Author response: Global phenotypic profiling identifies a conserved actinobacterial cofactor for a bifunctional PBP-type cell wall synthase

Article Figures and data Abstract Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Members of the Corynebacterineae suborder of Actinobacteria have a unique cell surface architecture and, unlike most well-studied bacteria, grow by tip-extension. To investigate the distinct morphogenic mechanisms shared by these organisms, we performed a genome-wide phenotypic profiling analysis using Corynebacterium glutamicum as a model. A high-density transposon mutagenized library was challenged with a panel of antibiotics and other stresses. The fitness of mutants in each gene under each condition was then assessed by transposon-sequencing. Clustering of the resulting phenotypic fingerprints revealed a role for several genes of previously unknown function in surface biogenesis. Further analysis identified CofA (Cgp_0016) as an interaction partner of the peptidoglycan synthase PBP1a that promotes its stable accumulation at sites of polar growth. The related Mycobacterium tuberculosis proteins were also found to interact, highlighting the utility of our dataset for uncovering conserved principles of morphogenesis for this clinically relevant bacterial suborder. Introduction Several medically important pathogens belong to the Corynebacterineae suborder of Actinobacteria, including Corynebacterium diphtheriae and Mycobacterium tuberculosis (Mtb). Like most bacteria, these organisms surround their cytoplasmic membrane with an essential cell wall matrix made of the heteropolymer peptidoglycan (PG). However, they uniquely modify the PG layer with an additional polysaccharide made of arabinan and galactan chains (Kieser and Rubin, 2014; Alderwick et al., 2015; Daffé and Marrakchi, 2019). This arabinogalactan (AG) layer is further modified by glycolipids called mycolic acids, forming a second membrane that is thought to function analogously to the outer membrane of Gram-negative bacteria (Kieser and Rubin, 2014; Alderwick et al., 2015; Daffé and Marrakchi, 2019). The overall envelope architecture of these bacteria is referred to as the mycolata cell envelope, and compounds that target its biogenesis are important components of current drug cocktails used to treat Mtb infections (Alderwick et al., 2015). Therefore, enhancing our understanding of the assembly mechanisms that construct the mycolata envelope has practical implications for anti-mycobacterial therapeutic discovery in addition to addressing a fundamental problem in microbiology. Like all other Actinobacteria analyzed thus far, members of the Corynebacterineae grow by inserting new envelope material at their cell poles (Flärdh, 2003; Daniel and Errington, 2003). The mechanisms that govern tip growth in these organisms are ill-defined, but the DivIVA (Wag31) protein has long been known to play a key role in the process (Flärdh, 2003; Letek et al., 2008; Nguyen et al., 2007). This protein is thought to assemble into a cytoskeletal-like matrix lining the inner face of the cytoplasmic membrane at the cell poles (Edwards and Errington, 1997; Ramamurthi and Losick, 2009; Lenarcic et al., 2009; Oliva et al., 2010). Similar to FtsZ polymers that underly the cytokinetic ring, these DivIVA assemblies are believed to function by promoting the recruitment of cell envelope synthases to the pole where they can promote surface elongation (Kang et al., 2008; Melzer et al., 2018). Indeed, both known classes of PG synthases have been found to localize to growing poles in several organisms (Valbuena et al., 2007; Sieger et al., 2013; Sieger and Bramkamp, 2014; Hett et al., 2010; Kieser et al., 2015a). These synthases include the bifunctional class A penicillin-binding proteins (aPBPs) (Sauvage et al., 2008) and the relatively recently characterized synthases composed of complexes formed between SEDS proteins and their class B PBP (bPBP) partners (Meeske et al., 2016; Rohs et al., 2018; Taguchi et al., 2019). Beyond a presumed DivIVA-requirement, it remains unclear how these PG synthases are recruited to the poles or how their activities are controlled and balanced with synthases involved in constructing the other envelope layers. Factors that mediate these important activities are likely to be encoded by genes of currently unknown function that are conserved among the Corynebacterineae. Phenotypic profiling has proven to be a useful strategy to identify phenotypes for genes of unknown function to help uncover their biological activity. The method originally took advantage of the ordered knockout collections of yeast and Escherichia coli (Nichols et al., 2011; Hillenmeyer et al., 2008). Profiles were generated by replica-plating the libraries on agar containing different drugs or other stresses and the fitness of each mutant under each condition was assessed based on measurements of colony size. Similar approaches utilizing transposon-sequencing have recently been employed to generate profiles for several bacterial species (Wetmore et al., 2015; Price et al., 2018), but an extensive analysis has not yet been carried out in the Corynebacterineae. Therefore, to better understand cell envelope assembly and polar growth in these organisms, we performed a global phenotypic profiling analysis of the model bacterium Corynebacterium glutamicum (Cglu). For the analysis, we used our recently generated high-density transposon mutant library of Cglu (Lim et al., 2019) and challenged it with a panel of antibiotics and other stresses. The fitness of mutants in each gene under each condition was then assessed by comparing the change in the proportion of mapped transposon insertions within a gene following ten generations of growth under a given condition. Clustering of the resulting phenotypic fingerprints for each gene revealed a role in surface biogenesis for several genes of previously unknown function. Further analysis of one such gene identified CofA (Cgp_0016) as a specific interaction partner of an aPBP-type PG synthase called PBP1a. CofA was shown localize to the cell pole and to be required for PBP1a to stably accumulate at these sites. Furthermore, we found that cognate CofA-PBP proteins from Mycobacterium tuberculosis and a pathogenic corynebacterium also participate in specific interactions. Thus, our overall results identify a conserved new component of the polar growth machinery within the Corynebacterineae and highlight the utility of our phenotypic profiling dataset for uncovering common principles of morphogenesis for this important group of bacteria. Results Phenotypic profiling of a high-density Cglu transposon library To generate phenotypic profiles, a Cglu transposon library of approximately 200,000 unique insertions was grown for eleven generations in the presence of drug or under a stress condition. For drug treatments, initial trials revealed that the best transposon-sequencing results were obtained when a drug concentration yielding a mild but observable decrease in growth rate was used. Typically these concentrations were 1/4 to 1/2 the measured minimal inhibitory concentration (MIC) for each drug (Figure 1—figure supplement 1A and Supplementary file 1). In several cases, two different treatment concentrations of a particular drug were used in the analysis. All total, 40 different growth conditions were analyzed with the collection of drugs used spanning all major targets, including PG and AG biogenesis, DNA replication, transcription, and translation (Figure 1A and Supplementary file 1). Figure 1 with 1 supplement see all Download asset Open asset Phenotypic profiling of a Corynebacterium glutamicum transposon mutant library. (A) Overview of the phenotypic profiling procedure. A transposon mutagenized library of Cglu MB001 was exposed to sub-MIC concentrations of the indicated antibiotics or to the listed stress condition for 11 generations prior to transposon sequencing analysis and the calculation of fitness scores for mutants in each gene under each condition. Several of the antibiotics were tested at two different concentrations such that a total of 40 different growth conditions were surveyed (see Supplementary file 1). (B) Scatterplot highlighting the reproducibility of the analysis for duplicate samples grown in the absence of drug. The calculated fitness scores for each gene in the two replicates are plotted. Scores were calculated by comparing the proportion of total transposon reads for each gene in the untreated samples grown for 11 generations relative to the reads mapped for the input library. (C) Pie chart summarizing results from the profiling analysis. Depicted are essential genes (453, gray), genes that displayed a strong phenotype (fitness value below 0.75 or above 1.25) in at least one condition (465, dark blue), genes that displayed a moderate phenotype (fitness value below 0.9 or above 1.1) in at least one condition (484, light blue) and genes that did not show a phenotype in any condition tested (1535, black). Following growth in each condition, genomic DNA was isolated from the cultures and transposon insertion profiles were analyzed by sequencing. Based on the results, a fitness score for mutants in each gene, referred for simplicity as gene fitness, was calculated by comparing the proportion of transposon reads mapped in the gene following one generation of growth in the absence of treatment relative to those mapped after eleven generations of growth in the treatment condition (see Materials and methods). Scores below 1.0 indicate reduced fitness relative to the population, whereas scores greater than 1.0 indicate greater fitness. Replicates of untreated cultures resulted in highly correlated fitness scores for each gene, with most genes yielding a fitness score near 1.0 as expected for such a comparison (Figure 1B). Overall, approximately 40% of all non-essential genes were found to have a moderate phenotype in at least one condition tested, indicated by a fitness score below 0.9 or above 1.1 (Figure 1C), with the vast majority of phenotypes observed being fitness defects. Hierarchical clustering of the phenotypic profiles generated from the analysis was used to identify sets of genes with similar fingerprints that may function together in the same biological pathway (Figure 2 and Figure 2—source data 1). As an indication that the clustering was accurately identifying factors with similar functions, genes encoding components of several characterized protein complexes were found to have highly correlated phenotypic signatures. For example, genes encoding the two cytochrome d oxidase subunits (cydA and cydB) and cytochrome transporter (cydC and cydD) cluster tightly together, primarily due to their hypersensitivity to the benzothiazinone BTZ043 (Figure 2). This finding is consistent with published literature demonstrating synergistic effects between electron transport chain inhibitors and benzothiazinones against Mtb (Lechartier and Cole, 2015). Additionally, genes encoding RipC, FtsE, and FtsX proteins that together form a complex required for proper cell wall remodeling at the division site (Lim et al., 2019; Tsuge et al., 2008; Maeda et al., 2016) were also found to have correlated profiles (Figure 2). Figure 2 with 1 supplement see all Download asset Open asset Clustering analysis of phenotypic profiles identifies genes with related functions. (A) Heatmap showing clustered phenotypic profiles for all non-essential genes in Cglu. The test conditions are oriented along the abscissa and are ordered by the stress or the physiological process affected by antibiotic treatment. The 2488 non-essential genes (excluding tRNAs, rRNAs, and transposons) are clustered along the ordinate with neighboring genes sharing similar fitness profiles across all conditions. The intensity of the red color indicates the magnitude of the fitness defect (dark red has a fitness value close to 0), white indicates a fitness value of 1, and blue color indicates a fitness advantages in a given condition. The full dataset is available in Figure 2—source data 1. (B) Expanded view of the profiles for select genes. See text for details. Figure 2—source data 1 Phenotypic profiling results. The gene fitness for each gene under each growth condition is listed. https://cdn.elifesciences.org/articles/54761/elife-54761-fig2-data1-v2.xlsx Download elife-54761-fig2-data1-v2.xlsx Another notable feature of the profiling analysis is that it successfully differentiated the biological function of genes from the same genetic locus. For example, genes in the putative cgp_3163–3168 operon clustered into two distinct groups. Mutants in four of these genes (cgp_3163, cgp_3165, cgp_3166, and cgp_3168) have been associated with defects in trehalose mycolate transport across the cytoplasmic membrane (Yamaryo-Botte et al., 2015). Accordingly, they were all found to cluster together in the profiling analysis (Figure 2). Gene cgp_3164 from this locus, on the other hand, was previously implicated in a different function. Mutants in this gene along with the unlinked mptA (cgp_2385) gene were found to have a defect in the elongation of the mannan backbone of membrane glycolipids (Cashmore et al., 2017). Consistent with this finding, cgp_3164 clusters with mptA (cgp_2385) in our analysis and has a phenotypic profile distinct from its neighboring genes in the cgp_3163–3168 locus (Figure 2). Therefore, the profiling and clustering results are not simply identifying functional groupings based on operonic organization. In addition to properly correlating the function of factors known to work together within the same complex or pathway, the profiling data also identified possible roles in cell surface biogenesis for genes of previously unknown function. For example, the putative cgp_3012–3020 operon stood out due to the specific hypersensitivity to bacitracin and vancomycin displayed by mutants in all of its genes except for cgp_3015 (Figure 2—figure supplement 1A–B). This hypersensitivity was confirmed by deletion of the entire locus or individual genes (cgp_3018 or cgp_3019) (Figure 2—figure supplement 1C). Also consistent with the phenotypic screen, these strains have a slight growth advantage on meropenem, but are not altered in their susceptibility to other drugs like ampicillin (Figure 2—figure supplement 1B–C). Vancomycin acts by directly binding the lipid II precursor for PG synthesis, which consists of the lipid carrier undecaprenol pyrophosphate (Und-PP) linked with the dissacharide pentapeptide monomeric unit of PG (Reynolds, 1989; Schneider and Sahl, 2010). Bacitracin similarly targets the lipid stage of PG biogenesis by blocking the dephosphorylation of Und-PP products of PG glycan polymerases such that Und-P is not regenerated for use in the synthesis of lipid II, causing PG synthesis to be inhibited (Schneider and Sahl, 2010). Thus, the specific hypersensitivity of mutants in the cgp_3012–3020 locus to drugs that interfere with lipid II biogenesis suggests that this large uncharacterized cluster of genes may be involved in undecaprenol synthesis or utilization to facilitate proper cell surface assembly in Cglu. Based on the overall accuracy of the functional connections made so far using the phenotypic profiling dataset, we anticipate that it will provide a useful resource for the discovery and characterization of new factors involved in cell surface biogenesis and other important biological processes within the Corynebacterineae. The profiles of cgp_0016 and ponA (cpg_0336) are highly correlated We continued mining the profiling data to identify genes of unknown function that had profiles that were highly correlated with those of factors known to play key roles in cell wall synthesis. This analysis peaked our interest in the gene of unknown function cgp_0016, which based on results presented below will henceforth be referred to as cofA (co-factor of PBP1a) (Figure 3A). The cofA gene encodes a small membrane protein of 114 amino acids with two predicted transmembrane domains and an N-in/C-in topology (Figure 3B). Its phenotypic profile clustered tightly with the ponA gene (Figure 3A) (correlation coefficient 0.91) encoding PBP1a, an aPBP-type synthase that helps build the PG layer. Like other aPBPs, PBP1a is a bitopic membrane protein with a small N-terminal cytoplasmic domain, a single transmembrane helix, and a large extracellular region (Figure 3B). The extracellular portion of the synthase contains both a glycosyltransferase (GTase, polymerase) domain that polymerizes lipid II into PG glycans and a transpeptidase (TPase) domain that forms the crosslinks between peptides of adjacent PG glycans in the wall matrix (Figure 3B). Cglu encodes a second aPBP called PBP1b, but its corresponding gene ponB (cgp_3313) had a phenotypic profile that was distinct from cofA and ponA (Figure 3A). Figure 3 with 1 supplement see all Download asset Open asset CofA is required for PBP1a accumulation. (A) Phenotypic profiles of cofA (cgp_0016), ponA (cgp_0336, encoding PBP1a), and ponB (cgp_3313, encoding PBP1b) displayed as in Figure 2. Note that cofA and ponA clustered tightly together in the analysis due to their similar profiles. Neither gene clustered near ponB, which is shown for reference. (B) Schematic showing the predicted membrane topology of CofA and PBP1a. (C) Cultures of wild-type Cglu and the indicated deletion mutants were grown, serially diluted, and plated as in Figure 2—figure supplement 1. The concentration of drugs used was 0.2 μg/mL ampicillin or 0.04 μg/mL meropenem as indicated. (D) Shown are mScarlet fluorescence (upper) and phase contrast (lower) micrographs of cells expressing the indicated fusion protein. Fusions were constitutively expressed from a construct integrated at the attB1 site. Translation of the fusions was controlled by the theophylline (riboE1) riboswitch and was induced with 0.3 mM theophylline in each case. The fusions were produced in strains deleted for the corresponding native untagged protein. Cells from an overnight culture were diluted 1:1000 in BHI supplemented with 0.3 mM theophylline and then imaged on CGX2 agarose pads after growth for 5.5 hr at 30°C. The brightness for the two mScar-PBP1a micrographs is normalized to allow for direct comparison. Bar equals 3 µm. (E) Bocillin labeling of PBPs in wild-type and mutant strains. Overnight cultures of the indicated strains were diluted 1:200 in BHI and grown until they reached an OD600 = 0.3. Cells were then treated with 10 μg/mL Bocillin-FL, and membrane fractions were isolated. Proteins (5 µg total) were then separated on a 10% SDS-PAGE gel and labeled bands were visualized using a Typhoon florescence scanner. Production of the mScar-PBP1a fusions was induced with 0.3 mM theophylline as for the microscopy analysis in panel D. Fluorescent band intensities for labeled PBP1a or mScar-PBP1a were quantified and normalized to the PBP2a band signal running just above 55 kDa. The PBP1a or mScar-PBP1a signal decreased by a factor of 5 in ΔcofA cells relative to the corresponding CofA+ strain. The high correlation of the ponA and cofA profiles suggested that CofA might be a cofactor specifically required for the function of PBP1a. Cofactors of aPBPs have been described in the Proteobacteria and Firmicutes (Typas et al., 2010; Paradis-Bleau et al., 2010; Greene et al., 2018; Fenton et al., 2018). However, no such cofactors have been identified as important for PG synthesis in the Corynebacterineae or other Actinobacteria. Given the critical roles played by these cofactors in the activation and/or control of PG synthesis by the aPBPs in other organisms, we decided to further investigate the connection between CofA and PBP1a in Cglu. To validate the profiling results, deletions of ponA and cofA were and the of the resulting mutants to antibiotics was As cells displayed hypersensitivity to ampicillin and relative to wild-type cells (Figure Cells deleted for cofA a similar to these but in both the defect was than that of cells (Figure ΔcofA mutants displayed a drug phenotype to the single mutant (Figure Thus, the mutant phenotypes of cells for PBP1a and CofA are However, the reduced of the phenotypes for CofA relative to PBP1a suggests that PBP1a function in the absence of This phenotypic is consistent with a model in which CofA promotes PBP1a function but is not required for its activity. CofA is required for the stable accumulation of PBP1a PBP1a has been shown to localize to the cell pole where it is likely in polar surface growth (Valbuena et al., 2007). We that one function of CofA might be the recruitment of PBP1a to the growing To investigate this we N-terminal fusions of the protein mScarlet to both PBP1a and fusions were functional as they the ampicillin hypersensitivity phenotype of the corresponding deletion (Figure supplement As expected based on prior results, mScar-PBP1a displayed a polar with additional signal at division (Figure and Figure supplement This did not PBP1a as mutant fusions for or both activities their polar recruitment signal (Figure supplement 1B–C). a similar (Figure and Figure supplement the polar of was in cells (Figure and Figure supplement The fusion displayed a signal of a membrane protein. In when mScar-PBP1a was assessed in ΔcofA the signal observed was reduced relative to cells CofA (Figure and Figure supplement This suggested that PBP1a may not stably accumulate in the absence of To investigate this we used the called Bocillin to PBP1a in wild-type and ΔcofA Like all Bocillin the site of can be used PBPs to their in membrane following protein and by of wild-type cells bands of (Figure The band was identified as PBP1a due to its absence in the profile of labeled cells (Figure As expected from the microscopy analysis, the intensity of the PBP1a band was reduced in samples from ΔcofA cells (Figure when cells expressing the mScar-PBP1a fusion were the PBP1a band was to a corresponding to the and this band was reduced in intensity when CofA was (Figure Given that the fusion protein was produced from a the of CofA is to be at the of ponA we from the data that CofA is most likely required for the stable accumulation of PBP1a. CofA directly with the transmembrane domain of PBP1a The accumulation defect of PBP1a in cells CofA suggested that the two proteins might and that this interaction is required to PBP1a. To their we used the recently in coli (Lim and 2019). For this a protein is to and a that the fusion to assemblies of the protein. The protein is expressed as an and interaction with the is assessed based on or not the is recruited to the polar this CofA was found to with PBP1a but not a control transmembrane domain fusion or a fusion to PBP1b, the other aPBP encoded by Cglu (Figure Additionally, CofA was found to (Figure we also observed that the transmembrane domain of PBP1a was and for a interaction in the (Figure We thus that CofA directly and specifically with PBP1a its transmembrane Figure Download asset Open asset CofA specifically with PBP1a. Shown are results from the with proteins expressed in coli proteins were with and the to target to polar assemblies of the protein. proteins were expressed as of the proteins or protein domains used as or are shown above the micrographs for reference. In each fluorescence of coli cells expressing the indicated and proteins are these are that protein a of For the the of fluorescence across at least cells was The resulting of fluorescence intensity for each cell were then to cell and to generate the A was used to the cells such that the cell pole with the fluorescence was on the of the at at (A) with transmembrane domain to as with with (B) with (C) to the domain of PBP1a with with a transmembrane domain with The interaction is required for the stable accumulation of PBP1a at the pole We tested the transmembrane domain of PBP1a was for its to accumulate in the absence of To we the native PBP1a transmembrane domain with the corresponding domain from coli PBP1a in the of the Bocillin labeling indicated that unlike wild-type this fusion protein in ΔcofA cells to that of the wild-type mScar-PBP1a fusion in CofA+ cells (Figure accumulation of PBP1a was by the transmembrane domain the fusion was functional in of to ampicillin to or ΔcofA cells (Figure the fusion was not recruited to the cell poles like wild-type but displayed a consistent with a membrane protein (Figure by the also found that the to with CofA (Figure supplement 1). The overall results with the PBP1a fusion are consistent with a model in which the interaction is for the stable accumulation of the complex at the cell Figure 5 with 1 supplement see all Download asset Open asset with CofA is required for polar of PBP1a. (A) Bocillin labeling of mScar-PBP1a in the indicated strains. Production of the fusions was induced with 0.3 mM theophylline as in Figure For this of total protein was for each to the PBP1a in which the transmembrane domain of coli PBP1a was used to the corresponding domain of native Cglu PBP1a. Fluorescent band intensities for labeled mScar-PBP1a was performed as in Figure The mScar-PBP1a band decreased in intensity by a factor of in ΔcofA cells relative to the corresponding CofA+ whereas the in ΔcofA cells was at of the mScar-PBP1a in CofA+ (B) Cultures of the indicated strains encoding or no fusion as indicated were grown and plated as in Figure 2—figure supplement 1. 0.3 μg/mL ampicillin with or 0.3 mM theophylline to the of the PBP1a fusions as indicated. (C) Shown are micrographs of or ΔcofA cells the indicated mScar-PBP1a or which was induced addition of 0.3 mM Cells were imaged on CGX2 with agarose showing fluorescence of the corresponding fusions a of least cells were analyzed for each Bar equals 3 µm. The interaction is conserved among the Corynebacterineae The CofA protein from including its two transmembrane consists of the domain of unknown function Proteins with this domain are found the Actinobacteria, and as in the cofA is in the same genetic directly of DNA and of (Figure