An attenuated Mycobacterium tuberculosis clinical strain with a defect in ESX-1 secretion induces minimal host immune responses and pathology

Although Mycobacterium tuberculosis (M.tb) DK9897 is an attenuated strain, it was isolated from a patient with extrapulmonary tuberculosis and vaccination with a subunit vaccine (H56) induced poor protection against it. Both attenuation and lack of protection are because M.tb DK9897 cannot secrete the EsxA virulence factor nor induce a host response against it. Genome sequencing identified a frameshift mutation in the eccCa1 gene. Since the encoded EccCa1 protein provides energy for ESX-1 secretion, it suggested a defect in the ESX-1 type VII secretion system. Genetic complementation with a plasmid carrying the M.tb H37Rv sequence of eccCa1-eccCb1-pe35 re-established EsxA secretion, host specific EsxA T-cell responses, and increased strain virulence. The ESX-1 secretion defect prevents several virulence factors from being functional during infection and therefore attenuates M.tb. It precludes specific T-cell responses against strong antigens and we found very little in vivo cytokine production, gross pathology or granuloma formation in lungs from M.tb DK9897 infected animals. This coincides with M.tb DK9897 being unable to disrupt the phagosome membrane and make contact to the cytosol.


M.tb DK9897 belongs to a lineage with few members. Since M.tb strains from different lineages can
induce variable host responses in macrophages, cell lines and mouse models [26][27][28] the genetic diversity among lineages could potentially influence the protective efficacy of TB vaccines. We, therefore, set out to test the ability of the H56 vaccine 29 to protect against aerosol infection with clinical M.tb isolates. H56 is a fusion protein of the M.tb proteins Ag85B, EsxA, and Rv2660c.
The M.tb DK9897 isolate was one of six clinical isolates selected from the M.tb strain collection at the International Reference Laboratory of Mycobacteriology harboring > ten thousands of clinical isolates cultured from individuals infected with mycobacteria. In our selection, we prioritized lineage coverage and sequence diversity but for safety reasons, we only included strains that were susceptible to standard anti-tuberculous treatment. M.tb DK9897 was originally isolated in February of 1998 from the cervical pus of a 92-year-old woman with tuberculous lymphadenitis. The isolate was susceptible to isoniazid, rifampicin, ethambutol, pyrazinamide and streptomycin. To investigate if M.tb DK9897 was part of a larger subgroup of mycobacterial isolates we genotyped the M.tb DK9897 strain along with the laboratory-adapted strains M.tb Erdman and M.tb H37Rv and an isolate belonging to the large Beijing family, M.tb DK9417. One quick and reliable marker commonly used for M.tb genotyping is the mycobacterial interspersed repetitive units (MIRU), located in variable number tandem repeats (VNTR) found at multiple loci scattered throughout the genome. The MIRU-VNTR genotyping data (Supplementary Table S1) was uploaded to the MIRU-VNTRplus database and a phylogenetical analysis was performed using a neighbor-joining algorithm and categorical distance coefficient using our four 24-locus MIRU-VNTR typing data and all isolates available in the MIRU-VNTR+ database as input. The results show that DK9897 does not belong to any of the established lineages but is a member of a new lineage with very few members that cluster between the M.tb Erdman and M.tb H37Rv (Fig. 1).
The H56 subunit vaccine induced poor protection against M.tb DK9897. Groups of mice were vaccinated with either H56 fusion protein in CAF01 adjuvant or BCG. Three weeks after the third H56 vaccination, the vaccine-specific T cell responses were investigated by flow cytometry in splenocytes. CD4 T cells that after stimulation with a pool of the three single antigens in H56 produced at least one of the cytokines IFN-γ , tumor necrosis factor-alpha (TNF-α or IL-2 were taken as being vaccine specific. Overall, the H56 vaccine-induced immune responses, in terms of T cell polyfunctionality and frequency of vaccine-specific cells, was at a similar level to what we have seen before 30 . On average 58.073 ( + /− 8704) of the spleen CD4 T cells recognized the vaccine (Fig. 2a). More than 60% of these were IL-2 producing T cells with high proliferative capacity and potential to become memory cells (Fig. 2b). The strongest T cell response was against Ag85B followed by EsxA and then a weak recognition of Rv2660c (Fig. 2c). Thus, we had a robust high-quality T cell response in the H56 vaccinated animals that previously have resulted in efficient protection against M.tb 30 . The expression of the Rv2660c gene has been questioned since no Rv2660c gene expression was found in cultures starved for nutrition's nor in mice infected with M.tb for up to 31 days with M.tb, instead strong expression of a small RNA on the opposite strand was found 31 . Yet, in two studies investigating human immune responses, more than 60% of persons with LTBI had T cell responses against Rv2660c and half of the TB patients responded specifically against a peptide pool covering the Rv2660c sequence 32,33 . Furthermore, immune mice and non-human primates have clear recall responses to Rv2660c emphasizing the availability of this antigen for recognition by primed T cells 30,34 . Thus, it is possible that the Rv2660c gene expression is tightly controlled and transcription only takes place under certain conditions such as e.g. immune stress. Six weeks after the third vaccination all mice were either aerosol challenged with M.tb Erdman or M.tb DK9897. The challenge dose was approximately 100 viable bacilli per animal for all infections. To compare M.tb strain virulence and vaccine protective efficacy the number of mycobacteria was determined in individual lungs from M.tb Erdman (Fig. 2d) and M.tb DK9897 infected (Fig. 2e)  SEM) whereas M.tb DK9897 grew to a level approximately 10-fold lower with less than 100.000 colony forming bacteria (CFU) per lung (Log 10 = 4.93 + /− 0.10 SEM) in non-vaccinated animals. In terms of protection, M.bovis BCG and H56 both protected quite effectively against the M.tb Erdman challenge with no statistical difference in protection. However, in M.tb DK9897 challenged animals H56 vaccination only reduced the CFU load 5-fold compared to non-vaccinated animals whereas the M.bovis BCG vaccinated group had a 50-fold reduction in bacterial number, and a significant difference (p < 0.001) compared to H56 vaccinated.
No EsxA secretion during M.tb DK9897 infection. M.tb DK9897 only multiplied from 100 to approximately 100.000 bacteria in non-vaccinated animals during the six-week infection period -10-fold less than M.tb Erdman. To investigate if the difference was because of a general longer doubling time relative to M.tb Erdman or it was specific for in vivo growth conditions we compared in vitro growth of the two strains, M.tb H37Rv and M.bovis BCG in rich medium. During the exponential phase, the virulent strains M.tb Erdman and M.tb H37Rv grow faster than M.tb DK9897 that again grows faster than the attenuated M. bovis BCG (Fig. 3a). Thus, also under in vitro conditions without nutrient limitations, immune pressure and an intracellular localization, we found a reduced growth rate for M.tb DK9897 compared to the more virulent M.tb strains. To understand why the H56 vaccine protects poorly against an attenuated M.tb strain we examined the vaccine-specific CD4 T cell response in lungs of infected animals. After six weeks of infection, a significant portion of the CD4 T cells (1 to 2%) recognized Ag85B regardless if the animals were infected with one or the other strain. As expected, T cells from M.tb Erdman-infected animals recognized EsxA (approximately 2% of the CD4 T cells) but we found no T cell recognition of EsxA in M.tb strain DK9897 infected animals (Fig. 3b). Even in animals vaccinated prior to M.tb DK9897 infection, there was barely a detectable CD4 T cell response specific for EsxA in the lung (Supplementary Figure S1). To investigate if M.tb DK9897 expresses and secretes the EsxA protein we grew the bacteria in parallel with M.tb Erdman in liquid cultures. During the exponential phase, we separated the cultures into bacteria pellet and culture supernatant and visualized the presence of the EsxA and Ag85B proteins in each fraction by western blot using specific antibodies (Fig. 3c). We found Ag85B in pellet and supernatant from both cultures confirming that both strains can express and secrete Ag85B correctly. In contrast, we only found EsxA in the pellet from both cultures. EsxA was undetectable in the culture supernatant from the M.tb DK9897 culture. Thus, both strains express EsxA, but M.tb DK9897 is unable to secrete the protein.   Figure S3). Using polyclonal antibodies raised against a 15mer peptide or a 101 amino acid fragment in EccCa1 we were not been able to get a positive signal with these antibodies in protein samples from M.tb DK9897 whereas both antibodies detected a protein with the correct size in M.tb Erdman samples. Thus, a truncated EccCa1 protein is likely either unstable or expressed in very low amounts. We cannot eliminate the possibility that the nucleotide insertion happened during culturing of the strain but we consider this unlikely since we kept the number of M.tb DK9897 passages at a minimum. We grew M.tb DK9897 once for initial typing and storage and subsequently used this stock to inoculate two separate cultures for sequencing which both gave identical results.

ESX-1 polymorphisms in clinical isolates.
In order to acquire a better understanding of how common mutations that are significantly enough to influence ESX-1 secretion are among clinical M.tb isolates we searched for polymorphisms of twelve genes known to be associated with and critical for ESX-1 secretion. We used an alignment-based search among all fully sequenced genomes of clinical isolates of the M.tb complex deposited in the NCBI database to evaluate the integrity of the orthologues genes in 143 genomes. Our search revealed that the twelve ESX-1 associated genes, in general, have an overall high degree of sequence conservation, with more than 99% nucleotides identical to their orthologues from the reference strain M.tb H37Rv. When aligned to the corresponding genes in M.tb H37Rv and evaluated for frameshift mutations (in/del) and stop codon loss/gain 18% (26 genomes) had one or up to six critical mutations in one of more of the investigated genes that were likely to prevent or diminish ESX-1 secretion (Supplementary Table S3).  showed this increase to be significant (Fig. 4c).     (Fig. 5c).

Discussion
The mycobacterial ESX-1 secretion system was linked with increased pathogenicity and host-pathogen interactions soon after its discovery 36,37 . Our results are in line with the growing pool of evidence showing the importance of the ESX-1 secretion system for the virulence of M.tb [38][39][40] and the generation of a host immune response exploited by the mycobacteria to increase growth and spreading 41 . Furthermore, our study shows that an attenuated M.tb strain lacking ESX-1 secretion can cause cervical tuberculous lymphadenitis. Although this is the first description of a clinical M.tb strain with a defect in ESX-1 secretion, there are cases of extra-pulmonary TB caused by Mycobacterium microti or Mycobacterium bovis BCG, both lacking ESX-1 function 42,43 . Most likely, these strains primarily cause disease in immune compromised individuals and since our case was a 92-year-old patient it is possible, she had developed one or more immune-deficiencies. Our MIRU analysis showed that M.tb DK9897 is a rare strain belonging to a mycobacterial cluster having very few members. However, mutations impairing EsxA secretion can occur independently in different phylogenetic lineages, we, therefore, searched the publically available M.tb complex genome sequences for mutations with major effect on the ORFs of ESX-1 related genes. To our surprise, we found 18% of the sequenced strains to have a mutation in one or more of the searched ESX-1 related genes. This is not in line with the importance of ESX-1 for virulence or that EsxA and EsxB based assays in numerous studies have diagnosed TB patients with sensitivities around 90% 44 . Most likely, the discrepancy is due to sequencing errors in the deposited genome sequences where several of the genomes have relatively low sequence coverage. Alternatively is ESX-1 secretion during a human infection not critically dependent on the same proteins that have been found to be important for ESX-1 secretion in mice. For our virulence investigation, we used a murine TB model to compare the ability of M.tb DK9897 and M.tb Erdman to grow after an aerosol infection. The model assesses the overall ability of a strain to enter the lung, infect alveolar macrophages, survive and replicate within macrophages. The low CFU load after six weeks infection suggest an attenuation of M.tb DK9897. We found similar numbers of bacilli in the lungs of M.tb Erdman and M.tb DK9897 infected animals shortly after aerosol infection. Thus, attenuation is unlikely to be due to a defect in establishing the initial infection and because of the slower in vitro growth rate, it is likely that one or more of the approximately 1200 mutations in the M.tb DK9897 genome sequence has a major influence on the doubling time of exponential growing M.tb DK9897. Since we know that ESX proteins induce T-cell responses 45 , the lack of a T cell response against EsxA that was restored by gene complementation confirmed the connection between in vivo attenuation and secretion of EsxA. Western blot analysis of mycobacterial pellets and supernatants from in vitro grown cultures revealed that M.tb DK9897, unlike M. bovis BCG, was able to express but not secrete EsxA. EsxA is transported across the bacterial membrane in a heterogeneous dimer complex with EsxB by the ESX-1 system (Fig. 6). Previous studies have shown that several genes, encoded from at least two chromosomal loci, are required for export of EsxA 16,18,37,46 . In order to identify a potential defect in any of the genes previously identified as essential for ESX-1 secretion, we sequenced the M.tb DK9897 genome and compared it to the M.tb H37Rv database sequence 35 . The comparative analysis revealed around 1200 polymorphisms to the published sequence, among which a single nucleotide insertion in the eccCa1 gene stood out as a probable explanation for the lack of ESX-1 activity. In virulent mycobacteria, eccCa1 encodes a 747 amino acid long protein involved in ATP cleavage and generation of energy to move proteins across the bacterial membrane through the ESX-1 system 47 . The eccCa1 insertion in M.tb DK9897 causes a frameshift, resulting in an EccCa1 protein lacking 305 amino acids at the C-terminus (Supplementary Figure S4). This is in agreement with an earlier study, showing attenuated growth of M.tb H37Rv in the human macrophage-like THP-1 cell line and in aerosol-infected mice and a lack of EsxA and EsxB secretion when the eccCa1 gene was disrupted 16 . Given the central function of EccCa1 in ESX-1 secretion, it is likely that ESX-1 does not secrete any proteins in M.tb DK9897 (Fig. 6). To confirm this hypothesis, we attempted to identify lack of another ESX-1 secreted protein, EspC, in the supernatant of M.tb DK9897 cultures using two polyclonal antibody preparations. Unfortunately, we were unable to get a signal using this approach, not even in positive control samples from fully virulent M.tb Erdman cultures.
To confirm that the defect in ESX-1 was due to the nucleotide insertion in the eccCa1 gene, we complemented M.tb DK9897 with a plasmid-borne version of the M.tb H37Rv eccCa1-eccCb1-pe35 genes 16 . This complementation restored secretion of, and T cell responses against, EsxA and the virulence of the recombinant mycobacteria increased slightly. This was quite similar to the modest increase in virulence that occurred after knock-in of the RD1 region into M.bovis BCG 15 . Thus, attenuation of M.tb DK9897 is partly due to a single nucleotide insertion in the eccCa1 gene disrupting the encoded proteins ATPase activity. This eliminates ESX-1 secretion of virulence factors and they become non-functional in bacterial survival and spreading. We did also identify a mutation in the pe35 gene from DK9897 compared to the M.tb H37Rv reference that potentially could be the culprit for the secretion defect (Supplementary Table S2). However, we find this unlikely because the mutation only affects the last amino acid in PE35, changing a glutamate codon into a STOP codon resulting in the expression of a PE35 protein lacking one amino acid at the C-terminal end. Furthermore, searching the NCBI database, we found that 83% (117 out of 141) of the sequenced genomes from the M.tb complex have exactly the same pe35 gene sequence as M.tb DK9897, the rest have the M.tb H37Rv sequence. It is possible that part of the EccCa1 associated  M.tb DK9897. (a) The core structure of the ESX-1 system is formed by the conserved components EccB1, EccCa1, EccCb1, EccD1 and EccE1, each possessing transmembrane region(s) spanning the mycobacterial plasma membrane 64 . The EccC is an ATP-driven translocase consisting of two subunits, EccCa1 and EccCb1, which are assembled once EccCb1 binds its target substrate, in this case, the EsxA/EsxB heterodimer, where EccCb1 interacts with the carboxy-terminal signal sequence of EsxB (marked as C in the figure) 65,66 . EsxB functions as a chaperone for EsxA secretion, which is a major ESX-1 virulence factor 21 . The secretion of EsxA/EsxB is co-dependent with the secretion of EspC/EspA and the C-terminus of EspC targets for the interaction with the cytosolic ATPase EccA1 67,68 . EspC polymerizes during secretion, indicating that EccA1 and EspA might function as cytosolic chaperones 69 . The polymerization of EspC results in the formation of a surface-exposed filamentous structure that spans the entire cell envelope 69 , possibly serving as a channel responsible for transporting ESX-1 substrates. (b) The truncated EccCa1 of M.tb DK9897 is unable to bind the EccCb1/EsxA/EsxB complex and is thus unable to secrete the EsxA/EsxB heterodimer. The secretion of EsxA/ EsxB and EspA/EspC are mutually co-dependent 67 , meaning that the secretion of EspC is presumably also disrupted in M.tb DK9897. attenuation could be due to an impairment of the cell wall integrity but we find this unlikely since we could not identify the truncated EccCa1 protein in M.tb DK9897 cultures 48 . Since complemented M.tb DK9897 is not as virulent as M.tb H37Rv, it is likely that one or more of the other sequence changes in M.tb DK9897 contributes to the strain attenuation via a different pathway that has an influence on both in vivo and in vitro growth.
To compare the host immune responses against M.tb Erdman and M.tb DK9897 we infected mice with different inoculums of the strains to allow them to reach the same CFU levels after six weeks infection. At this time point, there were approximately three-fold more T cells in M.tb Erdman-infected lungs than non-infected lungs and slightly fewer cells in M.tb DK9897 infected lungs. When lung CD4 T cells were stimulated in vitro, in the presences of excess antigen and co-stimulatory molecules, they were capable of responding and producing cytokines regardless of the challenge strain. However, in vivo, there was a big difference in cytokine expression. In mice infected with virulent M.tb, there was an expression of both pro-inflammatory and T cell cytokines whereas in M.tb DK9897 infected mice there was little or no cytokine production at all. Taken together, the in vitro and in vivo data suggest that there is limited antigen presentation and T cell activation in the chronic stage of an M.tb DK9897 infection, while the virulent strain induced an increased and more persistent response. The prolonged induction of immune responses leads to tissue destruction and, in accordance, we found many and large granulomas in lungs from mice infected with virulent M.tb whereas there was almost no gross pathology in lungs from M.tb DK9897 infected mice. Granuloma formation has been associated with M.tb escaping from the phagosome to the cytosol and induction of host cell death via necrosis 39,49 . We, therefore, investigated the difference in the two strains' ability to disrupt the phagosomal membrane and make a contact to the cell cytosol. M.tb Erdman clearly can establish this contact, whereas M.tb DK9897 and M. bovis BCG is incapable of this. Similar results were found in a previous study where pathogenic M.tb and Mycobacterium leprae made contact to the cytosol of the host cell and induced cell death whereas non-pathogenic M. bovis BCG did not 50 .
Thus, our data are in agreement with a model where virulent M.tb exploits ESX-1 to disrupt the phagosomal membrane and escapes into a more permissive environment, followed by a rapid expansion of M.tb and increased cell-to-cell spreading. The expansion and accumulation of high numbers of M.tb lead to host cell death, attraction, and activation of immune cells. Newly recruited immune cells become target cells for infection with free M.tb, while others, e.g. neutrophils, release cytokines and chemokines, thus increasing granuloma formation, pathology and further recruitment of cells. Faster replicating M.tb strains will induce more necrosis and accelerate the process 51 . In a human granuloma, there are few bacteria compared to a mouse granuloma and vacuolar-escaping is most likely a rare and transient event. However, it is still a critical step in bacterial expansion and cell-to-cell spreading, which we should understand in more detail to eliminate TB.

Materials and Methods
Animals. Six 16 . Six weeks after infection, mice were sacrificed and organs homogenized in PBS for bacterial enumeration, as described in 30 , and flow cytometry analysis.
Genotyping, Sequencing and Phylogenetic analysis. MIRU-VNTR 24-locus genotyping and DNA extraction from bacterial cultures were performed as previously described 52 . A minimum spanning tree was created with the MIRU-VNTRplus database website. MIRU-VNTR clusters were defined as strains with identical genotyping patterns and clonal complexes by a maximum difference of two loci 53  MiSeq system (Illumina, San Diego, California, USA) with 300 base pair (bp) paired-end reads from a Nextera DNA library, yielding a total of 2359 Mbp calls. Low-quality bases (Phred quality score < 20) and clipped reads shorter than 40 bp were removed using Trimmomatic (v0.32) 54 using the sliding window approach (size 4 bp) resulting in a mean read length of 225 bp. The trimmed reads were mapped to the M.tb strain H37Rv genome (GenBank ID: NC_000962.3) using the Burrows-Wheeler Aligner BACKTRACK (v0.6.2-r126) 55 resulting in a mean genome coverage of 365. Variant analysis was performed using VarScan (v2.3) 56 aligning reads with a quality of at least 1 (Phred quality score). The variants were classified into categories: LOW (synonymous variant), MODERATE (missense variant, in frame insertion/deletion) and HIGH (stop lost/gained, frameshift insertion/ deletion). Results were analyzed using Excel (v15.0.4885.1000). The complete annotation of the M.tb DK9897 genome is located under GenBank accession no. CP018778.
Alignment of ESX-1 associated genes from public available genomes. The presence and integrity of twelve ESX-1 related genes were analyzed in a panel of M.tb complex strains and clinical isolates for which the genomes had been completely sequenced. These organisms were selected from the Nucleotide collection database of the National Center for Biotechnology Information (NCBI), limiting the search by organism-Mycobacterium tuberculosis complex (taxid: 77643). The search was performed on the 7th of March 2017 using the nucleotide sequences for twelve genes from the reference strain M.tb H37Rv as input (espC, espA, ecca1, eccB1, eccCa1, eccCb1, pe35, esxA, esxB, eccD1, eccE1, and mycP1) and the BLASTN (v2.6.1+ ) program 57 . Homologous genes were identified and individually aligned against their reference orthologue from M.tb H37Rv to identify nucleotide substitutions, insertions, and deletions.
Multiplex cytokine assay. The Meso Scale Discovery multiplex mouse cytokine assay was used (IFN-γ , IL-1β, IL-2, IL-6, IL-5, IL-10, TNF-α , IL-12) and performed according to the manufacturer's instructions. The plates were read on the Sector Imager 2400 system (Meso Scale Discovery) and calculation of cytokine concentrations in unknown samples was determined by 4-parameter logistic non-linear regression analysis of the standard curve.

Cell cultures and infection. THP-1 cells (ATCC) were maintained in RPMI supplemented with
10% heat-inactivated focal calf serum (FCS). THP-1 cells were seeded in 12-well plates at a concentration of 0.2 × 10 6 cells/ml and treated with 20 ng/ml of Phorbol 12-Myristate 13-Acetate (PMA) for 72 h to induce their differentiation into monocyte-derived macrophages.
THP-1 derived macrophages were infected at various multiplicity of infection (MOI) of M.tb Erdman, M.tb DK9897 or BCG in antibiotic free complete RPMI for 2 h at 37 °C, 5% CO 2 . The infectious medium was removed and cells were complemented with fresh growth medium containing 10% FCS. At time course measurements cells were harvested with a cell scraper and moved to a 96 well plate for analysis. CCF-4 assay and flow cytometry. Cells were stained with a 1X final CCF4-AM solution (Life Technologies) in PBS according to the manufacturer's instructions for 2 h at room temperature (RT). Cells were washed three times with PBS and then stained with Live/dead reagent (APC-eF780, 1:1000) and fixed with 4% paraformaldehyde (PFA). Cells were analyzed in a BD LSRFortessa ™ using BD FACSDIVA ™ software (BD Bioscience). Data analysis were done with FlowJo v10.2 software. CCF-4 fluorescence was measured using 405 nm as excitation laser and 450/50 nm (blue) and 525/50 nm (green) emission filters (as illustrated in 62 ).
Statistical analysis. Prism 7 software (GraphPad v7.02) was used for all statistical analyses. Bacterial numbers were log-transformed before being analyzed using one-way ANOVA with Tukey's multiple comparisons test. In vivo cytokine levels were analyzed using two-tailed t-tests with Mann-Whitney correct. Statistically significant differences are marked by asterisks in figures and explained in the figure legends.