INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Corneal infection is one of the leading causes of visual loss (27). It is estimated that 30,000 cases of microbial keratitis occur annually in the United States (20) and that upwards of 100,000 cases occur annually worldwide (27). Infections often result in permanent corneal opacity, with loss of vision (1, 27), particularly in developing countries (2, 8). Bacterial keratitis is the most common form of suppurative corneal ulceration. Many organisms are capable of causing infection (7, 25, 27), and microbiologic examination of clinical specimens is required for diagnosis. Standard microbiology tests are successful in identifying a causative organism in up to 80% of cases (25). However, results are significantly compromised in cases in which the patient has received prior antibiotic treatment (7, 10).

At our institution, with a dedicated microbiology laboratory, positive culture rates vary from 40 to 60%, and only 8 to 15% of the cultures are polymicrobial. These rates are similar to those found at other clinical laboratories in the United States (16, 19, 26, 27). Algorithms for sequential restaining and reculturing of specimens have been proposed to increase the overall culture rate (9). More invasive techniques such as corneal biopsy are often undertaken for patients who continue to worsen clinically (15). Despite these measures, a significant proportion of cases remain without a microbiologic diagnosis. Clinical laboratories need a more sensitive diagnostic test that would increase the rate for identifying the etiologic organism(s) in bacterial keratitis, especially among patients who are culture negative, from whom samples were never obtained for culture but who are on antibiotics, or who have been treated without improvement.

A number of researchers have described success in identifying infectious agents in a variety of settings using culture-independent techniques (3-6, 11-14, 17, 18, 21, 24, 28). PCR has been shown to be especially suited to detecting small amounts of microbial DNA present in ocular specimens (3-5, 12, 14, 18, 24). This is particularly true for the diagnosis of intraocular viral eye disease (3, 14, 18, 24). A limited number of viruses are implicated in this setting, specifically, cytomegalovirus, herpes simplex virus types 1 and 2, and varicella-zoster virus, which permits a limited panel of PCR primers to be used to identify the etiologic agent (3, 4, 18, 24).

Use of PCR techniques for the identification of pathogens causing bacterial eye disease presents a challenge, given the large number of bacterial pathogens that are commonly encountered. Recently, the 16S subunit, or small subunit, of rRNA has been the target of PCR for the identification of bacterial pathogens in systemic diseases (6, 11-13, 17, 21-23, 28). The 16S rRNA contains regions of highly conserved sequences that are common among all previously studied bacteria interspersed with highly variable or divergent sequences that can differentiate one species from another (21). Primers that are complementary to conserved sequences of the gene and that flank variable regions can be used to amplify a portion of rRNA or its complementary ribosomal DNA (rDNA). The PCR product can then be sequenced to provide a unique identifier for the bacteria present in the specimen. This approach has been used to determine the microbial etiology of bacillary angiomatosis (22) and Whipple's disease (23) and has become a standard method for detecting bacterial pathogens (6, 28).

We investigated the possibility of using PCR amplification and sequence analysis of 16S rDNA to detect bacterial pathogens in patients with keratitis. By using a sequence alignment program, BLAST, organisms were identified by comparison of 16S rDNA sequences amplified from clinical specimens with those available in databases at the National Institutes of Health. Results of rDNA typing were then compared with those obtained by culture for patients with microbiologically documented bacterial keratitis.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Study populations and case definitions. Patients were recruited at the time of their initial presentation with keratitis to one of the ophthalmology facilities at the University of California at San Francisco, including the Francis I. Proctor Foundation, Beckman Vision Center, and San Francisco General Hospital. An ophthalmic history, including risk factors for corneal ulceration (ocular surface disease, previous ocular surgery, contact lens use, or trauma) and use of antibiotics or steroids, was obtained from each patient. An ophthalmologic examination, including slit-lamp examination of the external eye and assessment of intraocular inflammation, was also performed. Corneal specimens were collected at this visit as outlined below.

Specimen collection. All ocular specimens were collected by medically qualified personnel. Upon completion of the ocular examination and after instillation of topical anesthetic, a sterile kimura spatula or calcium alginate swab was used to scrape the area of infection. Scrapings were inoculated onto blood agar, chocolate agar, and cooked meat broth and were placed onto glass slides for staining with Gram and Giemsa stains. Additional specimens for the microbiology laboratory were collected on the basis of the methods used in that clinical setting, including, for some patients, Sabouraud's or nonnutrient agar with Escherichia coli overlay or calcofour white or acid-fast staining. A final specimen was collected with a calcium alginate swab moistened in sterile distilled water and was placed in transport buffer (10 mM Tris-HCl [pH 8.3] and 1 mM EDTA) for subsequent PCR. Specimens for the microbiology laboratory were processed immediately; organisms recovered from positive cultures were frozen at 20°C for future study. Specimens for PCR were stored at 20°C until use.

PCR of rDNA. In order to develop the proposed methodology, 12 common ocular pathogens were inoculated into transport buffer, rDNA was extracted, and PCR with oligonucleotide primers to conserved regions of the rDNA of E. coli was performed. Briefly, a calcium alginate swab was used to collect one colony from a culture plate of each of the following: Staphylococcus aureus, Staphylococcus epidermidis, alpha Streptococcus sp., group A Streptococcus, group B Streptococcus, Enterococcus fecalis, Pseudomonas aeruginosa, E. coli, Haemophilus influenzae, Enterobacter sp., Serratia marcescens, and Moraxella catarrhalis. The swab was placed in transport buffer and frozen at 20°C. Specimens were thawed, and the swab was centrifuged for 10 s in a second tube to recover the transport buffer. DNA was extracted and amplified by PCR as described previously (4, 5). Briefly, an equal volume of 1 M dithiothreitol was added to the sample to a final concentration of 40 mM. The sample was boiled for 10 min, and 1 µl of the lysed sample was used in a 100-µl volume of the PCR mixture as described previously (4). Two sets of primers were used to amplify two different segments of the 16S rRNA gene (Fig. 1; Table 1). These included primers 8FPL and 806R and primers 515FPL and 13B, which are complementary to conserved regions of the 16S rRNA of E. coli (22, 23). One microliter of the amplification product was then used in a heminested PCR with a modified upstream primer designed for automated sequencing, MF 91 or MF 515 (Table 1). A 10-µl aliquot of each PCR product was electrophoresed in an ethidium bromide-stained agarose gel with molecular weight standards for product size verification. The DNA was purified and cycle sequenced as described previously (4, 5) prior to electrophoresis on a 377 ABI automated sequencing system (Applied Biosystems, Foster City, Calif.).

View larger version (9K): [in this window] [in a new window]   FIG. 1.   Schematic diagram of the 16S rRNA gene of E. coli. Constant regions are shaded; variable regions are white. Approximate position of direction of replication of the primers are indicated by arrowheads.

Data analysis. Each 16S rDNA sequence was compared by using the BLAST alignment program with data available from GenBank at the National Institutes of Health. The computer alignment provides a list of matching organisms, ranked in order of similarity between the unknown sequence and the sequence of the corresponding organism from the database. The number of matched base pairs and the probability that the match occurred due to chance are also provided. The most closely matched sequence was determined to be that of the representative organism for the respective sequence. The percentage and absolute number of matched base pairs from each BLAST match were reported.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

The 12 ocular pathogens that were obtained from the laboratory for the purpose of developing our methodology were amplified by the rDNA primers. However, S. aureus was consistently amplified only by primer pair 515FPL and 13B.

Representative case patient. A 39-year-old female (patient 8; Fig. 2) with a history of herpes simplex keratitis managed with chronic low-dose fluorometholone suspension and trifluridine drops was referred to the Francis I. Proctor Foundation for management of infectious keratitis. She had been started on tobramycin (0.3%) drops and given a subconjunctival injection of tobramycin, and her steroid dose was reduced on the day prior to referral. Corneal scrapings had been collected that day, and these subsequently became positive for S. pneumoniae. When the patient was assessed at our institution, an inflamed eye and a stromal infiltrate (3.0 by 2.4 mm) with an overlying epithelial defect was noted. There was suppuration, with loss of corneal tissue. A sample from the patient was recultured at the time of referral after partial treatment: gram-positive diplococci were seen on Gram staining, but cultures were negative. Typing by 16S rDNA analysis revealed S. pneumoniae.

View larger version (133K): [in this window] [in a new window]   FIG. 2.   Photograph of affected eye of patient 8 showing area of corneal infection with a stromal infiltrate of 3.0 by 2.4 mm.

Typing by 16S rDNA analysis. A total of 11 corneal specimens were available from patients with microbiologically documented bacterial keratitis (Table 2). The patients ranged in age from 30 to 82 years. Eight patients had symptoms of 1 to 4 days in duration, but two patients (patients 3 and 4) had long-standing symptoms and one patient (patient 9) presented with an asymptomatic infiltrate on routine examination. All patients had predisposing risk factor(s) for development of bacterial keratitis. Five patients (patients 1, 5, 7, 10, and 11) had previously undergone penetrating keratoplasty, and three of these patients (patients 1, 7, and 11) were still using prednisolone acetate (1%) drops twice daily. Eight patients (patients 2 to 4, 6 to 9, and 11) were using antibiotic drops at presentation, ranging from commercially available broad-spectrum prophylactic medications to dual combinations of hourly fortified antibiotics for their keratitis. The infiltrate size ranged from 0.5 by 0.5 mm (patient 9) to 3.4 by 3.7 mm (patient 10); some patients had multifocal infiltrates, with the largest area of corneal involvement recorded. Visual acuity ranged from light perception to 20/80 in the affected eye. Three patients (patients 2, 4, and 8) had significant anterior chamber involvement with the presence of a hypopyon.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

We investigated the possibility of using PCR amplification and 16S rDNA sequence analysis to identify microorganisms in the setting of bacterial keratitis. These results were compared to those obtained by standard culture methods and showed that rDNA typing can be reliably used for the detection of pathogens in patients with bacterial keratitis.

To our knowledge, this is the first use of 16S rDNA typing for the detection of pathogens in patients with bacterial keratitis. Other researchers have used similar techniques for the diagnosis of other infectious diseases ever since the successful identification of the agents responsible for bacillary angiomatosis and Whipple's disease (11-13, 17). Jalava et al. (13) reported the use of PCR amplification of 16S rDNA to diagnose intra-amniotic infection in the setting of premature rupture of fetal membranes. This method was found to be fast, reliable, and more sensitive than standard bacterial culture. A variation of this technique has been reported by Hykin et al. (12) for the diagnosis of delayed postoperative endophthalmitis caused by Propionibacterium acnes among patients who were culture negative. Universal primers were used to amplify 16S rDNA from vitreous samples, followed by nested PCR with P. acnes-specific primers. The use of species-specific primers obviated the need for sequencing. Universal 16S rDNA primers have also been used in studies of premature labor (11) and pediatric sepsis (17) to detect the presence of bacterial rDNA. Although typing was not used to identify the specific organisms in those studies, the detection of bacterial DNA was found to have a sensitivity of 95% and a specificity of 98% (11) in determining the presence or absence of infection.

We found that two of the patients with microbiologically confirmed cases of infection in this study were PCR negative. This may be explained by the fact that the sample for PCR was obtained last so as to not compromise standard patient care. The sample may therefore have been inadequate, containing insufficient numbers of organisms for assay detection. It is also possible that not every organism will be amplified with similar efficiency by a given set of primers. We know from our preliminary studies that S. aureus was consistently amplified only by primer pair 515FPL and 13B. Attempts to amplify K. oxytoca, the organism cultured from patient 10, were unsuccessful with both sets of primers and were unsuccessful when a laboratory strain of K. oxytoca was used. Although the primers were chosen to be complementary to conserved regions of the 16S rRNA gene, some nucleotide sequence variation at the primer site may occur, resulting in unsuccessful amplification. Improvement in primer design should permit detection of all bacteria.

The 16S rDNA typing technique that we describe offers the potential to identify pathogens in patients who have negative cultures, patients from whom samples were never cultured but who are on antibiotics, or patients who have been treated without improvement. This is illustrated by patient 8, a patient who had been on antibiotics for 24 h prior to referral and who had negative cultures at the time of presentation at our institution. rDNA typing was able to identify S. pneumoniae. This correlated well with the culture result obtained prior to the commencement of antibiotic therapy and our Gram staining result. A major advantage of rDNA typing may be the ability to rapidly identify pathogens from patients who are reported as being culture negative.

PCR-based techniques offer several advantages over standard culture techniques. PCR collection materials are simple, inexpensive, and have a long shelf life. Results of PCR-based techniques can be available within 18 to 24 h, with the potential for shorter turn-around times as sequencing instrumentation is improved. However, the cost of the instrumentation required for sequencing will likely confine the availability of this technology to reference laboratories at major medical centers.

Some limitations of these techniques must be considered. The potential for contamination is a significant problem with PCR. Steps taken to reduce the risk of contamination in this study included UV treatment of all collection materials, tubes, pipette tips, buffer, and water; preparation of the PCR mixture in a laminar-flow hood; and use of separate rooms for sample preparation, DNA amplification, and analysis. Because no pathogenic bacteria were identified in any of our control specimens and no patient specimen was determined to be infected with a bacterium different from that identified by the microbiology laboratory, we feel that contamination was not a limiting factor in this study.

We compared our 16S rDNA sequences to the sequences in one database which may not contain rDNA sequences of unusual organisms and novel pathogens. Thus, a match may not always be possible, despite the use of a high-quality sequence. This may be partially remedied by searching other databases such as PDB, DDBJ, and EMBL. Additionally, nonbacterial pathogens such as fungi, protozoa, and viruses were not included in the analysis of corneal specimens with the primers and techniques that we describe in this report. Microbiologic studies or alternative techniques will continue to be required for a complete workup of patients with presumed infectious keratitis.

We have shown that PCR amplification and sequence analysis of 16S rDNA appears to be highly sensitive and specific for the identification of bacterial pathogens in the setting of keratitis. However, additional research is required to refine this technique in order to detect all potential bacterial pathogens of the eye. We hope that this approach will eventually become available in reference laboratories and will increase the rate of detection of etiologic microorganisms. The 16S rDNA typing technique has the potential to have a significant effect on the care of and visual outcomes for many patients with bacterial keratitis.