De Novo Cobalamin Biosynthesis, Transport, and Assimilation and Cobalamin-Mediated Regulation of Methionine Biosynthesis in Mycobacterium smegmatis

Alterations in cobalamin-dependent metabolism have marked the evolution of Mycobacterium tuberculosis into a human pathogen. However, the role(s) of cobalamin in mycobacterial physiology remain poorly understood.

S everal mycobacterial species have been identified among the subset of prokaryotes that possess the genetic capacity for de novo cobalamin biosynthesis (1)(2)(3)(4)(5). Included in this list of potential cobalamin producers is Mycobacterium smegmatis, the saprophytic mycobacterium commonly used experimentally as a surrogate for Mycobacterium tuberculosis, which causes tuberculosis (TB), a deadly respiratory disease claiming over 1 million lives globally every year (6)(7)(8). Cobalamin has one of the most complex structures of any of the biological cofactors, comprising a tetrapyrrole framework with a centrally chelated cobalt ion, dimethylbenzimidazole (DMB) as the lower axial base (a ligand), and an upper axial ligand (R-group; b ligand) (Fig.  1A). The nomenclature and catalytic activity of cobalamin depend on the b ligand. For example, in adenosylcobalamin (AdoCbl; also known as coenzyme B 12 ), the b ligand is a deoxyadenosyl group utilized by isomerases such as the methylmalonyl coenzyme A (methylmalonyl-CoA) mutase and class II ribonucleotide reductases. In methylcobalamin (MeCbl), which serves as the substrate of methyltransferases such as methionine synthase, the b ligand is a methyl group (Fig. 1A) (9,10).
Cobalamin biosynthesis is a complex multistep process requiring nearly 30 enzymecatalyzed biotransformations, including eight S-adenosylmethionine (SAM)-dependent methylations, ring contraction, six amidations, decarboxylation, cobalt insertion, aminopropanol attachment, and the assembly and attachment of the a ligand (11). Owing to the heavy energetic investment necessary to support the de novo pathway, this process is typically augmented in many organisms by the capacity for uptake and salvage (1). Interestingly, cobalamin-producing bacteria encode a much larger complement of genes involved in biosynthesis and salvage than the number of cobalamin-dependent metabolic pathways in those organisms. The preservation of de novo biosynthetic capacity therefore suggests the contribution of cobalamin to adaptation to specific lifestyles-an interpretation which is especially intriguing in the context of pathogenic, parasitic, and symbiotic bacteria (4).
Like most mycobacteria, M. smegmatis encodes several cobalamin-dependent enzymes (Table 1) (12), some of which appear redundant given the existence of isoenzymes or alternative mechanisms for the same metabolic pathway (13). Among these are the methionine synthases, MetH (5-methyltetrahydrofolate-homocysteine methyltransferase, EC 2.1.1.14) and MetE (5-methyltetrahydropteroyltriglutamate-homocysteine methyltransferase, EC 2.1.1.13), which catalyze the nonreversible transfer of a methyl group from 5-methyltetrahydrofolate to homocysteine in the final step in the biosynthesis of methionine, an essential amino acid required for translation initiation, DNA methylation, and cysteine biosynthesis (14,15). MetH requires cobalamin for activity, while MetE is a cobalamin-independent methionine synthase ( Fig. 1B) (9,16,17). We previously demonstrated that a cobalamin-sensing riboswitch located in the 59 untranslated region (59 UTR) of the metE gene in M. tuberculosis attenuated metE transcript levels in the presence of exogenous cyanocobalamin (CNCbl; vitamin B 12 ) (18). We also showed that M. tuberculosis CDC1551, a well-characterized isolate responsible for a TB outbreak in the United States (19), contains a natural truncation of the metH gene that renders the strain sensitive to exogenous CNCbl. This phenotype, which was recapitulated in an engineered M. tuberculosis H37Rv mutant containing an analogous metH truncation, suggested that the observed growth inhibition was due to methionine depletion resulting from the effective elimination of all methionine synthase activity in the CNCblreplete environment (18).
We postulated that metE would similarly be subject to riboswitch-mediated repression in M. smegmatis, given the inferred genetic capacity for de novo cobalamin biosynthesis in this organism (5). Moreover, the cobalamin-mediated repression of metE would render metH essential for growth of M. smegmatis in vitro. In this study, we provide direct biochemical confirmation that M. smegmatis constitutively produces cobalamin in vitro. We further show that M. smegmatis utilizes exogenous CNCbl and dicyanocobinamide ([CN] 2 Cbi) as precursors for the biosynthesis of the physiologically relevant cobalamin cofactor. However, our results indicate that the uptake of these corrinoid precursors by M. smegmatis is restricted. Finally, we demonstrate that the expression of metE in M. smegmatis is under constitutive repression by a cobalamin riboswitch, a finding that explains the essentiality of metH in this nonpathogenic mycobacterium under standard culture conditions.

RESULTS
De novo cobalamin biosynthesis pathway is functional in M. smegmatis. Genomic analyses indicate that M. smegmatis encodes the complete pathway for de novo cobalamin biosynthesis (5). To investigate the ability of M. smegmatis to synthesize cobalamin using the de novo pathway, we developed a liquid chromatography-tandem mass spectrometry (LC-MS/MS) method based on the derivatization of cobalamin to CNCbl by potassium cyanide (KCN). Then, utilizing multiple reaction monitoring (MRM) of two coeluting transitions corresponding to a-ribazole 59-phosphate and DMB (Fig. 1A), we identified cobalamin in the cell extracts as derivatized CNCbl, validated by two transitions coeluting at 2.79 min (Fig. 1C). Using this method, high-intensity CNCbl peaks were identified in cell extracts of wild-type M. smegmatis strain mc 2 155 grown aerobically to stationary phase in standard Middlebrook 7H9-oleic acid-albumin-dextrosecatalase (OADC) medium (Fig. 1D). To confirm de novo cobalamin production in M. smegmatis, we generated an unmarked, in-frame deletion of cobK (see Fig. S1A and C to E in the supplemental material). Homologs of this gene, which encodes a putative precorrin-6A reductase (5,11,20), have been shown in other organisms to be required for corrin ring synthesis (21)(22)(23). To ensure that all phenotypes reflected the consequences of the specific gene deletions and/or disruptions and were not confounded by off-site mutations, the parental strain and derivative mutants were subjected to whole-genome sequencing. Single nucleotide mutations (SNMs) in six genes were uniquely identified in the DcobK strain but not in the parental wild-type strain, but none of the SNMs could be linked to cobalamin or methionine metabolic pathways ( Table 2).
In contrast to wild-type cell extracts, the DcobK extracts lacked the dual coeluting peaks characteristic of CNCbl (Fig. 1E). This observation indicated the indispensability of CobK for cobalamin biosynthesis and provided further evidence that the cobalamin signal detected in the wild-type strain (Fig. 1D) resulted from de novo biosynthesis. We also confirmed the absence of cobalamin production in a double DmetE cobK::hyg knockout (KO) strain during growth in L-methionine-supplemented medium (Fig. 1F). This strain, in which the entire metE open reading frame (ORF) is deleted and cobK is disrupted by the insertion of a hygromycin (hyg) resistance marker ( Fig. S1B and F), is a methionine auxotroph that can be propagated only in media supplemented with methionine or CNCbl (Fig. 1G).
M. smegmatis assimilates exogenous CNCbl and (CN) 2 Cbi in vitro. The ability to propagate the DmetE cobK::hyg mutant in media containing methionine or CNCbl indicated that M. smegmatis can utilize exogenous methionine and CNCbl in the absence of an intact de novo cobalamin biosynthesis pathway, pointing to functional transport and assimilation pathways. We previously showed that M. tuberculosis could utilize dicyanocobinamide ([CN] 2 Cbi) during growth in vitro (12). To determine whether M. smegmatis could also assimilate this cobalamin precursor, we tested the ability of the DmetE cobK::hyg double mutant to grow in medium supplemented with (CN) 2 Cbi. First, the strain was grown to exponential phase with excess methionine (1 mM), after which 10-fold serial dilutions were spotted onto Middlebrook 7H10-OADC agar containing 10 mM (CN) 2 Cbi. After a 3-day incubation at 37°C, (CN) 2 Cbi uptake was qualitatively assessed by examining colony sizes ( Fig. 2A). For comparison, serial dilutions were also spotted on agar supplemented with 10 mM CNCbl. Interestingly, the growth of the DmetE cobK::hyg strain was very limited on agar supplemented with (CN) 2 Cbi ( Fig. 2A). In fact, growth was observed only in the most concentrated (undiluted) bacterial spots ( Fig. 2A). While this might have been as a consequence of methionine carryover, the fact that similar growth was not observed in the unsupplemented 7H10 plate ( Fig. 2A) suggested this was not the case. Instead, these observations implied the ability of M. smegmatis to utilize (CN) 2 Cbi, albeit to a much lesser extent than CNCbl ( Fig. 2A). To test the potential for (CN) 2 Cbi to support growth in liquid culture, an "MIC-type" alamarBlue assay (24) was performed in which growth from an inoculum of ;5 Â 10 3 DmetE cobK::hyg cells was determined in medium containing 2-fold serial dilutions of (CN) 2 Cbi (Fig. 2B). The DmetE cobK::hyg strain was viable at (CN) 2 Cbi concentrations higher than 7.5 mM (Fig. 2B), consistent with the ability to convert the corrinoid precursor to cobalamin.
To confirm the assimilation of (CN) 2 Cbi in M. smegmatis, we used LC-MS/MS to analyze cell extracts of wild-type, DcobK, and DmetE cobK::hyg strains grown to stationary phase in 7H9-OADC medium with or without excess (30 mM) (CN) 2 Cbi (Fig. 3). As a positive control for de novo cobalamin biosynthesis, we analyzed the wild-type strain grown in parallel without supplementation (Fig. 3B). For the DmetE cobK::hyg mutant, methionine supplementation was used to enable propagation in the absence of (CN) 2 Cbi (Fig. 3C). We observed that all the (CN) 2 Cbi-supplemented strains reached stationary phase simultaneously. However, the dual coeluting peaks characteristic of CNCbl were detected only in the (CN) 2 Cbi-supplemented DmetE cobK::hyg strain ( Fig.  3C and D), strongly suggesting uptake and conversion of (CN) 2 Cbi to cobalamin. Interestingly, cobalamin was not detectable in the (CN) 2 Cbi-supplemented DcobK strain ( Fig. 3E and F), which did not require supplementation for growth. The assimilation of (CN) 2 Cbi in the DmetE cobK::hyg strain was also accompanied by a distinct change in the color of the spent medium from purple to pale yellow (Fig. 3D, inset). By comparison, the color of the spent medium in the (CN) 2 Cbi-supplemented wild-type and DcobK cultures changed only slightly to a rusty hue ( Fig. 3B and F, insets), consistent with limited uptake in these strains.
MetE expression in M. smegmatis is regulated by a cobalamin-sensing riboswitch. We previously reported that the cobalamin-sensing riboswitch located in the 59 UTR of metE strongly attenuated the transcription of this gene in M. tuberculosis in the presence of exogenous CNCbl (18). To investigate whether the corresponding riboswitch in M. smegmatis operated similarly, we analyzed relative metE transcript levels during the exponential-growth phase in wild-type and DcobK strains grown in the presence or absence of CNCbl using droplet digital PCR (ddPCR). We found low but detectable levels of metE transcripts in the wild-type strain (Fig. 4A). By comparison, metE transcripts were 19Â more abundant in the DcobK strain ( Fig. 4A), supporting the notion that the abrogation of de novo cobalamin biosynthesis in the mutant released metE transcription from riboswitch-mediated repression. There was no significant difference in metE transcript levels between the CNCbl-supplemented and unsupplemented wild-type strain (Fig. 4A). In contrast, a small but statistically significant reduction (0.86Â) in metE transcript levels was observed in the DcobK strain in the presence of exogenous CNCbl (Fig. 4A). The absent to modest change in metE levels in these strains following CNCbl supplementation suggested that the uptake/assimilation of exogenous CNCbl might be restricted in M. smegmatis, consistent with the LC-MS/ MS results. Alternatively, these results could indicate selective repression of cobalamin uptake/assimilation systems in strains which do not require the cofactor for viability or growth.
To examine how cobalamin availability in M. smegmatis affected MetE protein content, we adapted a targeted MS method (25) to measure MetE peptide levels in wild-type and DcobK strains grown in the presence or absence of exogenous CNCbl. This analysis indicated that MetE protein levels were 3-fold more abundant in the DcobK mutant than in the parental wild-type strain (Fig. 4B). In the presence of exogenous CNCbl, the DcobK mutant exhibited a 2-fold decrease in MetE protein levels (Fig. 4B). Unexpectedly, exposure of the wild-type strain to exogenous CNCbl caused a 48-fold reduction in MetE protein levels (Fig. 4B). This result contrasted with the modest impact of CNCbl on metE transcript levels (Fig. 4A) and suggested that MetE expression was likely primarily controlled at the translational level by this riboswitch.
metH is essential in M. smegmatis. The inferred cobalamin-mediated repression of metE in turn implied that MetH function would be indispensable for the growth of wild-type M. smegmatis. To test this prediction, we attempted to generate an in-frame metH deletion mutant (Fig. S1A) by two-step allelic exchange mutagenesis (26). The DmetH construct was designed to mimic a naturally occurring metH truncation which partially disrupts the cobalamin-binding domain and eliminates the S-adenosyl-L-methionine (SAM)-binding domain of MetH in M. tuberculosis strain CDC1551 (18). Another metH KO construct containing a hyg marker (Fig. S1A) was also designed to enable the recovery of metH double crossover (DCO) mutants by "forced" selection on Hyg. Anticipating the loss of viability owing to metH essentiality, all media were supplemented with 1 mM L-methionine. Of 154 putative metH DCO recombinants screened by PCR, none (0/154) carried the DmetH allele; instead, all 154 colonies were wild-type revertants. Similarly, 60 putative hyg-marked DCOs were screened by PCR, none of which bore the DmetH allele. These results strongly suggested that metH was essential in M. smegmatis, consistent with recent genetic screens which identified metH among the subset of essential genes in M. smegmatis (27,28).
Conditional depletion by CRISPRi confirms metH essentiality. Since our attempts to delete metH in M. smegmatis were unsuccessful, we instead opted to generate a metH conditional knockdown (cKD). For this purpose, we employed the anhydrotetracycline (ATc)-inducible mycobacterial CRISPRi system (29), utilizing a panel of 13 short guide RNAs (sgRNAs) targeting different regions of the metH ORF and with different target complementarity scores (Table S3). An sgRNA targeting mmpL3, an essential gene involved in mycolic acid biosynthesis (30), was used as a positive control. Gene silencing in transformed cells was assessed by growth inhibition on ATc-containing selection medium. The induction of metH silencing by ATc inhibited colony formation in wild-type M. smegmatis, confirming the essentiality of metH under the conditions tested (Fig. 5A). Consistent with previous work (28,29,31), sgRNAs with higher complementarity scores displayed stronger gene silencing, leading to more robust growth inhibition than sgRNAs with lower scores (Fig. S2). ATc-dependent growth inhibition was rescued by supplementation with exogenous methionine (Fig. 5A), indicating that the lack of growth in the metH cKD strains resulted from methionine starvation.
To investigate this phenotype further, we traced the growth of the metH cKD strain at the single-cell level using microfluidics and time-lapse microscopy. A log-phase culture of cells carrying the metH cKD construct was preincubated with ATc for 6 h at 37°C and then loaded into the CellASIC ONIX2 microfluidic device and imaged in realtime over the course of 43 h with constant perfusion with 7H9-OADC medium containing ATc. In parallel, an uninduced (no-ATc) control was perfused with 7H9-OADC medium only. Analysis of the time-lapse images showed that the induced metH cKD strain shared similar morphological features of shape and size with the no-ATc control (Fig. 5B). Moreover, in both the metH cKD strain and the no-ATc control, cells divided by v-snapping (Movies S1 to S5, available at https://uct.figshare.com/s/65105b9914196c4b4654), which is typical of mycobacterial cell division (32,33). However, the growth rates of the metH cKD strain and the no-ATc control were markedly different (Fig. 5C). In the no-ATc control, microcolonies displayed an exponential-phase doubling time of 2.94 6 0.3 h (Fig. 5C). In this control, tracing of distinct cells was feasible only for the first 18 h of the experiment; by 24 h, microcolonies had attained confluence, occupying the entire field of view ( Fig. 5B; Movies S1 to S5 at the URL mentioned above). Consistent with methionine depletion in the metH cKD strain, this mutant exhibited a much slower replication rate, doubling every 5.23 6 0.4 h until the 18-h time point when the growth rate flatlined (Fig. 5C) and the expansion of microcolonies slowed and appeared to halt by 24 h (Fig. 5B; Movies S1 to S5 at the URL mentioned above). Abrogation of cobalamin biosynthetic capacity alleviates metH essentiality in M. smegmatis. The time-lapse microscopy data suggested that the essentiality of the metH gene might depend on both the presence of endogenous cobalamin and a functional metE riboswitch in M. smegmatis. Therefore, we reasoned it would be possible to create a metH deletion in the cobalamin-deficient DcobK strain. To test this hypothesis, we generated an unmarked in-frame metH deletion in the DcobK background and screened the resultant DCOs by PCR. As expected, PCR screening identified 9 out of 34 putative DCOs as DcobK DmetH double KO mutants (Fig. S1C to E), linking the essentiality of metH to endogenous cobalamin availability. We found that silencing of metH had no effect on the viability of the cobalamin-deficient DcobK mutant (Fig. 6A), confirming that endogenous cobalamin was required to block methionine biosynthesis via riboswitch-mediated repression of metE. Moreover, consistent with the limited impact of exogenous CNCbl on MetE protein levels (Fig. 4B), CNCbl supplementation had a negligible effect during the growth of the DcobK strain on solid medium following ATc-induced metH silencing (Fig. 6A).
Next, we investigated the impact of exogenous CNCbl during growth of the DcobK DmetH strain on solid versus liquid medium. To this end, a late log-phase (optical density at 600 nm [OD 600 ] ;1) culture was 10-fold serially diluted and spotted onto 7H10-OADC agar supplemented with or without 10 mM CNCbl (Fig. 6B). There was no impairment of growth of the CNCbl-supplemented DcobK DmetH mutant on solid medium (Fig. 6B). To determine if this phenotype was also observed in liquid medium, we seeded an inoculum of 2.5 Â 10 3 DcobK DmetH cells and analyzed cell proliferation after an overnight incubation at 37°C with or without 10 mM CNCbl using the alamarBlue assay (24) (Fig. 6C). Interestingly, CNCbl supplementation led to approximately 80% inhibition of the growth of the DcobK DmetH strain (Fig. 6C). This result confirmed the ability of M. smegmatis to assimilate exogenous CNCbl, although the uptake of the corrinoid was seemingly better in liquid than on solid medium.

DISCUSSION
Cobalamin production in M. smegmatis. The production of cobalamin by M. smegmatis was previously inferred indirectly from microbiological assays (34)(35)(36). Using a targeted LC-MS/MS approach, we provide direct proof of constitutive de novo cobalamin biosynthesis in M. smegmatis under aerobic conditions. The LC-MS/MS method optimized in this work utilized MRM of two mass spectra corresponding to DMB, the lower base in physiologically relevant cobalamin. Hence, we infer that M. smegmatis is able to synthesize and use DMB. Cobamides known as "pseudocoenzyme B 12 ," comprising an adenosine group as the a-axial ligand, rather than DMB, have been found in other bacteria (37,38). M. smegmatis CobT is homologous to the corresponding proteins in Sinorhizobium meliloti and Salmonella enterica, which have been implicated in the incorporation of adenine as the a-ligand of pseudocoenzyme B 12 under limited DMB availability (38). It is conceivable, therefore, that M. smegmatis might possess similar capacity for pseudocoenzyme B 12 synthesis. Since this remains unexplored, we cannot exclude the possibility that our DMB-dependent LC-MS/MS detection method did not capture the full variety of cobamides present in M. smegmatis. Nonetheless, these results indicated substantial cobalamin production in M. smegmatis, consistent with an early study which detected low-level cobalamin production in M. smegmatis using the Lactobacillus leichmannii tube assay (34). The disruption of cobK, encoding a predicted precorrin-6A reductase, abrogated cobalamin production, confirming that a single genetic lesion can cripple the entire pathway. It is noteworthy, therefore, that the loss of cobF, encoding a putative precorrin-6A synthase occurring immediately upstream of CobK in the pathway, has been identified as one of the defining molecular events in the evolution of pathogenic mycobacteria (39)(40)(41).
Corrinoid transport in M. smegmatis. Both CNCbl and (CN) 2 Cbi supported the growth of the DmetE cobK::hyg strain by enabling MetH-dependent production of methionine. These results demonstrate that M. smegmatis is capable of corrinoid transport and assimilation. However, our observation of strikingly restricted corrinoid transport in M. smegmatis raises questions about the corrinoid concentrations encountered in its natural habitats, and suggests the unlikelihood that exogenous corrinoids constitute a reliable source. CNCbl and (CN) 2 Cbi must undergo decyanation and adenosylation to produce adenosylcobalamin. While it is possible that M. smegmatis encodes enzymes for these reactions, work in other bacteria has demonstrated that decyanation can occur via the reduction of Co(III) to Co(II) without the need of specific decyanases (42). It has further been proposed that the steps involved in the reduction of Co(III) to Co(I), which is required for adenosylation, are not driven by reductases but, rather, are likely facilitated by electron transfer proteins (43). It seems probable, therefore, that bioconversion of CNCbl and (CN) 2 Cbi in M. smegmatis similarly occurs without the need for specific enzymes.
Interestingly, our data showed a seemingly enhanced capacity for (CN) 2 Cbi assimilation in the DmetE cobK::hyg strain compared to wild-type and DcobK strains. Since both these strains are able to grow without cobalamin owing to the presence of MetE as alternative methionine synthase, the reduced (CN) 2 Cbi uptake might suggest the potential for selective assimilation/transport as a function of methionine biosynthetic capacity. The peak intensity of the recovered cobalamin in the (CN) 2 Cbi-supplemented DmetE cobK::hyg strain was significantly lower than that of de novo-synthesized cobalamin in the wild type ( Fig. 3B and D), possibly indicating that much smaller amounts of cofactor are necessary to support growth.
In M. tuberculosis, the nonspecific ABC-type transporter BacA (Rv1819c) has been identified as the sole cobalamin and corrinoid transporter (12). Until recently, the mechanistic details of cobalamin transport by Rv1819c had remained elusive. However, the resolution of the crystal structure of Rv1819c (44) has provided key insights into its function in the uptake of hydrophilic molecules, suggesting that this protein passes a cargo slowly along its cavity via facilitated diffusion. Facilitated diffusion is a very low-efficiency process and, if the M. smegmatis homologue functions similarly in corrinoid uptake, it might explain the poor uptake of CNCbl and (CN) 2 Cbi. M. smegmatis also contains two predicted Rv1819c homologues, encoded by paralogous genes located at different genomic loci (MSMEG_3655 and MSMEG 4380). In addition, M. smegmatis contains an operon encoding putative homologues of BtuF (MSMEG_4560), BtuC (MSMEG_4559), and BtuD (MSMEG_4558), all components of the classic TonB-ExBD-BtuFCD cobalamin transport system in Gram-negative bacteria (45). Whether these genes encode functional transporters is still unknown and further research is needed to determine which proteins are involved in ferrying corrinoids and their precursors across the notoriously complex mycobacterial cell wall (46).
A riboswitch controls metE expression in M. smegmatis. We previously reported that a cobalamin-sensing riboswitch controlled metE transcription in M. tuberculosis (18). In that work, the level of metE transcript was decreased in the presence of exogenous CNCbl, leading to the conclusion that this riboswitch functioned as a transcriptional "off" switch. In the current study, we found that the levels of metE transcript were much lower in the cobalamin-replete wild-type M. smegmatis strain compared to the DcobK mutant. Since riboswitches sense ligand levels to attenuate expression (47), the low-level metE transcripts found in the wild-type strain likely reflects a physiological equilibrium between ligand-bound and unbound riboswitch states, which allows for limited gene expression. Therefore, although the uptake of exogenous CNCbl is restricted in both wild-type and DcobK strains, the low level of uptake was still enough to shift the endogenous ligand-riboswitch equilibrium more significantly in wild-type than in mutant cells, which exhibited elevated MetE protein content (Fig. 4). Unlike in M. tuberculosis, exogenous CNCbl was unable to exert significant changes to metE transcript levels in M. smegmatis, presumably owing to the limited uptake. While these results imply transcriptional regulation, we also observed an unexpected and dramatic reduction in MetE protein levels in the wild-type strain in the presence of exogenous CNCbl. These results suggested that this riboswitch might utilize a coupled translationaltranscriptional regulation mechanism by which the inhibition of translation initiation precedes transcription termination and mRNA instability (48)(49)(50). Future work will elucidate the precise mechanism of cobalamin-sensing riboswitches in mycobacteria.
Lack of MetH activity is detrimental to the growth of M. smegmatis. Our in vitro results support the conclusion that constitutive endogenous production of cobalamin compels M. smegmatis to rely on MetH for the biosynthesis of methionine. We found that the disruption of MetH activity retarded growth in the presence of cobalamin, ostensibly owing to methionine depletion. In contrast, in the absence of cobalamin, bacilli were relieved of riboswitch-mediated repression of MetE, allowing the alternative methionine synthase to substitute for the inactivated MetH. We predict that metH will be essential in all mycobacteria capable of de novo cobalamin biosynthesis, representing an important deviation from cobalamin-deficient pathogenic mycobacteria like M. tuberculosis. The corollary is that mycobacterial species that are incapable of de novo cobalamin biosynthesis will accommodate MetH inactivation. Indeed, metH-null M. tuberculosis mutants have been generated (18) and several naturally occurring, potentially inactivating mutations in metH have been found in circulating M. tuberculosis clinical isolates (13). Consistent with our findings, a recent Tn-screen identified metH among the subset of genes essential for the growth of M. smegmatis in vitro (27). Another recent study reported the inactivation of MetH in M. smegmatis, but in contrast to our findings, the authors did not observe any growth inhibition in their metH-null mutants in supplement-free media (36). The results presented here, together with our own independent Tn-seq and CRISPRi-seq analyses of M. smegmatis gene essentiality (28), demonstrate that metH cannot be disrupted in a cobalamin-replete strain without sacrificing viability. This apparent conflict might be explained by the possibility that the metE riboswitch in the parental strains used by Guzzo et al. to generate their metH-null mutants harbored inactivating mutations, which can accumulate spontaneously during the serial passage of mycobacterial cultures (51,52). To eliminate the potential confounding effect of mutations in our study, the wild-type strain and its derivative mutant strains were subjected to whole-genome sequencing.
In summary, we have shown that M. smegmatis, a nonpathogenic mycobacterium, is a constitutive producer of cobalamin in vitro. Surprisingly, the transport of corrinoids in M. smegmatis appears restricted despite the presence in the genome of multiple putative transporters. Notably, this study also revealed differences in the regulation of methionine biosynthesis between M. smegmatis and M. tuberculosis. These differences in cobalamin-dependent metabolism between an environmental mycobacterium and an obligate pathogen might be informative in understanding the selective pressures that have shaped M. tuberculosis metabolism for pathogenicity and host tropism.

MATERIALS AND METHODS
Bacterial strains and culture conditions. The bacterial strains and plasmids used in this study are described in Table S1 in the supplemental material. Unless specified, M. smegmatis cultures were grown in either Middlebrook (Difco) 7H9 broth supplemented with 10% oleic acid-albumin-dextrose-catalase (OADC) (Becton, Dickinson) and 0.05% Tween 80 or on Middlebrook (Difco) 7H10 agar supplemented with 10% OADC. For mycobacterial cultures, kanamycin (Km) and hygromycin (Hyg) were used at final concentrations of 25 mg/ml and 50 mg/ml, respectively. Escherichia coli was cultured in LB or LA with 50 mg/ml Km or 200 mg/ml Hyg, where appropriate. All cultures were incubated at 37°C. To generate growth curves, 50 ml of M. smegmatis cells were seeded at a concentration of 1 Â 10 6 CFU/ml in 96-well culture plates (Greiner Bio-One) and absorbance measurements were recorded every 1.5 h, over a period of 30 h, in a FLUOstar OPTIMA microplate reader (BMG Labtech).
Cloning. The oligonucleotides used for cloning and PCR are listed in Table S2. An in-frame, unmarked deletion in M. smegmatis cobK (MSMEG_3875) was generated by joining a 912-bp PCRgenerated fragment (FR1) containing 40 bp of the 59 end of cobK to a second 923-bp PCR-generated fragment (FR2) containing 107 bp of the 39 end of cobK in a three-way ligation reaction with a p2NIL backbone (Addgene plasmid number 20188) (53), using Asp718I, BglII, and HindIII restriction. The resultant vector (p3875K) contained a deleted 120-bp cobK allele. To generate an in-frame, unmarked deletion in M. smegmatis metH (MSMEG_4185), a 1,524-bp amplicon (FR1) of the 59 coding sequence of metH and another 1,480-bp amplicon (FR2) containing 354 bp of the 39 end of metH were joined in a three-way ligation reaction with p2NIL using Asp718I, HindIII, and BglII to produce the p4185K vector carrying a truncated metH allele of 1,848 bp. Counter-selection fragments carrying the lacZ, hyg, and sacB genes was excised from pGOAL19 (Addgene plasmid number 20190) (53) and cloned at PacI sites of p3875K and p4185K to generate the suicide vectors p3875K19 and p4185K19, respectively. To generate the hyg-marked metH construct, a hyg cassette was excised from the pIJ963 vector (54) and cloned into the BglII site of p4185K. A counter-selection cassette derived from pGOAL17 (Addgene plasmid number 20189) (53) was then inserted into p4185K to generate p4185K17. Constructs were validated by restriction enzyme mapping and Sanger sequencing using the primers listed in Table S2.
Isolation of allelic exchange mutants. M. smegmatis DcobK and DcobK DmetH mutants were generated by allelic exchange mutagenesis (26). A total of 100 ml of competent cells were incubated in a 1 mm cuvette with 1 to 8 mg DNA for 20 min on ice prior to pulsing in a GenePulser Xcell electroporator (Bio-Rad) with time constant and voltage settings at 5 ms and 1,200V, respectively. Single crossover (SCO) transformants were selected with Km and Hyg on 7H10-OADC plates. As colonies became visible, 50 ml of 2% (wt/vol) 5-bromo-4-chloro-3-indolyl-b-D-galactoside (X-gal) was underlain in each plate for blue/white screening of SCOs. PCR-verified SCOs were then cultured in antibiotic-free 7H9-OADC, followed by 10-fold serial dilutions and plating on 7H10-OADC containing 2% (wt/vol) sucrose. DCOs were screened by PCR and confirmed with Southern blotting or Sanger sequencing. For Southern blotting confirmation of DcobK, 2 to 3 mg DNA was digested overnight with StyI, separated on 1% agarose gel at 80 V, transferred and fixed onto a Hydrobond N1 membrane (Amersham), and hybridized overnight at 42°C with target-specific PCR-generated probes labeled with the ECL direct nucleic acid labeling and detection systems (Amersham). The target DNA fragments were visualized on Kodak hypersensitive X-ray films.
Cobalamin extraction. Wild-type or mutant M. smegmatis strains were cultured until stationary phase (OD 600 , ;2) in 50 ml 7H9-OADC supplemented with 3 mg/ml cobalt chloride. Cells were harvested by centrifugation at 4,000 Â g for 10 min at 4°C, resuspended in 8 ml of 50 mM sodium acetate buffer (pH 4.5), and stored at -80°C until needed. Once thawed, the cells were lysed by 5 min of sonication using a microtip sonicator set at 30 amplitude, 15 s pulse on and 15 s pulse off. Next, 16 ml of 100 mM KCN was added to the lysed cells and, with the extraction tube tightly closed, the samples were incubated at room temperature for 30 min in a chemical fume hood, followed by boiling at 90°C for 45 min inside the hood. The tube was then cooled on ice briefly and centrifuged at 4°C at 4,000 Â g for 10 min. The supernatant was filtered through a 0.22-mm filter and loaded onto a Sep-Pak C18 Plus light cartridge (Waters) which had been washed with 5 ml 75% (vol/vol) ethanol and conditioned with 10 ml of sterile water. Next, the cartridge was washed with 10 ml of water and eluted with 75% ethanol, collecting about 15 drops. The eluent was analyzed immediately by LC-MS/MS or stored in -20°C until needed.
When analysis was done on frozen samples, a centrifugation at 14,000 Â g for 10 min on a benchtop centrifuge was first performed, followed by chloroform purification.
Quantitative gene expression analysis by ddPCR. Droplet digital PCR (ddPCR) and data analysis was performed as described previously (55). Total RNA was extracted using the FastRNA Pro Blue kit (MP Biomedicals) and DNase treated with TURBO DNase (Ambion), after which 0.5 mg was used as the template for cDNA synthesis, using the High Capacity RNA to cDNA kit (Thermo Fisher Scientific). Primers and minor groove binder (MGB) TaqMan probes (Table S2) were designed using Primer Express 3.0 (Applied Biosystems). For duplexing, TaqMan MGB probes homologous to the target genes were labeled with 29-chloro-79-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC), whereas those binding the reference gene, sigA, were labeled with 6-carboxyfluorescein (FAM).
Targeted protein mass spectrometry. Triplicate cultures of M. smegmatis were grown to an OD 600 of ;1.2 in 7H9-OADC with or without 10 mM CNCbl. Cell lysis, fractionation, and the generation of tryptic peptides was done as previously described (25). Selected reaction monitoring (SRM) assays were developed in Skyline (version 4.1) using a spectral library generated from previous discovery MS data (25) with a cutoff score of 0.9. Skyline was set up to select two peptides per input protein, with the highest picked MS1 intensity in the discovery data, and then the top 5 most intense fragment ions for each of those peptides. A transition list was then generated for the Thermo Scientific triple stage quadrupole (TSQ) Vantage mass spectrometer. Samples were separated using a Thermo Accella LC system on a 10cm monolithic C 18 column (Phenomenex) with a 4.6-mm ID with a mobile phase that comprised a mixture of solvent A (water plus 0.1% formic acid) and solvent B (HPLC-grade acetonitrile plus 0.1% formic acid). The method run time was 45 min in total with a flow rate of 300 ml/min. The gradient program began with 3% B, followed by a gradient of 8% to 45% B from 5 to 25 min, and then an increase to 80% B at the 30-min mark for a 5-min wash, before returning to 3% B for the remainder of the method. The LC system was run in-line into a Thermo TSQ Vantage through a heated electrospray ionization (HESI) source. The source voltage was 13,500 V, with a capillary temperature of 300°C, a vaporizer temperature of 200°C, sheath gas of 30, and aux gas of 10. To determine the retention time for each peptide, methods were generated for the TSQ Vantage with a maximum of 20 transitions monitored per method. Since the original list contained 5 transitions, a total of 8 unscheduled methods were generated, with a cycle time of 5 s to maximize the amount of signal, a collision gas pressure of 1.5 mTorr, a Q1 peak width (FWHM) of 0.7, and collision energies as determined by Skyline. The unscheduled methods were then run with consecutive 2-ml injections of a reference sample to further refine the list of transitions and determine the retention time for each peptide. The reference sample was generated by pooling all samples. The unscheduled runs were analyzed in Skyline to determine the retention times for each peptide. Any transitions with no intensity, background-level intensity, interference, or ambiguous signal were removed from the method, and a minimum of 3 transitions per peptide were kept in the final list. The spectral library was used to further refine the assays and any peptides with a dotp score lower than 0.7 were removed from the final list.
Microplate alamarBlue assay. Cell viability was determined using the microplate alamarBlue assay (24) as follows: 50 ml of 1:1,000-diluted exponential-phase cultures (OD 600 , ;0.5) was added to 50 ml 7H9-OADC with or without 10 mM CNCbl in a 96-well plate. Plates were incubated overnight at 37°C, after which 10 ml of 100-mg/ml resazurin was added to each well. The plates were incubated for an additional 5 h at 37°C before fluorescence intensity measurements were taken using a FLUOstar OPTIMA microplate reader (BMG Labtech) using excitation and emission wavelengths of 485 nm and 508 nm, respectively.
Gene silencing using CRISPRi. Thirteen pairs of sgRNA oligonucleotides targeting the M. smegmatis metH ORF (Table S3) were designed as described previously (28). The oligonucleotides were annealed and cloned into the PLJR962 plasmid using BsmBI restriction sites in an overnight ligation reaction with T4 DNA ligase (NEB). Following ligation, the entire reaction mix (10 ml) was transformed into 50 ml of electrocompetent E. coli DH5a cells and selected on LB plates with 50 mg/ml Km. Plasmid DNA was extracted from single colonies and validated by Sanger sequencing using primer 1834 (Table S3). Next, competent M. smegmatis cells were transformed by electroporation with 200 ng of metH cKD constructs or an mmpL3 cKD control and selected on 7H10-OADC containing 25 mg/ml Km with or without 100 ng/ml ATc.
Whole-genome sequencing, genome assembly, and variant detection. Genomic DNA was extracted as described by van Helden et al. (56) from exponential-phase cultures of single colonies. Genomic libraries, prepared using the TruSeq Nano DNA (Illumina) sample preparation kit according to the manufacturer's instructions, were sequenced using a 150-bp paired-end strategy on an Illumina HiSeq 4000 instrument. Trimmomatic v0.35 (57) was used to remove adapters, leading or trailing bases with a quality score of ,3, reads shorter than 36 bp in length, and bases with an average quality score of ,15 based on a 4-base sliding window. BWA v0.7.12 (58,59) was then used to map paired-end reads to the M. smegmatis mc 2 155 reference genome (GenBank accession number CP000480.1). SAMtools v0.1.2 (61) was used to call bases. Sites that had Phred scores lower than 20 or coverage below 10-fold were removed from further analysis. SNPeff v4.1 (62), using the M. smegmatis mc 2 155 (uid57701) reference, was used to annotate variant positions.
Live-cell imaging and quantification of the growth of microcolonies. A 100-ml bacterial suspension of 2.0 Â 10 6 cells/ml was prepared and loaded on the four-chambered CellASIC ONIX B04A-03 microfluidic platform (Merck). Cells were trapped with the following pressure and flow time settings: channel A8 at 13.8 kPa for 15 s; channel A6 at 27.6 kPa for 15 s. Channel A6 was then rinsed at 6.9 kPa for 30 s. Untrapped cells were washed out by flowing inlet solution at 34.5 kPa for 5 min. 7H9-OADC medium containing 25 mg/ml Km with or without 100 ng/ml ATc was perfused continuously for 43 h. Livecell imaging was performed on a Zeiss AxioObserver using a 100Â, 1.4-numerical-aperture (NA) objective with phase contrast and a Colibri.7 fluorescent illumination system. Images were captured every 15 min using a Zeiss Axiocam 503 and analyzed using FIJI software (https://fiji.sc/). To quantify the growth of microcolonies, a threshold for the time-lapse images was set with a Yen filter in FIJI and the thresholded area over time was then quantified. For each strain type, data extraction and all subsequent analyses were performed on four independent fields of view. The data were analyzed using "R" software. Growth curves were generated by subtracting initial background objects from the size data over time and smoothed with a loess regression. Growth rates were predicted from fitting a linear model to the log 2 microcolony size obtained between 3 h and 18 h.
Data availability. Raw fastq files for whole-genome sequencing data for M. smegmatis mc 2 155, DcobK, and DcobK DmetH are available in the European Nucleotide Archive under the accession numbers ERS3716042, ERS3716043, and ERS3716041, respectively. Movies S1 to S5 can be accessed at https://uct.figshare.com/s/65105b9914196c4b4654.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 1 MB.