All trans (E) isomers of GPP, FPP, and GGPP were purchased from Isoprenoids, LC (Tampa, FL, USA). Unless otherwise noted, all other chemicals were purchased from Fisher Scientific (St. Louis, MO, USA) and all molecular biology reagents were purchased from Invitrogen (Carlsbad, CA, USA). Gas chromatography-flame ionization (GC-FID) detection was carried out using an Agilent 6890 GC-FID (Agilent Technologies, Santa Clara, CA, USA) using a previously described method (Schmidt and Gershenzon, 2008).

Genes of interest were identified via basic local sequence alignment (BLAST) against the M. tuberculosis H37Rv genome (Cole et al., 1998). Phylogenetic analysis of GGPP synthases (GGPPS) was accomplished via UPGMA modeling of amino acid sequences (Sneath and Sokal, 1973), using CLC Sequence Viewer 6.0 (CLC Bio, Cambridge, MA, USA). Complete operon analysis was performed against publically available sequences of Mtb diversity strains (Hershberg et al., 2008), M. canettii (Hershberg et al., 2008), M. marinum (Stinear et al., 2008), and M. ulcerans (Stinear et al., 2007), as well as previously published sequences of M. bovis (Garnier et al., 2003), and M. bovis BCG (Brosch et al., 2007), and the pre-publication reads of M. africanum (available via the Sanger Institute). Genomic regions were aligned using BLAST analysis and synteny mapping via the Tuberculosis Database (Reddy et al., 2009).

Genes of interest were cloned by PCR amplification from Mtb CDC1551 genomic DNA and confirmed via complete sequencing. These were transferred into pENTR/SD/D-TOPO (Gateway, Invitrogen) by directional recombination. Rv2173 and Rv0562 were modified to change the first codon to ATG to enable expression in Escherichia coli, for which purpose all genes were recombined into pDEST17 expression vectors for production with an N-terminal 6xHis tag.

LytB complementation was initially assessed by colony counting assays, carried out in three individual experiments, each in duplicate. 50 ng of plasmid (pDEST14) containing Rv3382c or Rv1110 were each electroporated into 100 μL electro-competent E. coli cells containing LytB under arabinose promoter control (McAteer et al., 2001) and were plated on NZY medium in the absence of arabinose, which allows only for growth of LytB complemented cells. Electroporated E. coli cells plated in the presence and absence of 0.2% arabinose were used as controls. More quantitative analysis of LytB activity was accomplished by comparison of relative growth rates, carried out in triplicate. Colonies from Plasmid (+) plates were grown to saturation in LB medium overnight. Cultures were diluted 1/100 into fresh LB medium containing 0.5 mM isopropylthiogalactoside (IPTG) and grown for 1 h, then diluted 1/50 into media containing 50 μg/mL carbenicillin and 0.2% (v/v) glucose to inhibit the arabinose promoter and thus completely inhibit any expression of E. coli LytB. After 220 min, cultures were diluted 1/50 into medium containing 0.2% (v/v) arabinose, thus activating the chromosomal LytB and allowing for comparative analysis of the relative health of both complemented strains (i.e., to demonstrate that expression of Rv3382c did not exert any toxicity relative to expression of Rv1110 instead). Culture growth was monitored every 20 min by measuring OD₆₀₀.

Putative GGPP synthases in pDEST17 were transformed into the C41 strain of E. coli (Lucigen, Middleton, WI, USA), and plated on NZY agar with carbenicillin (50 μg/mL) selection. Colonies were inoculated in 50 mL of NZY medium and grown to an optical density at 600 nm (OD₆₀₀) of 0.6 at 37^°C before being transferred to 16^°C for 1 h prior to induction with 0.5 mM IPTG, with subsequent 14–16 h incubation. Cells were harvested via centrifugation and lysed in 1 mL lysis buffer (10 mM Tris-Cl, pH 6.8, 10% glycerol, 1 mM dithiothreitol) via sonication. These lysates were clarified via centrifugation (10 min × 10,000g) and the resulting cell free extracts assayed for GGPP synthase activity. As E. coli does not harbor contaminating GGPP synthases (Burke and Croteau, 2002), any resulting production of GGPP was attributed to expression of the recombinant gene.

Assay buffer (50 mM sodium phosphate (pH 7.0), 10% glycerol, 5 mM MgCl₂, 1 mM dithiothreitol) was mixed with 20 μM IPP and the equivalent concentration of indicated allylic isoprenyl diphosphate. To each assay, 100 μL of purified protein (~1 mg/mL) expressed as a 6×His fusion in pDEST17 was added, totaling 1 mL total volume. Assays were conducted at 37^°C for indicated time and halted via incubation at 100^°C for 5 min. After cooling, 1.5 U calf intestinal alkaline phosphatase (New England Biolabs, Ipswich, MA, USA) was added according to manufacturer protocol. The phosphatase reaction was incubated overnight at 37^°C and extracted thrice with 1 mL hexanes. Samples were concentrated and analyzed via GC-FID and compared to similarly, dephosphorylated authentic GGPP.

The 6×His tagged versions of these enzymes were purified via Ni-NTA Superflow resin (Novagen, Merck, Germany) following the manufacturer directions. Protein was analyzed via denaturing gel electrophoresis (SDS-PAGE), demonstrating >90% purity. Enzymes were assayed as described against a substrate gradient (0–40 μM) with IPP and DMAPP/FPP for 5 min with 100 nM enzyme. Post-extraction, cembrene was added to a final concentration of 5 μM as an internal standard for quantification. Product concentration was calculated by integration of the product peak area and comparison to internal standard peak area. Michaelis–Menten curves were fit to the observed data (KaleidaGraph 4.0, Synergy Software, Reading, PA, USA), resulting in R² values ≥0.95.

To begin investigating the hypothesized biochemical redundancy, the uncharacterized Rv3382c and Rv3383c were analyzed and compared to the rest of the Mtb genome. Rv3382c is 990 bp long and encodes a protein of 329 amino acids (AA) that has been annotated as lytB1, while Rv3383c is 1053 bp long and encodes a 350 AA protein annotated as idsB (Camus et al., 2002). Both genes are non-essential (Sassetti et al., 2003), and neither is associated with pathogenicity in published genetic screens (Sassetti and Rubin, 2003; Pethe et al., 2004; Fortune et al., 2005; Joshi et al., 2006). Protein–protein BLAST analysis of Rv3382c and Rv3383c against the H37Rv genome revealed putative paralogs for both. Specifically, Rv3382c shares 50% AA identity with Rv1110, which is similarly annotated as lytB2 (Camus et al., 2002). There are four IDS paralogs to Rv3383c in the Mtb genome: Rv3398c, a previously identified FPP synthase annotated as idsA1 (Dhiman et al., 2004), which shares 35% AA identity with Rv3383c; Rv2173, annotated as idsA2, which shares 29% AA identity; Rv0562, annotated asgrcC1, which shares 30% AA identity; and Rv0989c, that we have recently demonstrated encodes a functional GPP synthase likely involved in decaprenyl diphosphate biosynthesis (Mann et al., 2011a), annotated as grcC2, which shares 28% AA identity.

LytB is a [4Fe–4S] protein that has been shown to be difficult to work with in vitro, due to complications from both oxygen corruption and substrate availability (Adam et al., 2002; Wolff et al., 2003). Accordingly, such enzymatic activity is most often demonstrated by a complementation approach (McAteer et al., 2001; Rohdich et al., 2002; Hsieh and Goodman, 2005). Briefly, this utilizes a previously developed strain of E. coli in which the endogenous LytB gene (EG11081) has been put under control of an arabinose promoter, such that bacterial growth requires either supplementation with arabinose or complementation with a functional LytB (McAteer et al., 2001). Both Rv3382c and Rv1110 were able to complement the growth of this strain of E. coli in the absence of arabinose, demonstrating functional LytB activity. Relative growth rate analysis indicates that complementation with Rv3382c results in somewhat less vigorous growth than does complementation with Rv1110, which does not appear to be due to a toxicity effect (Figure 2). This result then suggests that Rv3382c exhibits somewhat less robust LytB activity than Rv1110, which might be due to Rv3382c being 8 AA shorter at the N-terminus. However, this region does not contain any residues known to be involved in enzymatic activity, and the exact mechanism underlying reduced function of the LytB encoded by Rv3382c is unclear.

The production of GGPP is catalyzed by the condensation of IPP with allylic isoprenyl diphosphate acceptors, such as DMAPP (which then requires the addition of three molecules of IPP), GPP (e.g., that produced by Rv0989c, which then requires the addition of two molecules of IPP), or FPP (e.g., that produced by Rv3398c, which then only requires the addition of a single molecule of IPP; Figure 3). Thus, the putative GGPP synthases identified above were screened for such activity by assaying purified recombinant proteins expressed in E. coli, in the presence of IPP plus DMAPP, GPP, or FPP (Table 1). Both Rv0562 and Rv3383c were able to add IPP to FPP, producing GGPP (Figure 4). In addition, Rv0562 is able to produce GGPP using either DMAPP, GPP, or FPP as the allylic precursor, while Rv3383c can synthesize GGPP only when fed FPP (Table 1). By contrast, Rv2173 only exhibited intermittent GGPP synthase activity in these assays, indicating that its primary product may not be GGPP. Further studies are ongoing with this enzyme to clarify its function.

The commonly accepted evolutionary scheme for pathogenic mycobacteria indicates that the MTC emerged from a common ancestor, which was most recently shared with M. canettii, to create a pool of human and animal pathogens, including M. bovis and M. africanum, along with Mtb (Brosch et al., 2002). Early analysis of the Mtb genome isolated a putative five-gene operon involved in isoprenoid biosynthesis at coordinates 3790.85–3799.64 kb and containing loci Rv3377–Rv3383c. This operon was identified as potential horizontal gene transfer products from an exogenous source due to aberrations in GC content within this region (Becq et al., 2007). Rv3377c and Rv3378c are diterpene synthases that produce the immunomodulatory diterpenoid isoTb via cyclization of the diterpenoid precursor GGPP (Nakano et al., 2005a,b; Mann et al., 2009b; Maugel et al., 2010; Prach et al., 2010; Spangler et al., 2010; Hoshino et al., 2011; Mann et al., 2011b; Nakano et al., 2011). In our characterization of the diterpene cyclase encoded by Rv3377c, we noted that this gene was conserved in all four of the Mtb strains whose genomes were available at that time, while the M. bovis homolog contained an inactivating frameshift mutation that was conserved in the two sequenced strains, along with another that we investigated (Mann et al., 2009a). Based on the wider host range, but reduced infectivity in humans, of M. bovis, we then speculated that isoTb might play a role in the highly infectious nature of Mtb in humans. Given the current availability of genome sequences from a number of diverse Mtb strains (Comas et al., 2010), as well as several members of the MTC and other mycobacteria (Reddy et al., 2009), these were analyzed for the presence of all or part of the isoTb operon. Corresponding sequences could be found not only in the MTC and M. canettii, but also the much more distantly related M. marinum, although it is not present in the even more distantly related M. leprae, suggesting that the isoTb operon was acquired prior to separation of the lineages leading to M. marinum and the MTC/M. canettii. We further characterized the existing polymorphisms of not only the diterpene synthases [i.e., Rv3377c and Rv3378c, as previously reported (Mann and Peters, 2012)], but also the entire isoTb operon in these mycobacteria. Notably, very few mutations are found within the unique diterpene synthase encoding genes (i.e., Rv3377c and/or Rv3378c) in the sequenced strains of Mtb (Comas et al., 2010), with none in the highly infectious Beijing subgroup. By contrast, the other species within the MTC, as well as M. canettii, all have accrued frameshift and/or non-synonymous mutations in Rv3377c or Rv3378c. The range of observed mutations in these diterpene synthases is consistent with the current view of mycobacteria evolution, with only a single non-synonymous mutation found in the more recently derived M. africanum. By contrast, the more distantly related M. canettii contains both non-synonymous and frameshift mutations, wherein the frameshift mutation presumably precludes synthesis of isoTb, much as we previously demonstrated for M. bovis (Mann et al., 2009a). In addition, the much earlier diverging M. marinum does not contain the diterpene synthase genes at all, although it does have the other genes from this operon. Specifically, M. marinum has retained orthologs to the GGPP synthase encoded by Rv3383c, the LytB encoded by Rv3382c, and the pseudo-DXS encoded by Rv3379c, although there also has been an apparent transposon-mediated insertion within the operon. We suggest that this reflects loss of the diterpene synthases, rather than their later addition in the M. canettii/MTC lineage, as it is unclear why the other genes, which together produce the diterpene precursor GGPP, would have been assembled together in the absence of the downstream diterpene synthases (and prior to the acquisition of the carotenoid biosynthetic machinery, i.e., as seen in M. marinum). In the genome of M. ulcerans, which is derived from M. marinum (Stinear et al., 2007), the operon has been further reduced, with only the Rv3383c ortholog remaining (Figure 5). While this observation is consistent with the previously noted severe genome reduction of M. ulcerans (Stinear et al., 2000), M. marinum is believed to have undergone parallel evolution from a common ancestor with Mtb, but is generally considered to have retained and incorporated more genes, as expected for a more general pathogen, relative to the human host restricted Mtb (Stinear et al., 2007). Accordingly, isoTb production seems to be most highly conserved in Mtb, despite its comparatively specialized genome, which is consistent with our previous speculation positing a role for isoTb in the highly infectious nature of this human-specific pathogen (Mann and Peters, 2012).

Although the ability to produce isoTb has not been found and seems unlikely to occur outside of Mtb, it is notable that the GGPP producing IDS found in Mtb are subject to stronger conservation (Figure 6). Strikingly, the GGPP synthase encoding Rv3383c/idsB from the isoTb operon was retained even through the reductive evolution process leading to M. ulcerans where the remainder of the operon is lost (Figure 5). While the presence of a GGPP synthase is likely due to its role in production of carriers involved in cell wall synthesis, the retention of multiple GGPP synthases in these mycobacteria is striking, albeit somewhat puzzling. Rv0562/grcC1 is likely the ancestral GGPP synthase and is widely found in the family Mycobacterium, e.g., an ortholog is present in the fast-growing, non-pathogenic saprophyte M. smegmatis. However, those mycobacteria derived from the MTC and M. marinum ancestral strain also have retained the additional GGPP synthase encoded by Rv3383c/idsB, even upon loss of the other genes in the isoTb operon (e.g., in M. ulcerans). In the case of M. marinum, this might be due to its production of GGPP-derived carotenoids (Ichiyama et al., 1988), although M. marinum also has acquired multiple other IDS homologs. In any case, this rationale does not hold true for other mycobacteria, leaving unclear what selective pressure led to the wider conservation of Rv3383/idsB noted here. Interestingly, this specific conservation of dual GGPP synthase homologs across these diverse species of mycobacteria suggests a more significant role for GGPP than previously presumed, consistent with the previous observation of large pools of GGPP-derived geranylgeraniol in Mtb cellular extracts (Crick et al., 2000; Mann et al., 2011b). However, it should be noted that the increased sequence divergence of the Rv3383c/idsB orthologs in M. marinum and M. ulcerans also might reflect functional diversification.