A Mycobacterial Systems Resource for the Research Community

Rationale for selection of a core set of M. smegmatis genes.The genome of M. smegmatis is larger than those of the pathogenic species (6.9 Mb versus 4.4 Mb for M. tuberculosis), likely reflecting the more varied demands of its environmental niche. Therefore, we have focused on the most highly conserved proteins in the M. tuberculosis complex (MTBC), M. avium, M. abscessus, and M. leprae, since we predict that these proteins regulate and/or perform shared fundamental cellular processes. Such conserved genes are best studied in the most experimentally tractable mycobacterium, M. smegmatis, providing experimental insights applicable to the pathogenic species. M. smegmatis is an ideal host for this core-gene resource because it is a nonpathogenic, fast-growing model organism with developed genetic tools suitable for high-throughput analyses (27). The virtues of M. smegmatis as a model organism are further underscored from Tn-seq analyses, which indicate that 96% of essential and growth-impaired genes in M. smegmatis have a mutual ortholog in M. tuberculosis and the majority of those (90%) are essential in M. tuberculosis (28).

The 1,153 M. smegmatis protein-coding genes were selected using a best pairwise BLAST approach (<E⁻³⁰) with M. tuberculosis, M. leprae, M. abscessus, and M. avium genomes, with a secondary screen for synteny using MUMmer and DAGchainer (29, 30), which identifies genes that are in homologous syntenic groupings (Table S1). Similar results were obtained with independent algorithms such as Roary (Fig. 1a) (31) and OrtholugeDB, which measures phylogenetic distance ratios to predict orthologs (32, 33). The OrtholugeDB analysis confirmed that most of these genes have evident orthologs in many other sequenced mycobacterial genomes, beyond those we initially selected (Table S2). Even our lower ranked proteins, sharing over 50% identity across all the species listed, have high identity in pairwise comparisons, e.g., MSMEG_0004 (DciA) is 73% identical and 84% similar between M. smegmatis and M. tuberculosis (Fig. S1) (34). Broad conservation underscores the conserved function and, therefore, the relevance of this core gene resource for pan-mycobacterial studies. We note that the inclusion of M. leprae, which has undergone massive gene decay, reduces the overall number of conserved genes by about 600 genes, and further focuses the resource on proteins mediating fundamental core processes common to all of the species.

Many mycobacterial genes are misannotated because existing algorithms cannot accurately predict gene features in their G/C rich genomes and because ∼1/3 of mycobacterial genes are encoded from leaderless transcripts (35, 36). To ensure accurate gene cloning and precise disruption of genes, we utilized our previously published data that integrate both transcriptome sequencing (RNA-seq) and ribosome profiling (Ribo-seq) to accurately predict gene starts (https://www.wadsworth.org/research/scientific-resources/interactive-genomics) (36). Of the 1,153 genes within the MSR, 17 are annotated as pseudogenes in Mycobrowser because they contain nucleotide changes resulting in a frameshift, which are described as “not sequencing errors.” Sequence analysis of our individual plasmid clones shows that only three of these genes (MSMEG_2348, MSMEG_3479, and MSMEG_3831) are pseudogenes in our strain (Table S1). The remaining 14 genes encode full-length proteins (with the exception of MSMEG_4466, which uses an initiation codon 15 amino acids downstream of the misannotated start). This provides a cautionary note, indicating there are distinct genetic lineages among the laboratory strains of mc²155. Many of these pseudogenes are predicted to be essential in both M. smegmatis and M. tuberculosis (e.g., MSMEG_1400 encodes elongation factor G), which suggests that many of the reported sequence differences might not be correct and/or that the pseudogene-containing strains contain suppressor mutations.

Generation of a targeted knockout library.All conserved coding genes not identified as essential by Tn-seq were targeted for precise deletion by mycobacterial recombineering (28, 37). A high-throughput in vitro approach was developed to streamline the process of generating substrates for recombineering. This circumvented the need to generate cloned plasmid intermediates for each target gene, and it avoided complications presented by restriction sites present in the flanking DNA homology arms. Briefly, high-fidelity PCR was used to generate ∼300-bp flanking arms to the 5′ and 3′ side of each target gene, and these products were “sewn,” using overlap extension PCR, to a cassette encoding zeomycin-resistance (Zeo^r) flanked by loxP sites (Fig. S2). To reduce the chance of polar effects on downstream genes, the upstream and downstream flanking sequences were left intact. The loxP sites provide the option for precise excision of the Zeo^r gene by Cre recombinase, leaving behind a single loxP scar. The succession of substrate PCRs, electroporation, and colony screening PCRs were performed in matched 96-well format arrays, allowing complete sets of knockout substrates to be generated in a single day.

A total of 879 genes were subjected to gene-replacement recombineering and, of these, 569 genes were successfully targeted after two attempts (Fig. 1b and Table S1). Those genes not successfully replaced were subsequently targeted by CRISPRi.

Application of the deletion collection to identify genes required for biofilm formation.As an example of how this knockout collection can be used, we screened for mutants that are defective in biofilm formation. Strains were inoculated into liquid biofilm medium, and biofilm formation was assessed visually for up to 10 days of growth. Seventeen candidates were identified from an initial library of 468 deletion mutants. Of these, 9 had a reproducible, altered biofilm phenotype on subsequent screens (Table 1). The mutant phenotypes had one of three morphotypes (Fig. 2a): (i) hyper-pellicle formation (e.g., ΔMSMEG_5256), (ii) smooth pellicle with no or little architecture (ΔMSMEG_1824), or (iii) no biofilm (ΔMSMEG_4323). In addition to the biofilm morphotype, the colony morphology of each of the mutants was determined (Fig. 2b). Again, distinct and reproducible changes in colony morphology were observed, although two strains (ΔMSMEG_2760 and ΔMSMEG_5487) retained the wild-type colony morphotype (Table 1). Thus, through this simple screen we identified 9 genes that alter biofilm/colony morphology by either directly or indirectly altering the cell wall composition. As an independent validation of this screen, mutants of three of the genes had been previously associated with an altered biofilm. A transposon insertion in MSMEG_4323 was identified because it caused an unusual colony morphology; this mutant was subsequently shown to form defective biofilms (38). MSMEG_5439 encodes resuscitation promoting factor B (rpfB); when rpfB and rpfA (a paralog) mutants are combined, the resulting strain has altered colony and biofilm morphologies (39). Deletion of the M. tuberculosis hadC gene (Rv0637; MSMEG_1342 homolog shares 68% amino acid identity) reduced biofilm thickness and resulted in the loss of extra-long mycolic acids from cell wall fractions (40). Thus, the phenotype of the orthologous M. smegmatis hadC knockout (ΔMSMEG_1342) reproduces that described in M. tuberculosis, further demonstrating the potential of this type of screen to provide insight into other mycobacterial species.

• Open in new tab
• Download powerpoint

FIG 2

Biofilm (a) and colony morphotypes (b) of the indicated deletion mutants after 7 days of biofilm formation or 4 days of colony formation.

Generation of a CRISPRi library of essential genes.Defining functions for essential genes requires the ability to either conditionally repress or express the target gene. Here, we took advantage of the CRISPRi system optimized for mycobacteria, which allows gene-specific transcriptional repression (41). A total of 843 genes were targeted for transcriptional repression by CRISPRi. These included highly conserved essential genes from the MSR list, highly conserved orthologs that are essential in M. smegmatis or M. tuberculosis (28) and are not in the MSR gene list because they are not sufficiently conserved in all five mycobacterial species, and all genes for which a knockout clone was not obtained (Tables S1 and S3). We note that during the course of this study, a similar arrayed CRISPRi library targeting 272 essential M. smegmatis genes with direct M. tuberculosis orthologs was also generated (42). The MSR collection includes plasmids for all but 14 of the 272 essential genes targeted and, thus, this collection extends that library for CRISPRi applications in mycobacteria. Each gene sequence was screened to identify the optimal protospacer adjacent motif (PAM) guide sequences (5′-AGAAW-3′). Small guide RNAs (sgRNAs) were then designed 5′ to the identified PAM.

Testing CRISPRi efficacy of essential genes by viability assessment.To test the effect of CRISPRi on cell growth, 82 plasmids from a single 96-well plate were introduced into mc²155 by electroporation and individually tested by spotting cells on rich medium (TSA) with and without anhydrotetracycline (± Atc). Using a simple endpoint assay, 25/82 of the clones were completely inhibited for growth, while another 19 exhibited reduced colony size, consistent with repression of a gene that is either essential or is required for optimal growth under that condition (data not shown). The sensitivity of the spotting assay was further increased by spotting dilutions of cells onto defined Sauton’s medium (± Atc). Serial dilutions allowed for better side by side comparisons, including the effect of gene repression on colony size and nutritional requirements (Fig. 3a). Inhibition of two genes in the panel, MSMEG_0317 and MSMEG_1019, abolished growth (columns 3 and 10), while inhibition of MSMEG_0709 (dnaK), MSMEG_0832 (def), and MSMEG_0956 (hemB) severely reduced viability (columns 6, 8, and 9). This is in agreement with Tn-seq predicted growth-defect/essentiality calls (28). In these preliminary screens, ∼60% of transformants exhibited reduced growth on Sauton’s medium, further underscoring the efficacy of the CRISPRi system. One notable benefit of CRISPRi repression is that it can target multiple gene copies. MSMEG_1019 (Fig. 3a, column 10) is in a duplicated region of M. smegmatis (MSMEG_0991-1044 and MSMEG_2285-2337) and, thus, its function cannot be determined by Tn-seq or single-gene knockout. However, the CRISPRi sgRNA also represses the second gene copy, MSMEG_2299, and indicates they encode an essential gene product. Their ortholog in M. tuberculosis, Rv3053 (nrdE), is essential (43). Indeed, we have observed similar inhibition with CRISPRi clones targeting MSMEG_1017 (nrdH) and MSMEG_1033 (nrdG) in this same region, further underscoring the potential of CRISPRi inhibition (data not shown).

• Open in new tab
• Download powerpoint

FIG 3

Growth defects caused by expression of CRISPRi sgRNAs. (a) M. smegmatis containing pJR962 derivatives expressing sgRNAs targeting individual genes were grown to stationary phase in TSB in the absence of Atc. The cells were serially diluted 10-fold and 5 μl of each dilution was spotted onto Sauton’s medium without (left) or with (right) Atc, which induces expression of the CRISPRi system. The strain in column 5 is the control, wild-type mc²155, not containing a pJR962 clone. Genes targeted in each column are from left to right: 1-MSMEG_0029, 2-0244, 3-0317, 4-0482, 6-0709, 7-0789, 8-0832, 9-0956, 10-1019, 11-1066 and 12-1214. (b) Three clones (dnaN, gyrA, and trpG) were selected for further analysis. Cultures were diluted and spotted onto TSA or Sauton’s medium with or without Atc and growth was compared. The effect of repressing trpG can be seen in the more minimal Sauton’s medium. (c) Growth curves of cultures grown in TSB with or without Atc provide a more quantitative assessment on overall growth. MKD10 is the mc²155 parental control containing the empty vector pJR962. All cultures were diluted 100-fold before addition (or not) of Atc; the first readable data point was following overnight growth at 16.5 h.

Three gene targets were selected for further analysis: dnaN, gyrA, and trpG (MSMEG_0001, MSMEG_0006, and MSMEG_0029, respectively). A dilution series on solid rich medium indicated that dnaN and gyrA were essential for growth, while trpG was not essential in rich medium (± Atc, Fig. 3b). The trpG-targeted strain was shown to be slightly growth-inhibited on Sauton’s medium with Atc, indicating the condition-specific function of this gene (Fig. 3b). Growth curves of strains containing these three clones in liquid medium compared to wild-type mc²155 also showed that the expression of the CRISPRi clone targeted to dnaN and gyrA suppressed growth (Fig. 3c). The liquid growth assay provides a more quantitative view of suppression and identifies the window in which the repressed gene has the largest impact on relative growth rates (±Atc). We expanded the growth assays to a 96-well plate format and found that the relative growth rates (± Atc) were largely consistent with the spot endpoint assay (Fig. S3). However, observed growth inhibition differences dependent on the time of incubation with Atc suggest that the effects of CRISPRi-mediated suppression of some genes will be growth-rate dependent, presumably reflecting protein turnover and dilution following cell division. We envisage the 96-well assay providing a high-throughput format that can use different media and different cell stresses to reveal conditional sensitivities caused by induced repression of the target gene. Moreover, the plasmid resource can be introduced into other M. smegmatis derivatives, such as a specific gene knockout strain, to identify synthetic genetic connections obscured by redundant or compensatory pathways in a wild-type environment.

While we have not independently validated all 843 CRISPRi clones for their ability to silence genes, our preliminary data indicate that ∼60% will prevent or retard growth in standard culturing conditions. The lack of growth inhibition with some of the clones could be due to multiple reasons, including that knockdowns reduce target protein production as opposed to elimination by gene knockouts, that the gene targeted is essential only under specific conditions not tested here, that the gene was originally misassigned as essential, or that the gene was targeted by CRISPRi because it was refractory to conventional knockout. We used quantitative reverse transcriptase PCR (qRT-PCR) to verify CRISPRi inhibition in a subset of targeted genes (Fig. S4). We consistently saw a reduction of steady-state levels of mRNAs for the targeted genes upon CRISPRi induction, showing the CRISPRi system was functioning as intended, even in strains that were not growth inhibited. The qRT-PCR assay may be useful for individuals to assess CRISPRi efficacy provided that their gene of interest is reasonably well transcribed and does not have overlapping antisense transcripts, which the primers will also amplify. Culture conditions, including medium composition, nutrient restriction, antibiotic exposure, temperature, or many other extrinsic parameters may reveal genetic dependencies not observed in our cursory test.

Generation of a plasmid library of gene fusions for protein localization studies.Determining the localization of a protein within a cell can provide important insights into its function, especially if colocalization occurs with proteins of known function. We therefore created a library of fusions to the fluorescent protein Dendra (a monomeric, green-to-red photoactivatable monomeric fluorescent protein [44]), in order to both image the location of individual proteins and facilitate analysis of their colocalization with other core proteins. Individual core genes were PCR amplified with a high-fidelity DNA polymerase such that the amplicon ends were compatible with ligation-independent cloning. A plasmid was optimized for high-throughput gene cloning and controlled gene expression in mycobacteria (Fig. 4a). Briefly, the plasmid has a pUC origin of replication to allow growth in E. coli, and it encodes resistance to apramycin (Apr^r). It integrates site-specifically into the M. smegmatis chromosome at the L5 phage attB site, which ensures stable maintenance and a single-gene copy (45). The cloned gene is transcribed from the characterized, constitutive mycobacterial promoter P_smyc, which includes TetR operator sites to allow the option of precise control of gene expression via the Tet repressor protein (46). In addition, the plasmid provides T7 promoter and terminator sequences that allow overexpression in E. coli or M. smegmatis strains expressing T7 RNA polymerase. Unique restriction sites allow cloning in-frame with the dendra gene, connected via a flexible poly-alanine-glycine linker. The dendra gene was codon optimized for expression in mycobacteria and includes a C-terminal FLAG tag to also allow affinity purification or coimmunoprecipitation (co-IP) to screen for protein interactions (Fig. 4A). Restriction sites flanking the dendra gene allow customization of alternative C-terminal fusion tags of an MSR plasmid construct.

• Open in new tab
• Download powerpoint

FIG 4

(a) Schematic representation of the vector used for expression of Dendra gene fusions. Digestion with AseI and HindIII facilitates InFusion cloning of open reading frame amplicons, disrupting the ATTAAT AseI site to ATTAtg to form an ATG initiation codon for ORF expression. The AAGCTT HindIII site is regenerated and replaces the native ORF stop codon to allow continued translation into the alanine/glycine linker and Dendra. (b) Examples of protein localization to poles or septa with different Dendra fusion proteins expressed in M. smegmatis. Gene numbers and name are indicated.

A total of 1,116 genes were successfully cloned as a gene fusion with dendra (Table S1). Each cloned gene was sequence-verified using plasmid-based sequencing primers flanking the cloning site to read in from the 5′ and 3′ ends of each inserted MSR gene. Sanger sequence reads average 700 nt; thus, genes greater than ∼1.4 kb in length were not completely sequenced. Only two PCR-based mutations were detected in all sequence reads, consistent with the high fidelity of the Q5 DNA polymerase (New England BioLabs [NEB]). Thus, this low mutation rate indicates that undetected mutations are extremely unlikely in the longer genes.

Visualizing Dendra fusion protein localization.While the cloned library of genes has multiple uses, we chose to use it to determine protein localization by fluorescence microscopy. We examined the protein localization of 1,043 gene fusions in M. smegmatis and acquired over 7,000 high-quality fluorescent images. Of these, 761 strains yielded over 100 high-quality segmented single cells using a customized image analysis pipeline. The images indicated that many of the proteins have reproducible, characteristic positions in the mycobacterial cell, including polar, peripolar, envelope, mid-cell, septal, and as discrete cellular foci (Fig. 4b). Each of these locations suggests functional roles consistent with subcellular structures or regional multiprotein factories. Representative images for all gene fusions are available for viewing on our bioinformatics resource web site, along with summary plots of protein localization, heat maps, and the impact of fusion expression on cell length and shape based on hundreds of images per construct (https://msrdb.org/). A more detailed description of these analyses is in preparation (Zhu, J., Wolf, I.D., Dulberger, C.L., Rubin, E.J., and Fortune, S.M.).

As expected, some patterns of protein localization were consistent with those previously described in both M. smegmatis and M. tuberculosis. For example, proteins of the ESX-1 secretion system are known to be polar localized (47, 48). A Dendra fusion of the Esx1 protein EccA was polar localized (Fig. 4b). DivIVA is an essential protein in mycobacteria, as it plays a fundamental role in recruiting enyzmes required for cell wall synthesis at the cell pole (49). Our fluorescent images confirmed the extreme polar localization of DivIVA. Cells expressing DivIVA-Dendra were also misshapen and short, consistent with a dominant negative effect of expressing a properly folded fusion protein interfering with normal cell wall synthesis (Fig. 4b). In agreement with previous studies, we identified MSMEG_4287, MSMEG_0736 (TtfA), MSMEG_2690 (FtsK), and MSMEG_0035 (FhaA) as localized to the poles and septum (50–52) (Fig. 4b, and web site).

Colocalization of proteins in a cell often indicates proteins are interacting as a part of the cellular machinery, either directly or indirectly. The availability of a library of fluorescent images offers the ability to screen for similar patterns, which would be indicative of proteins colocalizing to the same site(s) in the cell. In our initial analysis, we identified a class of ∼30 proteins that had very similar patterns, in which two patches of fluorescent protein are located close to a pole (Fig. 5, e.g., MSMEG_0876). This was especially apparent using the heat maps generated from consolidated mapping data for 50 to 100 cells. Remarkably, many of these proteins had been previously identified by mass spectrometry as components of the intracellular membrane domain (IMD) (53). The IMD is concentrated in the polar region of growing cells and includes many proteins required for synthesis of cell envelope components (e.g., PimB’, MSMEG_4253, and MurG, MSMEG_4227) (Fig. 5). Subsequent analyses have confirmed that at least three of the colocalizing proteins are associated with the IMD, while the remaining seven are still under investigation (MSMEG_0988 [MenA] and MSMEG_0972) (Fig. 5) (54 and Y. Morita personal communication.). Thus, identification of similar protein localization patterns has independently validated the association of these proteins within the IMD and has identified additional putative IMD proteins that escaped detection by mass spectrometry. Many of these proteins are essential in both M. smegmatis and M. tuberculosis; hence, they likely play similar fundamental roles in cell wall precursor synthesis in all mycobacteria.

• Open in new tab
• Download powerpoint

FIG 5

A unique class of proteins is localized to the intracellular membrane domain (IMD) (53). Each Dendra fusion protein is shown expressed in M. smegmatis and characteristically features two prominent patches at the cell pole (top panel). The heat map below visually reinforces the similarity of protein localization for each fusion protein. The heat map is a representative cell with consolidated mapping data for 50 to 100 cells. Red indicates no fluorescence, while increasing intensity and protein localization is reflected by changes from yellow to green, blue, and purple. There were ∼30 protein fusions that had similar localization patterns. Gene numbers and name (if known) are given.