The diagnostic utility of nanopore targeted sequencing in suspected endophthalmitis

By reviewing the medical records from the hospital information system (HIS) of the Ophthalmic Center, a retrospective study was performed in the endophthalmitis patients who underwent NTS and culture simultaneously between January 1, 2018, and July 24, 2022. This study was carried out following the institutional guidelines and ethical standards of the 1964 Declaration of Helsinki and was approved by the Institutional Review Board of Renmin Hospital of Wuhan University (WDRY2019-K056). All patients provided their written informed consents to participate in this study.

The criteria included: 1. Recent history of eye surgery or penetrating ocular trauma, or with other predisposing factors; 2. Typical manifestations of significant loss of vision, ocular pain, ocular redness, etc.; 3. Marked intraocular inflammation like hypopyon and vitritis on ocular examination; 4. Intraocular fluid including aqueous humor (AH) or vitreous humor (VH) was collected and sent for NTS and microbiological culture simultaneously; 5. With a minimum 3-month follow-up.

The patients’ records were reviewed including demographic characteristics, best corrected visual acuity (BCVA) at their first visit, clinical features on slit-lamp examination, disease course and treatment, etiology test results (NTS and/or culture results), the time to confirmatory diagnosis, management details, length of hospitalization and BCVA at last follow-up.

The positivity rate of bacterial/fungal presence, diagnostic turnaround time, and frequency of changes in treatment protocol based on etiology test results (defined as any change directly attributable to sequencing or culture results) are used as primary outcome measures. Continuous data were presented as mean and standard deviation, whereas categorical data were presented as the number of suffered eyes and percentage. For statistical analysis, vision was reported as mean and median logMAR vision with Snellen conversion. Non-Snellen acuities were recorded in the following fashion: a visual acuity of 2/800 on Feinbloom’s low vision chart was considered equivalent to counting fingers (CF), and it was defined as 2.6 logMAR. Likewise, we used logMAR values of 2.7, 2.8, and 2.9 to represent the vision of hand movement (HM), light perception (LP), and no light perception (NLP), respectively.

To limit contamination, intraocular fluid (AH or VH) samples for NTS were obtained under strictly sterile conditions. AH samples were obtained through anterior chamber paracentesis, and VH samples were obtained during the biosurgery procedure. The operations were performed by the same operator, and purulent lesions and inflammatory exudate were cleared as much as possible during the operations. All clinical specimens were then sent to the clinical laboratory with specific pretreatment. The experimental procedures were performed by well-trained laboratory technicians in a qualified laboratory (Wuhan Dgensee Clinical Laboratory Co., Ltd. Wuhan 430,075, China).

Gram stain and KOH mount were routinely performed on aqueous humor or vitreous specimens. The remaining samples were inoculated on Columbia blood AGAR basal medium (for bacteria) and Sabouraud glucose AGAR medium (for fungi) using a BACTEC 9120 culture system (BD Diagnostics, Sparks, MD). For culture-positive cases, isolated fungi and/or bacteria were identified using the Vitek 2 Compact automated identification system (bioMerieux, Marcy L 'Etoile, Huang et al. 1061 France) and MALDI Biotyper mass spectrometry (Bruker, Marcy L 'Etoile, Huang et al. Madison, WI).

IBM SPSS Statistics Software Version 20 (SPSS, Inc, Chicago, Illinois, USA) was used for data analysis. Continuous variables like VA and diagnostic time were compared using a non-parametric, two-sided Wilcoxon rank sum test. The McNemar's test and the kappa statistic were used to compare the diagnostic positivity rates, the frequency of polymicrobial infection, and the frequency of treatment change between two methods. P value less than 0.05 was considered to be statistically significant.

Demographic characteristics, disease course, clinical features, management details, and visual outcomes were demonstrated in Table 1. A total of 30 males and 13 females were involved in our study and the mean age was 54.86 ± 18.29 years. Most cases (23/43, 53.49%) occurred after penetrating ocular injuries from metal objects, sticks, pencils, or stones. 16 patients (16/43, 37.21%) developed severe ocular inflammation after ophthalmic surgeries (14 after cataract surgery, 1 after pterygium excision, and 1 after implantable collamer lens (ICL) implantation). 1 patient had chronic comorbid conditions of liver abscesses and was suspected of binocular endogenous endophthalmitis. The mean interval between the insult (surgery/ trauma/ infection) and manifestation of the injury was 7.26 ± 7.79 (range, 1–30 days).

Presenting visual acuity (VA) results were as follows: no light perception (NLP; n = 2/44, 4.55%), light perception (LP; n = 9/44, 20.45%), hand motions (HM; n = 19/44, 43.18%), counting fingers (10/44, 22.73%), and undetermined VA (n = 4/44, 9.09%). These cases were characterized by severe anterior chamber inflammation and dense infiltration in the vitreous cavity. Hypopyon inflammatory/fibrinous exudation was observed in 33 of 44 eyes (75%). Visualization in the posterior segment view was poor in all 44 eyes due to severe vitritis (Fig. 1).

The visual outcome was defined as improvement, stabilization, and deterioration. Improved visual outcomes were found in 31 eyes (70.45%), and stabilized visual outcomes were found in 7 eyes (15.91%). 6 eyes (13.64%) had worse or deteriorated visual outcomes. 3 patients whose initial visions were LP, HM, and NLP respectively, underwent eye removal for severe inflammation involving the whole eye. Overall, the VA at the last follow-up was significantly improved compared with VA at the first visit (Z = −4.249, P < 0.001) (Table 2).

The results for NTS and culture are shown in Tables 3 and 4. In total, the positivity rate of NTS was 86.05% (37/43), and that of culture was 20.93% (9/43) (P < 0.001, kappa = 0.082). In all 9 culture-positive samples, the major pathogens were highly consistent with NTS results. Of the 37 NTS-positive samples, 22 showed polymicrobial infection. Whereas of the 9 culture-positive samples, only 1 showed polymicrobial infection. The frequency of polymicrobial results between NTS (22/43, 51.16%) and culture (1/43, 2.33%) was statistically significant (P < 0.001, kappa = 0.044).

Among 37 patients (38 eyes), a total of 45 species of bacteria and 11 species of fungi were identified by NTS, and no pathogens were detected in 6 eyes. These organisms were divided into pathogenic and non-pathogenic organisms. We have marked non-pathogenic organisms with an asterisk in Tables 3 and 4. The most frequently detected pathogens were Streptococcus spp., followed by Staphylococcus spp. and Enterobacter spp. Apart from well-recognized causative agents of endophthalmitis mentioned above, unusual and virulent pathogens were also revealed by NTS, including Sphingomonas paucimobilis, Mycobacterium abscessus, Stenotrophomonas maltophilia, Achromobacter spp., Aeromonas caviae, Morganella morganii, Acinetobacter junii, Colletotrichum spp., Moraxella osloensis, Bacillus cereus, Clostridium perfringens, Aeromonas veronii, and Citrobacter freundii.

A total of 20 microorganisms detected by NTS were identified as non-pathogenic. To our knowledge, some organisms have not been detected in any other clinical specimens: Cladosporium halotolerans, Paraburkholderia dipogonis, Yarrowia lipolytica, Sporidiobolus spp., Meyerozyma guilliermondii, Lactococcus spp., and Sac fungi. Whereas some were reported to colonize the oral cavity or the skin, and there were no reports of associated ocular infections caused by these organisms: Dialister spp., Micrococcus kristinae, Anaerococcus prevotii, Enterobacter cancerogenus, Corynebacterium tuberculostearicum, Atopobium parvulum, Corynebacterium jeikeium, Porphyromonas bennonis, Finegoldia magna, Corynebacterium confusum, Anaerococcus nagyae, Eubacterium tenue, and Lactococcus lactis.

All the patients obtained NTS results and determined treatment strategies within 1 or 2 days, with an average duration of 1.23 ± 0.43 days. The turnaround time for sequencing in the laboratory was around 8 h. The traditional culture required 3–4 days for bacterial detection and 5–7 days for fungal detection after collection of the specimens. The average turnaround time for culture was 3.58 ± 0.88 days. Thus, there was a significant difference between the two methods in terms of diagnostic time (Z = −5.805, P < 0.001). In this study, the mean duration of hospitalization in the Ophthalmic Center was 8.49 ± 2.45 days (range, from 1 to 16 days).

Specifically, in patients #2, #31, #34, #35 and #38, virulent fungi including Aspergillus gracilis, Aspergillus penicillioides, Candida albicans, Colletotrichum spp., and Candida parapsilosis were detected. Then they received additional intravenous and intraocular voriconazole (VCZ)/gentamicin (Gen) immediately. For patient #8, in whom an emerging multidrug-resistant gram-negative bacilli Enterobacter ludwigii was detected, levofloxacin (LEV) was intravenously administered instead. In addition, high abundance of Aeromonas caviae was found in patient #29. This is a rare and destructive gram-negative bacterium that often lives in sewage and seawater. This patient worked for leech farming and lived in a humid environment, which confirmed the source of the pathogen. Therefore, we added CIP to her systemic antibiotics. The management was similar in patients #12, #24, #28, #33, #37, #39, #40, #41, and #43, and the details of changes in antibiotic strategies were shown in Table 5. Most of these patients responded favorably, but 3 patients underwent enucleation because of severe inflammation involving the whole eye.

Patient #4 with EE was presented with liver abscess, high fever and sepsis. Later he developed eye pains and vision loss in both eyes (Fig. 2). He was treated with intravenous MEM + VAN and binocular intraocular CEF + IPM, but no sign of symptom improvement was observed in the right eye. Then, Phaco + PPV + SOT + IVT was performed in his right eye to remove the inflammatory lesions. AH samples was collected in both eyes for NTS and microbiological culture. 24 h later, NTS revealed Stenotrophomonas maltophilia in the right eye, and Mycobacterium abscessus + Stenotrophomonas maltophilia in the left eye. Culture of AH samples revealed no pathogen. Subsequent cultures of liver abscesses also reported the presence of Klebsiella pneumoniae + Stenotrophomonas maltophilia + Mycobacterium abscessus, confirming the diagnosis of EE.

In view of the severe condition and the newly detected pathogens by NTS, we switched the systemic antibiotic regimen to “MEM + ETM + CIP + CLR + MH”. Meanwhile, puncture and drainage of liver abscess was performed. Surprisingly, both systemic and ocular symptoms were alleviated considerably. His final BCVA maintained 20/200 in the right eye and 20/20 in the left eye.

Since the ONT first released the MinION to early users in 2014, many proof-of-concept studies have demonstrated its applications in infectious disease diagnostics [5]. For instance, the surveillance of emerging infectious diseases outbreak [6], identification of pathogen drug resistance [, , 5, 7, 8], and disease-related microbial community characterization [9]. Other studies have also provided clinical examples of the validation of NTS for pathogen identification in various samples, including aqueous humor or vitreous fluid [10], blood [11], and nasopharyngeal swabs [12]. In the field of ophthalmology, NTS is ideal for the analysis of microorganisms in AH or VH since a very limited sample volume (0.2 mL) is sufficient for detection [10].

As a third-generation sequencing technology, NTS has two unique advantages over NGS. First, it exhibits higher species-level resolution through a long-read sequencing strategy, which enhances accuracy by avoiding mis-assembly of genomic repeat regions. Second, nanopore-based technology is considered real-time as the data are generated read by read, whereas NGS results are not available until the end of the sequencing run. NTS has the potential to detect microorganisms within minutes of starting sequencing [13] and provide reliable results within 6 h of sample receipt [14]. Therefore, it is particularly useful for early antibiotic administration through timely detection of pathogens.

Similar to the study by Huang et al. [10], we were able to detect pathogens in intraocular fluid in a very short period of time with a high positivity rate. We also collated the culture and sequencing results of clinically suspected endophthalmitis from other researches in Table 6 (see Appendix A for full trans). Generally, there was a good correlation between NTS and standard culture results in double-positive cases. In this study, we also found that the main organisms identified by both methods were identical. However, the accuracy and sensitivity of the microbial profiles in culture results were poor. One potential explanation was frequent exposures to antibiotics before sample collection, which may have influenced bacterial cultivation. Also due to mutual inhibition mechanisms of bacteria, culture results often reported a single pathogen, indicating the risk of under-detection.

NTS technology has tremendous advantages in detecting multiple infections, especially in the case of mixed bacterial and fungal infections. In this article, polymicrobial results were reported in more than half of the patients by NTS. It allowed early identification of the uncultured and time-consuming microorganisms (e.g. anaerobes and fungi), regardless of prior use of broad-spectrum antibiotics [, 15, 16]. Therefore, NTS may function as a valuable supplementary to diagnostics when culture-based methods are flagged as negative.

Optimal clinical decision-making depends on identifying clinically relevant organisms present in the sample. However, conventional culture methods are always too slow and often fail to identify unusual or fastidious organisms. The average waiting times for the results of bacterial and fungal cultures were 48 and 72 h respectively, which is not conducive to guiding targeted antimicrobial therapy, especially for ocular emergencies like endophthalmitis.

While in the case of NTS, even with atypical and low-abundance pathogens, the turnaround time from sample to result was no more than 24 h [, 17, 18]. Thus, NTS enables early targeted therapy by reducing detection time and clinical turnaround time. When atypical and virulent pathogens are detected and inadequate therapy is given, NTS may save vision and reduce the risk of blindness by altering antibiotic therapy without delay [, 15, 16]. When no pathogens are detected or the detected microorganisms are determined to be non-pathogenic, this approach may contribute to an early de-escalation of broad-spectrum therapy, delaying antimicrobial resistance (AMR) [19]. Noticeably, as with the EE case in our study, NTS may be instructive in both topical and systemic medication.

Due to the presence of the blood-retinal barrier (BRB), NTS-guided antibiotic therapy is an effective complementary to patient management, but not a substitute for surgical treatment when persistent vitritis occurs [20]. We noticed the persistent symptoms and poor VA outcomes despite the coverage of broad-spectrum antibiotics in some patients, and PPV was needed to remove purulent lesions. This may be explained by the fact that visual outcomes in endophthalmitis are related to several factors, including presenting visual acuity, the presenting interval, and the promptness of appropriate therapy. Thus, further studies are required to clarify the role of NTS in altering the course of the disease and improving long-term VA. However, it has been noted in an array of literature that although the role of surgical and medical treatment in endophthalmitis varies, the most important intervention remains immediate intravitreal antibiotic injection [21]. Meanwhile, considering that poor visual prognosis of endophthalmitis is strongly associated with the type of pathogens involved, the identification of causative pathogens may still have important implications in predicting visual prognosis in the early stage of the disease.

NTS is a hypothesis-free approach and it has the potential to detect any unknown DNA-based microorganism in a clinical sample. This not only offers the promise of improved detection of traditional organisms, but also the ability to identify organisms not previously associated with endophthalmitis.

Comprehensive description of the microbial constituents may provide additional benefits: For one thing, multiple pathogens including less-common ones are assessed simultaneously during the initial sequencing run, thereby avoiding many rounds of testing. For another, it allows in-depth investigations of the ocular microbial community. This is vital both in maintaining ocular homeostasis [22] and in the pathophysiology of the disease. Changes in the eye microbiome have been confirmed to be linked with disease states like dry eye, diabetic retinopathy, glaucoma, macular degeneration, and infectious keratitis [23]. In recent studies, nanopore sequencing was proposed to monitor changes in the gut microbiome over time [, 24, 25]. Likewise, NTS could be adopted for monitoring the ocular microbiome in real-time and even function as a prognostic tool for ocular infectious and inflammatory conditions when validated further [26].

However, as with any sequencing technique, it has its limitations in determining which organisms are merely colonizers or contaminants, rather than pathogenic organisms. In response to this issue, researchers have applied variable cutoff values (e.g.  > 20 mapped read pairs per million read pairs (rM) [27],  > 50 reads [28],  > 10 reads per million (RPM) ratio metric [29] and  > 500 reads [30]) to limit the over-interpretation of low abundance microorganisms. A recent study using single gene targeted nanopore sequencing provided evidence that the samples having < 20 reads generally had a low load of pathogen [31]. Similarly, in our study, the cut-off value for the positive diagnosis was 20 reads. Besides, we have segregated the pathogenic microorganisms from those that are known to be commensals (as shown in Tables 2 and 3). In summary, clinicians need to evaluate NTS results carefully and avoid antibiotic abuse.