Wcg

SKa

b

a

ARRA

KGSPCCC

I

temelmm

Pf

h1

International Journal of Medical Microbiology 304 (2014) 1001–1010

Contents lists available at ScienceDirect

International Journal of Medical Microbiology

j ourna l h o mepage: www.elsev ier .com/ locate / i jmm

hole genome sequencing indicates Corynebacterium jeikeiumomprises 4 separate genomospecies and identifies a dominantenomospecies among clinical isolates

tephen J. Salipantea,∗, Dhruba J. Senguptaa, Lisa A. Cummingsa, Aaron Robinsona,yoko Kurosawaa, Daniel R. Hoogestraata, Brad T. Cooksona,b

Department of Laboratory Medicine, University of Washington, Seattle, WA 98195, USADepartment of Microbiology, University of Washington, Seattle, WA 98195, USA

r t i c l e i n f o

rticle history:eceived 27 March 2014eceived in revised form 11 July 2014ccepted 20 July 2014

eywords:enomicspeciationhylogenylinical testinglassificationorynebacterium jeikeium

a b s t r a c t

Corynebacterium jeikeium is an opportunistic pathogen which has been noted for significant genomicdiversity. The population structure within this species remains poorly understood. Here, we explorethe relationships among 15 clinical isolates of C. jeikeium (reference strains K411 and ATCC 43734, and13 primary isolates collected over a period of 7 years) through genetic, genomic, and phenotypic stud-ies. We report a high degree of divergence among strains based on 16S ribosomal RNA (rRNA) geneand rpoB gene sequence analysis, supporting the presence of genetically distinct subgroups. Wholegenome sequencing indicates genomic-level dissimilarity among subgroups, which qualify as four sep-arate and distinct Corynebacterium species based on an average nucleotide identity (ANIb) thresholdof <95%. Functional distinctions in antibiotic susceptibilities and metabolic profiles characterize two ofthese genomospecies, allowing their differentiation from others through routine laboratory testing. Theremaining genomospecies can be classified through a biphasic approach integrating phenotypic testingand rpoB gene sequencing. The genomospecies predominantly recovered from patient specimens doesnot include either of the existing C. jeikeium reference strains, implying that studies of this pathogen

would benefit from examination of representatives from the primary disease-causing group. The clin-ically dominant genomospecies also has the smallest genome size and gene repertoire, suggesting thepossibility of increased virulence relative to the other genomospecies. The ability to classify isolates to oneof the four C. jeikeium genomospecies in a clinical context provides diagnostic information for tailoringantimicrobial therapy and may aid in identification of species-specific disease associations.

ntroduction

Corynebacterium jeikeium is a clinically important opportunis-ic pathogen capable of causing a wide range of disorders includingndocarditis, septicemia, joint infection, pneumonia, osteomyelitis,eningitis, and soft tissue infection (Cazanave et al., 2012; Funke

t al., 1997; Ifantidou et al., 2010; Tleyjeh et al., 2005), particu-arly in immunocompromised patients or those with indwelling

edical devices. It is recognized as the most frequently recoverededically significant Corynebacterium species among patients in

∗ Corresponding author at: University of Washington Medical Center, 1959 NEacific, NW120, Seattle, WA 98195, USA. Tel.: +1 206 598 6131;ax: +1 206 598 6189.

E-mail address: stevesal@uw.edu (S.J. Salipante).

ttp://dx.doi.org/10.1016/j.ijmm.2014.07.003438-4221/© 2014 Elsevier GmbH. All rights reserved.

© 2014 Elsevier GmbH. All rights reserved.

intensive care facilities, with the capacity for nosocomial dissemi-nation (Funke et al., 1997; Tauch et al., 2005).

Previous work (Riegel et al., 1994) has sought to investigate therange of genomic, physiological, and phenotypic differences dis-played by this organism. Although DNA–DNA hybridization (DDH)studies revealed considerable genomic diversity among C. jeikeiumisolates, biochemical testing was unable to further delineate groupsamong C. jeikeium strains (Riegel et al., 1994). All isolates in thatstudy were consequently assigned to a single species under one offour “genomic groups.” Nevertheless, it was noted that some groupsdisplayed differences in their antibiotic susceptibilities (Riegelet al., 1994), hinting at dissimilarities in underlying physiology.To the best of our knowledge, the genomic diversity and popula-

tion structure of C. jeikeium has not been revisited in the 20 yearssubsequent to that publication.

Recently, whole genome sequencing technologies have madeit possible to more comprehensively explore the genomic content

1 of Medical Microbiology 304 (2014) 1001–1010

aaoqattKtatpctr(

M

I

oaujwdetaf

1

wuvpee

W

3(aBB0paAGcTpaCApCo(

cal i

sola

tes

and

asse

mbl

y

stat

isti

cs.

e

Sou

rce

Yea

r

isol

ated

Bas

esas

sem

bled

N50

stat

isti

c(b

p)

Nu

mbe

r

ofco

nti

gsEs

tim

ated

%

GC

Esti

mat

ed

fold

cove

rage

Gen

ban

kac

cess

ion

for

dra

ft

gen

ome

Gen

ban

k

acce

ssio

nfo

r

par

tial

16S

rRN

A

sequ

ence

Gen

ban

k

acce

ssio

nfo

r

par

tial

rpoB

sequ

ence

Blo

od(c

ath

eter

)20

06

2,42

1,04

2

72,2

98

139

61.4

5 15

4

JFC

M00

0000

00

KJ5

2626

8

KJ5

2629

0

Un

know

n

2009

2,33

6,58

2

119,

061

71

61.8

9

98

JFC

F000

0000

0

KJ5

2626

9

KJ5

2628

2U

nkn

own

2009

2,33

5,86

9

142,

412

68

61.7

1

113

JFC

G00

0000

00

KJ5

2627

0

KJ5

2628

8Sp

ine

2010

2,36

1,51

5

133,

002

72

61.8

105

JFC

I000

0000

0

KJ5

2627

1

KJ5

2628

1H

um

eral

mem

bran

e20

10

2,28

1,62

3

93,3

69

165

61.9

31

JFC

J000

0000

0

KJ5

2627

2

KJ5

2628

3

Blo

od(c

ath

eter

)20

11

2,51

3,21

7

101,

031

142

61.3

3

125

JFC

O00

0000

00

KJ5

2628

0

KJ5

2629

1

Un

know

n

2011

2,43

1,35

7

55,4

84

401

61.8

5

55

JFC

K00

0000

00

KJ5

2627

8

KJ5

2628

4V

erte

bral

dis

ksp

ace

2011

2,25

8,81

9

134,

848

58

62.9

5

64

JFC

R00

0000

00

KJ5

2627

4

KJ5

2628

9

Dec

ubi

tus

ulc

er

2011

2,38

6,73

3

80,5

71

121

61.6

3

116

JFC

N00

0000

00

KJ5

2627

3

KJ5

2629

3K

nee

burs

a

2012

2,30

4,26

5

257,

322

41

61.9

3

124

JFC

Q00

0000

00

KJ5

2627

5

KJ5

2628

5U

nkn

own

2012

2,31

0,14

3

170,

201

59

62.0

3

121

JFC

P000

0000

0

KJ5

2627

9

KJ5

2628

6Pl

eura

l flu

id20

12

2,39

3,24

9

175,

261

94

61.5

8

84

JFC

L000

0000

0

KJ5

2627

6

KJ5

2628

7Pl

eura

l flu

id

2012

2,43

0,05

6

100,

676

97

62.3

3

131

JFC

H00

0000

00

KJ5

2627

7

KJ5

2629

2A

TCC

Blo

od(e

nd

ocar

dit

is)

1987

2,42

6,46

1*

26,0

91

93

61.6

*

N/A

#A

CY

W01

0000

00

CJU

8782

3

AC

YW

0100

0000

*

K41

1

Axi

lla

1984

2,46

2,49

9*N

/A

N/A

61.4

0*

N/A

CR

9319

97

CR

9319

97*

CR

9319

97*

g

gen

ome

asse

mbl

y.bl

e.

002 S.J. Salipante et al. / International Journal

nd population structure of bacteria (Chan et al., 2012; Georgiadesnd Raoult, 2010), allowing for the robust classification of prokary-tes into meaningful taxonomic groups based on discrete anduantitative metrics (Richter and Roselló-Móra, 2009). Suchpproaches would be beneficial in the exploration of the evolu-ionary relationships among C. jeikeium strains, however, at thisime only the complete genome of C. jeikeium reference strain411(Tauch et al., 2005) and an incomplete genome of ATCC

ype strain 43734 (Jackman et al., 1987; Peterson et al., 2009)re available for such analyses. To better understand the popula-ion structure and diversity of C. jeikeium strains, here we haveerformed whole genome sequencing of 13 C. jeikeium primarylinical isolates. We use genomic and phenotypic data to explorehe relationships among available strains and to revisit the cur-ent classification of C. jekeium in light of genomic-era techniquesRichter and Roselló-Móra, 2009).

aterials and methods

solates

C. jeikeium strains were isolated from patients in our hospital orther hospitals in the US Pacific Northwest (Table 1), representingll C. jeikeium isolates sent to our laboratory for diagnostic molec-lar identification from the years 2006 to 2012. Reference strain C.

eikeium K411 (Kerry-Williams and Noble, 1984; Tauch et al., 2005)as obtained from the National Collection of Type Cultures (Lon-on, United Kingdom). C. jeikeium ATCC type strain 43734 (Jackmant al., 1987) was obtained from the American Type Culture Collec-ion (Manassas, Virginia, United States). All strains were culturederobically at 37 ◦C on sheep blood agar plates. DNA was extractedrom isolates using Ultraclean Microbial DNA Isolation kit (MoBio).

6S rRNA and rpoB gene sequencing

Taxonomically informative 16S rRNA and rpoB gene fragmentsere PCR amplified from bacterial genomic DNA and sequencedsing the Sanger method to establish gene sequences for 16S rRNAariable regions V1–V3 (first ∼500 bp) and a fragment of RNAolymerase subunit gene, rpoB as described elsewhere (Khamist al., 2005; Pottumarthy et al., 2003), or analogous sequences werextracted from published sequence data (Table 1).

hole genome sequencing

A total of 100 ng of each genomic DNA was digested for 2 h at7 ◦C in a 10 �L volume using 0.3 �L NEBNext dsDNA FragmentaseNew England Biolabs). DNA was simultaneously end-repairednd A-tailed in a 40 �L reaction containing 1× Rapid Ligationuffer (Enzymatics Inc.), 0.1675 mM each dNTP (New Englandiolabs), 0.1 �L E. coli DNA polymerase I (New England Biolabs),.5 �L T4 PNK (New England Biolabs), and 0.02 �L Taq DNAolymerase (New England Biolabs), incubated at 37 ◦C for 30 minnd 72 ◦C for 20 min. Annealed Y-adaptors (5′-ACACTCTTTCCCT-CACGACGCTCTTCCGATCT-3′ and 5′-P GATCGGAAGAGCGGTTCA-CAGGAATGCCGAG-3′, P = phosphorylation) were added at aoncentration of 0.2 �M and ligated at 25 ◦C for 20 min using4 DNA ligase in Rapid Ligation Buffer (Enzymatics Inc.). Afterurification with AmPure beads (Agencourt), the library was PCRmplified with KAPA HiFi HotStart ReadyMix using primer PRE-AP FWD AMP COMMON (5′-AATGATACGGCGACCACCGAGATCT-CACTCTTTCCCTACACGACGC-3′) and sample-specific barcoded

rimers (5′-CAAGCAGAAGACGGCATACGAGATXXXXXXXXCGGTCT-GGCATTCCTGCTGAACCG-3′), where Xs indicate the positionf an 8 bp sample-specific index. Cycling conditions were95 ◦C for 3 min, 10 cycles of 98 ◦C for 20 seconds, 65 ◦C for Ta

ble

1Pr

imar

y

clin

i

Isol

ate

nam

Cj4

7447

Cj1

4566

Cj1

6348

Cj1

9409

Cj2

1382

Cj4

7445

Cj3

0184

Cj3

0952

Cj4

7446

Cj3

7130

Cj3

8002

Cj4

7453

Cj4

7444

C.

jeik

eium

4373

4C.

jeik

eium

*Fro

m

exis

tin

#N

ot

app

lica

of Medical Microbiology 304 (2014) 1001–1010 1003

17b2Ao

D

(aasfiuwpI((Survtrmgvua(av

Ms

tBcPo

S

dthhS

R

P

tricfi

A.

B.

C.

0.002

0.005

Cj30952

Cj4744

44

3

Cj47447

Cj47445

C. jeikeium K411

C. jeikeium ATCC 43734

2

Cj47446C

j47453

Cj3

7130Cj30184

Cj3

8002

Cj14566

Cj16348

Cj2

1382

Cj19409

1

0.1

Cj4744

7

C. jeikeium K411

Cj47445

Cj30952

Cj47444Cj47

446

Cj37130

Cj30184

Cj38002

Cj1

4566

Cj16348

Cj2

1382

Cj19409

Cj47453

C. jeike

ium ATCC 43734

Cj47447

Cj47445

Cj47444

Cj47446

Cj30184

Cj37130Cj38002

Cj1

4566

Cj1

6348

Cj21

382

Cj19409C

j30952

C. jeikeium K411

C. jeikeium ATCC 43734

Cj4745

3

Fig. 1. Genetic and genomic phylogenies of primary C. jeikeium isolates and ref-erence strains K411 and ATCC 43734. (A) Phylogeny constructed from partial 16SrRNA gene sequence (regions V1–V2). (B) Phylogeny constructed from partial rpoBgene sequence. (C) Phylogeny constructed from whole genome variation. Brackets

S.J. Salipante et al. / International Journal

5 seconds, and 72 ◦C for 1 min, followed by one cycle of2 ◦C for 5 min). PCR product was purified with AmPureeads, pooled, and sequenced on a MiSeq (Illumina) using50 bp paired-end reads with a custom index primer (5′-GATCGGAAGAGCGGTTCAGCAGGAATGCCGAGACCG-3′). Allligonucleotides were synthesized by IDT.

ata analysis

Adaptors were trimmed using the program Fastq-Mcfhttp://code.google.com/p/ea-utils/) with skew filtering dis-bled and other parameters at their default. Draft genomes weressembled using AbySS v1.3.5 (Simpson et al., 2009). The N50tatistic (the length of which half of all contigs are equal to or larger)or assemblies was calculated using AbySS. Average nucleotidedentity, by BLAST, (ANIb) values were calculated for draft genomessing jSpecies v1.2.1 (Richter and Roselló-Móra, 2009). Variantsere called by aligning short read data from sequenced strains andublished sequence read data from C. jeikeium ATCC 43734 (SRA

D numbers SRX037287 and SRX 002393) against C. jekeium K411GenBank accession no CR931997.1) using BWA version 0.6.1-r104Li and Durbin, 2009) and SAMtools version 0.1.18 (Li et al., 2009).ingle nucleotide variant and indel calling were next performedsing VarScan version 2.3.6 (Koboldt et al., 2012) with a minimumead depth (defined as the number of sequence reads covering aariant site) of 5 and a minimum variant frequency (defined ashe fraction of reads contain a given variant) of 0.75. Estimatedead depth per strain was estimated by the Lander–Watermanethod (Lander and Waterman, 1988). Neighbor-joining phylo-

enetic trees for single gene targets and whole genome sequenceariants were generated by SplitsTree4 (Huson and Bryant, 2006),sing 1000 bootstrap replicates to estimate reliability. Compar-tive genomic analyses were visualized using BRIG version 0.95Alikhan et al., 2011) with default parameters. Gene prediction andnnotation for assemblies was performed using the RAST serverersion 4.0 (Aziz et al., 2008) with default parameters.

ass spectrometry, biochemical characterization, and antibioticusceptibility profiling

Matrix-assisted laser desorption/ionization time-of-flight spec-rometric classification of strains was performed using the MALDIiotyper system and Biotypes software (Bruker Daltonics). Bio-hemical testing of isolates was performed using the RapIDTM CBlus system (Remel). Antibiotic susceptibility testing was carriedut using the Etest® method (Biomerieux).

equence data availability

Sanger sequence from 16S rRNA and rpoB gene fragments, andraft genome assemblies for all isolates are publically availablehrough GenBank (Table 1). Sequence data generated for this studyave been submitted to the NCBI Sequence Read Archive (SRA;ttp://www.ncbi.nlm.nih.gov/sra) under study accession numberRP045192.

esults

hylogenetic analysis of 16S rRNA and rpoB gene sequences

We examined all C. jeikeium isolates identified by our labora-ory over a period of 7 years, which should therefore comprise a

elatively unbiased and representative sampling of the taxon ast is encountered clinically. Prior to initiation of these studies, thelassification of all isolates as C. jeikeium was confirmed to high con-dence (score ≥ 2.0) by mass spectrometry. We initially examined

indicate distinct genomospecies as defined in Table 2. Scale bars for each phylogenyare indicated below the corresponding letter code, and are expressed in “changesper site”. Nodes overlaid with a black dot indicate a bootstrap value of <75%.

the strain collection through targeted sequencing of taxonomi-cally informative genes, comparing our results to those of the fullysequenced reference strain K411 (Tauch et al., 2005) and partiallysequenced reference ATCC type strain 43734 (Jackman et al., 1987).The 16S rRNA gene is a useful target for classifying bacterial species(Clarridge, 2004), and phylogenetic analysis of partial 16S rRNAgene sequence data from our isolates suggested the presence of

genetically distinct C. jeikeium subdivisions (Fig. 1A), with one cladeencompassing most of the clinical isolates. Nevertheless, 16S rRNAgene sequence data can sometimes prove misleading or unreliablein inferring bacterial population structure (Georgiades and Raoult,

1 of Medical Microbiology 304 (2014) 1001–1010

2gagaPs(ta

W

asSnCmiA

pi(wba(wcsovaaCdts

tnMgeis2iclreatt

C

tGia

b

valu

es

base

d

on

wh

ole

gen

ome

dat

a.

Gen

omos

pec

ies

1

2

3

4

Isol

ate

Cj3

7130

Cj3

0184

Cj2

1382

Cj1

4566

Cj1

6348

Cj3

8002

Cj4

7453

Cj1

9409

Cj4

7446

C.

jeik

eium

K41

1

C.

jeik

eium

ATC

C43

734

Cj4

7447

Cj4

7445

Cj3

0952

Cj4

7444

ecie

s

1

Cj3

7130

100*

97.7

8

97.6

9

97.5

5

97.4

3

97.7

6

97.1

4

97.6

4

97.1

4

85.8

7

85.6

3

85.8

7

86.0

1

86.0

4

86.0

2C

j301

84

97.3

4

100

97.5

97.1

5

97.4

7

98.3

9

96.8

3 97

.34

97.0

3

85.7

7

85.5

2

85.7

8

85.9

4

86.1

9

86.1

8C

j213

82

97.6

97.7

1

100

97.4

3

97.5

4

97.5

1

96.9

6 97

.59

97.2

6

85.7

5

85.8

6

85.7

7

85.8

4

86

85.9

2C

j145

66

97.5

3

97.5

6

97.5

3

100

97.5

5

97.5

97.2

1

97.6

2

97.2

6

85.8

6

85.5

8

85.9

86.0

1

86.1

4

86.0

5C

j163

48

97.3

6

97.6

5

97.4

9

97.4

6

100

97.4

4

97.0

3

97.5

5

97.2

4

85.8

3

85.7

7

86.1

4

86.1

85.9

9

86.0

9C

j380

02

97.6

5

98.6

9

97.4

9

97.4

2

97.4

5

100

97.0

6

97.5

3

97.1

85.9

6

85.7

4

85.8

3

86.1

7

86.2

9

86.2

2C

j474

53

96.9

6

96.9

5

96.9

8

96.9

4

96.9

6

96.8

4 10

0

97.1

4

96.5

9

86.1

1

85.8

5

86.1

2

86.2

6

86.2

5

86.0

8C

j194

09

97.5

1

97.6

2

97.6

3

97.4

8

97.5

5

97.4

8

97.1

3

100

97.2

4

85.6

9

85.5

5

85.8

5

85.9

1

86.0

1

85.8

3C

j474

46

97.1

3

97.0

1

97.0

6

97.2

1

97.0

7

97.0

7

96.5

2

97.0

6

100

86.3

1

86.0

7

86.5

86.6

86.2

4

86.4

52

C.

jeik

eium

K41

185

.62

85.6

7

85.6

4

85.6

85.7

6 85

.71

85.8

4

85.5

9

86.4

100

96.3

5

96.0

8

96.2

3

91.6

4

91.9

7

C.

jeik

eium

ATC

C43

734

85.2

8

85.4

2

85.6

6

85.2

8

85.5

6

85.4

8

85.4

8

85.3

7

85.7

3

96.4

7

100

96.4

5

96.7

1

91.5

5

91.4

3

Cj4

7447

85.8

2

85.9

1

85.9

9

85.8

86.0

8

85.7

5

86.1

9

85.9

86.6

6

96.1

4

96.5

1

100

98.0

3

91.8

7

92.2

5C

j474

45

85.8

3

85.8

6

85.8

7

85.8

8

85.9

4

86.0

6

86.0

9

85.8

6

86.7

5

96.2

8

96.6

1

97.9

2

100

91.8

4

92.5

33

Cj3

0952

86.0

2

86.1

3

86.1

1

86.0

5

86.2

2

86.2

7

86.1

8

86.1

1

86.2

6

91.7

6

91.5

5

91.7

5

91.8

2

100

93.9

64

Cj4

7444

86.0

4

86.1

2

86.1

6 86

.09

86.3

1

86.1

7

86.1

7

86

86.4

9

92.1

5

91.5

3

92.3

9

92.5

7

93.9

9

100

cate

s

AN

Ib

valu

es

grea

ter

than

95%

004 S.J. Salipante et al. / International Journal

010), especially among Corynebacterium species, where 16S rRNAene diversity tends to be relatively low (Khamis et al., 2005). As anlternative approach, we therefore sequenced rpoB, a housekeepingene that has proven informative for the molecular identificationnd classification of Corynebacterium isolates (Khamis et al., 2005).hylogenetic analysis of rpoB also revealed substantial populationtructure suggesting the presence of genetically distinct groupsFig. 1B) although placement of strains to particular clades withinhe larger trees was not fully consistent when comparing 16S rRNAnd rpoB gene targets.

hole genome sequencing and identification of genomospecies

To more thoroughly explore the genomic differences displayedmong the collected strains, we next performed whole genomeequencing and de novo genome assembly of each clinical isolate.equencing was performed to an average read depth of 100 × (ando less than an average of 30 × per sample) with respect to the. jeikeium reference genome. For all isolates considered, the esti-ated GC content and number of nucleotides assembled for clinical

solates were similar to reported values for C. jeikeium K411 andTCC 43734 (Table 1).

A phylogeny based on variants (single nucleotide polymor-hisms and indels) identified across the whole genomes of strains

ndependently supported division of isolates among distinct taxaFig. 1C). Both broad partitions, as well as finer-scale relationshipsithin the partitions, were all supported by high confidence (>75%)

ootstrap values. In the genomic phylogeny, reference strains K411nd ATCC 43734 formed a clade along with two primary isolatesCj47447 and Cj47445). Two other strains (Cj30952 and Cj47444)ere related to this group, but together constituted a separate

lade. The remaining nine isolates, representing the majority oftrains examined, formed a distinct group more distantly related tother clades. The structure of the tree deduced from whole genomeariation closely resembled that of the rpoB-based tree (Fig. 1B),dding corroborating support for these partitions. Further, compar-tive genomic analysis of sequenced isolates against the completed. jeikeium K411 reference genome (Fig. 2A) demonstrated variableegrees of divergence, and variable regions of divergence, amonghe clinical strains which seemingly partitioned isolates among theame groups identified by the genomic phylogeny (Fig. 2B).

In order to formally circumscribe potential subdivisions amonghe isolates based on whole genome data, we performed an averageucleotide identity by BLAST (ANIb) analysis (Richter and Roselló-óra, 2009) (Table 2). Pairwise ANIb values of less than 95% are

enerally accepted as the cutoff for defining separate species (Gorist al., 2007; Richter and Roselló-Móra, 2009), and by that criterionsolates in this study were sharply delineated into four separatepecies, which we have designated C. jeikeium genomospecies 1,, 3, and 4. The species most genomically divergent from exist-

ng reference strains (genomospecies 1) is predominant among thelinical isolates we have encountered, representing 9 of the 15 iso-ates included in this study. The other 3 genomospecies have fewerepresentatives: the group containing both existing C. jeikeium ref-rence strains K411 and ATCC 43734 (genomospecies 2) bears twodditional clinical isolates from our study, whereas the remainingwo species (genomospecies 3 and 4) have only a single represen-ative each.

omparative gene content of C. jeikieum genomospecies

Differences among genomospecies were also manifested in

heir predicted gene content and core genome content (Table 3).enomospecies 1 harbored significantly fewer predicted cod-

ng sequences (2-tailed t-test, p value = 0.009) and functionallynnotated coding sequences (2-tailed t-test, p value = 3.3 × 10−7) Ta

ble

2Pa

irw

ise

AN

I

Gen

omos

p

*B

old

Ind

i

S.J. Salipante

et al.

/ International

Journal of

Medical

Microbiology

304 (2014)

1001–1010

1005

Table 3Comparative gene content of C. jeikeium genomospecies and core genome content.

Genomospecies 1 2 3 4

Isolate Cj37130 Cj30184 Cj21382 Cj14566 Cj16348 Cj38002 Cj47453 Cj19409 Cj47446 C.jeikeiumK411

C.jeikeiumATCC43734

Cj47447 Cj47445 Cj30952 Cj47444

Number codingsequences

2039 2136 1992 2063 2053 2036 2117 2080 2086 2211 2293 2148 2202 2001 2161

Functionally-annotatedcodingSequences

1382 1369 1374 1375 1365 1382 1378 1392 1374 1423 1435 1432 1434 1397 1452

Functionally-annotatedcodingsequencesshared withK411

1362 1341 1355 1353 1347 1363 1358 1367 1351 1423 1401 1412 1408 1385 1408

% K411functionally-annotatedcodingsequencespresent

95.71 94.24 95.22 95.08 94.66 95.78 95.43 96.06 94.94 100.00 98.45 99.23 98.95 97.33 98.95

Estimatednumber of coregenes forgenomospecies

1139 1384 N/A* N/A*

*Not applicable, only one representative of the genomospecies available.

1006 S.J. Salipante et al. / International Journal of Medical Microbiology 304 (2014) 1001–1010

K411GC content

GC skewNegativePositive

100% identity70% identity50% identity

Cj37130100% identity70% identity50% identity

Cj30184100% identity70% identity50% identity

Cj21382100% identity70% identity50% identity

Cj14566100% identity70% identity50% identity

Cj16348100% identity70% identity50% identity

Cj38002100% identity70% identity50% identity

Cj47453100% identity70% identity50% identity

Cj19409100% identity70% identity50% identity

Cj47446

100% identity70% identity50% identity

Cj47447

C. jeikeium K4112462499 bp

100% identity70% identity50% identity

ATCC 43734100% identity70% identity50% identity

Cj47445100% identity70% identity50% identity

Cj30952100% identity70% identity50% identity

Cj47444

1

2 3 4

A.

B.

2400 kbp

200 kbp

400 kbp

600 kbp

800 kbp

1000 kbp

1200 kbp1400 kbp

1600 kbp

1800 kbp

2000 kbp

2200 kbp

Fig. 2. Circular plot of genome diversity in sequenced C. jeikeium clinical isolates. (A) The map, GC content, and GC skew (either positive [enrichment of G over C] or negative[enrichment of C over G]) of the fully sequenced C. jeikeium K411 reference genome is depicted in the innermost three rings. The white and colored regions of outer ringsindicate sequences absent and present, respectively, in the draft genomes of clinical isolates and type strain ATCC 43734 relative to the C. jeikeium K411 reference. Intensityof coloration in outer rings indicates the degree of sequence identity relative to the reference genome. With the exception of the K411 reference strain, rings are groupedaccording to the genomospecies defined in Table 2. Strains are listed by genomospecies, from left to right, in the order that they occur when moving from the innermost tot a is bed .

tgicg

sosgsb

he outermost ring (key). (B) Example of a divergent genomic region. Enlarged areivergence is observed that distinguishes group 1 (inner 9 rings) from other groups

han isolates from genomospecies 2. The representative fromenomospecies 3 had the fewest predicted coding sequences of anysolate, regardless, the number of functionally annotated predictedoding sequences was within the range exhibited by members ofenomospecies 2.

The overlap in gene content between isolates and the fullyequenced K411 reference strain mirrored the calculated degreef genomic divergence measured by ANIb: isolates from genomo-

pecies 1 had the smallest percentage of predicted orthologousenes shared with K411 (Table 3), whereas members of genomo-pecies 2 had the highest. Although a core genome size could note estimated for genomospecies 3 and 4, because these groups

tween 800 and 1000 kbp in the K411 reference genome, as indicated. A pattern of

were represented by only a single isolate, the number of genesidentified in isolate Cj47444 of genomospecies 4 which werenot present in the sequenced reference strain (44 genes) wassignificantly higher than the number observed for other isolates(average of 21 genes, z-test p value = 1), suggesting a significantdegree of additional genomic content for that genomospecies.

Phenotypic characterization of C. jeikieum genomospecies

We next investigated whether any of the genomospecies weredistinguishable on the basis of phenotype. We performed a bat-tery of standard biochemical tests used for the clinical classification

S.J. Salipante

et al.

/ International

Journal of

Medical

Microbiology

304 (2014)

1001–1010

1007

Table 4Biochemical testing results.

Carbohydrate utilization Glycosidase substrates

Genomospecies Isolate Glucose Sucrose Ribose Maltose p-nitrophenyl-�-D-glucoside

p-nitrophenyl-�-D-glucoside

p-nitrophenyl-N-acetyl-�-D-glucosaminide

p-nitrophenylglycoside

o-nitrophenyl-�-D-galactopyranoside

1 Cj37130 + –/+* + – – – −/+ – –Cj30184 + – + – – – – – –Cj21382 + – + – – + – – -Cj14566 + – + – – – – – –Cj16348 + – + – – – – – –Cj38002 + – + – – – −/+ – –Cj47453 + – + – – – – – –Cj19409 + – + – – – – – –Cj47446 + – + – – −/+ – – –

2 C. jeikeium K411 + – – – – −/+ – – –C. jeikeium ATCC43734

+ – −/+ – – – – – –

Cj47447 + – + – – – – – –Cj47445 + – + – – −/+ – – –

3 Cj30952 – – – – – – – – –4 Cj47444 + – + – – −/+ – – –

Aminopeptidase substrates Single-enzyme or phenotypic tests

Genomospecies Isolate Proline-�-naphthylamide

Tryptophan-�-naphthylamide

Pyrrolidine-�-naphthylamine

Leucyl-glycine-�-naphthylamide

Leucine-�-naphthylamide

Urease Nitrasereductase

Catalase Yellowpigment

Phosphatase Esterase

1 Cj37130 – + – + + – – + – + +Cj30184 – + – + + – – + – + +Cj21382 −/+ + – + + – – + – + +Cj14566 – + – + + – – + – + +Cj16348 – + – + + – – + – + +Cj38002 −/+ + – −/+ + – – + – + +Cj47453 – + – + + – – + – + +Cj19409 – + – + + – – + – + +Cj47446 – + – + + – – + – + +

2 C. jeikeiumK411

– + – + + – – + – + +

C. jeikeiumATCC 43734

– −/+ – + + – – + – + +

Cj47447 – + – + + – – + – + +Cj47445 – + – + + – – + – + +

3 Cj30952 – −/+ – −/+ −/+ – – + – + +4 Cj47444 – + – + + – – + – + +

* Weak positive.

1 of Medical Microbiology 304 (2014) 1001–1010

olsIpog�ipw

cacicfadradagtgvor

D

d(asosdvima22pgivnts

“piiatgfco

scep

tibi

lity

pro

file

s.

ies

Isol

ate

Cef

tria

xon

e

Cip

rofl

oxac

in

Cli

nd

amyc

in

Eryt

hro

myc

in

Gen

tam

icin

Pen

icil

lin

Van

com

ycin

Tetr

acyc

lin

e

Trim

eth

opri

m-

sulf

amet

hox

azol

e

Cj3

7130

16#

0.25

>256

4

0.25

8

2

0.25

0.25

/4.8

Cj3

0184

16

0.25

2*

0.5

0.12

5

>32

1

0.5

1/19

Cj2

1382

16

0.25

>256

8 0.

125

>32

2

0.5

0.5/

9.5

Cj1

4566

32

0.25

>256

32

0.12

5

8

2

0.5

1/19

Cj1

6348

16

8

>256

8

0.12

5

16

1

0.5

0.5/

9.5

Cj3

8002

32

>32

4

0.25

0.25

8

2

0.5

0.12

5/2.

4C

j474

53

8

0.25

>256

4

2

8

2

0.5

1/19

Cj1

9409

16

0.12

5

>256

16

0.25

4

1

0.5

1/19

Cj4

7446

>32

>32

>256

64

1

>32

2

2

8/15

2C.

jeik

eium

K41

1

>32

0.5

>256

>256

>102

4

>32

1

8*

>32/

608

C.

jeik

eium

ATC

C

4373

4>3

2

0.25

256

4

>102

4

>32

2

>256

16/3

04C

j474

47

>32

>32

>256

>256

512

>32

1

64

>32/

608

Cj4

7445

>32

>32

>256

32

>102

4

>32

1

32

>32/

608

Cj3

0952

>32

>32

>256

>256

0.12

5

>32

1

0.5

1/19

Cj4

7444

>32

>32

>256

>256

0.12

5

32

1

0.5

>32/

608

por

ted

in

mg/

L.s

full

resi

stan

ce, *

ind

icat

es

inte

rmed

iate

resi

stan

ce.

008 S.J. Salipante et al. / International Journal

f bacteria in patient samples, comprising four carbohydrate uti-ization tests, five glycosidase substrate tests, five aminopeptidaseubstrate tests, and six single-enzyme or phenotypic tests (Table 4).solate Cj30952, the sole representative of genomospecies 3, dis-layed a markedly different biochemical profile compared withther isolates, being uniquely characterized by an inability to utilizelucose and reduced aminopeptidase activity against leucine--naphthylamide. These characteristics separated this particular

solate from all other isolates in all other genomospecies. Metabolicrofiles within and among the remaining genomospecies, however,ere not distinct.

Similarly, we sought to determine whether the antibiotic sus-eptibilities of the genomospecies could be used to differentiatemong them. All isolates tested were marked by resistance to peni-illin and ceftriaxone, and were found to be either resistant or haventermediate resistance to clindamycin, and were sensitive to van-omycin (Table 5). Genomospecies 2 was uniquely distinguishablerom other groups both in displaying resistance to gentamicin andlso elevated resistance to tetracycline. Although not a uniquelyefining characteristic, all members of genomospecies 2 also car-ied resistance to trimethoprim-sulfamethoxazole. Interestingly,lthough C. jeikeium strain K411 was initially identified as a multi-rug resistant strain from a hospitalized patient (Kerry-Williamsnd Noble, 1984), it appears that its resistance profile represents aeneral characteristic of genomospecies 2, rather than unique fea-ure of the original K411 isolate. It was not possible to define otherenomospecies based on their pattern of antibiotic susceptibilities:ariable levels of resistance to ciprofloxacin and erythromycin werebserved within genomospecies 1, and no consistent differences inesistance profiles distinguished among genomospecies 1, 3, and 4.

iscussion

The genus Corynebacterium, in general, is diverse and is notablyifficult to speciate using phenotypic identification approachesRoux et al., 2004), posing challenges for the task of distinguishingmong isolates using standard biochemical and phenotypic clas-ifiers. Further, available DNA–DNA hybridization data, althoughnce considered a “gold standard” for circumscribing species, areomewhat difficult to interpret for C. jeikeium because the proce-ure utilized was noted to yield lower-than-average hybridizationalues (Grimont et al., 1980; Riegel et al., 1994). Advancementsn genetic and whole genome sequencing technologies have now

ade it possible to infer intra-species population structure moreccurately based on quantitative genomic information (Chan et al.,012; Georgiades and Raoult, 2010; Richter and Roselló-Móra,009), enabling more comprehensive studies of Corynebacteriumopulation structure. In this work, we examined the genetic andenomic relationships among primary C. jeikeium clinical isolatesdentified through routine clinical workup in our laboratory, pro-iding an unselected sampling of pathogenic strains collected overearly a decade. Taken together, these data provide evidence thathe 15 isolates considered here constitute four separate and distinctpecies.

Our findings are consistent with the earlier description of 4genomic groups” within the C. jeikeium taxon (Riegel et al., 1994),roviding corroborating evidence for the conclusions presented

n the present work. Nevertheless, our use of genomic sequenc-ng data has enabled us to formally circumscribe these “groups”s true genomospecies based on quantitative metrics (Table 2),o precisely define the phylogenomic relationships among those

enomospecies (Fig. 1), and to explore the differences in genomiceatures that distinguishes them (Table 3). These investigations,oupled with phenotypic testing information and the generationf 13 new draft genome sequences derived from members of each Ta

ble

5A

nti

biot

ic

su

Gem

nop

ec

1 2 3 4

#A

ll

valu

es

reB

old

ind

icat

e

of Med

ggtat

lssostdcGbln2dtims

s(gfittaptic

ttcrstgspiot

phrhTfbcabaicdfiw

S.J. Salipante et al. / International Journal

enomospecies (Table 1), significantly expand knowledge of theenomic and phenotypic variation within the taxon. In light ofhese results, we advocate that C. jeikeium should be recognizeds four separate species in order to accurately reflect the taxon’srue population structure.

Genomospecies 1, which comprises the majority of clinical iso-ates, has both a significantly smaller genome and significantlymaller gene repertoire than that of the other C. jekieum genomo-pecies (Table 3). Evidence suggests that the virulence potentialf pathogenic bacteria is inversely proportional to overall genomeize and gene content (Georgiades and Raoult, 2011), an observa-ion which has been broadly demonstrated for phylogeneticallyissimilar intracellular (Moran, 2002; Wixon, 2001) and extra-ellular (Georgiades and Raoult, 2011) disease-causing agents.enome size itself has been shown to be a better predictor ofacterial virulence than the absolute number of classical “viru-

ence factor” genes in comparing pathogenic bacteria compared toon-virulent or less-virulent counterparts (Georgiades and Raoult,011; Merhej et al., 2013), likely reflecting the process of gene lossuring host adaptation (Merhej et al., 2009). Given this correla-ion, it is probable that genomospecies 1 is not overrepresentedn our dataset due to chance sampling, but rather because it

ore frequently causes human disease than do the other genomo-pecies.

Of particular note, both of the extant C. jeikeium referencetrains, K411 and ATCC 43734, belong to a single genomospeciesgenomospecies 2), which is distinct from the clinically dominantenomospecies identified in our study (genomospecies 1). Thisnding has two major implications. First, it suggests that much ofhe genomic diversity present among members of the C. jeikeiumaxon has not previously been represented through the two avail-ble reference strains. Secondly, it argues that studies of C. jeikeiumathogenesis may benefit from investigation of isolates classifiedo genomospecies 1, as existing C. jeikeium reference strains are notn a group responsible for the majority of infections encounteredlinically.

Available data suggest that two of the four genomically dis-inct species can be distinguished using routine clinical laboratoryechniques: the sole representative of genomospecies 3 is uniquelyharacterized through negative glucose utilization (Table 4) whileesistance to gentamicin can distinguish members of genomo-pecies 2 from others (Table 5). In the absence of additional testing,he only reliable primary mean to classify representatives fromenomospecies 1 and 4 is through the use of whole genomeequence analysis (Fig. 1C). However, a biphasic approach shouldrove tractable to many clinical laboratories: first, phenotypic test-

ng may be used to classify an unknown isolate to genomospecies 2r 3 (Tables 4 and 5), followed as necessary by rpoB gene sequencingo differentiate among species 1 and 4 (Fig. 1B).

Given that one genomospecies (genomospecies 1) has dispro-ortionately high representation among the clinical isolates weave studied, and that two of the other genomospecies are poorlyepresented in this survey, we cannot know at this time whether weave fully sampled all species previously classified as C. jeikeium.he isolates examined here represent a comprehensive collectionrom our laboratory, but are nevertheless limited in absolute num-er and potentially geographical scope. It is possible that genomicharacterization of supplementary isolates as they become avail-ble will delineate additional species, including those that maye less prevalent pathogens. Efforts should be taken to recognizend incorporate such groups into the Corynebacterium taxonomy,f they do indeed exist. Regardless, even now the proposed spe-

iation of the C. jeikeium taxon carries with it important clinicaliagnostic information about expected antibiotic resistance pro-les of isolates classified to a particular group (Table 5). Futureork may uncover additional clinically relevant distinctions among

ical Microbiology 304 (2014) 1001–1010 1009

these species, such as preferred sites of infection or the severity ofdisease course caused by infection, shedding light on other poten-tially characterizing biological properties of these agents that aretaxon specific and not possible to assay by conventional laboratorytesting.

Acknowledgment

This work was supported in part by the National Center forAdvancing Translational Sciences of the National Institutes ofHealth [grant number UL1TR000423].

References

Alikhan, N.F., Petty, N.K., Ben Zakour, N.L., Beatson, S.A., 2011. BLAST Ring ImageGenerator (BRIG): Simple prokaryote genome comparisons. BMC Genomics 12,402.

Aziz, R.K., Bartels, D., Best, A.A., DeJongh, M., Disz, T., Edwards, R.A., Formsma, K.,Gerdes, S., Glass, E.M., Kubal, M., Meyer, F., Olsen, G.J., Olson, R., Osterman, A.L.,Overbeek, R.A., McNeil, L.K., Paarmann, D., Paczian, T., Parrello, B., Pusch, G.D.,Reich, C., Stevens, R., Vassieva, O., Vonstein, V., Wilke, A., Zagnitko, O., 2008. TheRAST Server: Rapid annotations using subsystems technology. BMC Genomics9, 75.

Cazanave, C., Greenwood-Quaintance, K.E., Hanssen, A.D., Patel, R., 2012. Corynebac-terium prosthetic joint infection. J. Clin. Microbiol. 50, 1518–1523.

Chan, J.Z., Halachev, M.R., Loman, N.J., Constantinidou, C., Pallen, M.J., 2012. Definingbacterial species in the genomic era: Insights from the genus Acinetobacter. BMCMicrobiol. 12, 302.

Clarridge 3rd, J.E., 2004. Impact of 16S rRNA gene sequence analysis for identificationof bacteria on clinical microbiology and infectious diseases. Clin. Microbiol. Rev.17, 840–862.

Funke, G., von Graevenitz, A., Clarridge 3rd, J.E., Bernard, K.A., 1997. Clinical micro-biology of coryneform bacteria. Clin. Microbiol. Rev. 10, 125–159.

Georgiades, K., Raoult, D., 2010. Defining pathogenic bacterial species in the genomicera. Front. Microbiol. 1, 151.

Georgiades, K., Raoult, D., 2011. Genomes of the most dangerous epidemicbacteria have a virulence repertoire characterized by fewer genes but moretoxin–antitoxin modules. PloS One 6, e17962.

Goris, J., Konstantinidis, K.T., Klappenbach, J.A., Coenye, T., Vandamme, P., Tiedje, J.M.,2007. DNA–DNA hybridization values and their relationship to whole-genomesequence similarities. Int. J. Syst. Evol. Microbiol. 57, 81–91.

Grimont, P.A.D., Popoff, M.Y., Grimont, F., Coynault, C., Lemelin, M., 1980. Repro-ductibility and correlation study of three deoxyribonucleic acid hybridizationprocedures. Curr. Microbiol. 4, 325–330.

Huson, D.H., Bryant, D., 2006. Application of phylogenetic networks in evolutionarystudies. Mol. Biol. Evol. 23, 254–267.

Ifantidou, A.M., Diamantidis, M.D., Tseliki, G., Angelou, A.S., Christidou, P., Papa,A., Pentilas, D., 2010. Corynebacterium jeikeium bacteremia in a hemodialyzedpatient. Int. J. Infect. Dis. 14 (Suppl 3), e265–e268.

Jackman, P.J.H., Pitcher, D.G., Pelczynska, S., Borman, P., 1987. Classification ofcorynebacteria associated with endocarditis (group JK) as Corynebacteriumjeikeium sp. nov. Syst. Appl. Microbiol. 9, 83–90.

Kerry-Williams, S.M., Noble, W.C., 1984. Plasmid-associated bacteriocin productionin a JK-type coryneform bacterium. FEMS Microbiol. Lett. 25, 179–182.

Khamis, A., Raoult, D., La Scola, B., 2005. Comparison between rpoB and 16SrRNA gene sequencing for molecular identification of 168 clinical isolates ofCorynebacterium. J. Clin. Microbiol. 43, 1934–1936.

Koboldt, D.C., Zhang, Q., Larson, D.E., Shen, D., McLellan, M.D., Lin, L., Miller, C.A.,Mardis, E.R., Ding, L., Wilson, R.K., 2012. VarScan 2: Somatic mutation and copynumber alteration discovery in cancer by exome sequencing. Genome Res. 22,568–576.

Lander, E.S., Waterman, M.S., 1988. Genomic mapping by fingerprinting randomclones: A mathematical analysis. Genomics 2, 231–239.

Li, H., Durbin, R., 2009. Fast and accurate short read alignment withBurrows–Wheeler transform. Bioinformatics 25, 1754–1760.

Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abeca-sis, G., Durbin, R., 1000 Genome Project Data Processing Subgroup, 2009. TheSequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079.

Merhej, V., Georgiades, K., Raoult, D., 2013. Postgenomic analysis of bacterialpathogens repertoire reveals genome reduction rather than virulence factors.Brief. Funct. Genomics 12, 291–304.

Merhej, V., Royer-Carenzi, M., Pontarotti, P., Raoult, D., 2009. Massive compara-tive genomic analysis reveals convergent evolution of specialized bacteria. Biol.Direct 4, 13.

Moran, N.A., 2002. Microbial minimalism: Genome reduction in bacterial pathogens.Cell 108, 583–586.

Peterson, J., Garges, S., Giovanni, M., McInnes, P., Wang, L., Schloss, J.A., Bonazzi, V.,McEwen, J.E., Wetterstrand, K.A., Deal, C., Baker, C.C., Di Francesco, V., Howcroft,T.K., Karp, R.W., Lunsford, R.D., Wellington, C.R., Belachew, T., Wright, M., Giblin,C., David, H., Mills, M., Salomon, R., Mullins, C., Akolkar, B., Begg, L., Davis, C.,Grandison, L., Humble, M., Khalsa, J., Little, A.R., Peavy, H., Pontzer, C., Portnoy,

1 of Med

P

R

R

R

010 S.J. Salipante et al. / International Journal

M., Sayre, M.H., Starke-Reed, P., Zakhari, S., Read, J., Watson, B., Guyer, M., 2009.The NIH Human Microbiome Project. Genome Res. 19, 2317–2323.

ottumarthy, S., Limaye, A.P., Prentice, J.L., Houze, Y.B., Swanzy, S.R., Cookson,B.T., 2003. Nocardia veterana, a new emerging pathogen. J. Clin. Microbiol. 41,1705–1709.

ichter, M., Roselló-Móra, R., 2009. Shifting the genomic gold standardfor the prokaryotic species definition. Proc. Natl. Acad. Sci. U.S.A. 106,19126–19131.

iegel, P., de Briel, D., Prévost, G., Jehl, F., Monteil, H., 1994. Genomic diversityamong Corynebacterium jeikeium strains and comparison with biochemi-

cal characteristics and antimicrobial susceptibilities. J. Clin. Microbiol. 32,1860–1865.

oux, V., Drancourt, M., Stein, A., Riegel, P., Raoult, D., La Scola, B., 2004. Corynebac-terium species isolated from bone and joint infections identified by 16S rRNAgene sequence analysis. J. Clin. Microbiol. 42, 2231–2233.

ical Microbiology 304 (2014) 1001–1010

Simpson, J.T., Wong, K., Jackman, S.D., Schein, J.E., Jones, S.J.M., Birol, I., 2009. ABySS:A parallel assembler for short read sequence data. Genome Res. 19, 1117–1123.

Tauch, A., Kaiser, O., Hain, T., Goesmann, A., Weisshaar, B., Albersmeier, A., Bekel,T., Bischoff, N., Brune, I., Chakraborty, T., Kalinowski, J., Meyer, F., Rupp, O.,Schneiker, S., Viehoever, P., Pühler, A., 2005. Complete genome sequence andanalysis of the multiresistant nosocomial pathogen Corynebacterium jeikeiumK411, a lipid-requiring bacterium of the human skin flora. J. Bacteriol. 187,4671–4682.

Tleyjeh, I.M., Qutub, M.O., Bakleh, M., Sohail, M.R., Virk, A., 2005. Corynebacterium

jeikeium prosthetic joint infection: Case report and literature review. Scand. J.Infect. Dis. 37, 151–153.

Wixon, J., 2001. Featured organism: Reductive evolution in bacteria: Buchnera sp.,Rickettsia prowazekii and Mycobacterium leprae. Comp. Funct. Genomics 2,44–48.