A Nonribosomal Peptide Synthase Gene Driving Virulence in Mycobacterium tuberculosis

Over 10 million people developed tuberculosis (TB) in 2016, and over 1.8 million individuals succumbed to the disease. These numbers make TB the ninth leading cause of death worldwide and the leading cause from a single infectious agent. Therefore, finding novel therapeutic targets in Mycobacterium tuberculosis, the pathogen that causes most cases of human TB, is critical. In this study, we reveal a novel virulence factor in M. tuberculosis, the nrp gene. The lack of nrp highly attenuates the course of M. tuberculosis infection in the mouse model, which is particularly relevant in immune-deficient hosts. This is very relevant as TB is particularly incident in immune-suppressed individuals, such as HIV patients.

T uberculosis (TB) remains a leading cause of death by an infectious agent. Indeed, TB, caused by Mycobacterium tuberculosis, kills over 1.8 million people, afflicts over 10 million new individuals per year, and is estimated to exist in a latent form in one-fourth of the world's population (1). To reach the milestone of less than one TB case per million individuals, set by the World Health Organization (WHO) for 2050, numerous roadblocks will have to be overcome, notably, the development of more efficient vaccines and therapies. Success on these fronts requires a greater understanding of the mechanisms underlying the success of M. tuberculosis as a pathogen. In this regard, a variety of genetic strategies have been employed to unlock novel virulence factors in M. tuberculosis and hence druggable targets (2)(3)(4).
Nonribosomal peptide synthases (NRPSs) are large multifunctional proteins involved in the biosynthesis of short peptides in a manner that is distinct from the ribosomemediated translation of mRNA (5). The genes encoding NRPSs are predominantly found in bacteria and fungi that produce secondary metabolites (6). The short peptides produced by NRPSs usually form part of a larger more complex molecule. The genomes of mycobacteria, including the TB-causing M. tuberculosis, contain genes encoding NRPSs. While NRPSs in environmental bacteria are often associated with antibiotic and other secondary metabolite production (6), mycobacterial NRPSs are mainly involved in the production of components of complex cell wall lipids and siderophores (7,8). Rv0101 (nrp) is an M. tuberculosis H37Rv gene annotated as a NRPS (9). The gene encodes a putative bimodular NRPS containing seven catalytic domains (10,11), including an adenylation domain with predicted substrate specificity for lysine and phenylalanine. nrp is located in a cluster of genes with potentially linked functions: Rv0097 encodes a putative oxidase, Rv0098 encodes a long-chain fatty acyl-coenzyme A (fatty acyl-CoA) thioesterase (12,13), and FadD10 (Rv0099) is a fatty acyl-AMP ligase that catalyzes the transfer of fatty acyl chains to an acyl carrier protein encoded by Rv0100 (10). The knowledge about the precise function of nrp and its associated cluster is based solely on the structural characterization and in vitro enzyme assays with heterologously expressed and purified M. tuberculosis and Mycobacterium marinum proteins and suggests a role in the production of a novel lipopeptide with isonitrile functionality (10,14).
Importantly, several in vitro and in vivo infection studies have highlighted a potential role for nrp in the pathogenesis of TB. Specifically, nrp was reported as a nonessential gene for in vitro growth by transposon mutagenesis (15) but was otherwise required for growth in the spleens of C57BL/6J mice (16). Additionally, the gene cluster spanning Rv0096-Rv0101 was shown to be critical for survival in mice (17). Furthermore, Nrp was reported to be the most abundant M. tuberculosis protein in the lungs of infected guinea pigs by day 30 postinfection, while being undetected at 90 days postinfection (18). Overall, these reports indicate that the cryptic metabolite produced by the nrp cluster is likely to play a role in virulence, which calls for a deeper investigation of nrp in the context of M. tuberculosis biology and infection.
In this study, we sought to query the role of the nrp gene in the pathogenesis of TB in the mouse model of infection. Our strategy consisted of the generation of M. tuberculosis mutant strains lacking the nrp gene. We herein reveal an important role for nrp in the virulence of M. tuberculosis, likely mediating the initial interaction of the bacteria with the host macrophages. We show that nrp-deficient M. tuberculosis grows at a lower rate in immunocompetent and immunodeficient mice, causing less lung pathology and associated with significantly increased survival rates.

RESULTS
nrp cluster in mycobacteria. The M. tuberculosis nrp gene (Rv0101) encodes a bimodular NRPS with domains predicted to activate Lys and Phe (10,11,14). The gene is present in a cluster of 5 genes including nrp plus Rv0097 to Rv0100. A sequence comparison analysis of the nrp cluster among 72 genomes from species representative of the phylogeny of the genus Mycobacterium (19) revealed that this cluster is absent in most rapid-growing nontuberculous mycobacteria (NTM), commonly found in the environment and considered nonpathogenic (Fig. 1A). The NTM exceptions harboring genes of the nrp cluster with more than 40% identity score to the genes of M. tuberculosis were Mycobacterium peregrinum, Mycobacterium fortuitum, Mycobacterium mageritense, Mycobacterium tusciae, and Mycobacterium thermoresistible. Most slowgrowing mycobacteria (31 of 48 [64.6%]) harbor orthologs to the M. tuberculosis Rv0097-Rv0101 genes with identity scores above 45% (Fig. 1A). This includes other pathogenic species, such as Mycobacterium leprae (50.6% identity) and Mycobacterium ulcerans (55.2% identify) (Fig. 1A). A synteny analysis showed that the nrp cluster is located in a homologous genomic DNA region shared by the chromosomes of the species of the Mycobacterium tuberculosis complex (MTBC) and M. leprae, without major sequence rearrangements. The exception is the considerably smaller size of the nrp gene in M. leprae compared to the nrp genes found in species from the MTBC (4,206 bp in M. leprae NC_002677 versus 7,539 bp in M. tuberculosis NC_000962) (Fig. 1B). As a result of this difference, the structural prediction is that the nrp gene found in MTBC strains encodes a bimodular NRPS, whereas the homologues from other mycobacteria, such as M. leprae, encode a single NRPS module. A detailed analysis of the level of nucleotide diversity of these genes within the MTBC was obtained by the analysis of 220 M. tuberculosis genomes from 7 MTBC lineages (20) and showed that the level  (19). The genomics sequences include complete genomes and whole-genome shotgun sequences. The highest percentage (dark green) corresponds to the longest and most elevated sequence identities. (B) Synteny of the genome sections harboring the nrp cluster in M. tuberculosis, M. bovis, M. canettii, and M. leprae. The green shading indicate homologous DNA regions shared by the four chromosomes without major sequence rearrangements. The genes of the nrp cluster are highlighted in yellow. (C) Nucleotide diversity () level of the M. tuberculosis genes Rv0097, fcoT, fad10, Rv0100, and nrp and the complete genome across a set of 220 genomes, including strains from lineage 1 to 7 of the MTBC.
nrp Is a Virulence Factor for M. tuberculosis of nucleotide diversity of these genes is below the genome-wide level (Fig. 1C). Thus, the comparative genome analysis suggests that the nrp cluster is highly conserved among slow-growing members of the Mycobacterium genus, highlighting the larger size of the nrp gene as the most characteristic MTBC-associated feature.
Loss of the nrp gene results in attenuation of M. tuberculosis H37RV in mice. To investigate the role of nrp in the virulence of slow-growing mycobacteria, we first generated a null mutant of M. tuberculosis H37Rv nrp by using specialized transduction to replace nrp with a hygromycin resistance cassette. Subsequently, we generated a complemented strain (nrp-comp) by introducing an integrative shuttle cosmid vector (containing an M. tuberculosis H37Rv cosmid fragment spanning Rv0096 to Rv0109) into the Δnrp mutant (see Fig. S1A in the supplemental material). To control for possible polar effects of the Δnrp deletion on the adjacent gene Rv0102, the expression of this gene was monitored in either of the recombinant M. tuberculosis strains. We found that the expression patterns of Rv0102 were similar in mutant and complemented bacteria (Fig. S1B). Furthermore, the axenic growth of the two engineered M. tuberculosis H37Rv strains was identical and similar to that of the wild-type (WT) H37Rv (Fig. S1C), showing no major growth defects imposed by the absence of nrp. Additionally, the Δnrp strain retained its acid-fast properties (Fig. S1D).
To assess the role of nrp in the pathogenesis of M. tuberculosis in the murine model of infection, immunocompetent (C57BL/6) mice were infected with aerosols of WT, Δnrp, or nrp-comp strains of M. tuberculosis H37Rv. Bacterial burdens in the lungs and spleens of infected mice were analyzed at different time points postinfection ( Fig. 2A and B). At 3 weeks postinfection, fewer colonies of the Δnrp strain were found in infected lungs than of the WT or nrp-comp strains ( Fig. 2A). However, the lung bacterial loads of all three strains were similar at later time points ( Fig. 2A), indicating that the Δnrp strain eventually grew to a level comparable to that of the WT strain in the long term. The differences in bacterial loads in the spleen were stark between the strains, with considerably fewer bacteria cultured from the spleens of mice infected with the Δnrp strain (Fig. 2B). Furthermore, we observed a significant difference in the survival of mice infected with the different strains. While all mice in groups infected with WT or nrp-comp strains died prior to 325 days postinfection, those infected with the Δnrp strain survived for up to 700 days postinfection (Fig. 2C). Given that we were able to complement the growth phenotypes in C57BL/6 ( Fig. 2A to C), we chose to focus our comparison of the Δnrp and nrp-comp strains for all further studies described below in an effort to limit the number of mice used in our experiments.
To further explore the growth kinetics of the Δnrp strain in the early stages of infection, we repeated the infection of C57BL/6 mice to monitor CFU levels in lungs from day 14 to up to 90 days postinfection. The difference in CFU between Δnrp-strainand nrp-comp-strain-infected mice was readily visible on day 14 postinfection, corresponding to a 48-fold lower bacterial burden (1.683 log 10 average difference) in the case of mice infected with the Δnrp strain on day 20. From day 30 postinfection and up to day 90, the Δnrp strain reached lung bacillary counts comparable to those of the nrp-comp strain (Fig. 2D), as observed before ( Fig. 2A).
⌬nrp-strain-infected mice develop less lung pathology. In addition to the differences observed in the progression of bacterial burdens, a distinct phenotype was also observed in lung histopathological features during infection with the Δnrp strain versus the nrp-comp strain ( Fig. 2E and F). Specifically, by day 20 postinfection, no discernible lesions were observed in the lungs of Δnrp-strain-infected animals, whereas multiple small inflammatory infiltrates were already present in the lungs of animals infected with the nrp-comp strain ( Fig. 2E and F). This was likely due to the differences in bacterial burdens observed at this time point (Fig. 2D). By day 30 postinfection, the lesions in the lungs of mice infected with the nrp-comp strain were larger than those in mice infected with the Δnrp strain ( Fig. 2E and F), even though both presented the same bacterial burden (Fig. 2D). Remarkably, the histological differences were maintained throughout the chronic stage of infection, as on day 90 postinfection, the animals infected with the nrp-comp strain still presented a greater degree of pathology than those infected with the Δnrp strain ( Fig. 2E and F), despite the fact that the bacterial burdens had been similar in both cases for 60 days (Fig. 2D). This subdued pathology may explain the longer survival of mice infected with the Δnrp strain ( Fig. 2C).
Kinetics of the immune response is altered in ⌬nrp-strain-infected mice. So far, our data indicated that in the absence of nrp, mice had a better ability to control the infection at initial time points and showed a striking lesser pathology even at late time points of infection. These findings suggest that a differential immune response may be occurring in the absence of Nrp function. To test this hypothesis, we next investigated the dynamics of immune cell populations elicited during infection with Δnrp or nrp-comp strains. We focused on myeloid (alveolar macrophages, inflammatory monocytes, and neutrophils) ( Fig. 3A to C) and lymphoid (CD4 ϩ T cells, CD8 ϩ T cells, and B cells) (Fig. 3D to F) cells known to be relevant to the pathogenesis of TB. Whereas no substantial differences were found throughout the infection with either strain in The weights of the animals were monitored to determine survival curves. On days 20, 30, and 90 postinfection, lung pathology (E and F) was determined by hematoxylin and eosin (H&E) staining and morphometric analysis of the right upper lobes of infected lungs. Data are shown as the means Ϯ standard errors of the means (SEMs) from 5 independent animals in at least 2 independent experiments. The pictures in panel F are of one animal representative of the experimental group; ϫ40 magnification. Statistical analysis was performed with a two-way ANOVA using Sidak's test for multiple comparisons (A, B, D, and E) or with log-rank (Mantel-Cox) test for the Kaplan-Meier curves (C). * and ϩ refer to statistical differences between Δnrp and nrp-comp strains or Δnrp and WT strains, respectively. * or ϩ, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; **** or ϩϩϩϩ, P Ͻ 0.0001.   Fig. S2 in the supplemental material. (I) On day 20 postinfection, the expression of CCL2, CCL7, IFN-␥, and TNF was determined by real-time PCR, as described in Materials and Methods. Data are shown as the means Ϯ SEMs from 5 independent animals in at least 2 independent experiments. The initial bacterial burdens were log 10 1.86 Ϯ log 10 0.04460 and log 10 1.931 Ϯ log 10 0.0822 for Δnrp and nrp-comp strains, respectively. Statistical analysis was performed with a two-way ANOVA using Sidak's test for multiple comparisons. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001. by day 90 postinfection (Fig. 3B and D). In addition to the decreased numbers of inflammatory monocytes and CD4 ϩ T cells observed 20 days postinfection in the absence of nrp expression, the number of activated cells within these populations, as measured by the expression of surface markers major histocompatibility complex (MHC) class II and CD44, was also decreased ( Fig. 3G and H). The gating strategy using in the definition of these populations is presented in Fig. S2.
At the molecular level, the diminished recruitment of inflammatory monocytes seen in the Δnrp strain may be explained by a lower expression of CCL2 and CCL7, chemokines that play a role in the recruitment of myeloid cells to the site of infection (21), in the lungs of these animals (Fig. 3I). In line with a lower frequency of activated CD4 ϩ T cells in Δnrp-strain-infected animals, a decreased expression of gamma interferon (IFN-␥) was observed in the lungs of these animals compared to that in nrp-comp-strain-infected mice (Fig. 3I). Tumor necrosis factor (TNF), another inflammatory mediator during TB, is also expressed at a lower level upon infection with the Δnrp strain (Fig. 3I).
In all, the absence of nrp function in M. tuberculosis improves the host's ability to control bacterial growth early postinfection, which is reflected in a lower bacterial burden. This is accompanied by diminished inflammatory monocyte and CD4 ϩ T cell recruitment and activation, as well as decreased expression of inflammatory mediators, which in the long-term likely lower the progression of lung pathology. Thus, these data suggest Nrp as a novel virulence factor for M. tuberculosis, altering the initial phases of infection and the immune response, but with a long-lasting impact on the pathology developed at the site of infection.
⌬nrp mutant is attenuated for growth in immune-deficient mice. Our findings point to a role for nrp during the early stages of M. tuberculosis infection. These findings led us to investigate for possible differences in the course of infection with Δnrp or nrp-comp strains in immune-deficient mice. We first infected SCID mice with aerosols of WT H37Rv, Δnrp, or nrp-comp strains and observed a remarkably longer survival rate for mice infected with the Δnrp mutant (see Fig. S3). As with the above-described experiments performed with immunocompetent mice, given that the survival patterns were restored upon complementation, we performed all subsequent experiments with immunodeficient mice using the Δnrp and Δnrp-comp strains. Next, we infected recombination activating 2 (RAG2)-deficient (Ϫ/Ϫ) mice and IFN-␥ Ϫ/Ϫ mice. Both mouse strains are unable to control M. tuberculosis growth and typically succumb to infection 40 to 50 days after infection (22,23). In line with this, both RAG2 Ϫ/Ϫ and IFN-␥ Ϫ/Ϫ mice succumbed to infection with the nrp-comp strain ( Fig. 4A and B). Strikingly, both mouse strains infected with the Δnrp mutant registered a significantly higher survival rate (average survival of 114 days against 44.7 days in RAG2 Ϫ/Ϫ mice and 99 days against 45 days in IFN-␥ Ϫ/Ϫ mice). An analysis of lung CFU counts indicated that the Δnrp mutant grew at a lower rate in the lungs of infected RAG2 Ϫ/Ϫ mice (Fig. 4C). However, at the time of death, the lung bacterial burdens observed for each nrp strain were identical in both RAG2 Ϫ/Ϫ (Fig. 4C) and IFN-␥ Ϫ/Ϫ (Fig. 4D) animals. Furthermore, at the time of death, the lungs of RAG2 Ϫ/Ϫ (Fig. 4E) and IFN-␥ Ϫ/Ϫ (Fig. 4F) mice infected with the Δnrp strain displayed significantly less pathology than their nrp-comp counterparts. Therefore, our data are indicative that in the absence of acquired immunity or IFN-␥, the attenuated M. tuberculosis growth due to the loss of nrp function was much more prolonged, associating again with reduced pathology. This set of data supports a more effective control of bacterial burden by the host's innate immunity, in the presence of limited pathology, in the absence of nrp function. Primary macrophages are more efficient at controlling the growth of the ⌬nrp strain. Altogether, our findings suggest a more effective control of the Δnrp strain by the host during the innate immune response. Since macrophages are one of the first types of cells to be infected by M. tuberculosis and to exert microbicidal functions (24), we questioned whether the expression of nrp impacted the bacterial control by infected macrophages. For this, we resorted to an in vitro system where mouse bone marrow-derived macrophages (BMDM) or macrophages isolated from the peritoneal  cavity (pMac) were infected with either of the strains, and the intracellular bacterial burden was assessed 96 h later. This time point ensured macrophage viability, thus avoiding the phenotypes related to cell death, rather than intracellular bacterial growth. The growth of the nrp-comp strain at this time point, in either of the macrophage types, was higher than that observed for the Δnrp strain (Fig. 4G). We then investigated the impact of exogenous IFN-␥ on the ability of BMDM to control the nrp-deficient and -complemented M. tuberculosis strains. Exogenous IFN-␥ improved the control of either of the strains by the macrophages, ablating the differences observed in bacterial growth (Fig. 4H). Of note, the macrophages infected with either of the M. tuberculosis strains showed similar patterns of cell viability and of cytokine production (see Fig. S4), suggesting a specific phenotype linked to bacterial control in the absence of Nrp. Altogether, these data support a virulence-associated role for Nrp in M. tuberculosis infection during the initial interaction with host macrophages. However, intriguingly, once the macrophages become fully activated, the host advantage over the Δnrp M. tuberculosis strain was lost. These in vitro findings are in line with the in vitro data, where the attenuation of the nrp-deficient M. tuberculosis strain is particularly notable during the early phase of infection but lost once the acquired immune response is established.

DISCUSSION
TB remains a devastating disease to mankind, with high human and economic tolls (25,26). To efficiently tackle TB, an improvement of several current tools is needed, notably with regard to prevention, diagnosis, and treatment. Owing to the alarming drug resistance rates, which caused 480,000 cases of complicated TB in 2016 (1), there is a pressing need for the development of novel treatment strategies. In this context, a clearer understanding of the key virulence factors in M. tuberculosis may help in finding new vulnerable and druggable targets. Different experimental approaches have been pursued for the identification of essential M. tuberculosis genes (2, 4, 16), including target-based screening. In this study, we aimed at elucidating the role of the nrp gene in the pathogenesis of TB, resorting to a genetically engineered nrp-deficient mutant M. tuberculosis strain, as well as in vitro and in vivo models of infection.
Previous studies reporting that the nrp gene harbored by M. tuberculosis is essential for in vivo growth (16) and expressed during early time points postinfection (18) support a role for the cryptic metabolite produced by Nrp during infection. Furthermore, our in silico analysis showed that nrp is part of a cluster of genes that is rare among the mostly nonpathogenic rapid-growing NTMs and highly conserved among the slow-growing pathogenic species of the genus Mycobacterium. Overall, these findings support a role for the product of the nrp cluster in the pathogenesis of infection by Mycobacterium species.
A previous screen aimed at finding M. tuberculosis persistence mutants in isoniazidtreated mice identified a transposon mutant with an insertion upstream of Rv0097, a gene encoding a putative oxodoreductase and possibly also involved in the biosynthesis pathway of the cryptic metabolite associated with nrp (17). The infection of WT mice through an intravenous route showed that the transposon mutant was attenuated in lungs and spleen (17). This former study thus supports the notion that one or more genes in the Rv0096-Rv0101 cluster play a role in survival during in vivo infection. In addition to the findings obtained with the transposon mutant (17), the study presented herein characterizes this operon in a more thorough way from in silico to biological data. Indeed, by using the aerosol route of infection, we show that the nrp cluster is associated not only with a better transient control but also with altered lung pathologies and demonstrate its ability to cause delayed mortality in infected immunocompetent mice. Accordingly, we show that in immunocompetent hosts, the nrp-deficient M. tuberculosis strain showed a delayed growth pattern in the lungs during the first 20 days of infection compared to that of its WT H37Rv and complemented counterparts. This delay in growth at the site of infection was accompanied by the delayed dissemination of the nrp-deficient M. tuberculosis strain to the spleens of the infected mice, improved survival rates, and a lessened pathology even at late stages of infection and in immunocompromised hosts. This is important, as lung pathology is a major cause of TB pathogenesis, needed for bacterial transmission and associated with posttherapy complications (27,28). Thus, Nrp may be a bacterial modulator of lung pathology. It is possible that the decreased expression of inflammatory molecules, such as IFN-␥ and TNF, in lungs infected with the nrp-deficient M. tuberculosis strain contributes to the observed decreased pathology.
Our findings collectively support a role for nrp during the early phases of the immune response and demonstrate that M. tuberculosis has evolved to express genes that manipulate innate defenses. This early role for Nrp during infection fits with a previous report showing that the in vivo expression of nrp is restricted to the early phases of infection (18). Accompanying the lower bacterial burden in the lungs of M. tuberculosis Δnrp-infected mice, we found a delayed recruitment and activation of inflammatory monocytes and CD4 ϩ T cells. However, the recruitment of other immune cell populations was similar between the two infections, possibly suggesting that the differences observed for inflammatory monocytes and CD4 ϩ T cells did not result solely from the reduced bacterial burdens upon M. tuberculosis Δnrp infection. The differential course of infection in terms of dissemination and pathology caused by the M. tuberculosis Δnrp strain may be linked to the differential recruitment of inflammatory monocytes. Previous studies showed that the induction of CCL2 during mycobacterial infection is associated with the recruitment of permissive monocytes to the lesion, which then support bacterial growth (21,29). It is interesting to speculate that, as in the absence of nrp, CCL2 induction is lower and fewer permissive monocytes are recruited, and so a lower bacterial burden and less dissemination are observed. On day 30 postinfection, these differences are lost, and similar numbers of inflammatory monocytes are present in the lungs of mice infected with either of the M. tuberculosis strains.
The most notable and interesting finding on the role of nrp in TB pathogenesis came from the observation that the lack of this gene highly attenuates M. tuberculosis infection in immunodeficient hosts. We were surprised to discover that RAG2 Ϫ/Ϫ , SCID Ϫ/Ϫ , and IFN-␥ Ϫ/Ϫ mice infected with the M. tuberculosis Δnrp strain survived substantially longer than those infected with the complemented counterpart. As shown in the case of the RAG2 Ϫ/Ϫ mice, the growth of M. tuberculosis Δnrp over time was much slower than that of the complemented strain. This suggests that in the absence of acquired immunity, the growth of M. tuberculosis Δnrp remains in check, to the benefit of the host. A similar result is inferred from the infection of IFN-␥ Ϫ/Ϫ mice, as the bacterial burdens at the time of death were comparable between M. tuberculosis Δnrp and its complemented counterpart. Most importantly, independent of the genetics of the mouse host, the lung pathology resulting from infection with M. tuberculosis Δnrp was much less exuberant that that observed for the complemented strain. The bacterial burden phenotype of the M. tuberculosis Δnrp strain is most likely mediated by the bacterial interaction with the macrophage. Indeed, we showed that primary mouse macrophages control the bacterial growth more efficiently when nrp is absent. However, the mechanistic bases of this better control remain unknown but do not seem to depend on the altered production of, at least, TNF, interleukin 1␤ (IL-1␤), and interleukin 10 (IL-10). Further studies, perhaps encompassing global transcriptional analysis, are thus required to further explore this observation. As for the in vivo situation, once the macrophages became fully activated by IFN-␥, the impact of the absence of nrp is lost.
In all, our findings highlight a role for the nrp cluster in evading the basic microbicidal mechanisms of the macrophage, which is then overcome by the adaptive immune response. It is tempting to speculate that the mechanisms that enable this result from host-pathogen coevolution events. Several outstanding questions arise from this study: which is the mechanism hijacked by Nrp at the level of innate immunity? How does the acquired immune response abrogate the impact of Nrp in the macrophage? As an answer to these questions, the infection of mice deficient for Toll-like receptor 2 (TLR2), TLR4, IL-10, arachidonate 5-lipoxygenase (ALOX5), or arachidonate 15-lipoxygenase (ALOX15) with the M. tuberculosis Δnrp strain showed the same phenotype as WT mice (see Fig. S5 in the supplemental material). Therefore, basic recognition mechanisms, as well as an imbalanced anti-inflammatory response, do not seem to be associated with the nrp phenotype. The precise metabolite produced by the nrp cluster in M. tuberculosis remains unidentified, as we were unable to detect an nrp-related metabolite in lab-grown cultures of M. tuberculosis, and future efforts will likely focus on the identification of the product ex vivo in infected tissue. Though challenging, these experiments will enlighten the biological role of the nrp cluster as well as the best timing to target it from a therapeutic point of view. Despite the importance of pursuing these questions in the future, our study provides strong evidence linking the nrp cluster to M. tuberculosis attenuation, specifically in the context of an immunocompromised host. Considering the huge impact of the HIV epidemics on TB, our data call for immediate efforts to be put toward a further understanding of the biological role of Nrp and the development of therapies targeted to immunocompromised TB patients.
Construction of a M. tuberculosis H37Rv nrp null mutant. A null mutant of M. tuberculosis H37Rv nrp (Rv0101) was generated by Specialized Transduction using previously described protocols (33,34). Briefly, a derivative of temperature sensitive phage phAE159 was constructed to replace nrp (phΔnrp) with a hygromycin resistance gene (hyg). Hygromycin resistant colonies obtained following transduction of M. tuberculosis H37Rv were verified by Southern blot to confirm the chromosomal replacement of nrp with hyg. One such transductant was termed Δnrp and used in all subsequent experiments.
Generation of an nrp-complemented strain. A 28-kb M. tuberculosis H37Rv cosmid fragment (spanning genomic position bp 104933 to bp 133536) was packaged in the integrative shuttle cosmid vector pYUB412-Kan to generate pYUB2420. This recombinant cosmid spans a region that extends from Rv0096 to Rv0109. pYUB2420 was introduced into the Δnrp strain by electroporation, and six kanamycinresistant transformants were confirmed by Southern blot to contain chromosomally integrated pYUB2420. One of these transformants, termed nrp-comp, was selected for further studies.
To determine the concentration of M. tuberculosis in aliquots, 6 frozen aliquots were serial diluted and plated on Middlebrook 7H11 (BD Biosciences) agar plates supplemented with 10% OADC and 0.5% glycerol. Viable bacteria were determined by CFU enumeration after 21 to 28 days of incubation at 37°C.
Ethics statement. All animal experiments were performed in strict accordance with the recommendations of the European Union Directive 2010/63/EU. All experimental procedures were approved by the local animal ethics committee, and licensed by the Portuguese National Authority for Animal Health (DGAV, Portugal) with reference number 014811/2016-07-13. Animal procedures were conducted by laboratory personnel with accreditation for animal research given by DGAV. The mice were euthanized by CO 2 inhalation, with efforts to minimize suffering.
Mouse strains. C57BL/6 mice were purchased from Charles River Laboratory (Barcelona, Spain). RAG2-deficient mice were purchased from Instituto Gulbenkian de Ciência (IGC) or generously provided by Margarida Correia-Neves (ICVS), and IFN-␥ deficient mice were generously provided Susana Roque (ICVS). All transgenic mice were of a C57BL/6 genetic background. SCID mice were obtained from Charles River Laboratory (USA). Experimental infection. Mice were infected with M. tuberculosis H37Rv (WT), Δnrp, or nrp-comp strains via the aerosol route by using an inhalation exposure system (Glas-Col), according to previously published protocols (35,36), which resulted in the delivery of 100 to 200 viable bacteria to the lungs.
Survival. The weights of the infected mice were determined every week after infection for immunocompetent mice and every week after infection for the first 30 days and every 48 h from day 30 postinfection onwards for immunodeficient hosts. Mice were humanely euthanized if they lost 20% of their maximum weight or upon losing responsiveness to physical stimulation. Whenever possible, the lungs of moribund animals were harvested for histology and bacterial burden assessment.
Organ processing. At the indicated time points postinfection, the mice were euthanized by CO 2 asphyxiation, and the organs were aseptically excised and processed as described previously (35,36). Briefly, the lungs were perfused by applying 10 ml phosphate-buffered saline (PBS) through the right ventricle of the heart to flush blood cells, and single-cell suspensions were prepared in complete Dulbecco's modified Eagle medium (cDMEM; DMEM supplemented with 10% fetal bovine serum [FBS], glutamine, HEPES, and sodium pyruvate, all from Gibco). Single-cell suspensions were used for bacterial burden determination, flow cytometry analysis, and RNA extraction. The number of cells was counted using a Countess automated cell counter (Life Technologies).
Bacterial burden determination. The assessment of initial bacterial burden was performed 3 days postinfection by growing viable bacteria from whole-lung homogenates. For the other time points, lung and spleen single-cell suspensions were incubated with 0.1% saponin (Sigma-Aldrich) for 10 min to release intracellular bacteria. The CFU was determined by plating 10-fold serial dilutions of saponintreated cell suspensions on Middlebrook 7H11 agar plates supplemented as described in "Bacterial growth and stocks." BBL MGIT PANTA antibiotic mixture (BD Bioscience) was added to prevent the contamination of lung samples. Viable M. tuberculosis colonies were counted after 21 to 28 days of incubation at 37°C.
RNA extraction, reverse transcription, and real-time PCR. Total RNA from infected lungs was extracted with TRIzol reagent (Invitrogen), and cDNA was synthesized and analyzed by real-time PCR, as described previously (35,36). Target IFN-␥ and TNF mRNA expression was quantified using SYBR green (GrisP) and specific oligonucleotides and normalized to the ubiquitin mRNA levels. The expression of CCL2 and CCL7 was quantified using specific primer-probe sets (Applied Biosystems) and normalized to the expression of hypoxanthine phosphoribosyltransferase (HPRT).
Histology analysis. The right upper lobes of the lungs were fixed in PBS containing 3.7% formaldehyde, embedded in paraffin, sectioned in 2-to 3 -m-thick slices, and stained with hematoxylin-eosin (35). The lung surface area of inflammation was measured using ImageJ software (version 1.50e; NIH). The percentage of total lung area involved with inflammation was calculated by dividing the cumulative area of inflammation by the total lung surface area for each sample.
Infection of bone marrow-derived macrophages and peritoneal macrophages. BMDM were differentiated from bone marrow cells obtained from the femurs and tibiae of C57BL/6 mice and cultured in the presence of L-cell conditioned medium, as described previously (37). pMac were obtained by intraperitoneally injecting 1 ml of thioglycolate into WT C57BL/6 mice. After 4 days, the mice were euthanized and the peritoneums were washed with PBS. In both cases, 1 ϫ 10 6 cells were infected with Δnrp or nrp-comp strains at a multiplicity of infection (MOI) of 1 bacteria to 1 macrophage. Four hours after infection, the cells were washed 4 times with PBS (Gibco) to remove extracellular bacteria. The adherent cells were cultured at 37°C in 1 ml of cDMEM in the presence or absence of 100 U/ml of IFN-␥ for 96 h. Four or ninety-six hours postinfection, 0.1% saponin (Sigma-Aldrich) in PBS was added to the wells, and the cells were incubated at room temperature (RT) for 10 min to release intracellular bacteria. The number of viable bacteria was determined by plating 10-fold serial dilutions of the saponin-treated cell suspensions on supplemented Middlebrook 7H1 agar, as described in "Bacterial growth and stocks." Cytokine determination by ELISA. TNF, IL-1␤, and IL-10 concentrations in the supernatants of infected macrophages were determined 24 h postinfection by enzyme-linked immunosorbent assays (ELISAs) with commercially available kits (eBioscience), according to the manufacturer's instructions. Cytokine levels from uninfected cells were below the assay level of detection (not shown).
Cell viability assay. Cell viability was determined in noninfected or infected macrophages 2 or 4 days postinfection. Macrophages were gently detached from low-adherence tissue culture plates using a cell scraper. Cell viability was assessed by using trypan blue exclusion.
Statistical analysis. Data were analyzed using GraphPad Prism 6. The differences between groups were analyzed with two-way analyses of variance (ANOVAs) using Sidak's tests for multiple comparisons. Kaplan-Meier survival curves were analyzed with the log-rank (Mantel-Cox) test. Differences were considered significant at P values of Յ0.05.