Genomics of Corynebacterium striatum, an emerging multi-drug resistant pathogen of immunocompromised patients

Abstract

Objectives

Corynebacterium striatum is an emerging multi-drug resistant (MDR) pathogen of immunocompromised and chronically ill patients. The objective of these studies was to provide a detailed genomic analysis of disease causing C. striatum and determine the genomic drivers of resistance and resistance-gene transmission.

Methods

A multi-institutional and prospective pathogen genomics program flagged seven MDR C. striatum infections occurring close in time, and specifically in immunocompromised patients with underlying respiratory diseases. Whole genome sequencing was used to identify clonal relationships among strains, genetic causes of antimicrobial resistance, and their mobilization capacity. MALDI-TOF analyses of sequenced isolates provided curated content to improve rapid clinical identification in subsequent cases.

Results

Epidemiologic and genomic analyses identified a related cluster of three out of seven C. striatum among lung transplant patients who had common procedures and exposures at an outlying institution. Genomic analyses further elucidated drivers of the MDR phenotypes, including resistance genes mobilized by IS3504 and ISCg9a-like insertion sequences. Seven mobilizable resistance genes were localized to a common chromosomal region bounded by unpaired insertion sequences, suggesting that a single recombination event could spread resistance to aminoglycosides, macrolides, lincosamindes, and tetracyclines to naïve strains.

Conclusion

In-depth genomic studies of MDR C. striatum reveal its capacity for clonal spread within and across healthcare institutions and identify novel vectors that can mobilize multiple forms of drug resistance, further complicating efforts to treat infections in immunocompromised populations.

Graphical abstract

INTRODUCTION

Corynebacterium striatum is a Gram-positive, non-sporulating rod, and normal colonizer of the skin and mucous membranes [1]. However, in immunocompromised and chronically ill patients the organism can cause infections [2] including pneumonias [3], sepsis [4] and endocarditis [5]. Infections have been further linked to prolonged hospitalizations, repeated antibiotic exposures, and prolonged use of invasive medical devices [2, 6].

Corynebacterium striatum and other Corynebacterium species have developed resistance to multiple drug classes including β-lactams, aminoglycosides and fluoroquinolones [7–9]. Patient-to-patient transmission has also been reported in intensive care units [2, 9, 10]. The outbreak potential of this emerging pathogen highlights the need for improved monitoring and means to rapidly identify patient and nosocomial reservoirs.

A multi-institutional surveillance program [11] for multi-drug resistant (MDR) organisms identified C. striatum in immunocompromised patients with underlying pulmonary conditions and exposures to common procedures, including bronchoscopy. Whole-genome sequencing revealed previously un-suspected clonal associations, particularly among patients transferred from a common outlying institution. Analyses also defined the genetic determinants mediating antibiotic resistance and their capacity for mobilization. Our findings reveal a diverse repertoire of mobilizable forms of resistance that contribute to the emergence of MDR C. striatum, and potential for naïve patient strains to rapidly acquire resistance with prolonged exposures to antibiotics, as may occur in hospitalized patients.

METHODS

Ethics

Strains and data were collected under IRB protocol 2011-P-002287 (Bry, Partners Healthcare IRB) that covers prospective genomic surveillance for MDR organisms across 4 hospitals and 12 outpatient centers, including Brigham and Women’s Hospital (BWH), a 793-bed hospital in Boston, MA with large oncology and transplant services. Genomic analyses are conducted at the request of infection control staff to assist with clinical operations, activities that are not considered research. Findings were reported to infection control teams and were not recorded in the patient record. Analyses of genomic and clinical meta-data used de-identified datasets that were not linked to the clinical record. The study was deemed Non-human subject research (NHSR) by the local internal review board (IRB).

Bacterial strains and data

Corynebacterium striatum was isolated from 7 patients over a 7-month period in 2016 using the Crimson LIMS for prospective surveillance [11]. Microbiologic identification of C. striatum included colony morphology, Gram stain, catalase positivity, and results from the API Coryne Strip (BioMèrieux, France). In cases of inconclusive speciation by biochemical testing, 16S rRNA gene sequencing by the Sanger method was used to speciate strains. Briefly, the full 16S rRNA gene was amplified, sequenced and analyzed using the Pathogenomix 16S RipSeq database (Pathogenomix.com, Santa Cruz, CA) for identification, with species-level calls made by ≥99% identity and 0.8% difference from the next species. Kirby-Bauer disk diffusion testing for antibiotic resistance used Mueller–Hinton agar supplemented with 5% sheep’s blood, with the exception of Bactrim testing, which was performed on Mueller-Hinton agar without blood. Zone diameters were reported directly given the lack of Clinical and Laboratory Standards Institute (CLSI) approved cutoffs for C. striatum. The antibiotics tested included penicillin (10 μg), tetracycline (30 μg), chloramphenicol (30 μg), cefoxitin (30 μg), gentamicin (10 μg), vancomycin (30 μg), erythromycin (15 μg), sulfamethoxazole with trimethoprim, rifampin (5 μg), linezolid (30 μg), levofloxacin (5 μg), and clindamycin (2 μg) (Becton Dickinson, Sparks Glenco, Maryland).

Patient Demographic and Epidemiologic Studies

Patient metadata including demographics, admission-discharge-transfer data, clinical lab results, prior medical conditions, procedures, medications, and notes, were extracted from the Sunquest LIS, EPIC electronic health record, and Partners Healthcare’s Research Patient Data Registry ([12]). Patient epidemiologic analyses incorporated clinical and genomic findings in evaluating potential sources of clonal spread, including common procedures, providers and overlapping locations within and external to the hospital.

Library preparation, sequencing and assembly

Total DNA was isolated using the Qiagen EZ1 platform (Qiagen, Venlo, Netherlands). Illumina MiSeq libraries were prepared using the Nextera XT system (Illumina, San Diego, CA) and sequenced using the V2 (250-bp paired-end reads) kit (Illumina, San Diego, CA). 1% of PhiX control was included in each run as a control for machine performance. The average sequencing depth resulted in 90X coverage. MiSeq de novo assemblies used SPAdes (version 3.8.0-Linux) [13]. In addition to MiSeq short-read sequencing, isolates CORYNE-1 and CORYNE-2, identified as MDR isolates with potential involvement in separate clusters concerning for clonal associations, were also sequenced on the Pacific BioSciences (PacBio) RS II Sequencer, as described [14]. Size selection was performed with BluePippin (Sage Science, Beverly, MA). Analysis of the sequence reads used SMRT Link. De novo assembly was established with the PacBio Hierarchical Genome Assembly Process (HGAP4.0) program. The improved consensus sequence was uploaded in SMRT Link 5.01.9585. to determine the final consensus and accuracy scores using arrow consensus algorithm [15].

Genomic analyses

Genomic analyses followed the methods used in Pecora et al., 2015 [11]. Antimicrobial resistance genes were identified by at least 98% homology to the database of resistance genes compiled from CARD [16]. As a second check for AMR genes, proteins were also annotated using NCBI’s Pathogen Genome Annotation Pipeline (PGAP) [17] and then queried against translated BLAST using NCBI’s Bacterial Antimicrobial Resistance Reference Gene Database (PRJNA313047). Genetic determinants of virulence and pathogenesis were identified using the Virulence Factor Database on the PATRIC website [18]. Sequences were annotated on the PATRIC website [18] and with MacVector (Apex, NC). Transposons and mobile elements were annotated using BLAST (NCBI). Genomic data from isolates has been deposited into NCBI under project number PRJNA278886 (Supplementary Table 1)

Strain Clonality Studies

Single-nucleotide polymorphisms (SNPs) were called across total genomic content (chromosomal and mobile elements) in de novo-assembled contigs using kSNP3 [19], a method that does not require a reference strain. R package phangorn calculated Hamming distances between strain pairs using concatenated SNPs, and construction of an unweighted pair group method with arithmetic (UPGMA) tree. Trees were visualized in FigTree (http://tree.bio.ed.ac.uk/software/figtree/). Clonality results were confirmed using the NCBI Pathogen Detection system in PGAP [20]. Genome comparisons were visualized using the BLAST Ring Image Generator (BRIG) [21].

MALDI-TOF analysis

Genome-sequenced clinical isolates provided a well-characterized sample set to develop matrix-assisted linear desorption/ionisation-time-of-flight mass spectrometry (MALDI-TOF MS) spectra for use in clinical lab speciation of subsequent patient isolates. Analyses were performed on the VITEK® MS (bioMerieux, Marcy-L’Etoile, France). Isolates were cultured on 5% sheep’s blood agar plates (Remel) and analyzed after 24 h incubation at 37°C and 5% CO₂. Individual colonies were analyzed in duplicate by MALDI-TOF MS with α-Cyano-4-hydroxycinnamic acid as the matrix, and identified using FDA-approved IVD 2.0 database or Knowledge Base 2.0 database. Isolates were also analyzed in research use only mode and identified using the Saramis 3.0 database. The averaged mass spectra plots were plotted using Graphpad Prism software (GraphPad, LaJolla, CA).

RESULTS

Patient characteristics and bacterial isolates

Four MDR C. striatum (isolates CORYNE-1, -2, -3 and -4) occurred within a week period from respiratory cultures, and prompted investigations for a potential outbreak and monitoring of C. striatum from future respiratory cultures. Three additional cases occurred over the next 6 months. Of the seven patients, four were recently post-lung or bone-marrow transplant and receiving immunosuppressive therapy (Table 1). Four patients had received prophylactic broad-spectrum antibiotics at the time of C. striatum isolation and remained on this therapy. Six of the patients cleared C. striatum by their next visit. However, one patient (CORYNE-3) failed to clear the pathogen for a 9-months, spanning the period from the initial and subsequent cases.

Table 1.

Patient demographic data

	Sex/Age Range	Underlying Illness	Associated Micro	Procedures	Acute respiratory process	Number of Days in Hospital	Specimen Type (Quadrant growth^)	Treatment
CORYNE-1	M/70-80	Chronic bronchiectasis, H/O M. avium	None	None	None	Outpatient	Sputum (3+)	Levofloxacin
CORYNE-2	M/60-70	Lung transplant	None	Bronchoscopy	Lung transplant complications	3	Tracheal aspirate (4+)	Broad spectrum antibiotics *
CORYNE-3	M/50-60	AML, BMT and lung transplant with GVHD	P. aeruginosa	Bronchoscopy	Lung transplant complications	Outpatient	Bronchial alveolar lavage (3+)	Broad spectrum antibiotics **
CORYNE-4	F/50-60	AML, BMT with GVHD	None	Bronchoscopy, tracheostomy, intubated	Respiratory failure following BMT	37	Sputum (4+)	Vancomycin
CORYNE-5	F/60-70	Thoracic trauma, diabetes	None	Ventilator, thoracotomy	Hemothorax	12	Induced sputum (4+)	Broad spectrum antibiotics ***
CORYNE-6	M/70-80	Lung transplant	P. aeruginosa, K. pneumoniae	Bronchoscopy, tracheostomy	Lung transplant	9	Sputum (3+)	Broad spectrum antibiotics#
CORYNE-7	M/60-70	Lung transplant	None	Bronchoscopy	Lung transplant complications	22	Bronchial alveolar lavage (1+)	Bactrim (sulfamethoxazole w/ trimethoprim)

^{^}Quadrant growth is provided for an approximate bacterial quantification

^*vancomycin, levofloxacin, cefepime

^**vancomycin, cefepime, ceftazadime, pipercillin/tazobactam, doxycycline, colistin

^***vancomycin, levofloxacin, ceftazadime, meropenem

^#vancomycin, ciprofloxacin, ceftazadime, sulfamethoxazole-trimethoprim, doxycycline, meropenem

The isolates demonstrated multi-drug resistance phenotypes to three or more drug classes (Table 2), including resistance to clindamycin, erythromycin, levofloxacin, and bactrim (sulfamethoxazole/trimethoprim). Given poor performance of standard clinical microbiologic methods to speciate isolates, MALDI-TOF analyses of genomically confirmed C. striatum developed spectra to support rapid clinical identification of C. striatum in future samples (Supplementary Figure S1 and Table S2 [22]).

Table 2.

Genomic and phenotypic antibiogram of C. striatum isolates

Antibiotic Class Antibiotic Zone of Inhibition (mm)	CORYNE-1	CORYNE-2	CORYNE-3	CORYNE-4	CORYNE-5	CORYNE-6	CORYNE-7
Fluoroquinolones	gyrA mutation (S95T, D94A, E88A)	gyrA mutation (S95T, D94G, D87G)	gyrA mutation (S95T, D94G, D87G)	gyrA mutation (S95T, D94G, D87G)	gyrA mutation (S95T, D94G, D87G)	gyrA mutation (S95T, D94G, D87G)	gyrA mutation (S95T, D94A)
levofloxacin	6	6	6	6	6	6	6
Aminoglycosides	aac(3)-XI, aph(3″)-Ib, aph(6)-Id, aph(3′)-Ic	aac(3)-XI	aac(3)-XI	aac(3)-XI, aph(3″)-Ib, aph(6)-Id, aph(3′)-Ic		aac(3)-XI	aac(3)-XI
gentamicin	16	19	6	30	29	18	18
Phenicols	cmx			cmx
chloramphenicol	8	32	35	9	30	31	28
Macrolides	erm(X)	erm(X)	erm(X)	erm(X)	erm(X)	erm(X)	erm(X)
erythromycin	12	8	6	6	10	9	8
clindamycin	6	6	6	6	6	6	6
Tetracyclines	tet(W)	tet(W)	tet(W)		tet(W)	tet(W)	tet(W)
tetracycline	28	13	12	26	12	11	12
Beta-lactams	15	21	15	15	12	14	16
penicillin
penicillin	27	28	25	27	28	23	30
Other	23	25	31	25	23	24	24
vancomycin
sulfamethoxazole w/ trimethoprim	6	6	6	6	6	6	6
rifampin	36	36	40	36	34	36	36
linezolid	36	38	40	42	38	40	38

Kirby-Bauer disk diffusion testing. Bolded columns highlight the related sub-cluster of patient isolates CORYNE-2, -3, and -6.

Genomic clonality studies

Isolates underwent whole-genome sequencing and SNP analyses (Figure 1A) to evaluate SNP distances. PacBio sequencing of strains CORYNE-1 and CORYNE-2 was used to obtain closed reference genomes for analyses. SNP analyses identified CORYNE-2, -3 and -6 to be separated by 5–16 SNPS and suggestive of a clonal sub-cluster Figure 1 (Supplementary Table S3 and Figure S2). Resistance gene content and mobile-element analyses further assessed clonal relationships [23].

Figure 1. Corynebacterium striatum clonal relationships.

(A) UPGMA tree visualizing Hamming distances of total pairwise SNPs identified across strains. Scale bar shows distance of 500 SNPs. Boxed area (red) shows the related sub-cluster of patient isolates CORYNE-2, -3, and -6. (B) Visualization of genomic data aligned to the reference genome of CORYNE-2.

Genomic causes of multi-drug resistance phenotypes

Multiple resistance genes occurred in mobile transposable elements (Table 2). Analyses of the closed genome in CORYNE-1 identified a 60kb region between the chromosomal genes dppD and cgrA/B that harbored complex transposon insertions with their associated resistance genes (Figure 2). The complement of mobilizable resistance genes in strain CORYNE-2 also occurred in this region (Supplementary Figure 3). In both CORYNE-1 and CORYNE-2, the region was flanked by unpaired insertion sequences and harbored multiple integrases, suggestive of multiple transposition and recombination events, and that this region may represent a chromosomal hotspot for insertion of mobile vectors.

Figure 2. Genomic location of mobile antibiotic resistance genes.

in closed whole genome maps of CORYNE-1 and CORYNE-2. Insets show regions of mobile resistant vectors found in the chromosome. Open reading frames are in blue, and mobile elements (transposases or insertion sequences) in black. Brackets indicate mobile unit consisting of denoted resistance genes (tet(W) = yellow, cmx = teal, aminoglycoside resistance = green, erm(X)= purple).

Transposable elements bounded by IS3504 and ISCg9a insertion sequences, and with internal homology to Tn5432 (Figure 3), carried the macrolide and lincosamide resistance gene, erm(X), and the aminoglycoside resistance gene, aac(3)-XI. Mobile element insertion sites further confirmed clonal associations among the closely related strains CORYNE-2, -3, -6. The conserved 8 Kb region (Figure 3B) had identical insertion sites amongst the sub-cluster. Erm(X)-mediated resistance also occurred in strain CORYNE-5 within an IS3504 bounded element inserted near a putative chromosomal serine protease. In CORYNE-1 and -4, this element was bounded by a single ISCg9a (Figure 3D, 3E).

Figure 3. Genomic context of macrolide, lincoasmide and aminoglycoside resistance.

Reference sequence from Tn5432::erm(X), known to mediate macrolide and lincosamide resistance (Tauch 2000) (A) and genomic context of the erm(X) gene in the 7 patient isolates (B-E), with open reading frames in blue, mobile elements in black, aac(3)-XI in green and erm(X) in purple. Unlabeled arrows indicate hypothetical proteins. The size of the conserved mobile region is underlined and indicated below each mobile element.

A mobilizable tet(W) gene occurred in all strains except for CORYNE-4. The tet(W) gene offers ribosomal protection by releasing tetracyclines from the ribosome, thus allowing for tRNAs to bind and continue protein translation [24]. The tet(W) gene occurred in a conserved 11 Kb region flanked by IS3504 insertion sequences in CORYNE-2, -3, -6 (Figure 4A), and -7 (Figure 4B). However, the related cluster of CORYNE-2, -3, and -6, demonstrated highly homologous 5′ flanking regions that differed by 25 SNPs from CORYNE-7. In CORYNE-5, IS3504 insertion sequences flanked a shorter tet(W) region of 8.4 kB. In contrast to the other 6 isolates, the tet(W) gene in CORYNE-1 was disrupted by an integrase and transposase (Figure 4D).

Figure 4. Genomic context of tetracycline resistance.

Maps of the region containing tet(W) from patient isolates (A–D), with open reading frames in blue, mobile elements in black, tet(W) in yellow and erm(X) in purple. Unlabeled arrows indicate hypothetical proteins. CORYNE-7 (B) shared homology to CORYNE-2, -3 and-6 (A), but the region differed by 25 SNPs. The size of the mobile region is underlined and indicated below each mobile element.

Both CORYNE-1 and CORYNE-4 carried an additional aminoglycoside resistance gene aph(3′)-Ic, flanked by IS26-like insertion sequences. The IS26-like element disrupted a Tn5393-like transposon that carried aminoglycoside resistance genes aph(3″)-Ib, aph(6)-Id and chloramphenicol resistance gene cmx [7] (Supplementary Figure S4). Mobile resistance genes of publicly available genomes demonstrated similar patterns of insertion sequences (IS3504, IS26, Tn5393) and others not identified in this study (IS1628, TnAs3, IS407) (Supplementary Figure S5 and Table S4).

Non-mobile determinants of resistance

All strains carried chromosomal mutations in gyrA predicted to confer fluoroquinolone resistance [8], including identical variants found within the related cluster of CORYNE-2, -3 and -6 (Table 2). Finally, while all strains exhibited phenotypic resistance to sulfamethoxazole-trimethoprim, no obvious resistance-conferring variants were identified in the dhrf and related gene pathways, nor detection of sul resistance genes.

Pathogenesis

All strains demonstrated adhesive organelles including fimbriae and sortase-pilin machinery described in pathogenic Corynebacterium spp., including sortase A, sortase C, sortase E and the pilin subunit spaE [25] (Supplementary Table S5). The sortase-pilin machinery assists in adhesion to pharyngeal epithelial cells [26]. No toxins nor additional virulence factors were identified.

Epidemiologic analyses

Genomic SNP profiles and analyses of mobile elements identified strains CORYNE-2, -3 and -6 as most concerning for a clonal cluster, which directed further epidemiologic analyses of common procedures, equipment, providers, inpatient and outpatient locations. Of note, C. striatum was recovered from all three patients on respiratory specimens obtained on the day of admission, suggesting that the pathogen had been acquired during a prior admission to our hospital or from an outside facility. Analyses of local bronchoscopes and procedures showed no overlap in clinical teams or equipment within the related sub-cluster. However, all patients within the sub-cluster had been transferred from a common outlying institution where they had multiple common procedures, including surveillance bronchoscopies, intubations and oral care. The patient infected with CORYNE-3 failed to clear the infection for nine months and temporally overlapped at the outlying institution with both CORYNE-2 and CORYNE-6. Communication with the outlying institution noted the clonal relationships and concern for potential transmission of C. striatum and other species.

DISCUSSION

We present the first high-resolution genomic studies of Corynebacterium striatum, an emerging MDR pathogen of immunocompromised and chronically ill patients. Within hospitalized settings, C. striatum has the potential to acquire MDR vectors that further complicate efforts to treat underlying infections. Its emergence as a nosocomial pathogen emphasizes the need for rapid microbiologic identification in vulnerable patients, and the importance of genomic surveillance in identifying clonal outbreaks, including capacity for C. striatum to be transmitted through common medical procedures and equipment.

Risk factors for C. striatum infections relate to the patient’s immune status, prolonged hospitalizations, chronic exposure to antibiotics and use of invasive medical equipment. Within our cohort, nearly all patients had received broad immunosuppressive therapies per recent lung or bone marrow transplants. Findings of nosocomial transmission of C. striatum may also indicate broader breaches in sanitation of medical equipment or other procedures that could lead to transmission of additional pathogenic and MDRO agents [27].

The use of whole genome sequencing in clinical surveillance enables prospective monitoring of both sporadic and clonal C. striatum infections by analyzing strain relatedness. A combination of strain SNP analyses with evaluation of resistance gene and mobile element content further supported clonality studies. Analyses of transposon insertion sites revealed common carriage of two transposons at the same genomic coordinates for the sub-cluster of CORYNE-2, -3 and -6, a sub-cluster associated with acquisition at an outside institution. Genomic studies also supported clinical microbiologic efforts to improve rapid speciation of C. striatum using the genomically-confirmed isolates to develop spectra for clinical microbiologic MALDI-TOF platforms.

Genomic analyses provided new information regarding the complement of antibiotic resistance genes in MDR C. striatum, including the degree of acquired mobilizable resistance, raising concern for capacity to facilitate further spread to naïve strains or other Gram-positive species [28, 29]. Though of low pathogenic potential in an immunocompetent host, C. striatum may act as a nosocomial reservoir of mobilizable resistance to more pathogenic species. Genomic analyses also identified novel mobile resistance cassettes. For example, while erm(X) has been reported to be mobilized by IS1249 or IS3504 elements, we identified combinations of both elements, in addition to the ISCg9a element described in Corynebacterium glutamicum [30]. Additionally, the erm(X) transposon unit had acquired the aminoglycoside resistance gene aac(3)-XI gene. We also describe the first tet(W) mobile unit flanked by IS3504 elements.

A limitation of our study was the inability to assess whether C. striatum strains cultured from pulmonary samples were also present in expected commensal niches such as skin or nasopharyngeal sites. Among participating institutions, the surveillance program analyzes strains collected in the course of clinical care, as flagged by the local infection control teams for evaluation. Patients were not further evaluated for colonization at other sites.

We illustrate the capacity for pathogen genome sequencing to rapidly identify sub-clusters of strains, including species that target specific patient populations and may further provide underlying reservoirs of drug resistance. Integration of genomic analyses provides improved information to assist hospital epidemiology and diagnostic platforms in clinical microbiology laboratories.