Anaerobe xxx (2014) 1e6

Contents lists available at ScienceDirect

Anaerobe
journal homepage: www.elsevier.com/locate/anaerobe

Identification of antimicrobial resistance genes in multidrug-resistant

clinical Bacteroides fragilis isolates by whole genome shotgun
sequencing
Thomas Vognbjerg Sydenham a, *, Jo
zsef So

ki b, Henrik Hasman c, Mikala Wang d,
a
Ulrik Stenz Justesen , on behalf of ESGAI (ESCMID Study Group on Anaerobic Infections)
a
Department of Clinical Microbiology, Odense University Hospital, Odense, Denmark
b
Institute of Clinical Microbiology, Faculty of General Medicine, University of Szeged, Szeged, Hungary
c
The Microbial Genomics and Antimicrobial Resistance Group, National Food Institute, Technical University of Denmark, Lyngby, Denmark
d
Department of Clinical Microbiology, Aarhus University Hospital, Aarhus, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: Bacteroides fragilis constitutes the most frequent anaerobic bacterium causing bacteremia in humans. The
Available online xxx genetic background for antimicrobial resistance in B. fragilis is diverse with some genes requiring
insertion sequence (IS) elements inserted upstream for increased expression. To evaluate whole genome
Keywords: shotgun sequencing as a method for predicting antimicrobial resistance properties, one meropenem
Bacteroides fragilis resistant and five multidrug-resistant blood culture isolates were sequenced and antimicrobial resistance
Antimicrobial resistance
genes and IS elements identified using ResFinder 2.1 (http://cge.cbs.dtu.dk/services/ResFinder/) and a
Whole genome sequencing
custom BLAST database. Combinations of cfxA, cepA, cfiA, nimA, nimD, nimE, nimJ, tetQ, ermB, ermF, bexB,
linAn2 and mefEn2 genes were identified in the six isolates. blaOXA-347, an open reading frame predicted
to be a b-lactamase (Cheng et al., 2012), was identified in one strain. Full length IS elements were
identified directly upstream of four genes, but in most cases contigs terminated 100e150 bases upstream
of the gene in question. Even though partial IS elements were identified in these short sequences, certain
identification could not be ascertained. Full antiobiograms for B. fragilis from genetic data will most likely
require complete or nearly complete genomes. Current approaches to this are laborious and/or costly.
Emerging technologies such as nanopore based single DNA strand sensing could perhaps provide a so-
lution in the future.
© 2014 Published by Elsevier Ltd.

1. Introduction concerning b-lactam/b-lactamase inhibitor combinations have not

Bacteroides fragilis is the most frequently isolated anaerobe
bacterium in blood stream infections [1] and rising prevalences of
antimicrobial resistance and antibiotic resistance genes have been
observed in the past decades, presumably due to increased anti-
biotic pressure [2,3].
Several genes encoding antimicrobial resistance properties have
been identified [4]. cepA encodes a b-lactamase protein that hy-
drolyzes cephalosporins (except cefoxitin) and penicillins and cfxA
is associated with resistance towards cefoxitin. Mechanisms
* Corresponding author. Department of Clinical Microbiology, Odense University
Hospital, J.B. Winsløws Vej 21.2, 5000 Odense C, Denmark. Tel.: þ45 65414780;
fax: þ45 65414785.
E-mail address: Thomas.sydenham@dadlnet.dk (T.V. Sydenham).
been fully elucidated, but appear to be dependant on IS element
insertion directly upstream of cepA and concomitant membrane
modifications such as porin loss [5]. Carbapenem resistance is
caused by expression of a metallo-b-lactamase encoded by the cfiA
gene, and metronidazole resistance is associated with the presence
of the nitroimidazole resistance genes nimA-H and nimJ [4,6]. While
ermB, ermF, and ermG genes that encode a macro-
lideelincomycinestreptogramin (MLS) mechanism are the main
causes of clindamycin resistance, msrSA and mefA-mediated mac-
rolideelincosamid resistance, and lincomycin and erythromycin
resistance conferred by a linA type gene (linAn2) have also been
demonstrated in Bacteroides spp. [4]. tetQ confers tetracycline
resistance in the majority of Bacteroides isolates with tetX and tetX1
being associated with tetracycline resistance and a higher tigecy-
cline MIC [4]. The bexA gene encodes an efflux pump that possibly is

http://dx.doi.org/10.1016/j.anaerobe.2014.10.009
1075-9964/© 2014 Published by Elsevier Ltd.

Please cite this article in press as: Sydenham TV, et al., Identification of antimicrobial resistance genes in multidrug-resistant clinical Bacteroides
fragilis isolates by whole genome shotgun sequencing, Anaerobe (2014), http://dx.doi.org/10.1016/j.anaerobe.2014.10.009
2 T.V. Sydenham et al. / Anaerobe xxx (2014) 1e6
responsible for fluoroquinolone resistance in B. fragilis and bexB
encodes a putative efflux pump also though to mediate fluo-
roquinolone resistance [4,7].
The genetic background for antimicrobial resistance in B. fragilis
is diverse in that insertion of sequence (IS) elements upstream of
some resistance genes confer increased expression and antimi-
crobial resistance [8]. Notable examples of this are the cfiA, cfxA,
nim, and erm genes. IS elements carry outwards facing promoters
that increase expression when inserted upstream of resistance
genes [8]. However, antimicrobial susceptibility cannot always be
predicted by the absence of an IS element upstream of a gene. In
some resistant strains no IS elements upstream of cfiA or cfxA have
been identified suggesting the presence of other mechanisms
governing expression than IS element induced promoter activity
[9e11].
For other bacterial species, antimicrobial susceptibility predic-
tion by whole genome sequencing and resistance gene identifica-
tion has been explored with sensitivities and specificities
comparable to phenotypical susceptibility testing [12,13]. In the
case of B. fragilis, where the presence of IS elements upstream of
some key resistance genes is required for a high level antimicrobial
resistance, and several different genes are associated with pheno-
typical resistance, whole genome sequencing and identification of
resistance genes with easy-to-use websites or programs could be
an attractive alternative to traditional identification with multiple
laborious PCR and Sanger sequencing experiments required. We
sought to ascertain whether antimicrobial resistance genes rele-
vant to B. fragilis could be identified using whole genome shotgun
sequencing (WGS) and easy to use web-based tools. The ability to
identify IS elements upstream of resistance genes from WGS data
was also explored. Preliminary results for three of the six isolates
included in this study were presented at the 24th European
Congress of Clinical Microbiology and Infectious Diseases in Bar-
celona 2014 (Poster P1193).
2. Material and methods
2.1. Isolates and their antibiotic resistance determination
One isolate resistant to meropenem and five multidrug-
resistant B. fragilis blood culture isolates were included in the
study; O17 (DCMOUH0017B), O18 (DCMOUH0018B), S01
(DCMSKEJBY0001B), O42 (DCMOUH0042B), O67
(DCMOUH0067B), and O85 (DCMOUH0085B). A case report has
previously been published for isolate O18 [14] and a case report
concerning isolate S01 is being prepared (Ank et al., in preparation).
Species identification was confirmed using Matrix Assisted
Laser Desorption and Ionization Time of Flight Mass Spectrometry
(MALDI-TOF-MS) on the Biotyper platform (Bruker Daltonics, Bre-
men, Germany). Antimicrobial susceptibility testing was performed
using the Etest gradient method (BioMe
rieux, Craponne, France) on
Brucella blood agar supplemented with 5 mg/L hemin and 1 mg/L
vitamin K1 (Becton, Dickinson GmbH, BD Diagnostics, Heidelberg,
Germany) according to the manufacturer's instructions and inter-
preted according to EUCAST clinical breakpoints version 4.0 (www.
eucast.org/clinical_breakpoints/).
2.2. Whole genome shotgun sequencing and de novo assembly
Genomic DNA was purified using the MasterPure DNA Purifi-
cation kit (Epicentre Biotechnologies, Madison, WI, USA) following
the manufactures instructions and libraries were prepared using
the Nextera XT kit (Illumina, Essex, United Kingdom). 250 bp (base-
pair) paired-end sequencing was performed using the Illumina
MiSeq platform. Read quality metrics was evaluated using the
Please cite this article in press as: Sydenham TV, et al., Identification of antimicrobial resistance genes in multidrug-resistant clinical Bacteroides
fragilis isolates by whole genome shotgun sequencing, Anaerobe (2014), http://dx.doi.org/10.1016/j.anaerobe.2014.10.009
FASTX-toolkit and de novo assembly was performed using the
SPAdes 3.0.0 assembler implemented on the Orione Galaxy web
server (http://orione.crs4.it/) [15,16]. K values were set to 21, 33, 55,
77, 99, and 127 as suggested in the SPAdes 3.0 manual. Contigs
under 200 bp in length were removed using tools on the Orione
Galaxy server. During the submission process to the NCBI Whole
Genome Shotgun submission portal, contigs tagged as contamina-
tions by the contamination screen performed by NCBI staff were
removed. Contigs were trimmed of sequences tagged as adapter
sequences, and genomes were annotated using the NCBIs Pro-
karyote Genome Annotation Pipeline (PGAAP) version 2.0 [17].
Assembly metrics were calculated using the Quast web server
(http://quast.bioinf.spbau.ru/) [18] and the crude read depth was
calculated by dividing the total number of sequenced nucleotides
with the genome length reported by Quast.
2.3. Identification of genes and IS elements
The assemblies were submitted to ResFinder 2.1 (http://cge.cbs.
dtu.dk/services/ResFinder/) which identifies acquired antimicrobial
resistance genes using local nucleotide BLAST [19]. Percent identity
(%ID) threshold and minimum length settings were 90% and 60%
respectively. Genes not included in the ResFinder database (bexA,
bexB, mefE, and linAn2) and known IS elements relevant to Bacter-
oides identified in the literature [8] and by searching the NCBI
nucleotide database were downloaded from NCBI GenBank and
incorporated in a custom BLAST database. For included IS elements
please see Supplementary Table 1. This database was searched us-
ing the draft assemblies as queries using nucleotide BLAST in CLC
Main Workbench version 7.0.3 (CLC bio, Aarhus, Denmark). Iden-
tified IS elements upstream of resistance genes were noted. In
several cases contigs ended approximately 100e150 bases up-
stream of resistance genes. Several fragments of various IS ele-
ments were identified with high scores in these short upstream
regions. The IS element fragment with the highest %ID similarity
score and the longest hit length was noted as a possible IS element.
3. Results
3.1. Whole genome shotgun sequencing
Six isolates were sequenced using the MiSeq platform. Contigs
per assembly ranged from 186 to 367. Draft genome lengths varied
from 5,117,284 to 5,509,612 bp with crude read depths of 39e171
and contig N50 values ranging from 101,349 to 184,070 bp. DDBJ/
EMBL/GenBank accession numbers, assembly versions used in this
paper, and SRA accession numbers can be found in Supplementary
Table 2. Please see Supplementary Table 3 for WGS quality and
assembly metrics specified for each isolate.
3.2. Antimicrobial susceptibility and identification of resistance
genes
Main results from antimicrobial susceptibility testing and
identification of known resistance genes and IS elements are pre-
sented in Table 1. Five strains were resistant towards meropenem
with four of these isolates also displayed resistance to imipinem.
Four strains were resistant towards metronidazole, two were
resistant towards piperacillin/tazobactam, and four showed resis-
tance towards clindamycin. Combinations of cfxA, cepA, cfiA, nimA,
nimD, nimE, nimJ, tetQ, ermB, ermF, blaOXA-347, bexB, mefEn2, and
linAn2 were identified in the six strains at >98 %ID.
 T.V. Sydenham et al. / Anaerobe xxx (2014) 1e6 3

Table 1
Antimicrobial susceptibility testing and resistance gene and IS element identification for the six Bacteroides fragilis strains.

Antibiotic susceptibility testinga Resistance gene and IS element identificationb

Strain Antibiotic Etest MIC Result Gene Upstream IS %ID HSP/query Hit Contig Contig Hit start Hit end Hit length
(mg/L) element length (%) strand length (bp) (nt) (nt) (bp)

O17 MEM >32 R cfiA 99.6 99.2 M Contig177 12,656 11,774 12,517 744
IPM >32 R *IS614B 100.0 7.9 P Contig177 12,656 12,530 12,656 127
MTZ >32 R nimJ 99.4 100.0 M Contig16 1653 1017 1514 498
*IS614B 94.4 7.8 P Contig16 1653 1529 1653 126
tetQ 99.3 99.3 P Contig184 63,401 1889 3849 1961
bexB 91.2 100.0 M Contig194 180,804 964 2328 1365
CLI 0.094 S
PTZ >256 R

O18 MEM >32 R cfiA 100.0 100.0 M Contig167 101,303 100,384 101,133 750
IPM 16 R *ISBf12 100.0 7.9 P Contig167 101,303 101,178 101,303 126
MTZ 16 R nimD 99.2 100.0 P Contig251 7349 6749 7243 495
IS1169 99.4 99.8 P Contig251 7349 5429 6740 1314
CLI 6 R ermF 99.8 72.0 M Contig168 106,254 88,121 88,697 577
IS4351 99.8 100.0 M Contig168 106,254 88,698 89,879 1181
LinAn2 100.0 100.0 P Contig39 77,145 15,558 16,070 513
tetQ 99.8 100.0 M Contig246 125,751 8958 10,883 1926
bexB 91.1 100.0 M Contig162 34,537 967 2331 1365
mefEn2 99.8 100.0 P Contig39 77,145 16,095 17,300 1206
PTZ >256 R

S01 MEM >32 R cfiA 99.2 100.0 M Contig109 31,468 30,547 31.296 750
IPM 16 R *IS1187 94.8 13.9 M Contig109 31,468 31,297 31,468 163
MTZ 64 R nimE 100.0 100.0 M Contig125 8458 7612 8046 435
*ISBf6 100.0 40.6 M Contig125 8458 8073 8458 386
CLI >32 R ermF 99.5 100.0 M Contig104 22,876 21,944 22,744 801
*IS1187 100.0 9.9 M Contig104 22,876 22,751 22,866 116
tetQ 90.5 100.0 M Contig104 22,876 19,732 21,657 1926
tetQ 99.8 100.0 P Contig115 10,163 75,973 77,898 1926
bexB 91.1 100.0 P Contig111 74,184 63,700 65,064 1365
PTZ 6 S

O42 MEM 0.094 S

IPM 0.25 S
blaOXA-347 100.0 100.0 P Contig74 2087 1131 1955 825
cepA 100.0 100.0 P Contig208 53,110 36,310 37,212 903
MTZ 8 R nimA 98.3 100.4 P Contig1 8433 3373 3905 533
ISBf13 100.0 100.0 M Contig1 8433 2121 3339 1219
CLI >256 R ermF 99.5 100.0 P Contig74 2087 128 928 801
*IS613B 100.0 8.0 P Contig74 2087 1 127 127
LinAn2 100.0 100.0 P Contig104 107,515 58,444 58,956 513
tetQ 100.0 100.0 P Contig212 222,964 44,695 46,668 1974
bexB 99.1 100.0 M Contig213 422,215 219,072 220,436 1365
mefEn2 99.8 100.0 P Contig104 107,515 58,981 60,186 1206
PTZ 0.38 S

O67 MEM 8 R cfiA 100.0 100.0 P Contig301 135,251 66,868 67,617 750
IPM 0.5 S *IS1187 95.5 9.5 P Contig301 135,251 66,728 66,839 112
MTZ 0.19 S
CLI 0.38 S LinAn2 100.0 100.0 M Contig292 39,473 1682 2194 513
tetQ 100.0 100.0 P Contig112 175,335 116,174 118,099 1926
bexB 90.9 100.0 P Contig23 83,642 30,107 31,471 1365
mefEn2 99.8 100.0 M Contig292 39,473 452 1657 1206
PTZ 2 S

O85 cfxA 99.9 100.0 M Contig55 193,255 186,708 187,673 966

IS614B 99.0 93.2 P Contig55 193,255 187,686 189,181 1497
MEM 32 R cfiA 100.0 100.0 P Contig55 193,255 76,886 77,635 750
IPM 1 S *IS1188 93.6 9.1 M Contig55 193,255 76,673 76,858 186
MTZ 0.25 S
CLI >256 R ermB 99.7 98.8 P Contig54 249,202 206,597 207,325 729
tetQ 90.5 100.0 M Contig171 42,614 13,157 15,082 1926
tetQ 99.8 100.0 M Contig60 172,757 14,304 16,229 1926
bexB 90.9 100.0 M Contig51 455,838 370,536 371,900 1365
PTZ 2 S
a
Breakpoints (S/R) defined by EUCAST: Meropenem (MEM) 2/>8, imipinem (IPM) 2/>8, metronidazole (MTZ) 4/>4, clindamycin (CLI) 4/>4, and piperacillin/tazo-
bactam (PTZ) 8/>16. Other abbreviations: MIC ¼ minimal inhibitory concentration; R ¼ resistant; S ¼ susceptible.
b
Identified genes are displayed next to relevant the antibiotic. IS elements identified upstream of a gene follow the respective gene in the table. Partial IS elements are
marked with an asterisk (*). Hit strand indicates the strand orientation on which the identified gene/IS element is located by M (Minus) or P (Plus). Contig indicates the
assembled contig on which the gene/IS element is located. Hit start and hit end indicate the first and last nucleotide of the BLAST hit corresponding to the Plus strand of the
contig. Other abbreviations: IS ¼ insertion sequence; %ID ¼ percent identity; HSP ¼ high scoring segment pairs; bp ¼ base pairs; nt ¼ nucleotide position.

Please cite this article in press as: Sydenham TV, et al., Identification of antimicrobial resistance genes in multidrug-resistant clinical Bacteroides
fragilis isolates by whole genome shotgun sequencing, Anaerobe (2014), http://dx.doi.org/10.1016/j.anaerobe.2014.10.009
4 T.V. Sydenham et al. / Anaerobe xxx (2014) 1e6
3.3. Identification of IS elements
Full length IS elements were identified in the upstream regions
of four genes; IS1169 and IS4251 upstream of nimD and the trun-
cated ermF-gene in isolate O18, ISBf13 upstream of nimA in isolate
O42, and IS614B upstream of cfxA in isolate O85. ISBf6 was identi-
fied with 100.0 %ID and an HSP/Query length of 40.6%. The partial
length hit was due to termination of the contig 386 bp into the IS
element. Upstream of cepA in isolate O42, cfiA in O67 and O85, and
ermB in isolate O85, the contigs continued thousands of bases up-
stream of the resistance genes and no full length IS elements were
identified. Concerning the genes where partial or whole IS ele-
ments were identified upstream of the gene, 62% (8/13) of contigs
terminated just approximately 100e200 bp upstream of the iden-
tified antimicrobial resistance gene. Partial BLAST hits for IS ele-
ments were recorded upstream of several resistance genes with
upstream regions terminated by contig ends. A partial hit for IS1187
with a hit length of 112 bases was noted upstream of cfiA in isolate
O67 and a partial hit of 186 bases of IS1188 was found directly
upstream of O85 where the contigs do not terminate just upstream
of the cfiA gene. Several of the IS elements and partial IS elements
were identified on the non-sense/opposite strand of the gene, but
still “upstream” of the gene.
3.4. Correlation between resistance profile and identified genes and
IS elements
cfiA genes were identified in all five meropenem resistant iso-
lates. In two meropenem resistant isolates, O67 and O85, cfiA was
identified and the uninterrupted upstream region did not contain a
full length IS element. In the other three meropenem resistant
isolates, possible IS elements were identified in the short sequences
upstream of the cfiA genes. nim genes were identified in the four
metronidazole resistant isolates and not in the two susceptible
isolates. In strains O18 and O42 full length IS elements were
identified upstream of the nim gene, in strain S01 a partial ISBf6 that
was cut short after 386 bp due to contig termination was identified
upstream of nimE. In strain O17 the contig with nimJ terminated
140 bp upstream of the gene, with a partial IS614B identified in the
short upstream sequence.
Of the six isolates, four were resistant to clindamycin, with erm
genes identified in all four isolates. A full length IS4351 was found
upstream of the truncated ermF gene in isolate O18. In isolates S01
and O42 partial IS elements were identified in the short upstream
sequences, and in isolate O85 no IS element was identified in the
uninterrupted upstream region of ermB. linAn2 was identified in the
clindamycin resistant isolates O18 and O42, but also in the sus-
ceptible isolate O67. No IS elements was identified upstream of the
linAn2 in the resistant isolates. The two piperacillin/tazobactam
resistant isolates were also highly carbapenem-resistant and all
harbored cfiA-genes and not cepA.
3.5. Novel resistance gene alleles
In strain O18 ermF was identified with 99.8 %ID and 72% hit
length (557/801 nucleotides) on the complement strand of con-
tig168. On closer examination another partial 238 bp ermF was
discovered 1309 bp downstream on the complement strand.
Joining the two partial ermF hits and submitting the 815 bp result to
BLAST-n resulted in a 98.2 %ID hit to ermF in several accessions also
including ermF in GenBank: M17124. BLAST alignment showed a
gap of 15 bases in an otherwise perfectly aligned sequence (not
shown). This led us to the conclusion that an insertion event could
have occurred disrupting the ermF gene into two parts. Indeed
PGAAP annotation had resulted in a predicted transposase gene
Please cite this article in press as: Sydenham TV, et al., Identification of antimicrobial resistance genes in multidrug-resistant clinical Bacteroides
fragilis isolates by whole genome shotgun sequencing, Anaerobe (2014), http://dx.doi.org/10.1016/j.anaerobe.2014.10.009
(GenBank: KFX74121.1) spanning almost the entire region dividing
the two partial ermF sequences. O18 is resistant to clindamycin at a
lower level (MIC ¼ 6 mg/L) than isolates S01, O42, and O85 that
contain full length erm-genes, so it is possible that the partial ermF
gene is expressed and the resulting protein product confers clin-
damycin resistance albeit at a lower level than a full length protein.
Along with a certain ID of tetQ in strains S01 and O85, hits with
90.5 %ID and 100% high scoring segment pair (HSP)/query length to
tetQ were found. The submission of the tetQ-like gene nucleotide
sequences to NCBI BLAST-n resulted in two hits with 99 %ID; a gene
titled elongation factor Tu GTP binding domain protein in Tannerella
forsythia ATCC 43037, complete genome (GenBank: CP003191.1
location 1,331,023e 1,332,960) and a gene titled small GTP-binding
protein domain in Bacteroides xylanisolvens XB1A draft genome
(GenBank: FP929033.1 location 4,261,675e4,4263,600). According
to the accessions, both genomes were annotated using prediction
methods. Searching the NCBI BLAST nr protein sequence database
using blastx resulted in 100% ID hits to GenBank: WP_008645586.1,
a non-redundant multispecies tetQ protein sequence, as the highest
hit, making it apparent that strains S01 and O85 harbor two copies
of tetQ.
bexB was identified in strain O42 with 99.1 %ID and hits were
noted with 90.9e91.2 %ID to bexB in the other five genome se-
quences. Aligning the five bexB-like sequences using the alignment
feature in the CLC Main Workbench, resulted in a perfect gapless
alignment with almost 100% conservation at all nucleotide posi-
tions (not shown). The five sequences shared 97.1e99.9 %ID, and the
bexB-like genes of O18, S01, O67 and O85 showed very high simi-
larity with 99.5e99.9 %ID. The two highest %ID scores when sub-
mitting the sequence of the possible bexB-like gene of strain O17 to
NCBI nucleotide BLAST were 91.3 %ID with B. fragilis 638R genome
(GenBank: FQ31204.1 location 2,664,849e2,663,485) and 91.2 %ID
to the complete bexB gene in B. fragilis strain 86-5443-2-2 (Gen-
Bank: AY375536.1 location 5963e4599), the latter being the bexB
sequence used in the custom BLAST database used in this study.
Searching the NCBI BLAST nr protein sequence database using
blastx resulted in 100 %ID hits to GenBank: WP_005794219.1
multidrug transporter MatE [Bacteroides] for the bexB-like genes in
all five strains. This protein shares 100% homology to GenBank:
AAQ92937.1, the protein product of the bexB sequence used in the
custom BLAST database in this study.
A gene annotated as blaOXA-347 was identified by ResFinder in
strain O42. The blaOXA-347 sequence referenced by ResFinder is
GenBank: JN086160.1 (location 1583e2407). The predicted b-lac-
tamase gene in JN086160.1 was identified using amoxicillin
screening of a fosmid metagenomic library generated from gut
microbiota cloned into Escherichia coli, and the closest match to the
predicted protein was with 53% identity to a class D b-lactamase
from Riemerella anatipestifer [20]. This is to our knowledge the first
published identification of a possible blaOXA gene in Bacteroides
spp.
4. Discussion
The significant antimicrobial resistance genes relevant to
B. fragilis could be identified using WGS and the ResFinder website.
bexA, bexB, linA and mefE are not included in the ResFinder 2.1
database, but could be added in future versions. In the six isolates
included in this study, four to eight resistance genes were identified
in each isolate. cfiA and nim genes were identified in all isolates
with meropenem and metronidazole resistance respectively. erm
genes were identified in the four clindamycin resistant isolates.
linAn2 was identified in one clindamycin resistant isolate, but was
also found in a clindamycin susceptible isolate. Research data on
the degree of linA contribution to clindamycin resistance in
 T.V. Sydenham et al. / Anaerobe xxx (2014) 1e6
Bacteroides is scarce. Our results do indicate that the presence of
linA alone does not confer clindamycin resistance.
In several cases IS elements or partial IS elements were identi-
fied on the opposite strand of the resistance gene. Previous studies
have identified IS elements in the upstream regions of the resis-
tance gens, and shown that the presence of the IS elements confer
increased expression of the genes [8,21]. The “direction” of the IS
element is defined by the coding strand orientation of the reading
frame encoding the transposase, but some IS elements are known
to hold an outward oriented promoter oriented in the opposite
direction of the transposase gene [8]. The outward facing promoter
is located in the inverted regions, making it possible that the pro-
moter motifs are also present on the opposite/complement strand
[22]. Indeed, in strain O42, the promoter motif TAnnTTTG can be
found upstream of nimA, where ISBf13, is on the complement
strand of the nimA upstream region. The same phenomenon has
also been observed by Husain and colleagues while describing a
conjugative transposon harboring nimJ [23]. The consequence of an
IS element presence on the opposite strand of the resistance genes
needs to be explored experimentally.
In most cases, the identification of IS elements was hampered by
the termination of contigs relatively close to identified antimicro-
bial resistance genes. Partial hits to known IS elements were
observed in these short upstream regions, however partial IS
element hits were also recorded upstream of cfiA in two cases
where the contigs were not terminated just upstream of the genes.
Unfortunately, partial IS element hits upstream of a gene must
therefore be interpreted with caution and cannot necessarily be
attributed to actual presence of an IS element. It is known that
repetitive sequences impede de novo genome assembly using short
reads and result in fractured assemblies [24]. In testing assemblies
from simulated reads derived from a finished reference genome
25%e77% of non-reconstructable genes were tagged as trans-
posase- or IS-related, depending on the read length simulated [25].
Taking this into account, a terminated contig at a position where a
potential IS element has been identified by BLAST, could be taken as
further indication that an IS element is indeed present at that
position.
The identification of new alleles in B. fragilis with approximately
90 %ID to tetQ and bexB, the possible blaOXA gene, and the ermF
gene disrupted by a transposase are notable findings resulting from
simple analysis of the WGS assemblies. It is possible that these
genes would not have been identified using specific primer based
PCR screening, as the primer bindings sites could have changed in
the new tetQ and bexB alleles, and a PCR product of the disrupted
ermF gene would differ from the known ermF product being notably
longer and might go unnoticed. Interestingly, the isolate with the
disrupted version of ermF was still clindamycin resistant, but at a
much lower level than isolates carrying full length erm genes. This
raises the question whether the protein product of the disrupted
ermF gene has lower, but still functional, ribosomal RNA methyl-
ation activity than the full length version.
Several authors predict that whole genome sequencing and
prediction of antimicrobial resistance from the genetic sequences
will be implemented as a routine test in the near future in clinical
microbiology [26e29]. However, a better understanding of the
genetic determinants for resistance is required before correct pre-
diction of antimicrobial resistance properties in B. fragilis from WGS
data will be feasible. ResFinder and the custom BLAST database
constructed for this study only identifies known acquired resis-
tance genes and IS elements. Point mutations, inversions, deletions
and genomic rearrangements that could alter antimicrobial sus-
ceptibility phenotypes are not identified using this approach. Flu-
oroquinolone and rifampicin resistance has been shown to be
influenced by point mutations in B. fragilis [30,31], but we did not
Please cite this article in press as: Sydenham TV, et al., Identification of antimicrobial resistance genes in multidrug-resistant clinical Bacteroides
fragilis isolates by whole genome shotgun sequencing, Anaerobe (2014), http://dx.doi.org/10.1016/j.anaerobe.2014.10.009
5
include phenotypic susceptibility testing for these antibiotics as
they are not currently used for treatment of B. fragilis infections.
Studies on E. coli show complicated relationships between resis-
tance determinants with some resistance mutations conferring
increased sensitivity to other antibiotics and cross-resistance de-
terminants influencing the resistance to multiple antibiotics [32].
An even more detailed “genetic antiobiogram” for each isolate
could be generated from a method designed to also incorporate
these possible mechanisms and inter-dependencies once they are
discovered for B. fragilis.
cfiA mediated carbapenem-resistance is one of the most studied
areas of antimicrobial resistance in B. fragilis, but no explanation of
the resistant phenotypes without the presence of IS elements up-
stream of the gene has yet been found. IS elements upstream of a
resistance gene are almost always associated with antimicrobial
resistance, but as we have found in this study, there is an inherent
confounding issue concerning the termination of contigs due to
probable IS sequences. Therefore, to accurately predict resistance
for B. fragilis, it is likely that complete, or almost complete, genome
sequences are required. The results presented in this study are
based on 2*250 bp paired-end sequencing on the Illumina MiSeq
platform, offering fragmented draft assemblies with hundreds of
contigs per genome at a fairly low cost (approximately 220 USD)
and generated in a reasonable time frame. Short read paired-end
sequencing on the Illumina platform was chosen as this approach
was used in previous proof-of-concept studies comparing WGS
results to phenotypic resistance testing in various aerobic bacteria
[12,13]. If WGS is to be used for rapid clinical diagnostics and
antiobiogram generation, genome finishing requiring multitudes of
Sanger sequencing experiments is too laborious and costly to be
applicable. Sequencing mate-pair libraries on the Illumina platform
and utilizing these with the paired-end sequences in the de novo
assembly would result in a “higher quality” assemblies with sig-
nificant reduction in the contig numbers [33]. This would possibly
result in upstream regions encompassing the IS elements in the
final assemblies. However, high quality mate-pair libraries are up to
seven times more expensive than paired-end libraries and can be
difficult to produce. The preparation and sequencing of an addi-
tional library would significantly increase the costs as well as the
processing time per isolate, which makes this method unlikely to
be the first choice in a clinical setting. Other technologies such as
single molecule, real-time DNA sequencing, as offered by Pacific
Biosciences [34], or the emerging nanopore based technologies [35]
offer the prospect of finished single chromosome microbial ge-
nomes, but using the Pacific Bioscience platforms is more expensive
and in some instances error prone than the Illumina platform and
nanopore based sequencing is still under development.
5. Conclusions
Genes encoding proteins with antimicrobial resistance proper-
ties and IS elements could be identified using WGS data from six
clinical blood isolates, the easy to use web-based tool ResFinder and
custom BLAST database. Complete IS elements were identified
upstream of four genes, but in most cases, only short upstream
sequences were available due to contig termination. However,
partial IS elements could be identified at these sites, and the
termination of the contigs at these sites is a possible indication of IS
element presence.
As the fragmented assemblies resulting from paired end data
inhibits certain identification of IS elements upstream of identified
resistance genes, complete, finished genome sequences are most
likely required for full “genetic antiobiograms” for B. fragilis and the
required technologies to easily achieve this are still under
development.
6 T.V. Sydenham et al. / Anaerobe xxx (2014) 1e6
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.anaerobe.2014.10.009.
References
[1] Wexler HM. Bacteroides: the good, the bad, and the nitty-gritty. Clin Microbiol
Rev 2007;20:593e621. http://dx.doi.org/10.1128/CMR.00008-07.
[2] Boyanova L, Kolarov R, Mitov I. Recent evolution of antibiotic resistance in the
anaerobes as compared to previous decades. Anaerobe 2014. http://
dx.doi.org/10.1016/j.anaerobe.2014.05.004. Epub ahead of print.
[3] Nagy E, Urba
n E, Nord CE. Antimicrobial susceptibility of Bacteroides fragilis
group isolates in Europe: 20 years of experience. Clin Microbiol Infec 2011;17:
371e9. http://dx.doi.org/10.1111/j.1469-0691.2010.03256.x.
[4] Eitel Z, So
ki J, Urba

n E, Nagy E. The prevalence of antibiotic resistance genes in
Bacteroides fragilis group strains isolated in different European countries.
Anaerobe 2013;21:43e9. http://dx.doi.org/10.1016/j.anaerobe.2013.03.001.
[5] Dubreuil L, Odou MF. Anaerobic bacteria and antibiotics: what kind of unex-
pected resistance could I find in my laboratory tomorrow? Anaerobe 2010;16:
555e9. http://dx.doi.org/10.1016/j.anaerobe.2010.10.002.
[6] Husain F, Veeranagouda Y, Hsi J, Meggersee R, Abratt V, Wexler HM. Two
multidrug resistant clinical isolates of Bacteroides fragilis carry a novel
metronidazole resistance nim gene (nimJ). Antimicrob Agents Chemother
2013. http://dx.doi.org/10.1128/AAC.00386-13.
[7] Franco AA. The Bacteroides fragilis pathogenicity island is contained in a pu-
tative novel conjugative transposon. J Bacteriol 2004;186:6077e92. http://
dx.doi.org/10.1128/JB.186.18.6077-6092.2004.

ki J. Extended role for insertion sequence elements in the antibiotic resis-
[8] So
tance of Bacteroides. World 2013;3:1e12.

ki J, Gonzalez SM, Urb

[9] So an E, Nagy E, Ayala JA. Molecular analysis of the
effector mechanisms of cefoxitin resistance among Bacteroides strains.
J Antimicrob Chemother 2011;66:2492e500. http://dx.doi.org/10.1093/jac/
dkr339.
[10] So
ki J, Edwards R, Urba
n E, Fodor E, Beer Z, Nagy E. Screening of isolates from
faeces for carbapenem-resistant Bacteroides strains; existence of strains with
novel types of resistance mechanisms. Int J Antimicrob Agents 2004;24:
450e4. http://dx.doi.org/10.1016/j.ijantimicag.2004.06.017.
[11] Johansson A, Nagy E, So
ki J, ESGAI (ESCMID Study Group on Anaerobic In-
fections). Detection of carbapenemase activities of Bacteroides fragilis strains
with matrix-assisted laser desorption ionizationetime of flight mass spec-
trometry (MALDI-TOF MS). Anaerobe 2014;26:49e52. http://dx.doi.org/
10.1016/j.anaerobe.2014.01.006.
[12] Zankari E, Hasman H, Kaas RS, Seyfarth AM, Agersø Y, Lund O, et al. Geno-
typing using whole-genome sequencing is a realistic alternative to surveil-
lance based on phenotypic antimicrobial susceptibility testing. J Antimicrob
Chemother 2013;68:771e7. http://dx.doi.org/10.1093/jac/dks496.
[13] Stoesser N, Batty EM, Eyre DW, Morgan M, Wyllie DH, Del Ojo Elias C, et al.
Predicting antimicrobial susceptibilities for Escherichia coli and Klebsiella
pneumoniae isolates using whole genomic sequence data. J Antimicrob Che-
mother 2013. http://dx.doi.org/10.1093/jac/dkt180.
[14] Hartmeyer GN, So
ki J, Nagy E, Justesen US. Multidrug-resistant Bacteroides
fragilis group on the rise in Europe? J Med Microbiol 2012;61:1784e8. http://
dx.doi.org/10.1099/jmm.0.049825-0.
[15] Cuccuru G, Orsini M, Pinna A, Sbardellati A, Soranzo N, Travaglione A, et al.
Bioinformatics 2014. http://dx.doi.org/10.1093/bioinformatics/btu135.
[16] Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, et al.
SPAdes: a new genome assembly algorithm and its applications to single-cell
sequencing. J Comput Biol 2012;19:455e77. http://dx.doi.org/10.1089/
cmb.2012.0021.
Please cite this article in press as: Sydenham TV, et al., Identification of antimicrobial resistance genes in multidrug-resistant clinical Bacteroides
fragilis isolates by whole genome shotgun sequencing, Anaerobe (2014), http://dx.doi.org/10.1016/j.anaerobe.2014.10.009
[17] Angiuoli SV, Gussman A, Klimke W, Cochrane G, Field D, Garrity GM, et al.
Toward an online repository of standard operating procedures (SOPs) for
(meta) genomic annotation. OMICS J Integr Biol 2008;12:137e41. http://
dx.doi.org/10.1089/omi.2008.0017.
[18] Gurevich A, Saveliev V, Vyahhi N, Tesler G. QUAST: quality assessment tool for
genome assemblies. Bioinformatics 2013;29:1072e5. http://dx.doi.org/
10.1093/bioinformatics/btt086.
[19] Zankari E, Hasman H, Cosentino S, Vestergaard M, Rasmussen S, Lund O, et al.
Identification of acquired antimicrobial resistance genes. J Antimicrob Che-
mother 2012;67:2640e4. http://dx.doi.org/10.1093/jac/dks261.
[20] Cheng G, Hu Y, Yin Y, Yang X, Xiang C, Wang B, et al. Functional screening of
antibiotic resistance genes from human gut microbiota reveals a novel gene
fusion. FEMS Microbiol Lett 2012;336:11e6. http://dx.doi.org/10.1111/j.1574-
6968.2012.02647.x.
[21] Bayley DP, Rocha ER, Smith CJ. Analysis of cepA and other Bacteroides fragilis genes
reveals a unique promoter structure. FEMS Microbiol Lett 2000;193:149e54.
http://dx.doi.org/10.1111/j.1574-6968.2000.tb09417.x.
[22] Cerden ~ o-Ta

rraga AM, Patrick S, Crossman LC, Blakely G, Abratt V, Lennard N,
et al. Extensive DNA inversions in the B. fragilis genome control variable gene
expression. Science 2005;307:1463e5. http://dx.doi.org/10.1126/
science.1107008.
[23] Husain F, Veeranagouda Y, Boente R, Tang K, Mulato G, Wexler HM. The Ellis
island effect. Mob Genet Elem 2014;4. http://dx.doi.org/10.4161/mge.29801.
[24] Ricker N, Qian H, Fulthorpe RR. The limitations of draft assemblies for un-
derstanding prokaryotic adaptation and evolution. Genomics 2012;100:
167e75. http://dx.doi.org/10.1016/j.ygeno.2012.06.009.
[25] Kingsford C, Schatz MC, Pop M. Assembly complexity of prokaryotic genomes
using short reads. BMC Bioinform 2010;11:21. http://dx.doi.org/10.1186/
1471-2105-11-21.
[26] Ko€ser CU, Ellington MJ, Cartwright EJP, Gillespie SH, Brown NM, Farrington M,
et al. Routine use of microbial whole genome sequencing in diagnostic and
public health microbiology. PLoS Pathog 2012;8:e1002824. http://dx.doi.org/
10.1371/journal.ppat.1002824.
[27] Didelot X, Bowden R, Wilson DJ, Peto TEA, Crook DW. Transforming clinical
microbiology with bacterial genome sequencing. Nat Rev Genet 2012;13:
601e12. http://dx.doi.org/10.1038/nrg3226.
[28] Bertelli C, Greub G. Rapid bacterial genome sequencing: methods and appli-
cations in clinical microbiology. Clin Microbiol Infect 2013;19:803e13. http://
dx.doi.org/10.1111/1469-0691.12217.
[29] To€ro
€k ME, Peacock SJ. Rapid whole-genome sequencing of bacterial pathogens
in the clinical microbiology laboratorydpipe dream or reality? J Antimicrob
Chemother 2012;67:2307e8. http://dx.doi.org/10.1093/jac/dks247.
[30] Onodera Y, Sato K. Molecular cloning of the gyrA and gyrB genes of Bacteroides
fragilis encoding DNA gyrase. Antimicrob Agents Chemother 1999;43:2423e9.
[31] Ko KS, Kuwahara T, Haehwa L, Yoon Y-J, Kim B-J, Lee K-H, et al. RNA poly-
merase b-subunit gene (rpoB) sequence analysis for the identification of
Bacteroides spp. Clin Microbiol Infect 2007;13:48e54. http://dx.doi.org/
10.1111/j.1469-0691.2006.01553.x.
[32] Lazar V, Nagy I, Spohn R, Csorgo B, Gyorkei A, Nyerges A, et al. Genome-wide
analysis captures the determinants of the antibiotic cross-resistance interac-
tion network. Nat Commun 2014;5. http://dx.doi.org/10.1038/ncomms5352.
[33] Wetzel J, Kingsford C, Pop M. Assessing the benefits of using mate-pairs to
resolve repeats in de novo short-read prokaryotic assemblies. BMC Bioinform
2011;12:95. http://dx.doi.org/10.1186/1471-2105-12-95.
[34] Chin C-S, Alexander DH, Marks P, Klammer AA, Drake J, Heiner C, et al.
Nonhybrid, finished microbial genome assemblies from long-read SMRT
sequencing data. Nat Methods 2013;10:563e9. http://dx.doi.org/10.1038/
nmeth.2474.
[35] McGinn S, Gut IG. DNA sequencing e spanning the generations. New Bio-
technol 2013;30:366e72. http://dx.doi.org/10.1016/j.nbt.2012.11.012.