Clinical and genomic features of Corynebacterium macginleyi-associated infectious keratitis

Infectious keratitis is a potentially sight threatening ophthalmological emergency. Contact lens wear is a common risk factor. Diagnostic advances such as MALDI-TOF MS provides new insights into the spectrum of corneal pathogens and on microbes previously considered as commensals. Corynebacterium macginleyi was described in 1995, and in 2018, the genomic features of three isolates were reported after whole-genome sequencing. Here we describe the clinical characteristics of patients with infectious keratitis (n = 29) presumably caused by Corynebacterium macginleyi, and analyze the genomic features of C. macginleyi (n = 22) isolated from the corneal ulcers of these patients. The disease course was uneventful apart from minor interventions such as corneal cross-linking and amniotic membrane transplant. Genome sequencing and comparison revealed a highly conserved core genome of C. macginleyi. Based on the analyses of single nucleotide polymorphisms, the population could be divided into two main clades that also differed in a few clade-specific genomic islands. Patients infected with an isolate belonging to the minor clade (n = 7) presented a more severe disease. Comparisons with other corynebacterial species clearly separated C. macginleyi. C. macginleyi may be considered a corneal pathogen; genomic analysis provided insights into its population structure and disease-causing potential.


Results
We identified 29 patients from January 2004 to June 2018 with suspected infectious keratitis whose corneal cultures displayed growth of C. macginleyi. Median age was 55 years, and the most common risk factors for infectious keratitis were contact lens wear (19/29 cases; 66%) and ocular surface disease (3/29 cases; 10%) ( Table 1). Polymicrobial growth was seen in 13 (45%) of the 29 patients (Table S1). All but one patient (28/29; 97%) fulfilled the clinical criteria for infectious keratitis 20 . The patient who did not fulfill the clinical criteria was a contact lens wearer (monthly replaceable contact lenses with extended wear, i.e. continuous wear both day and night) with two weeks of pain and redness in one eye. The onset of ocular symptoms was preceded by a common cold. The patient had continued to wear contact lenses until two days before the first consultation. At the first visit, the best corrected visual acuity was 0.4 Snellen decimals. Examination in a slit lamp revealed a mixed conjunctival injection and multiple infiltrates, both centrally and in the periphery, with stromal thickening and epithelial edema. No epithelial defect could be detected with fluorescein staining. After corneal culture, treatment was initiated with topical levofloxacin (5 mg/mL) hourly with an addition of ciprofloxacin ointment (3 mg/g) for the night when the levofloxacin treatment was tapered during night time. At the first follow up 2 days after the first visit, the patient's discomfort had decreased, the conjunctival injection was diminished, the corneal infiltrates were thinner and smaller and the epithelial edema was reduced and so was the stromal thickening. Best corrected visual acuity was 0.7 Snellen decimals. The levofloxacin treatment was gradually tapered, no topical steroid or general pharmacological therapy were prescribed and no surgical intervention was performed. At follow up visit number 4 (the last visit, 16 days after the first visit), the patient was asymptomatic. Best corrected visual acuity was 1.0 Snellen decimals, and examination in a slit lamp revealed no conjunctival injection; instead subepithelial haze/scarring with some thinning of the stroma could be observed. The topical antibiotic treatment was discontinued after 17 days of treatment.
Most of the patients (26/29; 90%) received topical treatment with fluoroquinolones. Of these, 18 patients received a fluoroquinolone antibiotic only, 15 patients were treated with levofloxacin (5 mg/mL). One of these patients also received additional ciprofloxacin ointment (3 mg/g) for the night as previously described above, and three patients received moxifloxacin (5 mg/mL). A combination of a fluoroquinolone antibiotic and an additional topical antibiotic was given to 8 patients. Of these, three patients received levofloxacin (5 mg/mL) in combination with chloramphenicol as ointment (10 mg/g). Two patients received levofloxacin (5 mg/mL) in Table 1. Clinical characteristics of patients with Corynebacterium macginleyi-associated infectious keratitis, January 2004 to June 2018. a = 3 patients with no information on duration of symptoms (2 with monomicrobial growth, 1 with polymicrobial growth). b = 1 patient with polymicrobial growth information missing on BCVA at first visit. c = 2 patients with monomicrobial growth information missing on BCVA at last visit. d = Mann-Whitney U-test, 2-tailed. e = chi 2 test, 2-sided. f = Fisher's exact test, 2-sided. g = 2 patients received an amniotic membrane transplant, and 1 patient had a Gundersen. h = 2 patients received amniotic membrane transplant and 1 patient later had the eye eviscerated. BCVA = best corrected visual acuity. www.nature.com/scientificreports/ combination with fusidic acid as ointment (1%); one patient received levofloxacin (5 mg/mL) in combination with fortified vancomycin drops (50 mg/mL); one patient received levofloxacin (5 mg/mL) in combination with tobramycin (3 mg/mL) and one patient received moxifloxacin (5 mg/mL) in combination with chloramphenicol as ointment (10 mg/g). The remaining three patients, not treated with a fluoroquinolone antibiotic as initial treatment, received a combination of fortified topical vancomycin (50 mg/mL) and ceftazidime (50 mg/mL). Median duration of topical therapy was 15 days. There were no statistically significant differences in median duration of topical treatment between the group of patients with monomicrobial and polymicrobial growth (data not shown). Additional intervention such as corneal cross linking, amniotic membrane transplant, or evisceration was reported in 7/29 cases (24%) ( Table 1). The eviscerated patient was an elderly patient with recurrent herpetic keratitis who had undergone a penetrating keratoplasty in the past with a best corrected visual acuity of 0.01 Snellen decimals prior to the episode of infectious keratitis.
In 23 of the 29 patients, the isolated C. macginleyi strains were stored; one of these isolates was recovered very late and therefore not subjected to whole-genome sequencing. There were no statistically significant differences between patients with monomicrobial (n = 16) and polymicrobial growth (n = 13) regarding any of the background characteristics, best corrected visual acuity at the first or last visit, or treatment ( Table 1).
All tested C. macginleyi isolates were susceptible to benzylpenicillin (n = 25), vancomycin (n = 25), gentamicin (n = 25), ciprofloxacin (n = 27), and moxifloxacin (n = 23). Breakpoints for chloramphenicol, ceftazidime, and levofloxacin are currently lacking. MIC determination for chloramphenicol was performed for 23 strains of C. macginleyi; six had a MIC of ≤ 3 mg/L, 16 had MIC values ranging from 4 to 8 mg/L, and one had a MIC of 128 mg/L. MIC determination for ceftazidime was performed in 22 strains; 19 had a MIC of ≤ 8 mg/L and 3 had a MIC of 12 to 24 mg/L. Low MIC values (≤ 0.1 mg/L) were seen for levofloxacin in all 23 isolates tested ( Table 2). Table 2. Antibiotic susceptibility pattern of Corynebacterium macginleyi strains isolated from patients with infectious keratitis, determined as minimum inhibitory concentration. *Stored but retrieved very late therefore not sequenced. In 2 patients susceptibility testing was not performed at the time of disease episode, and later determination was not possible since the isolates were not stored or were too slow-growing.   Table 3 presents the sequencing and genome statistics for each genome, including number of pairedend reads after quality filtering, average read length, number of contigs after assembly, coverage, total genome size, GC content, and GenBank accession number. The GC content was almost identical in each genome, ranging from 57.1 to 57.2%. Draft genome sizes varied between 2325 and 2484 kb (average: 2403 kb), giving a maximum size difference between genomes of 159 kb. GC content and genome sizes were in agreement with the three previously sequenced genomes of C. macginleyi 18 . The number of CDS predicted in the genomes by the RAST annotation pipeline ranged from 2411 to 2662 (average: 2505 CDS). About 37% of all CDS were annotated as hypothetical or conserved hypothetical proteins; this reflects the current lack of functional knowledge regarding ocular corynebacteria. A KEGG analysis could assign 45% (1179 CDS in strain 12T220) of all CDS to a KEGG pathway; among these, 20 CDS were assigned to lipid metabolism. A large number of repeat regions were found (87 regions in strain 12T220).
Phylogenomic comparison of C. macginleyi. The genome sequences of the 22 strains and the three previously sequenced strains (Table 3) were phylogenetically compared by calling single nucleotide polymorphisms (SNPs) within the core genome. According to a Parsnp analysis, the core genome comprised 82% of the reference genome (strain CCUG 32361), with a total number of 75,158 SNPs in the core genome (Fig. 1). Each of the strains was individual; that is, they carried many SNPs. Even the closest relatives, strains 150801 and 160806, still had 563 SNPs in the core genome (comprising 91% of each genome). The SNP analysis further revealed that C. macginleyi strains could be separated into two main clades: a major clade (hereafter called clade I) containing 18 strains and a minor clade (clade II) containing 7 strains (Fig. 1). Subgrouping the patients according to the clade assignment of the isolated C. macginleyi strains revealed that patients with a clade II strain (n = 7) were significantly older than patients with a clade I strain (n = 15), with median ages of 71 years and 50 years, respectively (p = 0.002). Patients with a clade II strain had lower median best corrected visual acuity at the first visit (0.04 vs. 0.85; p = 0.041) and last visit (0.4 vs. 1.0; p = 0.029) than patients colonized with a clade I strain (Table 1).
In order to determine the presence of genomic islands, we conducted a total genome sequence comparison with BRIG (Fig. 2). The analysis revealed a strong synteny among the strains; only a few genomic islands were found, including two islands that were present in most clade I strains but missing in all clade II strains. One island of 18 kb was present in all clade I strains except NML 080212, encoding many hypothetical proteins, mobile element proteins, and an endonuclease. Another island of 14 kb was present in all clade I strains except NML 080212, CCUG 32361, and 180208, encoding hypothetical proteins and a few proteins associated with Comparison with other corynebacterial species. Comparative analyses showed that the closest relatives of C. macginleyi were C. accolens and Corynebacterium segmentosum; the latter two should be considered as one species as judged from average nucleotide identity (ANI) determination (Fig. 3). The ANI between C. macginleyi and the two species C. accolens and C. segmentosum was around 88% in both cases. Other related corynebacterial species were Corynebacterium striatum and Corynebacterium aurimucosum, each of which had an ANI of 73% with C. macginleyi. Three strains assigned to C. aurimucosum were wrongly classified into this species; they actually belonged to the species Corynebacterium pseudogenitalium, which had an ANI of 77% with C. macginleyi.

Discussion
C. macginleyi was first described by Riegel et al. 1 . Until recently, infections caused by C. macginleyi have only been described as case reports, mainly of ocular infections 11,12 . With recent diagnostic developments such as MALDI-TOF MS, determination to species level has become faster, easier, cheaper, and hence more frequently performed. This offers new possibilities to identify corneal pathogens among bacteria previously considered as commensal and/or low-virulence. In this article, we report clinical characteristics of bacterial keratitis presumably caused by C. macginleyi and genomic traits of C. macginleyi isolated from infected corneal ulcers. Contact lens wear was the most common risk factor, noted in 66% (19/29) of the cases. This high rate of contact lens wear in the present cohort may indicate that biofilm formation of the bacterial causative agent is involved, as suggested by both Ruoff et al. 11 and Suzuki et al. 12 . The clinical course of infectious keratitis caused by C. macginleyi could generally be described as uncomplicated, with a median best corrected visual acuity of 0.8 Snellen decimals at the first visit and 1.0 Snellen decimals at the last visit and no need for corneal transplantation. This may be due to both the lack of virulence traits and the antibiotic susceptibility pattern of C. macginleyi. Indeed, genome sequencing of 22 strains of C. macginleyi confirmed the absence of virulence factors known for a few other corynebacterial species such as Corynebacterium diphtheriae and Corynebacterium pseudotuberculosis. However, the evaluation of the pathogenic potential of C. macginleyi is difficult due to the absence of functional studies and the fact that a large portion of the genome harbors genes that encode proteins with unknown function (around 37% of all CDS); this underlines the lack of knowledge concerning ocular corynebacteria.
Our study describes for the first time the population structure of C. macginleyi on the basis of extensive genome sequencing and subsequent comparisons. It was found that the species can be divided into two phylogenetically distant lineages, here called clade I and clade II.
Before our study, only three C. macginleyi strains were genome sequenced 18 : two (NML 080212 and NML 120205) which caused ocular infections in Canada 21 and one (CCUG 32361) which was derived from a culture collection. All three strains belong to clade I. No previous study has reported information regarding clade II strains or patients infected with these strains. In the present study, patients infected with clade I and clade II strains had different clinical characteristics and visual outcomes, with those infected with clade II isolates seeming to have a more serious infection as judged from best corrected visual acuity at first and last visit and the need for more than one surgical intervention. We have no explanation for this; it might be due to the fact that these patients were considerably older, or it could be due to a small sample size. Several genomic differences exist between clade I and clade II strains. For example, clade I strains harbor a genomic island that encodes among other things, a glycerophosphodiester phosphodiesterase, which is a protein that catalyzes glycerophosphodiester macginleyi 12T220) could be aligned to the genomes of C. accolens and C. segmentosum. According to an ANI analysis, C. accolens and C. segmentosum should be considered as one species; they had an ANI of > 95%. Only ca. 7% and ca. 6% of the reference genome could be aligned with the genomes of C. aurimucosum and C. striatum, respectively. A few mistakes in corynebacterial species assignments were uncovered: three strains assigned to C. aurimucosum (620_CAUR, 911_CAUR, 1237_CAUR) were wrongly classified into this species but actually belong to the species C. pseudogenitalium, and one strain assigned to C. accolens (AH4003) is a close relative of C. striatum (ANI of 83%) and should be renamed. www.nature.com/scientificreports/ hydrolysis. In Haemophilus influenzae, a homolog of this protein has been shown to contribute to adhesion and evasion of the host immune response 22 .
The isolates of C. macginleyi in our study were fully susceptible to fluoroquinolones, which were the primary treatment for uncomplicated cases of infectious keratitis during the study period of 2004-2018. This is in contrast to an earlier finding that Etest showed a high level of fluoroquinolone resistance which correlated to mutations in the quinolone resistance-determining region (QRDR) of the gyrA gene 23 . The sequenced strains in our study had no mutations in the QRDR of the gyrA gene, which is in line with their phenotypic susceptibility to fluoroquinolones. In addition, susceptibility testing to benzylpenicillin, chloramphenicol, vancomycin, ceftazidime, and gentamicin showed a very low rate of reduced susceptibility among tested isolates of C. macginleyi.
Almost half of the patients displayed polymicrobial growth, with growth of additional bacteria (such as Staphylococcus aureus, coagulase-negative staphylococci, Moraxella catarrhalis, and Cutibacterium acnes) or fungi. The lack of any statistically significant differences between the groups of patients with mono-and polymicrobial growth regarding either background characteristics or outcome parameters may support the notion that C. macginleyi was the causative agent of infectious keratitis in the cases reported in this study. However, we cannot exclude the possibility that the lack of any significant differences between the two groups was due to unknown factors or to the small sample size.
Due to the retrospective design of this study and the lack of routine storage of isolates, we also compared the group of patients whose isolates were stored and sequenced (n = 22) with those whose were not (n = 7). We could not find any statistically significant differences in outcome measures and clinical features (data not shown). This does not support a selection of more severe cases in the group of patients whose isolates were stored, but we cannot exclude the possibility that this lack of difference was due to a small sample size. Nevertheless, it is plausible that the 22 strains of C. macginleyi stored and sequenced in this study were a representative cohort of C. macginleyi isolates from patients with infectious keratitis.
One limitation of the present study is its retrospective design, which may have resulted in loss of cases and isolates, since not all isolates of Corynebacterium spp. and diphtheroid rods were determined to species level at the time of the disease episode and/or stored for later analyses. Despite the small sample size, this study includes to date the largest number of patients with infectious keratitis caused by C. macginleyi. These sequences of 22 genomes can provide a foundation for meaningful population studies and functional investigations of hostinteracting properties and the pathogenic potential of C. macginleyi.
C. macginleyi can be considered a corneal pathogen. Despite a mostly uneventful disease course of infectious keratitis, whole-genome sequencing revealed two different clades of C. macginleyi, one of which was previously undescribed and was associated with a more severe disease course.

Materials and methods
Patient population. Patients with suspected infectious keratitis and a positive corneal culture displaying growth of C. macginleyi were identified by a search of the database at the Department of Laboratory Medicine, Clinical Microbiology, Örebro University Hospital, covering January 2004 to June 2018. The search criterion was corneal cultures with growth of Corynebacterium spp., diphtheroid rods, or C. macginleyi. As previously described 24 , an additional search for the diagnosis code for keratitis (H16.9 in the International Classification of Disease, version 10) was performed for the years 2004-2014. Identified patients were eligible for inclusion if corneal cultures were positive for Corynebacterium spp., diphtheroid rods, or C. macginleyi. Stored isolates of Corynebacterium spp. or diphtheroid rods were determined to species level by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) (Microflex LT and Biotyper 3.1, Bruker Daltonik, Bremen, Germany). Of the 29 patients identified with growth of C. macginleyi, 23 had stored isolates; however, one isolate was recovered very late, and so only 22 isolates were included in the genomic analysis.
To describe the clinical appearance of infectious keratitis in this study, we applied the clinical criteria for microbial keratitis proposed by Stapleton et al. 20 : stromal infiltration with overlying epithelial defect in combination with at least one of lesion within or overlapping the central 4 mm of the cornea and/or uveitis and/or pain. For further analysis, the patients were divided into groups according to the presence or absence of additional microorganisms isolated from the corneal ulcer (polymicrobial growth), and according to the phylogenetic clade of the isolated C. macginleyi strains.
Annotation and phylogenomics. Gene prediction and annotation of all genomes were performed with RAST 29 , and functional annotation was performed with BlastKOALA 30 . For phylogenomic analyses, the core genome was identified and aligned with the Parsnp program from the Harvest software package 31 . Reliable coregenome SNPs identified by Parsnp were used for reconstruction of whole-genome phylogeny. CSI phylogeny and JSpeciesWS were used for interspecies comparisons with other corynebacterial species 32 . Phylogenetic trees were visualized using the Interactive Tree Of Life (iTOL; https ://itol.embl.de/). The BRIG program 33 was used for visualization and for comparative genome analyses including the three previously published genomes of C. macginleyi 18  www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.