M. bovis AF2122/97 was maintained on modified Middlebrook 7H11 solid medium containing 0.5% lysed defibrinated sheep blood, 10% heat inactivated foetal bovine serum and 10% oleic acid-albumin-dextrose-catalase (OADC) (27). Liquid cultures of M. bovis were grown in Middlebrook 7H9 medium containing 75 mM sodium pyruvate, 0.05% Tween^®80 and 10% albumin-dextrose-catalase (ADC). M. tuberculosis H37Rv and Mycobacterium smegmatis mc²155 were maintained on Middlebrook 7H11 solid medium supplemented with 0.5% glycerol and 10% OADC. Liquid cultures were grown in Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.05% Tween^®80 and 10% ADC unless stated otherwise. MycomarT7 Phagemid was propagated on M. smegmatis mc²155 lawns grown on Middlebrook 7H10 solid medium supplemented with 0.5% glycerol and 10% OADC in a 0.6% agar overlay. The strains and plasmids used or made in this study are given in Table 1.

Transposon libraries in M. bovis (AF2122/97) and M. tuberculosis (H37Rv) were generated using the previously described MycomarT7 phagemid system as per Majumdar et al. with modifications (29). Briefly, 50 ml cultures of M. bovis and M. tuberculosis at OD₆₀₀≅1 were washed twice with MP buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM MgSO₄ and 2 mM CaCl₂) at 37°C, and then incubated with ~10¹¹ pfu of ϕMycoMarT7 phage for 16–18 h at 37°C without rolling. Transduced bacteria were washed in pre-warmed PBS + 0.05% Tween^®80 to remove extra-cellular phage and plated on Middlebrook 7H11 solid medium containing 0.5% lysed defibrinated sheep blood, 10% heat inactivated foetal bovine serum, 10% OADC, 25 μg/ml kanamycin and 0.05% Tween^®80. Cultures were allowed to grow for 5–6 weeks. Concurrent CFU plating was performed to estimate transduction efficiency. Approximately 15–20 colonies from each library were used for validation of random insertion using a nested PCR strategy followed by Sanger sequencing, method and data are shown in Supplementary Material S1. Libraries were scraped from the plates and incubated in liquid medium at 37°C with hourly vortexing for 3 h to homogenise. Homogenised mutants were distributed to cryovials and stored at −80°C for further selection or gDNA extraction.

Unless stated otherwise, reagents were acquired from Sigma Aldrich. Genomic DNA from harvested libraries was isolated by a bead beating procedure (mechanical lysis) or using de-lipidation followed by enzymatic lysis as previously described by Long et al. (30) and Belisle et al. (31). Briefly, for mechanical lysis, library aliquots were disrupted using 0.1 mm glass beads and bead-beating by 3 × 15 s bursts (5000 rpm) interspersed with 2 mins on ice using a beat-beater (Biospec). For enzymatic lysis, libraries were de-lipidated with equal volumes chloroform-methanol (2:1) for 1 h with agitation every 15 mins, suspension was centrifuged at 3,488 × g for 10 mins the bacterial pellet allowed to dry for 2 h after removal of both solvent layers. De-lipidated bacteria were suspended in TE buffer and incubated with 100 μg/ml lysozyme in the presence of 100 mM TrisBase (pH 9.0) at 37°C for 12–16 h. Bacterial lysis was completed by incubating for 3 h at 55°C in the presence of 1% SDS and 100 μg/ml proteinase K (NEB). Lysates from both methods were extracted twice with equal volumes of phenol–chloroform–isoamyl alcohol (25:24:1). The aqueous layer was harvested by centrifugation at 12,000 × g for 30 mins and DNA was precipitated with 0.1 volumes of 3M sodium acetate (pH 5.2) and one volume of ice-cold isopropanol overnight at −20°C. DNA pellets were washed several times in ethanol. DNA was re-suspended in water and quantity and quality were determined using a DeNovix Spectrophotometer (DeNovix Inc, USA), agarose gel electrophoresis and fluorometry using Qubit4 (Invitrogen).

Two μg of extracted DNA libraries were resuspended in purified water and sheared to approximately 550 bp fragments using a S220 focussed-ultrasonicator (Covaris), according to the manufacturer's protocol. Sheared DNA was repaired using NEBNext blunt-end repair kit (New England Biolabs) and purified using Monarch PCR clean-up kit (New England Biolabs). Blunted DNA was A-tailed using NEBNext dA-tailing kit and column-purified. Custom transposon sequencing adaptors, or “TraDIS tags,” (Table 2) were generated by heating an equimolar mix of adaptor standard primer and adaptor P7+index to 95°C for 7 mins and then allowed to cool to room temperature. Adaptors were ligated to A-tailed library fragments using NEBNext quick ligase kit. Transposon-containing fragments were enriched by PCR using ComP7 primers ComP5 using Phusion DNA polymerase (New England Biolabs) in a 20-cycle reaction. Library fragments were subsequently cleaned up with AMPureXP purification beads (Beckman).

Indexed libraries were combined, spiked with 20% PhiX, and sequenced on the Illumina HiSeq 3000 platform, using v2 chemistry, generating single-end reads of 150 bp. Raw FASTQ sequence files were pre-processed using the TPP utility of TRANSIT python package (32) including removing TRADIS tags and adapter sequences and mapping using BWA-MEM algorithm [32], to generate insertion files in.wig format. Custom annotations, “prot tables,” were created from the M. bovis strain AF2122/97 annotation file (NCBI Accession Number LT708304, version LT708304.1) and for the M. tuberculosis strain, H37Rv [NCBI Accession Number AL123456, version AL123456.3, assembly build GCA_000195955.2 (ENA)]. TRANSIT software was run on both M.bovis and M. tuberculosis files using the default normalisation (TTR), which normalises by trimming the top and bottom 5% of read counts and normalising to the mean read count. The TRANSIT HMM algorithm (33) was used to make calls of essentiality for each TA insertion site, and for each gene based on annotated gene boundaries. Data files (fastq) are deposited in SRA (PRJNA754037).

We utilised dCas9 from Streptococcus pyogenes (dCas9_Spy) for silencing as previously described (28). sgRNA targeting wag31_Mtb/Mb and Rv2182c/Mb2204c were designed according to the parameters derived from Larson et al. (34). Protospacer adjacent motif (PAM) sites, “NGG,” were chosen and putative sgRNAs 20 bp downstream of the PAM were selected. All sgRNAs designed targeted the coding non-template strand. The probability of complementarity to any other region of the genome and predicted secondary structure of the sgRNA transcript was analysed using a basic local alignment search tool (BLAST) and M-fold, respectively (35, 36). Complementary forward and reverse primers using the sequence (without the PAM) with appropriate ends for ligation to the pRH2521 vector were designed (Table 2). Oligos were annealed and cloned into pRH2521 using BbsI as previously described (28, 37). One microgram of pRH2502 was electroporated at 25 kV, 25 μF with 1000 Ω resistance into electrocompetent M. bovis and M. tuberculosis to generate strains expressing dCas9_Spy (Mtb__dCas9/Mb__dCas9). These strains were grown and further electroporated with 1 μg of pRH2521 expressing sgRNAs targeting wag31_Mtb/Mb and Rv2182c/Mb2204c or pRH2521, the sgRNA –ve plasmid.

Cultures were grown to OD₆₀₀ ≅ 0.1–0.2 and the CRISPRi machinery induced with 200 ng/ml of anhydrotetrcycline (aTc) for 24 h. Total RNA was extracted as previously described (38). Briefly, cultures were centrifuged at 3,488 × g at 4°C for 10 mins. Pellets were resuspended in 1 ml of TRIzol containing 0.1 mm glass beads and were disrupted by three cycles of 30 s pulses at 6,000 rpm using a Precellys homogenizer. RNA was purified using a Qiagen RNeasy kit combined with on-column DNase digestion according to the manufacturer's instructions. Quantity and quality were determined using a DeNovix Spectrophotometer (DeNovix Inc, USA) and agarose gel electrophoresis.

To remove traces of contaminating DNA, RNA samples were treated with RNase-free DNase I (Invitrogen) according to the manufacturer's instructions. cDNA was synthesised from 100 ng of RNA using Superscript III Reverse transcriptase according to manufacturer instructions. qPCRs were performed using PowerUp SYBR Green Master Mix with 1 μl of cDNA and 0.3 μM of either sigA primers or gene specific primers (Table 2) in a final volume of 20 μl. Samples were run on a BioRad CFX96 analyser at 50°C for 2 mins, 95°C for 2 mins, followed by 40 cycles of 50°C for 2 mins, 95°C for 2 mins, followed by 40 cycles of 95°C for 15 s, 72°C for 1 min and 85°C for 5 s at which point fluorescence was captured. A melt curve analysis was also carried out for each run at 65–95°C in increments of 0.5°C. Gene expression data was analysed using the 2^−ΔΔCT method (39). Reverse transcriptase –ve samples were used as a control to ensure removal of gDNA. All results were normalised against the house keeping gene sigA. Two or three biological replicates were run, with each measured in duplicate, unless otherwise stated.

The Mycomar transposon inserts randomly into TA sites in bacterial genomes (40). There are 73,536 and 74,604 TA sites present in the M. bovis (AF2122/97) and M. tuberculosis (H37Rv) genomes, respectively. The smaller number of TA sites in M. bovis is likely to be reflective of a smaller genome. We successfully generated transposon libraries in M. bovis and M. tuberculosis containing 39,987 (M. bovis) and 29,919 (M. tuberculosis) unique mutants, representing 54 % (M. bovis) and 40 % (M. tuberculosis) saturation (Table 3). This corresponded to an insertion in 3,625/3,989 (91%) coding sequences in M. bovis and 3,554/4,018 (86%) coding sequences for M. tuberculosis. The distribution of transposon insertions in the two species is shown in Figure 1.

Himar1 transposase has been previously suggested to exhibit local sequence preferences rendering ~9% of possible TA sites non-permissive to insertion (23) and others have also observed TA insertion cold spots within the M. tuberculosis genome. Using the non-permissive sequence pattern, “SGNTANCS” (where S is either G/C), we identified 6657 non-permissive sites in both M. bovis and M. tuberculosis genomes (data not shown). Taking a similar approach to Carey et al., we found that removing these sites prior to determining gene essentiality as described below did not affect the gene calls (25).

We examined in vitro gene essentiality in M. bovis and M. tuberculosis using the TRANSIT HMM method (33). This approach classifies genes into four categories; those that are essential for growth and cannot sustain a transposon insertion (ES), those where the transposon insertion results in a growth defect (GD) and those where the transposon insertion results in a growth advantage (GA). Those that show no impact as a result of the transposon insertion are considered non-essential (NE). From this analysis, 527 genes were classified as ES (15.3%), 176 genes were classified as GD (5.1%) and 131 as GA (3.8%) in the M. bovis genome. In M. tuberculosis 477 genes were classified as ES (13.7%), 179 genes were classified as GD (5.1%) and 1 gene as GA (0.03%). A complete list of calls for the genes that are conserved between both species is given in Supplementary Table S1. The status of the genes that are M. bovis specific are also included in the table.

Early sequencing and functional annotation of the genome of M. tuberculosis categorised genes into several different functional classes with an uneven distribution of genes across the classes (4, 5). We examined the distribution of the genes classified as ES in M. tuberculosis (477) and M. bovis (527) across the functional classes to determine if: (i) ES genes are over-represented in any particular functional class when compared to the genome as a whole; (ii) there are differences between the two species. The results are shown in Table 4. Chi squared testing showed that the distribution of ES genes across the functional classes was significantly different to the distribution of all orthologues (p ≤ 0.01). ES genes in both species are over-represented in “information pathways” and “intermediary metabolism and respiration” and under-represented in “conserved hypotheticals” and “PE/PPE” functional classes. Our data are in line with previous reports; Griffin et al. noted that the distribution of ES genes across the different functional classes were different compared to the genome as a whole (22). DeJesus et al. also noted that insertions in PE/PPE genes were under-represented likely due to GC rich sequences and an increased proportion of non-permissive TA sites in the PE/PPE genes (23). There were no major differences in distribution of ES genes across the functional classes when M. tuberculosis and M. bovis were compared with each other except for “insertion sequences and phages” which did not contain any genes classified as ES in the M. bovis genome.

Genes categorised as ES in this study were compared between the two species and also compared to previously reported studies (21–23, 26, 42) (Supplementary Table S1). We found that the M. bovis dataset generated in our study shared 370 (70%) of genes classified as ES with M. tuberculosis in vitro (this study; Figure 2A) and up to 86% overlap with three key published M. tuberculosis data sets: DeJesus et al. (23) (71%), Griffin et al. (86%) and Minato et al. (79%) indicating good correlation with previous reports (Figure 2D). Similarly, the M. tuberculosis dataset generated in our study shared good overlap with other published datasets (Figure 2C). When comparing M. bovis genes classified as ES with those reported by Butler et al. (26) we found that 220 (42%) genes were shared between these data sets (Figure 2B). Butler et al. reported a total of 318 genes to be essential in M. bovis in vitro prior to selection in Dictyostelium discoideum compared to 527 reported in this study. Both libraries showed similar saturation levels (58 vs. 54% in this study) and use similar analysis methods, therefore differences might be due to the conditions under which the libraries were generated (although both studies used Middlebrook 7H11 solid medium supplemented with lysed sheep blood, heat inactivated foetal bovine serum and OADC) or between laboratory variation as might be expected for whole genome techniques such as Tnseq. It should also be noted that the similarities between the studies increases when GD genes are considered, for instance of the 307 genes that appear to be uniquely ES in our study, 212 of these are classified as GD in the study by Butler et al., indicating a debilitating impact of the transposon insertion.

Genes uniquely classified as ES in either species are of interest to determine potential genetic insights for phenotypic differences between these closely related mycobacterial species. In this study 157 genes were uniquely ES in M. bovis when compared to the M. tuberculosis (Figure 2A), however, of these 157, 61 were classified as GD in M. tuberculosis. The remaining 96 were classified as NE in M. tuberculosis (Supplementary Table S2). The existence of multiple datasets allows for a robust meta-analysis and so we compared across datasets and found that there were 42 genes that were either ES or GD in this study and the study by Butler et al., and were classified as NE in M. tuberculosis in this study and the study by DeJesus et al. (Supplementary Table S3). Included in this subset of genes is Rv3543c (fadE19), Rv3541c and Rv3540c (lpt2), genes which are encoded on the same operon (Rv3545c-Rv3540c—based on intergenic gaps) regulated by kstR and involved in cholesterol catabolism. This study and the Butler et al., study indicates that insertional mutagenesis of this operon has a debilitating impact in M. bovis but not in M. tuberculosis.

Data for the entire kstR regulon is given in Supplementary Table S4. Interestingly, the media used in this study and the study by Butler et al., contains traces of cholesterol due to the presence of lysed sheep blood, although there is no evidence that cholesterol presented a selective pressure (for M. tuberculosis) in this study as there is little overlap of the M. tuberculosis dataset with the study by Griffin et al., In addition to the Rv3545c-Rv3540c operon considered above, several orthologues in the kstR regulon were classified as ES in M. bovis; Mb3538 (Rv3508), Mb3568 (Rv3538) and Mb3581 (Rv3551), and Mb3595 (Rv3565). Others such as Mb3541 (Rv3511) and Mb3574c (Rv3544c) were classified as GD. Interestingly insertions in the genes belonging to the mce4 operon and required for growth on cholesterol mostly confer a growth advantage for M. bovis. These observations might reflect a difference in the requirement for cholesterol catabolism in vitro in a complex carbon mixture compared to M. tuberculosis.

One of the key metabolic differences between M. bovis and M. tuberculosis is the inability of M. bovis to utilise carbohydrates. Genes in the glycolytic pathway (Supplementary Table S5) such as, enolase (eno), pyruvate kinase (pykA) and pyruvate carboxylase (pca) might be expected to be NE in M. bovis as pykA is non-functional in M. bovis (43). The datasets show that eno is ES in M. bovis as well as M. tuberculosis perhaps indicating that its essentiality is linked to a role other than glycolysis. Similarly, the suggestion that a transposon insertion in pykA confers a growth advantage (this study only) is counter-intuitive and might suggest a non-glycolytic role for this enzyme. Only our dataset suggests that a transposon insertion in icl1, an enzyme required for growth on fatty acids, confers a growth advantage in M. bovis.

The two-component system PhoPR has been shown to control the biosynthesis of sulfolipid (SL-1) and di- and poly-acyltrehaloses (DAT and PAT) and also secretion of ESAT-6 by regulation of the espACD gene cluster (44, 45). It is of particular interest because a non-synonymous SNP in the sensor histidine kinase phoR in M. bovis renders signalling through the system defective and M. bovis lacks SL-1, however, the existence of compensatory mutations that restore ESAT-6 production obscures the role of the regulon in M. bovis (46). Additionally, it is known that genes associated with the synthesis of PDIM are over-expressed in M. bovis AF2122/97 (11) although the mechanism by which this occurs is not entirely known. There is some evidence that PhoPR indirectly controls the expression of PDIMs (16). Of the genes in the PhoPR regulon (Supplementary Table S6) only Rv3778c seems to be consistently required across species and studies. Genes in the redox sensing WhiB family are included in the regulon (whiB1, whiB3 and whiB6) but only whiB1 is ES in M. bovis in our study.

Genes encoding for the transport of inorganic sulphate; subI, cysW, cysW and cysA, were classified as ES in M. bovis in this study. These genes are involved in reductive sulphur assimilation and are conserved across the MTBC (47). This pathway was found to be essential in vitro for BCG Pasteur, but not BCG Danish, when grown on the same media (48). In our study, they were found to be essential in vitro for M. bovis but this was not corroborated in the study by Butler et al., Of the studies with M. tuberculosis, these genes have been found to be essential in vitro in some studies (22, 23) but not others (21). These discrepancies may reflect differences in the ability of the species (or strains) to acquire sulphate from organic sources such as methionine, different affinities of the transporter between species (or strains) or alternatively may reflect differences in the oxidative stress levels of the environment under which the screens are performed. Sulphated metabolites, such as mycothiol play a key role in the protection against oxidative stress (49). The metabolism of fatty acids can increase the cytoplasmic pool of reducing equivalents leading to a redox imbalance, therefore central carbon metabolism might also play a role in the differences observed across the studies.

Finally, as the electron transport chain and ATP synthesis is a relatively new therapeutic pathway, we chose to examine ES more closely in these pathways (Supplementary Table S7). These pathways are targets of recently introduced drugs such as bedaquiline (ATP synthase) and those in development e.g., Q203 which targets the terminal cytochrome bc₁-aa₃ oxidase (50). Unsurprisingly, the genes encoding the ATP synthase are largely ES in both species in all studies (Rv1304-Rv1311) with the exception of Rv1304 (atpB). The genes that encode a sub-unit of the terminal cytochrome bc1-aa3 oxidase complex (qcrCAB) the target of Q203 are classified as either ES or GD. One interesting observation is that both our study and the study by Butler et al., suggests that a growth defect occurs as a result of an insertion in nuoG but this is not observed in any of the M. tuberculosis studies. nuoG forms part of the multi-subunit NADH reductase-I complex in the respiratory chain and transfers electrons to the menaquinone pool while simultaneously contributing to the proton gradient through its proton pumping function. Menaquinone biosynthesis itself has been a long standing drug target in M. tuberculosis (51–53). Menaquinone is synthesised from chorismate by a series of enzymes (MenF, MenD, MenH, MenC, MenE, MenB, MenA). The enzyme that catalyses the first committed step in this biosynthetic pathway is encoded by menD and is essential in all datasets. Genes in this biosynthetic pathway (menA, menB and menC) are also ES in M. bovis in this study, illustrating that targets of this pathway might also be useful in the treatment of M. bovis infections.

Wag31 is required for peptidoglycan synthesis and several Tnseq datasets have classified wag31 in M. tuberculosis as ES in vitro (21–23). Additionally, essentiality of wag31 in M. tuberculosis has been verified using conditional mutants (54). The Tnseq data obtained in this study classified wag31 in M. bovis as ES, but NE in M. tuberculosis. The study by Butler et al., assigned wag31 as NE in M. bovis. In order to probe this discrepancy with the literature and to establish CRISPRi silencing in M. bovis this gene was chosen for silencing. Early CRISPRi studies in M. tuberculosis performed by Singh et al. successfully utilised two plasmids encoding sgRNAs guides targeting +26 bp to +45 bp and +144 bp to +163 bp downstream of the annotated start codon of wag31_Mtb (Table 1; Figure 3A). We utilised these plasmids to make strains of M. tuberculosis and M. bovis where wag31 has been silenced. There are no differences in sequence in the area complementary to the sgRNAs between the two species, therefore the plasmids originally designed by Singh et al., for use in M. tuberculosis could be used in M. bovis. M. bovis AF2122/97 was transformed with pRH2502 to create a strain expressing dcas9_Spy(Mb__dCas9). Mb__dCas9 was then transformed with plasmids expressing the sgRNA guides. An identical strategy was used to make the equivalent strain in M. tuberculosis. Strains were cultured to exponential phase and serial dilutions were spotted onto agar containing 200 ng/ml aTc. Controls (without aTc, without sgRNA) were also included. The results, presented in Figure 3B, show that silencing wag31_Mb in both M. bovis and M. tuberculosis results in a severe growth defect, visible at 10⁻¹ dilution with complete cessation of growth at 10⁻² dilution. This supports the consensus in the literature that wag31 is an essential gene in M. tuberculosis. It also supports the classification of wag31 as ES in M. bovis rather than NE as reported by Butler et al., Dysgonic growth of M. bovis on this medium can be observed in the figure. This has been previously reported (43, 55).

Rv2182c/Mb2204c is annotated as a 1-acylglycerol-3-phosphate O-acyltransferase (agpat) and involved in glycerophospholipid metabolism. It is thought to synthesise diacylglycerol-3P through the addition of acyl chains to monoacylglycerol-3P. It is classified as ES in M. tuberculosis in this study and by others (22, 23, 56). It is classified as ES in M. bovis in this study but NE in the study by Butler et al., Strains of M. tuberculosis and M. bovis were constructed expressing sgRNAs targeting +2 bp to +21 bp and +40 bp to +59 bp downstream of the annotated start codon of Rv2182c/Mb2204c (Table 2; Figure 4A). The impact of inducing the system on expression of Rv2182c/Mb2204c was measured using RT-qPCR. The results, which are shown in Figure 4B show that dCas9_Spy is similarly induced in both M. tuberculosis and M. bovis with 150 to 350-fold induction of expression in the presence of aTc. Additionally, the results show that, in the presence of the sgRNA, there is a clear reduction in expression of Rv2182c/Mb2204c in both species. These data demonstrate effective gene silencing of Rv2182c/Mb2204c in both M. tuberculosis and M. bovis, respectively.

To determine the impact of induction of the guides, strains were cultured to exponential phase and serial dilutions were spotted onto agar containing 200 ng/ml aTc. The results (Figure 4C) show that silencing Rv2182c in M. tuberculosis results in a severe growth defect, with almost complete cessation of growth at 10⁻¹ dilution. However, the consequence of silencing Mb2204c on the growth of M. bovis is far less pronounced with a small reduction visible at the lowest dilution 10⁻⁴. This demonstrates that, unlike wag31, silencing of Rv2182c and its orthologue Mb2204c in M. tuberculosis and M. bovis respectively, has a differential impact on growth, with M. tuberculosis being more vulnerable and showing a greater growth defect. These results do not support the classification of Mb2204c as an ES gene in M. bovis but they clearly highlight that there are different phenotypic consequences as a result of silencing the orthologue in both species.