Supplementary Figures S1 to S8. 

For the contralateral eyes, five samples had no bacterial results detected by Nanopore sequencing. One patient did not have enough tear sample collected to allow for Nanopore analysis. The final four patients had the greatest number of reads, respectively, for the following bacterial species: S. saccharolyticus (participants 2 and 6), Streptococcus sanguinis (participant 3), and Staphylococcus hominis (participant 4). Detailed results featuring taxonomic classification for these four cases, including all bacterial species with a minimum abundance cutoff of 3%, are provided in Supplementary Figures S9 to S12. 

Information from Supplementary Figures S1 to S12, including the number of bacterial species with a minimum abundance cutoff of 3%, total reads, and percentage of reads attributed to the most predominant bacterial species, is summarized in Table 3. 

Table 3.

 

Summary of Nanopore Taxonomical Results From Ulcerated and Contralateral Eyes

 

Discussion

Our findings demonstrate the efficacy of PCR amplification of the 16S gene directly from tears, followed by Nanopore sequencing, in identifying the causative bacterial agents of corneal ulcers. This method demonstrated high sensitivity, successfully detecting bacterial pathogens in all samples positive by traditional culture methods. Notably, it also identified bacterial DNA in cases where traditional cultures failed, suggesting that tear-based Nanopore sequencing may offer greater sensitivity compared to standard culture-based methods. This noninvasive approach eliminates the need for DNA extraction and delivers results within hours, which is crucial for the timely management of corneal ulcers. 

In our study, Nanopore sequencing demonstrated complete concordance with traditional culture methods in identifying bacterial pathogens in the four cases in which cultures were positive. Specifically, S. saccharolyticus was identified by sequencing in a case in which culture grew coagulase-negative Staphylococcus (participant 1). Given that S. saccharolyticus is a coagulase-negative species, this finding aligns with the culture results and provides species-level identification. In participant 2, sequencing identified S. roterodami, a species within the S. aureus complex, corresponding to the culture result of S. aureus.¹⁶ It is worthwhile noting that conventional biochemical tests used in clinical microbiology laboratories may not differentiate between S. aureus and closely related species like S. roterodami. The ability of Nanopore sequencing to precisely identify S. roterodami highlights the enhanced resolution of sequencing-based diagnostics in distinguishing closely related bacterial species. For participants 3 and 4, both culture and Nanopore sequencing identified P. aeruginosa, a common pathogen in corneal ulcers. The rapid identification of P. aeruginosa is clinically significant due to its propensity for rapid progression and the necessity for prompt, appropriate antimicrobial therapy.¹⁷ Of note, both participants 3 and 4 were contact lens users, and P. aeruginosa is known to be the most common pathogen in contact lens–related infections, which carry a high risk for developing hypopyon and potential vision impairment.^18,19 Therefore, rapid diagnosis and appropriate antimicrobial therapy are particularly crucial in contact lens–associated corneal ulcers, a need that Nanopore sequencing addresses. Indeed, Nanopore sequencing provided results within hours, whereas culture results typically required at least 1 day, demonstrating the potential of this method to expedite clinical decision-making. In short, concordance between Nanopore sequencing and culture in these cases underscores the reliability of the sequencing method in accurately detecting the causative pathogens in corneal ulcers. 

Traditional culture-based diagnostics are often limited by high false-negative rates, which range from 32.6% to 79.4% and significant false-positive rates of approximately 10%.^20–22 This limitation stems from several factors, including variability in sample collection techniques and the fastidious nature of certain pathogens that do not grow under standard culture conditions. In contrast, our tear-based sequencing approach, which relies on detecting bacterial DNA, should in theory circumvent these hurdles. Expectedly, in our study, this method identified bacterial reads in two of the four ulcers that did not exhibit growth on culture, indicating higher sensitivity than traditional methods, although there is no definitive gold standard for comparison in these culture-negative cases. Notably, one organism detected in these culture-negative cases was C. acnes, formerly known as Propionibacterium acnes, and is considered a fastidious organism due to its stringent growth requirements.^23,24 

We detected bacterial counts in tears from four of the nine contralateral eyes. The species identified are considered commensal bacteria, as they are commonly associated with the normal ocular microbiota or are prevalent colonizers of human skin. Staphylococcus and Streptococcus have been found in normal ocular microbiota, including the tear film, eyelids, and the conjunctiva.^25–27 Specifically, S. hominis and S. saccharolyticus are known skin commensals that can be present on the eyelids and surrounding skin.^28,29 Additionally, the presence of these bacteria could also be attributed to potential contamination during tear collection, especially if the Schirmer strips inadvertently contacted the eyelid skin or eyelashes. This underscores the importance of meticulous collection methods to minimize contamination and accurately interpret sequencing results when using tear-based diagnostics. It should also be noted that while C. acnes was the predominant bacterial species in the ulcer of participant 6, it was also identified in two other ulcerated eyes and all four contralateral eyes. This highlights its dual role as both a potential cause of corneal ulcers and a commensal skin bacterium, and differentiating between the species as a true pathogen or a contaminant remains challenging.³⁰ 

A major advantage of our tear-based approach is its usefulness in clinical settings where access to an ophthalmologist is not feasible or laboratory resources are limited, as it requires no specialized equipment or training for sample collection, compared with the traditional culture techniques, which require corneal scrapings by an ophthalmologist using a slit lamp. It should be noted that the properties of the tool used for tear collection can influence DNA recovery.⁴ We used Schirmer strips for tear collection due to their low cost, ease of use, and noninvasive nature.³¹ This method eliminates the discomfort and potential complications associated with corneal scrapings, but alternative collection methods may be explored in future studies. A previous study has demonstrated that conjunctival swabbing can be used to detect and identify bacterial pathogens in microbial keratitis.³² Our tear-based method has the advantage of ease of collection and eliminates the need for DNA extraction. Eliminating the DNA extraction step not only reduces the time and resources required for microbial identification but also minimizes the need for specialized laboratory equipment and expertise. This makes the method particularly suitable for use in resource-constrained environments, where rapid and accurate diagnosis is essential but laboratory infrastructure may be limited. 

Another advantage of this sequencing-based method is its ability to detect low-abundance pathogens and provide insights into the polymicrobial nature of corneal ulcers. In our study, causative organisms identified by Nanopore included pathogens commonly associated with corneal ulcers such as Staphylococcus, Pseudomonas, and C. acnes. Notably, K. rhizophila has not been previously reported as a cause of corneal ulcers. It belongs to the Kocuria genus, a group of gram-positive bacteria that, in rare cases, can cause infectious keratitis in immunocompromised patients.^33–35 Nanopore sequencing was particularly effective at identifying bacterial agents in corneal ulcers that were too small for conventional scraping and culture, highlighting its utility in detecting the complex polymicrobial composition of these infections. Of the eight eyes that yielded positive results on Nanopore, seven showed reads from more than one bacterial genus. This finding aligns with previous reports on the polymicrobial nature of infection-induced corneal ulcers. For instance, in one study, among 81 corneal ulcers analyzed via culture, 43% yielded more than one bacterial organism.³⁶ Other studies have estimated that between 1.9% and 25% of corneal ulcers are polymicrobial in nature.^37–39 Our findings suggest that polymicrobial infections may be more common than previously recognized, possibly due to the enhanced detection capabilities of this new method. These findings also indicate that in corneal ulcers, a favorable environment may be conducive to the growth of multiple bacterial species. 

This study is limited by its small sample size and the fact that it was conducted at a single center, which may restrict the generalizability of the findings. Additionally, the risk of contamination during the tear collection and sequencing processes cannot be completely ruled out, particularly in cases with low read counts. Some detected bacterial species might represent contaminants from the normal ocular flora, as the ocular surface naturally harbors commensal microorganisms. Although we used contralateral eyes as controls to help distinguish between pathogenic bacteria and normal flora, meticulous sample collection techniques are essential to minimize contamination. 

In conclusion, while this method succeeded in identifying the causative organisms, its accuracy needs to be validated across a broader spectrum of bacterial, fungal, and viral pathogens.⁴⁰ Future research should also explore integrating this technology with antibiotic susceptibility testing to enhance its clinical utility and inform more targeted treatment strategies. 

Acknowledgments

Supported through a grant from the Connecticut Lions Eye Research Foundation. MFB is supported by NIH Research Grant P30CA016359 from the National Cancer Institute, R21EY035090 from the National Eye Institute, and the Office of the Assistant Secretary of Defense for Health Affairs through the Melanoma Research Program under Award No. HT9425-23-1-1070. CYB is funded by American Heart Association Award 857722 and K23 DK129836 from the National Institute of Diabetes and Digestive and Kidney Diseases. 

Disclosure: M. Dibbs, None; M. Matesva, None; D. Theotoka, None; C. Jayaraj, None; B. Metiku, None; P. Demkowicz, None; J.S. Heng, None; Y. Wang, None; C.Y. Bakhoum, None; J. Chow, None; M.F. Bakhoum, None