Rapid tests for the detection of the Mycobacterium abscessus subsp. bolletii strain responsible for an epidemic of surgical-site infections in Brazil

A single strain of Mycobacterium abscessus subsp. bolletii , characterised by a particular rpoB sequevar and two highly related pulsed field gel electrophoresis patterns has been responsible for a nationwide outbreak of surgical infections in Brazil since 2004. In this study, we developed molecular tests based on polymerase chain reaction restriction-enzyme analysis (PRA) and sequencing for the rapid identification of this strain. Sequences of 15 DNA regions conserved in mycobacteria were retrieved from GenBank or sequenced and analysed in silico . Single nucleotide polymorphisms specific to the epidemic strain and located in enzyme recognition sites were detected in rpoB , the 3’ region of the 16S rDNA and gyrB . The three tests that were developed, i.e., PRA- rpoB , PRA-16S and gyrB sequence analysis, showed 100%, 100% and 92.31% sensitivity and 93.06%, 90.28% and 100% specificity, respectively, for the discrimination of the surgical strain from other M. abscessus subsp. bolletii i solates, including 116 isolates from 95 patients, one environmental isolate and two type strains. The results of the three tests were stable, as shown by results obtained for different isolates from the same patient. In conclusion, due to the clinical and epidemiological importance of this strain, these tests could be implemented in reference laboratories for the rapid preliminary diagnosis and epidemiological surveillance of this epidemic strain.

A single strain of Mycobacterium abscessus subsp.bolletii, characterised by a particular rpoB sequevar and two highly related pulsed field gel electrophoresis patterns has been responsible for a nationwide outbreak of surgical infections in Brazil since 2004.In this study, we developed molecular tests based on polymerase chain reaction restriction-enzyme analysis (PRA) and sequencing for the rapid identification of this strain.Sequences of 15 DNA regions conserved in mycobacteria were retrieved from GenBank or sequenced and analysed in silico.Single nucleotide polymorphisms specific to the epidemic strain and located in enzyme recognition sites were detected in rpoB, the 3' region of the 16S rDNA and gyrB.The three tests that were developed, i.e., PRA-rpoB, PRA-16S and gyrB sequence analysis, showed 100%, 100% and 92.31% sensitivity and 93.06%, 90.28% and 100% specificity, respectively, for the discrimination of the surgical strain from other M. abscessus subsp.bolletii isolates, including 116 isolates from 95 patients, one environmental isolate and two type strains.The results of the three tests were stable, as shown by results obtained for different isolates from the same patient.In conclusion, due to the clinical and epidemiological importance of this strain, these tests could be implemented in reference laboratories for the rapid preliminary diagnosis and epidemiological surveillance of this epidemic strain.
More than 2,000 cases of patients showing surgicalsite infections caused by rapidly growing mycobacteria (RGM) were reported to the Brazilian National Health Surveillance Agency (ANVISA) between 2003-2008.This was a sudden increase relative to previous years with only 13 cases reported between 2000-2003(ANVISA 2009)).Most of the infections were related to laparoscopic or arthroscopic surgeries and were caused by RGM with a Mycobacterium abscessus type 2 polymerase chain reaction (PCR) restriction-enzyme analysis (PRA)-hsp65 pattern (BstE II: 235 and 210 bp; Hae III: 200, 70, 60 and 50 bp) (Devallois et al. 1997).The identification of these isolates was confirmed by rpoB sequencing (Adékambi et al. 2003).All of the obtained sequences were highly similar to the rpoB sequence of Mycobacterium massiliense CCUG 48898 (GenBank accession AY593981.2) and had two base substitutions (GenBank accession EU117207) (Leão et al. 2010).A total of 152 isolates obtained from surgical patients in different states belonged to a single strain and showed two closely related pulsed field gel electrophoresis (PFGE) patterns with a one-band difference when using either DraI or AseI restriction enzymes (Cardoso et al. 2008, Viana-Niero et al. 2008, Duarte et al. 2009, Leão et al. 2010).This strain was classified as M. abscessus subsp.bolletii according to the revised taxonomic status of the members of the Mycobacterium chelonae-M.abscessus group (Leão et al. 2009(Leão et al. , 2011)).One representative isolate from this strain was deposited in the collection of the National Institute for Quality and Health Control (INCQS), Rio de Janeiro (RJ), Brazil, under the accession INCQS 00594.
The epidemiological importance of this strain, which is herein referred to as the epidemic strain, is attributed to the extensive geographic and temporal occurrence of the epidemic of surgical-site infections (ANVISA 2011).Additionally, this strain was isolated from sputum, bronchoalveolar lavage fluid (BAL) and urine from patients unrelated to the outbreak (Matsumoto et al. 2011).The reports of RGM infections related to surgical procedures diminished significantly after 2008, although in 2010 additional surgical patients infected by the same strain were reported in the states of Amazonas (AM) and Rio Grande do Sul (RS), demonstrating that these infections may become endemic to Brazil.
Currently, the identification of this particular strain is based on rpoB sequencing and typing by PFGE.We have recently demonstrated that PFGE is a valid typing method for M. abscessus subsp.abscessus (Sampaio et al. 2006) and M. abscessus subsp.bolletii (Matsumoto et al. 2011).Sequencing and PFGE are not available in most public laboratories in Brazil and the epidemiological surveillance of this outbreak has suffered from the lack of a simple and rapid test to screen M. abscessus subsp.bolletii isolates, particularly when these isolates are derived from patients submitted to invasive procedures, such as laparoscopic, arthroscopic and plastic surgeries.
The objective of this study was to develop rapid tests for the specific identification of the epidemic strain among RGM isolates showing the M. abscessus type 2 PRA-hsp65 pattern.

SUBJECTS, MATERIALS AND METHODS
Mycobacterial isolates -A total of 116 mycobacterial isolates showing the M. abscessus type 2 PRA-hsp65 pattern were included in this study.Isolates were classified as belonging to the epidemic strain when they showed one of the two PFGE patterns that were previously defined as being characteristic of this strain (Leão et al. 2010).
Isolates showing highly similar PFGE patterns (differing from the epidemic strain patterns by up to three bands) corresponded to subtypes of the epidemic strain, according to van Belkum et al. (2007), and were categorised as genetically related (Matsumoto et al. 2011).
The type strains M. bolletii CCUG 50184 and M. massiliense CCUG 48898, both of which were reclassified as M. abscessus subsp.bolletii by Leão et al. (2011), and 55 clinical isolates from our collection were used in evaluation phases Ia and Ib (described in the Results section) (Table I).Only one isolate per patient was included.These isolates had been previously characterised by rpoB gene sequencing, PFGE-DraI and PFGE-AseI (Viana-Niero et al. 2008, da Costa et al. 2009, Duarte et al. 2009, Leão et al. 2010, Matsumoto et al. 2011).
Twenty coded isolates from 15 patients from the Institute of Microbiology Professor Paulo de Góes (Federal University of Rio de Janeiro) were sent to our laboratory (Mycobacteria Laboratory, Federal University of São Paulo) and used for the evaluation of the PRA-rpoB in evaluation phase IIa (Fig. 1).
The 59 isolates used in phases IIa and IIb were later characterised by gyrB and rpoB sequencing, PRA-16S and PFGE-DraI.
PFGE typing -PFGE typing was performed as described in previous publications from our group (Leão et al. 2010, Matsumoto et al. 2011).PFGE patterns were analysed using the BioNumerics v. 5.1 program (Applied Maths, Sint-Martens-Laten, Belgium).
PCR and DNA sequencing -The primers used for the PCR amplification and sequencing of rpoB, the 3' region of the 16S rDNA and gyrB genes are listed in Table II.The rpoB amplification was performed using primers MycoF and MycoR, as described by Adékambi et al. (2003).Two primers described in previous publications (Adékambi & Drancourt 2004, Gomila et al. 2007) were used for the amplification and sequencing of the 3' region of the 16S rDNA.The gyrB amplification primers were designed based on the sequence of the M. abscessus ATCC 19977 genome (GenBank accession NC_010397).
PCR products were purified by gel filtration (HiYield Gel/PCR DNA Extraction kit -RBC Biosciences Corp, Taipei, Taiwan).Dideoxy sequencing was performed using a BigDye ® Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and an ABI PRISM 3100 DNA Analyzer (Applied Biosystems) according to the manufacturer's instructions.
DNA sequence analysis -Sequences were aligned using the CLUSTALW Multiple Alignment program (Thompson et al. 1994) included in the BioEdit v. 7.0.5.3 package (Hall 1999).Single nucleotide polymorphisms (SNPs) present exclusively in the INCQS 00594 sequences were identified.For the development of the PRA tests, SNPs located in enzyme recognition sites were identified using the NEBcutter v.2.0 (Vincze et al. 2003).
All of the nucleotide positions described in the text correspond to positions in the M. abscessus ATCC 19977 genome (GenBank accession NC_010397).
PRA-rpoB -For the PRA-rpoB determinations, the primers MycoF (Adékambi et al. 2003) and MycoR2 (Table II) were used to amplify a 221 bp fragment of the rpoB gene.The 15 µL reactions were composed of the following reagents: 1.5 µL of 10X Buffer (150 mM Tris-HCl, pH 8.75, 500 mM KCl, 1% Triton X-100 and 20 mM of MgCl 2 ), 140 µM of each dNTP, 0.4 µM of each primer, 0.3 U of Taq DNA Polymerase (RBC Biosciences Corp, Taipei, Taiwan) and 2 µL of DNA.The PCR was performed using the following conditions: an initial denaturation step at 95ºC for 3 min, 35 cycles of 95ºC for 30 sec, 65ºC for 30 sec and 72ºC for 1 min and a final extension at 72ºC for 5 min.The digestion of 10 µL aliquots of the PCR product using DpnII (New England Biolabs, Ipswich, MA, USA) was carried out in a final volume of 20 µL according to the manufacturer's instructions.
PRA-16S -For the PRA-16S determinations, the primers 800F (Adékambi & Drancourt 2004) and 16R1492 (Gomila et al. 2007) were used to amplify a 733 bp fragment.The same reagents, volumes and amplification conditions described above were used in the 16S rDNA PCR with the exception of changes in the MgCl 2 concentration (1.5 mM final concentration) and the annealing temperature (60ºC).A 10 µL aliquot of the PCR product was digested with XbaI (Promega, Madison, WI, USA) in a 20 µL reaction according to the manufacturer's instructions.
gyrB sequence analysis -For the gyrB sequence analysis, a 524 bp fragment was amplified using the primers GYRB7 and GYRB8 (Table II) in a 25 µL reaction using the same reagents and concentrations described for PCR-rpoB but with a 60ºC annealing temperature.The primer GYRB7 was used for sequencing.

RESULTS
The development and evaluation of the tests for the rapid detection of the epidemic strain were performed in four consecutive phases (Ia, Ib, IIa and IIb) (Table I).
In phase Ia, the sequences of three genes (rpoB, the 3' region of the 16S rDNA and gyrB) were analysed.The INCQS 00594 sequences were aligned to the corresponding sequences from M. bolletii CCUG 50184, M. massiliense CCUG 48898 and seven randomly se- lected isolates that did not belong to the epidemic strain (B67, IAL 002, IAL 014, IAL 028, P7, P13 and P53).
The seven non-epidemic strains were not isolated from surgical-site infections and their rpoB sequences and PFGE patterns were different from those of the epidemic strain.All of the isolates showed the M. abscessus type 2 PRA-hsp65 pattern.
The following SNPs that occur only in INCQS 00594 and are located in restriction enzyme recognition sites were identified: an A/G SNP at position 3,916,702 in rpoB, located in a DpnII restriction site (AATC in the epidemic strain and GATC in isolates from other strains), a T/C SNP at position 1,463,374 in the 16S rDNA, located in a XbaI recognition site (TCTAGA in isolates belonging to the epidemic strain and CCTAGA in isolates that do not belong to the epidemic strain), and an A/C SNP at position 8,148 in gyrB, located in a TauI recognition site (GCSGC, where S indicates C or G; GAGGC in the epidemic strain; GCGGC in non-epidemic strain).
Three tests were developed in phase Ib, two tests are based on PRA (PRA-rpoB and PRA-16S) and one test is based on gyrB sequencing.
The 752 bp rpoB fragment described by Adékambi et al. (2003) contains multiple DpnII recognition sites and a new primer, MycoR2 (Table II), was designed for the amplification of a 221 bp fragment that contains only the DpnII site of interest.This DpnII restriction site is absent in the epidemic strain due to an A/G substitution and the corresponding amplicon was not cut with this enzyme.In contrast, this DpnII restriction site was present in the analysed isolates that are not related to the epidemic strain and digestion of the PCR products from these isolates with DpnII generated two fragments of 168 bp and 53 bp.
For the PRA-16S determination, primers 800F and 16R1492 (Table II) were used to amplify a 733 bp fragment of the 16S rDNA gene and the amplicon was digested with XbaI.The XbaI restriction site was present exclusively in the strain INCQS 00594 and two fragments of 503 bp and 230 bp were produced by the PRA-16S test.The 16S amplicons of the strains unrelated to the epidemic strain were not digested.
Additional TauI sites in gyrB are present 20 bp downstream from the SNP of interest, thereby complicating the design of a PRA test.The relatively small size difference between the digested and non-digested fragments would not be detectable using agarose gels.As an alternative approach, the outbreak strain was discriminated using sequence analysis.
Of the 47 isolates used in the three tests in phase Ib, 13 belonged to the epidemic strain, nine were retrieved from surgical-site infections (307597, B52, CRM 0019, CRM 0184, IAL 040, IAL 041, IAL 042, IAL 043 and RS 700345) and four were isolated from other specimens (IAL 020 and P10 from sputum, IAL 034 from BAL and IAL 036 from urine).Two of the sputum isolates (CRM 270 and CRM 273) showed PFGE patterns that differed in two bands when compared to the patterns of the epidemic strain isolates and were thus considered genetically related to the epidemic strain (Matsumoto et al. 2011).The nine isolates from surgical-site infections and isolates IAL 020 and IAL 034 from the other specimens were correctly identified by the three tests as belonging to the epidemic strain.Two isolates that are genetically related to the epidemic strain (CRM 270 and CRM 273) tested positive by PRA-rpoB and PRA-16S, but did not show the SNP that is characteristic of the epidemic strain according to gyrB sequence analysis.Two isolates from strains other than the epidemic strain (IAL 003 and P14) were identified by PRA-16S as belonging to the epidemic strain (i.e., false positives).False negative results from the gyrB sequence analysis were observed for isolates IAL 036 and P10, which belong to the epidemic strain (Table I).PRA-rpoB was selected for the evaluation phases IIa and IIb.Additionally, PRA-16S and gyrB sequence analyses were performed to complement the previously obtained results.
A total of 20 coded isolates from 15 patients were used in a blind evaluation of the PRA-rpoB test (phase IIa).Nine isolates showed positive results and were re-sistant to digestion with DpnII, which is characteristic of the epidemic strain.Seven of these isolates (CRM 311, CRM 312, CRM 313, CRM 353, CRM 572, CRM 575 and CRM 619) were obtained from surgical-site infections and two were isolated from BAL (CRM 552 and CRM 622).Eleven isolates, none of which were from surgical patients, tested negative (i.e., the rpoB amplicon was cut with DpnII).The results obtained from the gyrB sequence analysis were consistent with the PRA-rpoB results.The PRA-16S test identified an additional isolate from a sputum sample (CRM 573) as belonging to the epidemic strain.The characterisation of these isolates by PFGE confirmed that the nine isolates that tested positive by the three developed tests belonged to the epidemic strain and that CRM 573 was a false positive result of PRA-16S (Fig. 1, Table I).
In the final evaluation (phase IIb), the PRA-rpoB determinations were performed in a different laboratory.The PRA-rpoB protocol and primers were sent to the laboratory at CRPHF and 39 isolates from 26 patients with the M. abscessus type 2 PRA-hsp65 pattern were tested at CRPHF.Eight isolates from six patients tested positive by PRA-rpoB.Of these eight isolates, only two isolates (CRPHF 3464 and CRPHF 4636) were recovered from patients that had undergone surgical procedures.Five isolates (CRPHF 4760, CRPHF 6026, CR-PHF 6027, CRPHF 6492 and CRPHF 6493) from three patients were obtained from sputum samples and one isolate (CRPHF 3761) was obtained from BAL.The remaining 31 isolates tested negative for the epidemic strain (Table I).These results were compared to the results obtained from the PRA-16S and gyrB sequence analyses.All but one (CRPHF 4760) of the PRA-16S results were consistent with the results obtained by PRA-rpoB.The gyrB SNP that is characteristic of the epidemic strain was detected only in the two surgical isolates and in the BAL isolate that tested positive by PRA-rpoB and PRA-16S.The comparison of these results with the results of the rpoB sequencing and PFGE-DraI determinations showed that the two surgery isolates and the BAL isolate that tested positive according to all three of the developed tests (CRPHF 3464, CRPHF 3761 and CRPHF 4636) belonged to the epidemic strain and four out of five of the sputum isolates that tested positive by PRA-rpoB and PRA-16S (CRPHF 6026, CRPHF 6027, CRPHF 6492 and CRPHF 6493) were genetically related to the epidemic strain (Fig. 2).The fifth sputum isolate (CRPHF 4760) was not related to the epidemic strain and therefore was a false positive PRA-rpoB result.All of the remaining isolates, including three isolates from surgical-site infections (CRPHF 3851, CRPHF 4817 and CRPHF 6979), tested negative according to the three developed tests and belonged to strains other than the epidemic strain (Fig. 2, Table I).
Of the 116 isolates included in this study, 95 isolates corresponded to one isolate per patient, one environmental isolate and the M. bolletii and M. massiliense type strains.Of these 98 isolates, 26 belonged to the epidemic strain and 72 belonged to other strains (including the isolates that are genetically related to the epidemic strain).The PRA-rpoB test showed 100% (26/26) sensitivity, i.e., all isolates belonging to the epidemic strain were correctly differentiated.The specificity of this test was 93.06% (67/72) due to five false positive results.The respective sensitivity values for the PRA-16S and gyrB sequence analyses were 100% (26/26) and 92.31% (24/26) and the specificity values were 90.28% (65/72) and 100% (72/72), respectively (Table III).pulsed field gel electrophoresis (PFGE) was considered the gold standard method to discriminate the epidemic strain.N: different strain; P: epidemic strain.

DISCUSSION
In this study, we developed three tests for the rapid discrimination of the epidemic strain among isolates previously identified as M. abscessus type 2 according to PRA-hsp65 assays.
Initially, the sequences of 15 DNA different regions that are conserved in mycobacteria were studied: hsp65, rpoB, gyrA, gyrB, the 5' and 3' regions of the 16S rDNA, 23S rDNA, the 16S-23S internal transcribed sequence and genes previously used in multilocus sequence analysis (Macheras et al. 2011), argH, cya, glpK, gnd, pgm, pta and purH (data not shown).The sequences of these genes from M. massiliense and M. bolletii were retrieved from GenBank and aligned to the corresponding sequences of INCQS 00594.Additionally, certain isolates included in phases Ia and Ib were sequenced.SNPs and enzyme restriction sites specific to the epidemic strain were identified in silico after sequence alignment.This paper describes the analysis of three DNA regions, rpoB, the 3' region of 16S rDNA and gyrB, that contained SNPs that were exclusive to the epidemic strain and were located in enzyme recognition sites.The other studied regions, isolates, accessions and primers are provided in the supplementary data.The in silico analysis of glpK and pta showed that these genes could represent additional targets for the differentiation of the epidemic strain, although they were not used in further studies.
An A/G SNP at nucleotide position 3,916,702 corresponds to the first base of the 857th amino acid (Leu) of rpoB and is a synonymous mutation.This SNP is one of two SNPs found in the rpoB sequence from INCQS 00594 (GenBank accession EU117207) that define the epidemic strain sequevar.An A/C SNP at nucleotide position 8,148 corresponds to the first base of the 634th gyrB codon (Arg) and is also a synonymous mutation.The T/C SNP at nucleotide position 1,463,374 detected in the 16S rDNA gene does not occur within hypervariable region A or B, which are important for the discrimination of mycobacterial species (Tortoli 2003).
Of the 22 isolates obtained from surgical patients in this study, 19 epidemic strains and three non-epidemic strains (CRPHF 3851,CRPHF 4817 and CRPHF 6979) were correctly identified by the three methods.
Thirteen isolates from 11 patients (CRM 270, CRM 273, CRM 552, CRM 622, CRPHF 3761, CRPHF 6026, CRPHF 6027, CRPHF 6492, CRPHF 6493, IAL 020, IAL 034, IAL 036 and P10) were identified as belonging to the epidemic strain or genetically related according to the PFGE results, although these isolates were obtained from patients that were not submitted to surgical procedures.Interestingly, all of the genetically related isolates (CRM 270, CRM 273, CRPHF 6026, CRPHF 6027, CR-PHF 6492 and CRPHF 6493) had rpoB sequences that were 100% similar to that of the epidemic strain INCQS 00594.In addition, all of these genetically related isolates were obtained from sputum, thereby indicating that a strain that is closely related to the epidemic strain can cause pulmonary disease.The 13 non-surgical isolates tested positive by PRA-rpoB and PRA-16S, although 8/13 (61.54%) did not display the gyrB SNP that is characteristic of the epidemic strain.
During the development and evaluation of the PRA-rpoB test, new surgical-site infections occurred in Manaus (AM) and Carazinho (RS) in 2010.Six isolates from AM and eight from RS were screened using the PRA-rpoB test and were identified as belonging to the epidemic strain.These results were corroborated by rpoB sequencing and PFGE analyses (data not shown).Therefore, PRA-rpoB proved useful for the rapid preliminary screening of the epidemic strain in field situations.
Using one isolate per patient (98 isolates, including 1 environmental isolate and the 2 type strains), we observed discordant results at frequencies of 5.10% (5/98), 7.14% (7/98) and 2.04% (2/98) for the PRA-rpoB, PRA-16S and gyrB sequence analyses, respectively.The 100% sensitivity of the PRA-rpoB test was anticipated because it uses one of the two SNPs that are characteristic of the epidemic strain rpoB sequevar.The sensitivity of the gyrB sequence analysis (92.31%) was lower than the sensitivity values of the other two tests (100% each), although no false positive results were obtained in the gyrB test and the specificity (100%) was higher than those of the PRA-rpoB and PRA-16S tests.It is important to note that the PRA-rpoB and the PRA-16S tests, but not the gyrB sequence analysis, identified the genetically related isolates (6 isolates from 4 patients) as belonging to the outbreak strain.Therefore, the gyrB sequence analysis appears to be more specific to the identification of the outbreak strain, whereas the other two tests can also be used to identify genetically related isolates.The choice of specific tests could, therefore, depend on the objective of the analysis.
For 11 patients, multiple isolates from each patient were studied.The results of PRA-rpoB, PRA-16S and gyrB sequence analysis were consistent for all of the isolates from the same patient, confirming the stability of the three developed tests.
The PRA-rpoB test proved to be an efficient tool for the differentiation of the epidemic strain.The PRA-16S and the gyrB sequence analyses were shown to be useful as additional tests.The rapid tests developed here can be performed in one day and the interpretation is straightforward for both the PRA tests (i.e., the presence or absence of digestion) and sequence analysis (i.e., the detection of one SNP).In contrast, PFGE requires approximately two weeks for completion, including the necessary culture period.Moreover, PFGE requires technical training, expensive equipment and software for analysis if a large number of isolates are assayed.The PRA-rpoB test can be performed in a shorter amount of time than PFGE and is easily executed, especially in laboratories that already use the PRA-hsp65 test for mycobacteria identification.The developed tests could be implemented for the epidemiological surveillance for the epidemic strain among surgical patient isolates that show the M. abscessus type 2 PRA-hsp65 pattern.

Fig. 1 :
Fig. 1: pulsed field gel electrophoresis (PFGE)-DraI patterns of isolates from Federal University of Rio de Janeiro used in evaluation phase IIa.Nine isolates belonging to the epidemic strain are boxed.BAL: bronchoalveolar lavage.Asterisks mean Matsumoto et al. (2011).

Fig. 2 :
Fig. 2: pulsed field gel electrophoresis (PFGE)-DraI patterns of isolates from National Reference Laboratory Professor Hélio Fraga (CR-PHF), state of Rio de Janeiro, used in evaluation phase IIb.A: isolates belonging to the epidemic strain; B: genetically related isolates; BAL: bronchoalveolar lavage.

TABLE I
Mycobacterium abscessus type 2 isolates used in phases Ia, Ib, IIa and IIb and results obtained with polymerase chain reaction restriction-enzyme analysis (PRA)-rpoB, PRA-16S and gyrB sequence analysis