One hundred thirty-five MTBC clinical strains were randomly selected from a collection of MTBC strains isolated at the Catholic University of the Sacred Heart in Rome between 2007 and 2011. Each clinical isolate was grown in Middlebrook 7H9 medium supplemented with 0.2% glycerol, 10% ADC and 0.05% Tween 80 at 37°C. Genomic DNA was extracted from liquid cultures by using the CTAB method, as previously described (van Embden et al., 1993).

The genomic DNA of each MTBC clinical strain was genotyped by 15-loci MIRU-VNTR method (Gutierrez et al., 2005) and spoligotyping (Kamerbeek et al., 1997), as previously described. For MIRU-VNTR typing, each multiplex PCR set included the genome of Mtb H37Rv and water, as positive and negative controls, respectively. Similarly, in each experimental session of spoligotyping, both genomes of Mtb H37Rv and M. bovis BCG were used as positive controls and water as negative control. All multiplex PCRs for MIRU-VNTR typing were analyzed by capillary electrophoresis by using Applied Biosystems 3130xl Genetic Analyzer and PCR fragments size was estimated with the GeneMapper 4.0 software (Applied Biosystems). On MIRU-VNTRplus database, where possible, best-matches based on similarity search were inferred to the 15-loci MIRU-VNTR and spoligotyping patterns by setting a distance cut-off of 0.4, then a tree-based identification was performed by using default distance measures (www.miru-vntrplus.org; Gutierrez et al., 2005; Allix-Béguec et al., 2008). A UPGMA tree was eventually generated and rooted to the two reference strains of M. canettii available on the database.

The pe_pgrs33 gene from each MTBC clinical strain was amplified with two primers previously described, PE_PGRS33-F1 and PE_PGRS33-R1 (Wang et al., 2011; Supplementary Table 1), by using the Expand High Fidelity PCR System (Roche) kit. Sanger sequencing was performed in Applied Biosystems 3130xl Genetic Analyzer. All identified polymorphisms were confirmed by double-strand DNA sequencing, as previously described (Talarico et al., 2005; Supplementary Table 1). Each DNA sequence was assembled respect to pe_pgrs33 of Mtb H37Rv by using SeqMan (Lasergene 7). The genetic relationships between the 19 pe_pgrs33 alleles identified in this study were investigated using MEGA version 6 (Tamura et al., 2013), by inferring a Maximum Likelihood tree under HKY model and setting 1000 bootstrap resampling. The selective pressure acting on all pe_pgrs33 alleles was evaluated by calculating the dN/dS ratio with DnaSP (Rozas et al., 2003).

Five of the 19 pe_pgrs33 alleles identified in this study and their native promoters were amplified from the genome of one respective MTBC clinical isolate with two primers previously described, PG335Hn-338bp and 18c3AXb (Palucci et al., 2016; Supplementary Table 1), by using the Expand High Fidelity PCR System (Roche) kit. Each pe_pgrs33 allele under the control of its native promoter was cloned into the integrative plasmid pMV306 in-frame with the HA epitope sequence (Palucci et al., 2016). The previously generated MtbΔ33 mutant strain (Palucci et al., 2016) was then complemented with all constructs, according to procedures already described (Cascioferro et al., 2011; Palucci et al., 2016). To assess PE_PGRS33 expression, recombinant mycobacterial mid-log phase cells were homogenized in lysis buffer (10 mM Tris-HCl, 5 mM EDTA, protease inhibitors cocktail, pH 9.5) and 0.1 mm zirconia/silica beads (Biospec Products), as previously described (De Maio et al., 2014). Protein lysates, devoid of cellular debris and unlysed cells, were separated by SDS-PAGE on 12% polyacrylamide gels and were then transferred to nitrocellulose membrane by western blot. HA-tagged PE_PGRS33 was detected by using anti-HA monoclonal antibody HA.11 (1:1,000) (Covance), as primary antibody, and anti-mouse IgG peroxidase conjugate (1:4,000) (Sigma-Aldrich, Saint Louis, MO), as secondary antibody. Since, as expected, the protein encoded by 33^all3 was undetectable by western blot, two control PCRs were carried out by using the genomic DNA of Mtb H37Rv and the pMV306 plasmid engineered with the 33^all3 allele under the control of its native promoter (subsequently used to complement the MtbΔ33 mutant) as positive controls and water as negative control (Supplementary Figure 1; Supplementary Table 1). As further control, polymorphisms in 33^all3 of the MtbΔ33::33^all3 strain were reconfirmed by Sanger sequencing (data not shown).

Mice manipulation was performed prior approval by the Ethics Committee of the Catholic University of the Sacred Heart in Rome (Prot. Number: n° T21/2011). Peritoneal murine macrophages (pMMOs) were isolated from 9 to 14 weeks old female C57BL/6 mice (Harlan), as previously described (Palucci et al., 2016). pMMOs were seeded at 1.2·10⁶cells/ml in 48-well plates in RPMI medium supplemented with 10% FBS and 1% of both L-glutamine and sodium pyruvate and were incubated overnight at 37°C in a 5% humidified atmosphere. After removing non-adherent cells, pMMOs were infected for 1 h with Mtb strains at a multiplicity of infection (MOI) of 1:10 and then incubated in complete RPMI medium. Four hours later, intracellular mycobacteria were determined by CFUs enumeration, as previously described (Palucci et al., 2016). Buffy coats of 4 male donors were processed to collect human peripheral blood mononuclear cells (PBMCs) and monocyte-derived macrophages (MDMs). PBMCs were isolated from buffy coats by using Ficoll, seeded for mycobacterial infections at 1.2·10⁶ cells/ml in 48-well plates in X-VIVO™ 15 medium (Lonza) with human serum type AB 2% (Lonza) and incubated overnight at 37°C in a 5% humidified atmosphere. MDMs were isolated by positive selection from unplated PBMCs by using CD14 MicroBeads, following the manufacturer's instructions (Miltenyi Biotec). MDMs were seeded at 1.2·10⁶ cells/ml in 48-well plates, as indicated for PBMCs and were incubated for 6–7 days at 37°C in a 5% humidified atmosphere until MDM differentiation. Both PBMCs and MDMs were infected with Mtb strains at a MOI of 1:1. According to the method described above for pMMOs, MDMs were infected and intracellular mycobacterial CFUs were determined at 4 and 72 h post-infection (Palucci et al., 2016). For PBMCs infections, mycobacterial inocula at a MOI of 1:1 were directly added to cell culture supernatants and at 72 h post-infection total mycobacterial CFUs were enumerated (Palucci et al., 2016). Specifically, cell pellets obtained prior centrifugation of supernatants from PBMCs infection were lysed and resuspended in PBS with 0.05% Tween 80, where MDMs lysed in 0.1% Triton X-100 were previously added, then total lysates were serially diluted and plated on 7H11/OADC agar medium (Palucci et al., 2016). Supernatants from PBMCs and MDMs at 72 h post-infection were collected and stored a −80°C until being assayed for cytokine ELISA.

Supernatants collected from MDMs and PBMCs at 72 h post-infection were filtered to remove mycobacteria and cytokines were analyzed by using a Cytometric Bead Array (CBA) (BD Biosciences, San Jose, USA) and a FACS CANTO II (BD Biosciences, San Jose, USA), according to the manufacturer's instructions. CBA results were generated by FCAP Array™ software (BD Biosciences, San Jose, USA). For supernatants collected from PBMCs at 72 h post-infection, IFN-γ release was also evaluated by using ELISA QuantiFERON TB-Gold (Qiagen, Hilden, Germany), according to the manufacturer's instructions.

Animal studies were carried out in strict accordance with the Amsterdam protocol on animal protection and welfare, the Directive 2010/63/EU of the European Parliament and the Council of 22 September 2010 on the protection of animals used for scientific purposes and the French Decree 2013-118. The protocol was approved by the Minister of Higher Education and Research after favorable opinion of the Ethics Committee (CEEA Nord-Pas de Calais/INSERMU1019 n° 00579.01 from 23/07/2014). All efforts were made to minimize suffering of the animals. Six-weeks-old female BALB/c mice (Janvier) were challenged with mycobacterial strains via the intranasal route with 10 μl/nostril. Challenge suspensions were adjusted in order to obtain an inhaled dose of approximately 1,000 CFU/lungs. At day 0, 28, and 49 post-infection, lungs from euthanized mice were collected and homogenized by using an MM300 apparatus (Qiagen) and 2.5-mm diameter glass beads. Ten-fold serial dilutions of lung homogenates were plated on 7H11/OADC agar medium and CFUs enumerated, as previously described (Palucci et al., 2016).

Murine lungs were fixed with 10% paraformaldehyde and then embedded in paraffin for sectioning, according to standard methods. In order to correlate the presence of histologic lesions with acid-fast bacilli (AFB), replicas 3 μm sections were cut and stained with both Haematoxyline and Eosin (HE) and Ziehl–Neelsen (ZN), following standard techniques. The lesions morphology and distribution were evaluated by light microscopy. At least 6 lung sections for all mice of the 5 groups were analyzed in different points (24 lung sections per experimental group in total). For each section the number of ZN positive cells, the total surface area and the area with lesions were measured at 400x magnification and averages calculated for each section and group. Slides were imaged using Nikon Eclipse 80i microscope and digital computer images were recorded with a Nikon DS-L2 camera control unit and the Nikon dedicated software 3422.1001.1798.080117. Cellular automatic count and histological measurements were carried out using the dedicated software Axiovision ver. 4.4 (Zeiss) by two independent researchers on two independent photo series.

This assay was carried out by using two monoclonal antibodies (mAb), purified biotin anti-F4/80 Clone BM8 (Caltag Labs, cat. n. MF48000) and purified anti-MT 16 kDa antigen (Santa Cruz Biotechnology, cat. n. sc-58169). Signals were revealed with Streptavidin Alexa Fluor® 555 and Streptavidin Alexa Fluor® 488. For immunomicroscopy, slides obtained from fixed-lung tissues were also processed, as previously described, and mounted on positively charged Superfrost slides (Fisher Scientific). Deparaffinization, rehydration and antigens retrieval of tissue sections were performed by using Dewax and HIER Buffer L (Thermo Fisher Scientific). To prevent non-specific bindings, slides were incubated in PBS containing 2% BSA, stabilizing protein and 0.015 mol/L sodium azide (Protein Block Serum-Free, Dako). For each target, tissue sections were incubated in two different steps with primary mAb (overnight at 4°C), Ab (1 h at room temperature) and the respective fluorophore. Slides were counterstained with Hoechst blue and then covered. Images were acquired by using a Leica TCS SP 5 confocal microscope (Leica Microsystems, Germany) and processed with LAS AF Lite application software developed by Leica Microsystems CMS GmbH for contrast and brightness adjustments. Negative controls prepared by omission of primary antibodies did not show any fluorescence under the conditions described above.

Experiments were conducted in triplicate and replicated at least three times. In vivo experiment was performed one time in quintuplicate. Statistical analysis was performed by using GraphPad Prism version 6 (GraphPad software, CA, USA). For pMMOs infections, CFUs were expressed as mean ± SD and analyzed by one-way ANOVA, followed by Dunnett's multiple comparison test. For infections of human PBMCs and MDMs, results concerning CFUs and cytokine levels were expressed as median and analyzed by Kruskal-Wallis one-way ANOVA, followed by Dunnett's multiple comparison test. Results obtained from in vivo experiments were expressed as mean ± SD and analyzed by performing two-way ANOVA, followed by Dunnett's multiple comparison test. Histopathological results were analyzed by Student's t-test.

One hundred thirty-five MTBC clinical strains isolated in Rome from 2007 to 2011 were genotyped by using 15-loci MIRU-VNTR (Gutierrez et al., 2005) combined with spoligotyping (Kamerbeek et al., 1997). The in silico analysis performed on MIRU-VNTRplus database (Gutierrez et al., 2005; Allix-Béguec et al., 2008) allowed to identify 125 isolates (92.6%) and assign the respective MTBC superlineage (Supplementary Figure 2). The majority of isolates (104 isolates, 77%) belonged to the modern MTBC superlineages, which comprised the superlineages 2 (East-Asian, 2 isolates, 1.5%), 3 (East-African Indian, 7 isolates, 5.2%) and 4 (Euro-American, 95 isolates, 70.4%). This latter was composed of the Cameroon (4 isolates, 4.2%), Haarlem (39 isolates, 41.1%), LAM (11 isolates, 11.6%), S (12 isolates, 12.6%), Uganda I (4 isolates, 4.2%), Uganda II (1 isolate, 1.1%), X (2 isolates, 2.1%) lineages and unassigned strains (22 isolates, 23.2%, indicated in Supplementary Figure 2 as “?”) which could plausibly belong to the T-specific lineage of the Euro-American superlineage (Allix-Béguec et al., 2008). Although to a lesser extent (21 isolates, 15.6%), our collection of MTBC strains embraced three ancient superlineages: the most representative was the superlineage 1 (Indo Oceanic, 17 isolates, 12.6%), followed by the superlineages 5 and 6 (West African 1 and West African 2, respectively, 1 isolate, 0.7% each) plus the animal superlineage (2 isolates, 1.5%). The phylogenetic relationships between all 135 genotyped MTBC clinical strains were inferred by constructing an UPGMA tree rooted by two reference strains of M. canettii on MIRU-VNTRplus database (Figure 1; Gutierrez et al., 2005; Allix-Béguec et al., 2008). According to the phylogeographical distribution described for the MTBC by Gagneux and colleagues (Gagneux, 2012), our results showed how all MTBC superlineages are represented in the metropolitan area of Rome, except for the recently described ancient superlineage denominated 7 or Ethiopian (Firdessa et al., 2013), and that the Euro-American superlineage is the most prevalent.

The genetic variability of the pe_pgrs33 gene within the MTBC population analyzed in this study was assessed by Sanger sequencing. As shown in Supplementary Table 2, a total of 19 pe_pgrs33 alleles were identified. Apart from the allele referred to as 11, which corresponded to pe_pgrs33 of Mtb H37Rv and was found in 42 MTBC clinical isolates (Figure 1), we identified 11 alleles previously described in literature (Talarico et al., 2005, 2007; Wang et al., 2011; McEvoy et al., 2012) and 7 alleles containing new genetic variations (alleles 3, 7, 8, 9, 14, 16, and 17) compared to pe_pgrs33 of the H37Rv reference strain. Among all pe_pgrs33 alleles, we identified a total of 13 SNPs (9 non-synonymous and 4 synonymous), mainly occurring in the PGRS region. All indel events were exclusively located in the PGRS domain, the majority of which were in-frame and varied from a minimum of 9 bp to a maximum of 72 bp. The largest mutations were in-frame deletions found in 11 EAI isolates (allele 1), in the only isolate of M. africanum belonging to the West African 1 superlineage (allele 6) and in all the 7 Delhi/CAS isolates (alleles 18 and 19) (Figure 1). The only frameshift mutation identified was 1 bp deletion in position 1,014 bp and was detected in 3 pe_pgrs33 alleles (alleles 1, 2, and 3) all belonging to the EAI superlineage (Figure 1). Specifically, the altered frame in these latter alleles results in the introduction of a premature stop codon, which translates into the loss of the last 160 amino acids from the C-terminal of the PGRS domain and the “acquisition” of a 36-amino acids stretch, rich in proline residues (approximately 20%) and short antigenic epitopes, as revealed by the in silico antigenicity prediction of the protein (Supplementary Figure 3). The 3 pe_pgrs33 alleles, sharing the only frameshift mutation identified in this study, were more evolutionarily distant from the other alleles in the unrooted tree shown in Supplementary Figure 4. Moreover, a strict association between these alleles and the EAI superlineage was revealed by the 15-loci MIRU-VNTR and spoligotyping-based UPGMA tree color-coded by the pe_pgrs33 alleles (Figure 1). Similarly, a “clustered” association within the MTBC superlineages was quite fulfilled in the same tree also for the other alleles. In agreement with the results of Copin et al. (2014), the dN/dS ratio of 0.64 obtained for the pe_pgrs33 alleles identified in this study supports the evidence that a purifying selection is acting on pe_pgrs33 to preserve the gene sequence from deleterious SNPs.

Until recently, a possible role of PE_PGRS33 in the antigenic variability of Mtb has been entrenched with the identification of numerous polymorphisms in the pe_pgrs33 gene among Mtb clinical isolates (Talarico et al., 2005, 2007; Wang et al., 2011; McEvoy et al., 2012). However, a recent study questioned this hypothesis, suggesting an ongoing purifying selection on pe_pgrs33 and a limited impact on the functionality and antigenicity of PE_PGRS by indel events occurring in the PGRS domain (Copin et al., 2014). To elucidate the biological meaning of natural polymorphisms occurring in pe_pgrs33 and their impact on the pathogenesis and virulence of Mtb, 5 of the 19 pe_pgrs33 alleles identified in this study, designated from here as 33^allx (where “x” referred to the number of the respective allele), were selected based on the heterogeneity of their genetic variations. Among the selected alleles, 4 (33^all3, 33^all5, 33^all6, and 33^all18) were characterized by SNPs, in-frame indels and a single frameshift deletion occurring in the region encoding the PGRS domain of PE_PGRS33 compared to the 33^all11 allele, which corresponds to pe_pgrs33 of Mtb H37Rv (Table 1; Figure 2A). Moreover, no polymorphisms were detected in the native promoter sequence of the 5 pe_pgrs33 alleles used in this study. These alleles were cloned in the integrative plasmid pMV306, under the control of their native promoter and upstream the HA epitope sequence, and all plasmids used to complement the previously generated MtbΔ33 mutant strain (Palucci et al., 2016; Figure 2). Transcriptional analysis by RT real-time PCR was used to confirm expression in the complemented strains and higher expression of pe_pgrs33 was observed in the complemented strains compared to the wild type (+2,5-fold, data not shown), as expected when the gene is inserted in the attB site following complementation with the integrative plasmid pMV306 (Palucci et al., 2016). Apart from MtbΔ33::33^all3, for all complemented strains, PE_PGRS33 expression was assessed on protein lysates by western blot using an anti-HA antibody (Figure 2B). For MtbΔ33::33^all3, PCR (Supplementary Figure 1) and Sanger sequencing (data not shown) allowed to confirm both strain complementation and the altered frame in 33^all3.

In previous studies, large in-frame sequence variations in the pe_pgrs33 gene were associated to peculiar epidemiological phenotypes, like non-cavitary pulmonary TB or extrapulmonary TB in children (Talarico et al., 2005, 2007; Wang et al., 2011). To investigate the pathogenic impact of large sequence variations occurring in pe_pgrs33 among MTBC clinical isolates, peritoneal murine macrophages (pMMOs) were infected at a MOI 1:10 with the Mtb H37Rv, MtbΔ33 mutant and complemented strains with all 5 pe_pgrs33 alleles selected in this study (MtbΔ33::33^all11, MtbΔ33::33^all3, MtbΔ33::33^all5, MtbΔ33::33^all6, and MtbΔ33::33^all18). Since we recently showed that the MtbΔ33 mutant is impaired in its ability to enter into macrophages (Palucci et al., 2016), the capacity to restore this phenotype by the complemented strains was assessed at 4 h post-infection by enumerating intracellular CFUs (Figure 3). Similar to the MtbΔ33::33^all11 strain, MtbΔ33::33^all5, MtbΔ33::33^all6, and MtbΔ33::33^all18 were able to entry into pMMOs as efficiently as the parental strain. The significant reduction in this capacity for the MtbΔ33 mutant (p < 0.01) was also markedly observed for MtbΔ33::33^all3 (p < 0.001), indicating that whilst frameshift mutation in pe_pgrs33 alleles impair the cell entry phenotype of Mtb, small and large in-frame sequence variations as well as nsSNPs do not affect the PE_PGRS33-mediated entrance of Mtb into murine macrophages.

The impact of pe_pgrs33 polymorphisms was further investigated in vitro by infecting human monocyte-derived macrophages (MDMs) isolated from healthy donors with the Mtb H37Rv, MtbΔ33 mutant and complemented strains at a MOI 1:1 (Figure 4). MtbΔ33::33^all11 and MtbΔ33::33^all5 were able to complement the entry phenotype into MDMs compared to the parental strain (Figure 4A). Similar to the MtbΔ33 mutant (p < 0.01), MtbΔ33::33^all3 (p < 0.001), MtbΔ33::33^all6 (p < 0.05), and MtbΔ33::33^all18 (p < 0.05) were unable to fully restore the ability of Mtb H37Rv to entry into macrophages. To assess the impact of pe_pgrs33 polymorphisms on the capacity of Mtb to replicate intracellularly, macrophages infected as above were harvested at 72 h to determine CFUs and results expressed as logarithmic ratio between intracellular CFUs at 72 and 4 h. As shown in (Figure 4B), no statistically significant differences were observed for the MtbΔ33 mutant and complemented strains compared to Mtb H37Rv. Based on previous evidences on the immune modulating properties of PE_PGRS33 (Basu et al., 2007; Zumbo et al., 2013), two pro-inflammatory cytokines, TNF-α and IL-1β, were evaluated by performing a cytometric bead array (CBA) assay on supernatants collected at 72 h from MDMs infection. In (Figure 5), results of TNF-α and IL-1β levels are reported and show a reduced ability for the MtbΔ33 mutant and complemented strains to elicit TNF-α secretion compared to Mtb H37Rv, even if this difference did not reach statistical significance. To further assess whether pe_pgrs33 genetic polymorphisms could elicit peculiar phenotypes during infection of human primary cells, in terms of replication capacity of Mtb and modulation of host immune responses by the pathogen, we used the panel of selected mycobacterial strains to infect peripheral blood mononuclear cells (PBMC) isolated from the same donors. Also in this cellular model, mycobacterial CFUs enumerated at 72 h post-infection did not result in significant differences in the replication capacity of all recombinant Mtb strains compared to Mtb H37Rv (Figure 6A). Moreover, no significant perturbations in the ability of Mtb to induce a TNF-α, IL-1β, and IFN-γ-mediated proinflammatory response were detected in PBMCs infections, although MtbΔ33::33^all3 and MtbΔ33::33^all5 showed a slight decreased capacity to induce both TNF-α, and IL-1β (Figure 6B) secretion in line with what observed in MDMs infections.

To assess the role of PE_PGRS33 on the virulence of Mtb, 5 groups of BALB/c mice were intranasally infected with Mtb H37Rv, the MtbΔ33 mutant and 3 complemented strains selected from those previously assessed in vitro, one expressing the allele corresponding to pe_pgrs33 of Mtb H37Rv (MtbΔ33::33^all11) and two expressing pe_pgrs33 alleles characterized by major genetic variations (MtbΔ33::33^all3 and MtbΔ33::33^all6) (Figure 7). At day 0 post-infection, the number of CFUs enumerated in murine lungs did not significantly vary, indicating that the Mtb H37Rv, MtbΔ33 and complemented strains were similarly able to infect mice. At day 28 post-infection, which is considered the hallmark of active disease in mice, no variation of the bacterial burden in the lungs was found for both MtbΔ33::33^all6 and MtbΔ33 compared to the parental strain. Conversely, a reduced bacterial load was detected for MtbΔ33::33^all11 and for MtbΔ33::33^all3 strains compared to Mtb H37Rv. Interestingly, at day 49 post-infection, lung CFUs from mice infected with MtbΔ33 were higher (p < 0.05) than those from mice infected with Mtb H37Rv and increased significantly (p < 0.001) from the CFUs enumerated at day 28 post-infection, indicating that the lack of PE_PGRS33 results in an enhanced ability of Mtb to replicate/persist in the lung tissue of mice during the chronic phases of the disease. A similar trend was also found for MtbΔ33::33^all3, with lung CFUs significantly higher at day 49 compared to day 28 post-infection (p < 0.01). Conversely, bacterial loads were significantly reduced for MtbΔ33::33^all11 (p < 0.001) and MtbΔ33::33^all6 (p < 0.01) compared to the parental strain at day 49 post-infection.

For all 5 strains tested in in vivo infections (Mtb H37Rv, MtbΔ33::33^all11, MtbΔ33::33^all3, MtbΔ33::33^all6, and MtbΔ33), both quantitative and qualitative histopathological examinations were carried out on lung sections of mice sacrificed at day 28 and day 49 post-infection (Figure 8). At day 28 post-infection, no major differences were observed between the Mtb strains tested, in terms of number and extension of granulomas and net pulmonary area with lesions (Figures 8A–C). Similarly, at day 49 post-infection, apart from MtbΔ33::33^all3 (p < 0.01) and to a lesser extent MtbΔ33::33^all6, no significant differences were found between all other strains analyzed (Figures 8A–D). However, a higher, but not significant number of granulomas was enumerated for MtbΔ33::33^all11, MtbΔ33::33^all6, and MtbΔ33, in particular, compared to the parental strain (Figure 8A). Conversely, the lungs of mice infected with MtbΔ33::33^all3 were characterized by a reduction in the number of granulomas compared with those of mice infected with Mtb H37Rv (p < 0.01; Figure 8A). Surprisingly, based on the size of granulomas and net pulmonary area with lesions, MtbΔ33::33^all3 caused more extensive tissue damage (p < 0.01) compared to the parental strain, with major pulmonary confluent lesions which justify the low number of granulomas detected in the lungs (Figures 8B,D). At both day 28 and day 49 post-infection, the distribution and percentage of cells with acid-fast bacilli was similar for all analyzed strains, apart from MtbΔ33::33^all3, which was associated with a higher number of AFB compared to the parental strain at day 49 post-infection. This last observation was consistent with the features and extent of histopathological lesions shown in Supplementary Figures 5, 6. Taken together, these results suggest an interesting increased capacity of Mtb to evoke lung tissue damage in murine model, when the PE_PGRS33 protein is truncated or completely absent.

At higher magnification, lung sections from mice infected with MtbΔ33, MtbΔ33::33^all3, and MtbΔ33::33^all6 were remarkably characterized by a high number of foamy macrophages at day 49 post-infection compared with the sections obtained from the other groups of mice (Supplementary Figure 6). Moreover, AFB positive cells remained stable in Mtb H37Rv, MtbΔ33::33^all11, MtbΔ33::33^all6, and MtbΔ33 strains, but MtbΔ33::33^all3 showed a higher number of lesions, a reduced percentage of positive cells and a high number of AFB found extracellularly (Supplementary Figure 6). To better characterize bacilli localization of MtbΔ33::33^all3 compared to Mtb H37Rv and MtbΔ33 in infected tissues, immunofluorescence was performed by using two markers, anti-F4/80 and anti-MT 16 kDa antigen, to point out macrophages and bacilli, respectively. The intra and extracellular position of bacilli was then evaluated by comparing the results of immune-colocalization with those obtained from observation of Ziehl-Neelsen (ZN)-stained lung sections (Figure 9). As shown in Figure 9B, a yellow signal was repeatedly observed in lung sections of mice infected with Mtb H37Rv and derived from overlapping between red and green signals, which label mycobacteria and macrophages, respectively. An evident separation between the red and green fluorescence of MtbΔ33::33^all3 and macrophages, respectively, revealed the predominant extracellular position of this strain (Figure 9D). Also for MtbΔ33, it was demonstrated that the two fluorescence signals were detected within the same area although did not overlap regularly (Figure 9F). Moreover, further differences, in terms of immune cells composition, were also observed in lung tissue sections, as showed in ZN images (Figures 9A,C,E). Particularly, a prevalence of macrophages and some lymphocyte were found in lung tissue of mice infected with Mtb H37Rv (Figure 9A). Similarly, lungs of mice infected with both MtbΔ33::33^all3 and MtbΔ33 were characterized by a predominance of foamy macrophages, although the former also contained apoptotic cells and the latter polymorphonuclear cells (Figures 9C,E). These further investigations allowed to point out a mainly extracellular localization for the MtbΔ33::33^all3 and MtbΔ33 strains, showing how the lack or truncation of PE_PGRS33 may affect the cellular localization of Mtb during the chronic persistent TB disease stages.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.