A total of 175 unique isolates that represent all of the presently described members of the MTC were included in the analysis and were derived from three strain collections, maintained at different institutions. One set of strains (n = 47) came from the National Reference Center for Mycobacteria in Forschungszentrum, Borstel, Germany and was collected in 2001-2003 from patients with pulmonary TB in Ghana. This set of Ghana strains was provided in a non-random blinded fashion but was known to contain both M. africanum and M. tuberculosis (as controls). All strains were previously characterized using the GenoType MTBC^® assay, as per the manufacturer's instructions, and these results were provided subsequent to the derivation of species identity using RD markers. A complete listing of the Ghana collection isolates by strain number accompanies a recent article by Wirth et al. 24 (excepting all M. bovis from Ghana and the non-M. bovis strains 10514/01, 1473/02, and 5357/02) and was recently made available as part of the MIRU-VNTRplus database http://www.miru-vntrplus.org/MIRU/index.faces 27 . Another 124 isolates were of a well-described strain collection from the Weill Medical College of Cornell University, New York. The extensive molecular characterization of the Cornell collection, and a complete listing by MTC species, unique identifier, and origin, was previously reported 7 . Only one isolate from that collection (M. tuberculosis strain W) was not included in the current evaluation. This sampling was composed of "M. canettii" (n = 5), M. tuberculosis (n = 44), M. africanum West African-1 (n = 12) (note: given previously as M. africanum subtype Ib), M. africanum West African-2 (n = 18) (note: given previously as M. africanum subtype Ia), the dassie bacillus (n = 4), the oryx bacillus (n = 2), M. microti (n = 10), M. pinnipedii (n = 7), M. caprae (n = 1), M. bovis (n = 14), and M. bovis BCG (n = 8). Lastly, 15 DNA samples were provided from the collection of the National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands 28 . These included strains of M. tuberculosis (strains

13

22

128

2

In previous reports we described 2 , and then expanded upon 7 , a PCR-based protocol for the differentiation of the various MTC species on the basis of genomic deletions. This MTC PCR-typing panel targets eight independent loci for amplification (16S rRNA, cfp32 [Rv0577], MiD3 [IS1561'], RD4 [Rv1510], RD7 [Rv1970], RD1 [Rv3877-Rv3878], RD9 [Rv2073c], and RD12 [Rv3120]), each of which either results in an amplicon of an expected size or fails, depending upon the genomic content of the MTC strain being evaluated. The resulting band-pattern that is observed following agarose gel electrophoresis is indicative of MTC species identity. Of note, the RD12 target region in M. bovis and M. caprae overlaps a specific LSP in "M. canettii" (RD12^can), while the RD1 target region in M. bovis BCG overlaps a specific LSP in the dassie bacillus (RD1^das). With this protocol, the pattern of bands for M. microti and M. pinnipedii are identical, while the pattern of bands for the orxy bacillus is the same as that of M. africanum West African-2. The MTC PCR-typing panel has been successfully applied to collections of MTC strains from Rio de Janeiro, Brazil, and Kampala, Uganda, in order to characterize the diversity of MTC species within these locales 21 31 .

Purified DNA was prepared for PCR as previously described 2 . For some strains, culture thermolysates (80°C for 30 min) were used as the source of DNA in PCR amplifications. The primers used for the MTC PCR-typing panel, the RD^Rio flank multiplex, RD174, RD701, RD702, RD711, RD713, in addition to targets containing the the pks15/1 micro-deletions and SNPs at aroA ²⁸⁵, 3'cfp32 ³¹¹, gyrA ⁹⁵, gyrB ¹⁴⁵⁰, hsp65 ⁵⁴⁰, katG ²⁰³, katG ⁴⁶³, PPE55 ²¹⁴⁸, PPE55 ²¹⁵⁴, narGHJI ^-251, RD13¹⁷⁴, rpoB ¹⁰⁴⁹, rpoB ¹¹⁶³, Rv1510 ¹¹²⁹, and TbD1¹⁹⁷, were the same as described earlier 7 32 33 . For analysis of the loci RD8, RD9, RD10, RD701, and TbD1 additional new site-specific 3-primer combinations were designed for each, similar to as previously detailed 32 , and each included two deletion flanking primers and one primer internal to the deletion. The 3-primer PCRs were each designed to amplify a product of one size when the target locus is intact or to produce a different band size when a known LSP is present. New primers were also designed to amplify a 1069-bp nat gene fragment and the SNP-containing targets in nat ⁷⁵¹ and Rv1332 ⁵²³. New primers, along with expected band sizes and the PCR program used to amplify, are listed in Table 1. The general PCR protocol was identical to that used previously 2 7 . PCR amplification from purified DNA was performed using the following cycling conditions: Program 1a (with an initial denaturation step of 5 min at 94°C, followed by 45 cycles of 1 min at 94°C, 1 min at 60°C, and 1 min at 72°C, and ending with a final elongation step for 10 min at 72°C) or program 2a (similar to program 1a but with an annealing temperature of 65°C). PCR testing of DNA thermolysates was performed in a similar manner using the following cycling conditions: Program 1b (with an initial denaturation step of 5 min at 94°C, followed by 45 cycles of 1 min at 94°C, 1 min at 60°C, and 4 min at 72°C, and ending with a final elongation step for 10 min at 72°C) or program 2b (similar to program 1b but with an annealing temperature of 65°C). Programs 1b and 2b were also used to amplify from purified DNA when potential target PCR fragments were greater than 1,250 bp. PCR products were visualized as previously described by agarose gel electrophoresis 2 . Negative or unexpected positive PCR results were repeated at least once for confirmation. Importantly, all PCR tests included parallel samples containing DNA of M. tuberculosis strain H37Rv (ATCC 27294^T) and either M. africanum West African-1 strain Percy16, M. africanum West African-2 strain ATCC 25420^T, or M. bovis strain ATCC 19210^T, where appropriate, as controls. All controls consistently provided the expected results for each particular marker screened. Negative control PCRs, lacking input DNA, were also included to control for DNA contamination.

PCR of the nat gene (1069 bp) was performed by a slightly different protocol. PCR Program 2a and a PCR reaction mix in 50 μl, with 40 pmol of each primer, 5 mM MgCl₂, 0.2 mM dNTPs, 1U Taq polymerase (Invitrogen, Brazil), PCR-buffer (10 mM Tris-HCl, 1.5 mM MgCl₂, 50 mM KCl, pH 8.3) (Invitrogen, Brazil), 10% glycerol, and 10 ng of target DNA were used in this case.

It should be noted that the M. africanum West African-1- and M. africanum West African-2-restricted LSPs were amplified by RD flanking primers 23 and analyzed as previously described 7 with the results based upon a size estimation of the PCR products on agarose gel. PCR amplification of RD713 in M. africanum clade 1 strains typically yields a 2,798 bp amplicon, while amplification of this locus in other MTC strains either results in a 4,248 bp product (PGG2 and PGG3 M. tuberculosis) or no PCR product (PGG1a MTC species with the partially overlapping RD7 deletion and PGG1b M. tuberculosis which possess additional genomic content at this locus 7 ). In PCR amplification of RD711, most, but not all, M. africanum clade 1 strains are expected to yield a 944 bp amplicon while the remaining M. africanum West African-1 strains and MTC species amplify a 2,885 bp product. With respect to RD701, all M. africanum West African-2 strains are expected to generate a 340 bp amplicon while strains from the other MTC species amplify a 2,081 bp PCR fragment. Likewise, for RD702, all M. africanum West African-2 strains are expected to amplify a 732 bp product while strains from the other MTC species produce a 2,101 bp PCR fragment. For this study, RD9 (as part of the MTC PCR typing panel), TbD1, and RD701 were evaluated by both 2-primer and 3-primer PCR tests.

Characterization of the SNPs at gyrA ⁹⁵, gyrB ¹⁴⁵⁰, hsp65 ⁵⁴⁰, katG ²⁰³, katG ⁴⁶³, narGHJI ^-251, rpoB ¹⁰⁴⁹, rpoB ¹¹⁶³, and Rv1510 ¹¹²⁹ was performed by PCR-RFA 7 33 . For the analysis of the SNPs in aroA ¹⁷⁷, 3'cfp32 ³¹¹, mmpL6 ⁵⁵¹, nat ⁷⁵¹, and TbD1¹⁹⁷ novel PCR-RFA procedures were developed, similar to those previously detailed 2 7 . The restriction enzymes and expected digest band sizes for each PCR-RFA are listed in Table 2. Amplified products from M. tuberculosis H37Rv and a second appropriate MTC species (see above) were included in all digest reactions as controls. All unexpected digestion results were repeated least once for confirmation. For each PCR-RFA evaluation, the PCR fragments from at least one strain of each digest pattern were sequenced in order to confirm the presence or absence of the target SNP.

Because it was not possible to develop a PCR-RFA based approach for characterization of the SNPs at PPE55 ²¹⁴⁸, PPE55 ²¹⁵⁴, RD13¹⁷⁴, and Rv1332 ⁵²³, SNP analysis for these markers was performed by direct sequencing of the PCR products. The same procedure was used for verification of micro-deletions in the pks15/1 locus 34 . In most cases, the primers for PCR amplification primers were also used for sequencing, as previously described 2 7 , with the exception of the 1069 bp nat fragment which was also sequenced using internal primers Table 1). Sequencing was performed using the BigDye Terminator kit (PE Applied Biosystems) on an ABI 3730 DNA Analyzer, either at the Cornell University BioResource Center (Ithaca, NY) http://www.brc.cornell.edu or at the Oswaldo Cruz Foundation (PDTIS DNA Sequencing Platform/FIOCRUZ, Rio de janeiro, RJ.); http://www.dbbm.fiocruz.br/PDTIS_Genomica/) and the results were analysed as previously described 2 7 .

Gene fragment sequences containing novel SNPs were submitted to GenBank for M. africanum West African-1 (Rv1332 ⁵²³; accession number FJ617580) and M. africanum West African-2 (nat ⁷⁵¹; accession number FJ617579). Previously identified polymorphic gene fragment sequences are now available for M. africanum West African-1 (aroA ²⁸⁵ [FJ617581] and TbD1¹⁹⁷ [FJ617582]) and M. africanum West African-2 (hsp65 ⁵⁴⁰ [FJ617583]; Rv1510 ¹¹²⁹ [GU270931]; and rpoB variants [FJ617584, FJ617585, FJ617586]).

For this study we applied the MTC PCR-typing panel to a blinded, M. africanum-enriched, challenge collection of MTC strains isolated from patients with TB in Ghana (n = 47). As a result, 18 M. tuberculosis isolates, 20 strains of M. africanum West African-1, and 9 M. africanum West African-2 strains were putatively differentiated 7 . Strains were identified as M. tuberculosis by the successful amplification of targets internal to the RD9 and RD12/RD12^can loci. Strains were identified as M. africanum West African-1 on the basis of failure of amplification of the RD9 locus but the successful amplification of the RD7 target region, while M. africanum West African-2 strains were putatively identified on the basis of failure of amplification of the RD9 and RD7 loci but the successful amplification of regions within the RD1^bcg/RD1^das, RD4, and RD12 loci. No M. bovis strains (which would have shown a pattern lacking in amplicons for RD4, RD7, RD9, and RD12) or other MTC species were identified (see ref. 7 for the expected MTC PCR typing panel patterns of "M. canettii", M. microti, M. pinnipedii, and the dassie bacillus). Of note, all strains amplified for the cfp32 (Rv0577) gene, a target that has been previously proposed to be MTC-restricted and may be necessary for pathogenesis 2 7 35 . The segregation of M. tuberculosis from M. africanum in this collection by the MTC PCR typing panel paralleled the results derived from the GenoType MTBC^® assay, which assigned these isolates as either M. tuberculosis (n = 18) or M. africanum subtype I (n = 29). These identifications were consistent with independently derived data for this strain set 24 . Fig. 1 illustrates a typical MTC PCR-typing panel profile for M. tuberculosis, M. africanum West African-1, M. africanum West African-2, and M. bovis. A summary of all molecular test results derived in this study is provided in Table 3 and illustrated schematically in Fig. 2. With respect to the RD markers interrogated above, note their phylogenetic positions in Fig. 2 at nodes 1, 6, 9, 14, and 16-19.

An exception to the common M. tuberculosis MTC PCR-typing panel profile occurred with 9 M. tuberculosis strains from Ghana, which failed to amplify the IS1561' target (see Fig. 1B). Previously, strains with this particular band pattern were found to share a clonal deletion called RD^Rio that defines a major, newly recognized, lineage of M. tuberculosis that is the predominant cause of TB in Rio de Janeiro, Brazil, and that has disseminated to many countries around the world 7 31 32 . However, multiplex PCRs for both the RD^Rio LSP and the coincident RD174 deletion 32 showed that these Ghanaian strains were not RD^Rio genotype M. tuberculosis. Rather, data from the MIRU-VNTRplus website identified these strains as being of the RD726-harboring Cameroon genotype (ST61 and variants) and lists the strains as lacking IS1561' 27 . The Cameroon genotype therefore appears to possess an undefined LSP of IS1561' that overlaps RD^Rio (Fig. 2; see node 4) and the MiD3 locus in M. microti and M. pinnipedii (Fig. 2; see node 16) 7 31 .

In addition to the MTC PCR-typing panel, some PCR targets used in SNP analysis, as will be described below, amplify from genomic regions that are deleted is some MTC species or lineages 7 . The successful amplification of the 3'cfp32 and RD13 loci in all the strains of the Ghana collection confirmed the species distribution obtained using the MTC PCR-typing panel, as these targets are deleted in either "M. canettii" (Fig. 2; see node 1) or both M. caprae and M. bovis (Fig. 2; see node 17), respectively 7 . Furthermore, PPE55 is located proximal to IS1561' and so the failure to amplify PPE55 from the 9 Cameroon genotype M. tuberculosis isolates is consistent with a single genomic deletion in the region of IS1561' (Fig. 2; see node 4). Lastly, TbD1 is an important phylogenetic marker that categorically divides M. tuberculosis into two major lineages 1 . All M. tuberculosis isolates in the Ghana collection failed to amplify from targets internal to TbD1 (Fig. 2; see node 2), while all M. africanum clades 1 and 2 strains yielded an amplicon of the correct size, consistent with the previous finding that isolates from the M. africanum→M. bovis evolutionary tract are all TbD1-positive and likely a branch off of a TbD1-positive M. tuberculosis lineage 1 24 .

We next evaluated the Ghana strain collection by PCR (using LSP flanking primers) for RDs that have been described previously as being either specific to M. africanum West African-1 (RD713), restricted to a subgroup of M. africanum West African-1 (RD711), or specific to M. africanum West African-2 (RD701 and RD702) 7 23 . All M. africanum West African-1 strains (n = 20) yielded amplification products for RD711 and RD713 of shorter band sizes that were consistent with amplicons that bridge a deletion (Fig. 2; see nodes 7 and 8). All M. tuberculosis strains (n = 18) contained the RD711 and RD713 regions, while each M. africanum West African-2 strain (n = 9) yielded PCR fragments suggestive of intact RD711. Each M. africanum West African-2 strain (n = 9) also failed to produce any amplification products from the RD713 locus region, as expected, owing to the overlapping RD7 7 . Likewise, all M. africanum West African-2 strains produced shortened RD701 and RD702 amplicons (Fig. 2; see node 11), while each M. tuberculosis and M. africanum West African-1 strain exhibited PCR fragments representative of intact sequences within these loci. The M. africanum clade-specific bridge-deletion PCR results were therefore congruent with the MTC PCR-typing panel data.

A drawback, however, of the MTC PCR-typing assay as it was designed is that overlapping polymorphisms may occur in the target regions of the panel. Such hypothetical LSPs would therefore have the potential to cause a failure in amplification and to confuse the interpretation of banding patterns which may, in turn, lead to erroneous species determinations. To begin to address this issue, with respect to loci relevant to the species within the current Ghana collection, we developed new 3-primer combination sets for RD8, RD9, RD10, RD701, and TbD1 (Table 1). As was expected from previous phylogenetic evaluations 1 3 7 , each of the test loci were found to be intact in the Ghana collection PGG2 M. tuberculosis strains, excepting TbD1. Moreover, excepting RD9, each of the studied RDs were intact in the M. africanum West African-1 strains, while in the M. africanum West African-2 strains only TbD1 remained intact, i.e. the RDs 8-10 and RD701 were deleted. Overall, no inconsistencies were observed with respect to species identification within the Ghana MTC strain collection across the different strategies for PCR deletion analysis that were employed.

For the second stage of this study we screened the Ghana MTC collection for known phylogenetically relevant SNPs. With respect to the M. tuberculosis strains, we determined that all were PGG2 (n = 19) (Fig. 2; see nodes 3 and 5). Consistent with this determination, the 7-bp pks15/1 micro-deletion was observed in all the M. tuberculosis strains; this polymorphism is positioned at the same point along the MTC evolutionary tree as the katG ⁴⁶³ CTG→CGG SNP that marks PGG2 M. tuberculosis strains (Fig. 2; see node 3). Likewise, an SNP in the narGHJI operon promoter (-215 C→T), that is phylogenetically coincident with TbD1 33 was also present in all of the Ghanaian M. tuberculosis isolates evaluated (Fig. 2; see node 2). Lastly, the gyrB ¹⁴⁵⁰ G→T polymorphism (also a target of the GenoType MTBC^® assay 14 15 16 ) is known to coincide with the RD9 deletion and likewise segregated the M. tuberculosis isolates from the strains of the M. africanum strains (Fig. 2; see node 6).

The following considers SNPs that inform the phylogenetic interrelationships among most of the non-M. tuberculosis MTC species. First, all the M. africanum strains (n = 28) were PGG1. Previously, an ACC→ACT SNP at katG ²⁰³ has been used to segregate PGG1 strains into PGG1a and PGG1b 30 . Huard et al. 7 reported that this SNP is present in M. africanum West African-2 and all downstream species in the MTC evolutionary tree (Fig. 2; see node 9). As expected, the Ghana collection M. africanum West African-1 strains were determined to be PGG1b, while the M. africanum West African-2 strains were PGG1a by katG ²⁰³ analysis. Additional inter-species-specific SNPs that colocalize with the katG ²⁰³ SNP and segregate the M. africanum clades (and are also notably coincident with RD7, RD8, and RD10) have also been reported at 3'cfp32 ³¹¹ (G→A), PPE55 ²¹⁴⁸ (A→G), PPE55 ²¹⁵⁴ (A→G), and RD13¹⁷⁴ (G→A), in addition to a 6-bp pks15/1 micro-deletion (Fig. 2; see node 9) 7 34 . These loci were interrogated and indeed found to partition the M. africanum West African-2 strains from the M. africanum West African-1 and M. tuberculosis strains of the Ghana collection, consistent with previous reports 7 34 . Lastly, we also screened for an inter-species-specific SNP in mmpL6 ⁵⁵¹ (AAC→AAG) 1 7 that is not observed in M. africanum West African-1, M. africanum West African-2, nor the dassie bacillus, but is present in all of the remaining distal species along the oryx bacillus→M. bovis evolutionary track of the MTC phylogenetic tree 1 7 26 . As was expected, we found mmpL6 ⁵⁵¹ to be unaltered in the M. africanum West African-1 and West African-2 strains of the Ghana MTC collection (Fig. 2; see node 15). The mmpL6 ⁵⁵¹ SNP occurs within a TbD1 locus gene and was thus deleted in the TbD1-negative M. tuberculosis strains of the Ghana collection.

We then investigated SNPs that have been previously described to be restricted to either M. africanum West African-1 or M. africanum West African-2 within the MTC 7 . SNPs at aroA ²⁸⁵ (G→A) and TbD1¹⁹⁷ (C→T) were found to be limited to the M. africanum West African-1 strains of the Ghana MTC collection, thereby coinciding with the M. africanum West African-1-specific LSP RD713 (Fig. 2; see node 7). Point mutations at Rv1510 ¹¹²⁹ (G→A), hsp65 ⁵⁴⁰ (C→G), and rpoB ¹¹⁶³ (C→T) were also screened and found to be restricted to the M. africanum West African-2 strains (Fig. 2; see nodes 10-12); a previously noted sublineage-specific SNP at rpoB ¹⁰⁴⁹ (C→T) was not observed (Fig. 2; see node 13). However, from previous data 7 , only hsp65 ⁵⁴⁰ has been shown to be truly M. africanum West African-2-specific and to associate pylogenetically with RD701 and RD702. In fact, Rv1510 ¹¹²⁹ was previously found to be an inter-species-specific SNP that M. africanum West African-2 shares with the dassie bacillus, and is indicative of a common ancestor between these species, while not all M. africanum West African-2 strains possess the rpoB ¹¹⁶³ and rpoB ¹⁰⁴⁹ SNPs 7 . These latter point mutations appear to have been acquired in a step-wise sequential order and to define the branch points of sublineages within the M. africanum West African-2 species. All Ghana M. africanum West African-2 strains evaluated in this study therefore fell into the second of three potential rpoB sequence-based sublineage branches. Overall, each of the known MTC inter-species-specific, species-specific, and sublineage-specific SNPs for which the Ghana MTC collection was evaluated were entirely consistent with the current RD analyses and showed a species distribution that paralleled previous descriptions 7 .

In the process of sequencing the RD711 bridge amplicon to confirm its correct amplification in an M. africanum West African-1 strain, we noted a nonsynonomous G→T SNP in the region 5' of the RD711 deletion breakpoint and within the Rv1332 gene, affecting nucleotide 523 (Rv1332 ⁵²³; V175L). To investigate the distribution of this Rv1332 ⁵²³ SNP amongst the MTC species, we generated a new primer pair to amplify the SNP-containing region upstream of RD711. We then performed PCR and sequence analysis of the amplified products upon samples from select MTC strains of the Cornell collection representing each of the MTC species and major M. tuberculosis lineages, i.e., "M. canettii" (n = 2), TbD1-positive M. tuberculosis PGG1 (n = 2), TbD1-negative M. tuberculosis PGG1 (n = 2), M. tuberculosis PGG2 (n = 2), M. tuberculosis PGG3 (n = 3), M. africanum West African-1 (n = 12), M. africanum West African-2 (n = 2), the dassie bacillus (n = 2), the oryx bacillus (n = 2), M. microti (n = 2), M. pinnipedii (n = 2), M. caprae (n = 1), M. bovis (n = 2), and M. bovis BCG (n = 2). Only the 12 M. africanum West African-1 strains possessed the Rv1332 ⁵²³ substitution. When the Ghana collection was subsequently evaluated (n = 47), the Rv1332 ⁵²³ SNP was likewise restricted to the 20 M. africanum West African-1 strains. In total, 85 MTC isolates were screened, 32 of which were M. africanum West African-1. The data thus supported that the Rv1332 ⁵²³ SNP is a specific marker for M. africanum West African-1 and is only the third such polymorphism reported to date (Fig. 2; see node 7) 7 .

Previously, the nat (Rv3566c) gene product arylamine N-acetyltransferase has been investigated as a potential contributor to reduced isoniazid susceptibility in M. tuberculosis 36 . In the course of those investigations, SNPs were identified in the nat gene that were restricted to different M. tuberculosis lineages. We found a novel nonsynonomous G→A SNP in two M. africanum West African-2 strains at nat nucleotide 751 (nat ⁷⁵¹; E251K) upon amplification and sequencing of a 1069-bp nat fragment using samples from a subset of MTC representative strains (RIVM collection; n = 15). Test sequencing of the 1069-bp nat amplicon from 16 MTC strains from the Cornell collection supported the limited distribution of the nat ⁷⁵¹ SNP. We then developed a PCR-RFA protocol for the nat ⁷⁵¹ SNP, amplifying a shorter product using new primers and employing the restriction enzyme BcgI, and applied the protocol to all strains of both the Cornell (n = 124) and Ghana collections (n = 47). Consistent with the preliminary test results, all MTC isolates amplified nat successfully. However, only the 27 M. africanum West African-2 strains possessed the nat ⁷⁵¹ polymorphism, as determined by PCR-RFA. The West African-2 strains showed a 4-band digest pattern on agarose gel electrophoresis as opposed to the remaining MTC strains that showed a 3-band digest pattern (see Table 2). Thus, this SNP appears to be a specific marker for M. africanum West African-2 (n = 175 unique MTC strains evaluated in total) and is only the second SNP reported to be restricted to this clade (Fig. 2; see node 11) 7 . Of note, both the nat ⁷⁵¹ and hsp65 ⁵⁴⁰ M. africanum West African-2-specific SNPs are present in the genomic sequencing project of M. africanum strain GM041182 that is currently nearing assembly completion http://www.sanger.ac.uk/sequencing/Mycobacterium/africanum/.