Mycobacterium marinum SecA2 Promotes Stable Granulomas and Induces Tumor Necrosis Factor Alpha In Vivo

Abstract

SecA2 is an ATPase present in some pathogenic Gram-positive bacteria, is required for translocation of a limited set of proteins across the cytosolic membrane, and plays an important role in virulence in several bacteria, including mycobacteria that cause diseases such as tuberculosis and leprosy. However, the mechanisms by which SecA2 affects virulence are incompletely understood. To investigate whether SecA2 modulates host immune responses in vivo, we studied Mycobacterium marinum infection in two different hosts: an established zebrafish model and a recently described mouse model. Here we show that M. marinum ΔsecA2 was attenuated for virulence in both host species and SecA2 was needed for normal granuloma numbers and for optimal tumor necrosis factor alpha response in both zebrafish and mice. M. marinum ΔsecA2 was more sensitive to SDS and had unique protrusions from its cell envelope when examined by cryo-electron tomography, suggesting that SecA2 is important for bacterial cell wall integrity. These results provide evidence that SecA2 induces granulomas and is required for bacterial modulation of the host response because it affects the mycobacterial cell envelope.

INTRODUCTION

Although Mycobacterium tuberculosis was identified as the causative agent of tuberculosis in humans over 100 years ago, many of the mechanisms leading to the success of this pathogen are still unclear. M. tuberculosis has infected one-third of all humans (19, 23), yet only 5 to 10% of infected people develop active disease. Nonetheless, humans are the only reservoir for M. tuberculosis, and individuals with active disease are key to the wide dissemination of the organism. Therefore, it appears that the bacterial strategy to fill its ecological niche is to limit severe pathology to those few people who will act as important vectors of dissemination. It has been known for some time that these individuals have compromised cellular immunity due to old age, malnutrition, or intercurrent illness. Thus, the ability of the vast majority of individuals to mount an immune response that contains the infection and limits disease is key to the success of M. tuberculosis as a human pathogen. Identifying host and pathogen factors that achieve this balance of susceptibility versus immunity to M. tuberculosis is of great basic biological interest and is important for efforts aimed at eradicating this disease.

M. marinum is a model species for M. tuberculosis infection that causes a chronic granulomatous disease in ectotherms that shares many features with tuberculosis (7). Less commonly, it is able to cause localized disease in humans in the form of skin lesions (22) and disseminated disease in immunocompromised humans (28, 36). M. marinum is the closest genetic relative to the M. tuberculosis complex and shares virulence factors with M. tuberculosis (40). Additionally, zebrafish and mice infected with M. marinum develop caseating granulomas that are very similar to those found in active human pulmonary disease (3, 39, 41). Granuloma formation, maturation, and maintenance are complex processes that are thought to benefit the host by containing the bacteria but may also benefit the bacteria by creating a niche for growth (9, 11, 16). Because of the genetic and pathological similarities between M. tuberculosis and M. marinum infections, as well as the fact that both can cause both primary progressive and chronic disease, it is thought that the two organisms share similar mechanisms of establishing disease and modulating the host immune response.

Several protein secretion systems are required for growth and virulence of mycobacteria (10). SecA2, found in all mycobacteria and some Gram-positive bacteria (31), is required for the secretion of several proteins, including the antioxidant enzymes SodA and KatG (2, 14). Mutation of secA2 in M. tuberculosis has revealed a role for SecA2 in the growth, survival, and virulence of the pathogen in vivo prior to the onset of adaptive immunity (2). In addition, studies in murine bone marrow-derived macrophages (BMDMs) suggest that SecA2 is required to inhibit the immune response (21). Additional studies have shown that loss of SecA2 enhances priming of antigen-specific T cells in vivo, and vaccination with a ΔsecA2 mutant consequently increases protection against M. tuberculosis infection in mice and guinea pigs (15, 30), suggesting that SecA2 plays a role in modulating host immunity.

The mechanism of immune modulation by SecA2 in vivo is not well understood, either from the perspective of host immunity or in terms of how SecA2 affects bacterial physiology. We have studied M. marinum infection in both zebrafish and mice to investigate this question. Our data suggest that SecA2 modulates adaptive immunity to promote granuloma stability, perhaps through induction of tumor necrosis factor alpha (TNF-α). SecA2 also is required for normal bacterial cell envelope function and appearance, suggesting that this may be the mechanism through which it modulates host immune responses.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Wild-type M. marinum strain M and subsequent mutant strains used in this study were derived from a human clinical isolate (ATCC BAA-535). The ΔkasB strain has been described previously (13). The ΔsecA2 strain, an M. marinum secA2::Kan^r mutant with an insertion mutation in the MMAR_2698 gene, was constructed using a previously published double-selection strategy (12). Flanking upstream sequence was amplified using PCR with genomic DNA template and primers SecA2U5 (5′-TAATACTAGTTGAACAGCACATTCAGTC-3′) and SecA2U3 (5′-TATAGATATCTAGGCCAGGTTTGATCGG-3′), where genomic sequences are italicized. Likewise, downstream sequence was amplified using primers SecA2D5 (5′-TAATGATATCTGGCACCGATATCCGGTT-3′) and SecA2D3 (5′-TTAATGCATACTAGTACAGACCCCAGATCAGAAACG-3′). These two flanking sequences were sequentially cloned into pCR 2.1-TOPO (Invitrogen). A kanamycin resistance gene derived from pUC4k (GE Healthcare) was then inserted between the upstream and downstream segments using EcoRV. This entire piece containing the upstream-Kan^r-downstream segment was finally cloned into pLYG304 (12). The mutant was then isolated after double selection, and the mutated genomic region was PCR amplified and tested with restriction digestion (see Fig. S1 in the supplemental material) and confirmed by DNA sequencing. The ΔsecA2 strain was complemented with secA2 (MMAR_2698) to produce the SA2^comp strain, by integration of this gene and the mycobacterial hsp60 promoter into the ΔsecA2 chromosomal attB site using the pMV306.hyg shuttle vector (35). For experiments, strains were grown to mid-log phase at 30°C and 105 rpm in Middlebrook 7H9 broth (Difco) supplemented with 0.2% glycerol, 0.05% Tween 80, and 10% Middlebrook albumin-dextrose-catalase enrichment (BBL) or at 32°C on Middlebrook 7H10 agar plates (Difco) supplemented with 0.5% glycerol and 10% Middlebrook oleic acid-albumin-dextrose-catalase enrichment (BBL). Medium was supplemented with 10% sucrose, kanamycin (30 μg/ml), zeocin (50 μg/ml), or hygromycin (50 μg/ml) as appropriate.

Macrophage infections.

Murine BMDMs were cultured from C57BL/6 wild-type mice and infected with M. marinum as described previously (24). For all experiments, 4 × 10⁵ BMDMs were seeded in 12-well plates overnight and then infected at a multiplicity of infection (MOI) of 1. For activated-macrophage experiments, BMDMs were incubated with recombinant mouse gamma interferon (IFN-γ; 30 ng/ml) for 24 h before infection and included in the medium for 4 h postinfection. The TNF-α concentration was measured by enzyme-linked immunosorbent assay (eBioscience) at 24 h postinfection.

Zebrafish Infections.

Wild-type zebrafish (AB strain) were purchased from the Zebrafish International Resource Center (Eugene, OR) and maintained in static tanks under conditions and standards specified by the IACUC at Genentech, Inc., and outlined previously (8). For infections, fish were anesthetized with 0.015% tricaine methanesulfonate (MS-222) for 3 to 5 min and then injected intraperitoneally with 10⁴ M. marinum bacteria in 50 μl of phosphate-buffered saline (PBS) or 50 μl of PBS alone for controls. To assay the bacterial load in the whole fish, fish were euthanized in 0.05% MS-222 for 20 min, bathed in 70% ethanol, and then homogenized in 0.1% Triton X-100 (1.5 ml) using disposable tissue grinders (VWR 47732-450). Each fish homogenate (250 μl) was decontaminated of normal flora using BBL Mycoprep reagent (BD). Tenfold serial dilutions were plated on 7H10 plates for enumeration of colony forming units (CFU). Viability of the M. marinum strains was unaffected by incubation with 0.1% Triton X-100 (1 h) or Mycoprep reagent (25 min). For histopathological analysis, fish were euthanized at 7 days postinfection, fixed in 10% formalin for 7 days, and then processed whole for paraffin-embedded histology. Serial pairs of parasagittal sections with maximal representation of pancreatic and liver tissue were stained by hematoxylin-eosin (H&E) and Ziehl-Neelsen stains. Total pancreatic foci of three or more clustered acid-fast bacteria were enumerated, as were those clusters specifically associated with recognizable granulomas.

Mouse infections.

All studies were compliant with the ILAR guidelines and were approved by the IACUC at Genentech, Inc. Twelve-week-old, female C57BL/6 mice (Jackson Laboratory-West) and Rag2-knockout mice (B6.129S6-Rag2tm1Fwa N12; Taconic) were inoculated with 10⁷ M. marinum bacteria in 100 μl of PBS via injection into the tail vein. For analysis of visible tail lesions, the lengths of individual lesions were measured and combined for each tail. For analysis of bacterial load in the tail, tails were severed at the tail base, weighed, cut into 5-mm pieces, and homogenized in 3 ml 0.1% Triton X-100 in Dulbecco's modified Eagle's medium, using an AHS 200 homogenizer (VWR) with sawtooth adaptors (10 × 105 mm; Troemner). Tenfold serial dilutions of homogenates were plated on 7H10 plates for enumeration of CFU, which were calculated as the number of CFU per gram of tissue. For cytokine analysis, tails were severed from the tail base and immediately placed on dry ice. While on dry ice, tails were cut into 5-mm pieces, frozen in liquid nitrogen, and pulverized with a biopulverizer (Biospec Products) that was chilled in liquid nitrogen. The tissue was incubated on ice for 1.5 h in 1 ml of PBS supplemented with complete, EDTA-free, protease inhibitor cocktail (Roche) and then centrifuged twice at 20,000 × g and 4°C for 20 min. Supernatants were collected for Bradford analysis (Bio-Rad) to determine total protein content for normalization of each tail sample and Luminex analysis (Bio-Rad) to quantify cytokine amounts in each tail. For histology, tails were fixed in 10% formalin, minimally decalcified in Immunocal decalcifier (Decal Chemical Corp.), trimmed into at least five 3-μm cross sections, embedded in paraffin, sectioned, and stained by H&E and Ziehl-Neelsen stains. Additionally, immunohistochemical staining for F4/80 (macrophages), CD3 (T cells), and B220 (B cells) was performed.

Quantitative RT-PCR analysis.

For analysis of secA2 expression levels in M. marinum, cultures were grown to mid-log phase in 7H9 broth. Cultures were incubated for 5 min with lysozyme and RNA was isolated using an RNeasy kit (74104; Qiagen). A second DNase treatment was performed per the manufacturer's instructions (M0303S; New England BioLabs). Quantitative reverse transcription-PCR (RT-PCR) was performed using a QuantiFast SYBR Green PCR kit (Qiagen 204054), the gene-specific primers listed in Table 1, and an ABI 7500 RT-PCR system. The fold induction in the ΔsecA2 and SA2^comp strains was calculated relative to that of the wild-type bacteria, using the ΔΔC_T (C_T is threshold cycle) model with SodA as the reference gene and no template and no reverse transcriptase reactions for controls. For zebrafish cytokine analysis, previously published primer sets were used (17, 32, 42, 43). Whole fish were euthanized in 0.05% MS-222 for 20 min and then homogenized in 1 ml TRIzol (Invitrogen) per 50 mg of tissue. RNA was extracted from tissues according to product instructions and stored at −80°C. cDNA was synthesized using Anchored Oligo(dT)₂₀ primer (12577-011; Invitrogen) and Superscript III reverse transcriptase (18080-044; Invitrogen). Quantitative RT-PCR was performed as stated above. Fold induction of experimental groups was calculated relative to that of the PBS control group, using the ΔΔC_T model with β-actin as the reference gene and no template and no reverse transcriptase reactions for controls.

Table 1.

Primer sets used to quantitate gene expression by quantitative RT-PCR^a

Host and cytokine or gene	Forward primer (5′–3′)	Reverse primer (5′–3′)	Reference or source
Zebrafish
β-Actin	TGCTGTTTTCCCCTCCATTG	TTCTGTCCCATGCCAACCA	Rojo et al. (32)
IL-1β	TGGACTTCGCAGCACAAAATG	GTTCACTTCACGCTCTTGGATG	Watzke et al. (43)
IL-12	TCTAACTTCAGCGCAGTGGA	TGCGGTGGTGTAGTGAGTG	Ito et al. (17)
IFN-γ	CTTTCCAGGCAAGAGTGCAGA	TCAGCTCAAACAAAGCCTTTCGCT	Vojtech et al. (42)
TNF-α	GATGGTGTCCAGGAGGAAAG	CAGAGTTGTATCCACCTGTT	Ito et al. (17)
M. marinum
secA2	GTGGGTCAACTCGTCAAA	CTCTTCGGAGTACAGATCGA	This work
sodA	ACAAGCTGCTGATCTTCCAGGTCT	TGGCGAAGTCGACCTTCACATTCT	This work

^aThese primers were used for quantitative RT-PCR using QuantiFast SYBR Green. Sequences for all zebrafish cytokine primers were obtained from published studies, while those for M. marinum were developed here.

SDS sensitivity assay.

M. marinum (2.5 × 10⁵ ml⁻¹) grown to mid-log phase in 7H9 broth was inoculated into 7H9 containing sodium dodecyl sulfate (SDS; 0.01, 0.05, 0.1, and 0.25%, vol/vol) and incubated for 20 h at 30°C and 105 rpm. CFU were enumerated on 7H10 agar plates.

Cryo-electron tomography.

M. marinum grown to mid-log phase in 7H9 broth was prepared for microscopy as described previously (5).

Statistical analyses.

Significance was determined using one-way analysis of variance (ANOVA) testing across all the samples analyzed in a given experiment. When calculated F values were greater than critical F values in these analyses, Fisher's least-significant-difference post hoc tests were applied to determine pairwise significant differences. Significance is denoted in the figures (*, P < 0.05; **, P < 0.01).

RESULTS

SecA2 is needed for virulence and growth in zebrafish.

Previous studies have shown that M. tuberculosis lacking SecA2 is partially attenuated for growth and virulence in a mouse model of infection (2, 21). To investigate the role of SecA2 in modulating the host immune response, we created an M. marinum strain with secA2 disrupted by a kanamycin-resistance cassette (ΔsecA2) via homologous recombination (12, 29) (Fig. 1A). We verified that the ΔsecA2 strain contained the expected gene-disrupted secA2 by PCR amplifying the mutated region with primers annealing outside the regions found in the knockout construct, diagnostically checking the product with restriction enzyme digestion (see Fig. S1 in the supplemental material), and sequencing this product. To complement the mutant, secA2 driven by the mycobacterial hsp60 promoter was stably inserted into the ΔsecA2 strain's genomic DNA at the attB site (35). The ΔsecA2 plus secA2 complemented strain (SA2^comp) grown in 7H9 broth expressed 2.25-fold more secA2 mRNA than wild-type M. marinum, whereas secA2 was not detected in the ΔsecA2 strain by quantitative RT-PCR (Fig. 1B), indicating that the SecA2 protein indeed is not present in the mutant strain and is reconstituted in the complemented strain. When grown in liquid medium, the ΔsecA2 strain grew similarly to the wild-type and SA2^comp strains (Fig. 1C), consistent with the reported in vitro growth of the M. tuberculosis ΔsecA2 strain (2). ΔsecA2 also grew similarly to the wild-type and SA2^comp strains within isolated BMDMs, regardless of whether the macrophages were in an activated state (Fig. 1D). This lack of defects in vitro differs from mouse macrophage infections with the M. tuberculosis ΔsecA2 strain (21, 38) and likely reflects differences in bacterial physiology between M. marinum and M. tuberculosis. For instance, in the 96-hour time course of our macrophage infections with M. marinum here (Fig. 1D) and previously (18, 24), we measure 4 to 10 doublings of M. marinum, whereas 2 doublings have usually been monitored in the M. tuberculosis ΔsecA2 studies (21, 38).

Fig 1.

The ΔsecA2 strain is attenuated for virulence and growth in zebrafish. (A) Schematic diagram of construction of the M. marinum ΔsecA2 strain. Approximately 1.6 kb of secA2 was replaced with a kanamycin-resistance cassette by homologous recombination. (B) Quantitative RT-PCR analyses of secA2 mRNA expression in the M. marinum ΔsecA2 and SA2^comp strains relative to wild type. 7H9 broth cultures grown to mid-log phase were analyzed. (C) Growth of the M. marinum cultures in 7H9 broth. The optical density (OD) at 600 nm was measured over time. The ratio of the generation time of the ΔsecA2 and SA2^comp strains to that of wild type averaged 0.96 ± 0.09 (range, 0.92 to 1.09) and 1.06 ± 0.1 (range, 0.96 to 1.16), respectively, in four independent experiments. SEM values are given. (D) Unactivated (top) or activated (bottom) BMDMs from C57BL/6 mice were infected with M. marinum wild type, ΔsecA2, or SA2^comp at an MOI of 1, and intracellular bacterial growth was determined by enumeration of CFU from macrophage lysates at various times postinfection. Shown are data from one representative experiment, in which duplicate samples were analyzed. (E) Zebrafish were infected with various M. marinum strains via intraperitoneal injection in three independent experiments, and their survival was observed over time (wild type, n = 10; ΔsecA2 strain, n = 12; and SA2^comp strain, n = 11). (F) Bacterial burden in the whole fish was quantified by enumeration of CFU at multiple times postinfection. The values represent the mean ± SEM (n = 3 per time point for wild type; n = 6 per time point for the ΔsecA2 and SA2^comp strains). Data represent two independent experiments with similar results for the ΔsecA2 and SA2^comp strains. One of these experiments also analyzed wild type. The results for the ΔsecA2 and SA2^comp strains are significantly different (*, P < 0.03).

To assess whether SecA2 is required for virulence of M. marinum in vivo, we infected adult zebrafish. All zebrafish infected with 10⁴ wild-type bacilli died at between 12 and 25 days postinfection, as did fish infected with the SA2^comp strain (Fig. 1E). At the time of death, the abdomens of the zebrafish were visibly hemorrhaging, indicating an inflammatory response. However, no hemorrhaging was detected in zebrafish infected with 10⁴ ΔsecA2 bacilli, and all fish survived until 38 weeks postinfection, at which point they were euthanized. Thus, SecA2 in M. marinum is required for virulence in zebrafish. To determine whether the ΔsecA2 strain fails to kill the fish because it is more easily cleared by the host, CFU were enumerated from the whole fish at multiple times postinfection (Fig. 1F). The wild-type and SA2^comp strains showed steady growth throughout the first 11 days of infection. In contrast, the ΔsecA2 strain was comparable to the wild-type and SA2^comp strains in growth until 7 days postinfection but was cleared thereafter. These data suggest that SecA2 is required for continued growth and survival of mycobacteria in vivo, and this requirement is perhaps related to the onset of acquired immunity. These kinetics are consistent with the findings of a previous study in which adaptive immunity effects began to appear at between 8 and 11 days postinfection, when a similar dose of M. marinum was used to infect adult zebrafish (39).

SecA2 markedly enhances the inflammatory response in zebrafish.

Histopathological analyses of the liver and pancreas were performed at 7 days postinfection, the time point at which the bacterial burdens were indistinguishable (Fig. 1F). While zebrafish infected with both the SA2^comp and wild-type strains had equivalent numbers of granulomas in the liver and pancreas per zebrafish (6.6 ± 3.23 [standard error of the mean {SEM}] for the SA2^comp strain and 6 ± 0.89 for wild type), granulomas were rarely detected in ΔsecA2 mutant-infected zebrafish. In fact, granulomas were found in only 1 of 10 ΔsecA2 mutant-infected fish. In addition, ΔsecA2 mutant-infected fish had less overall inflammation, although the quality of the inflammation was similar (Fig. 2). Thus, the ΔsecA2 strain induced less inflammation and fewer granulomas.

Fig 2.

SecA2 is needed for normal granuloma numbers in zebrafish. At 7 days postinfection of zebrafish with different M. marinum strains in two independent experiments, histopathological analysis was performed by H&E and Ziehl-Neelsen (acid-fast) staining. The granuloma burden per fish was enumerated from liver and pancreas sections.

SecA2 is important for disease and granuloma stability in mice.

To determine whether SecA2 had a similar role in a distant host, mice were infected with M. marinum as previously described (3). When C57BL/6 mice were infected with 10⁷ wild-type, ΔsecA2, or SA2^comp bacteria via the tail vein, disease was localized to cooler areas of the mouse, like the tail, ears, paws, and nose, presumably because the optimal growth temperature of M. marinum is ∼32°C. The burden of visible granulomatous tail lesions in each mouse was quantified over 7 weeks by measuring and combining the length of all lesions on each tail (Fig. 3A). Lesions began to appear at about 9 days postinfection in all infected mice. For the first 3 weeks, the disease caused by the ΔsecA2 strain was indistinguishable from that caused by the wild-type and SA2^comp strains (Fig. 3A). However, after 3 weeks, ΔsecA2 mutant-infected mice had less disease. In fact, lesions caused by the ΔsecA2 strain but not the wild-type or SA2^comp strain seemed to resolve (Fig. 3B). Hence, the ΔsecA2 strain is able to cause disease early in infection but is attenuated for virulence after 3 weeks. Histopathological analyses of individual tail lesions were performed at 2 and 7 weeks postinfection to analyze the granuloma architecture (data not shown and Fig. 4). There was no difference at either time point in the character of the inflammation of the lesions induced by the wild-type, ΔsecA2, or SA2^comp strain. Granulomatous structures rich in macrophages and T cells, with occasional B cells, were observed at both time points with H&E staining (data not shown and Fig. 4) as well as immunohistochemical staining (data not shown), regardless of the infecting bacterial genotype. Thus, in mice as well as zebrafish, SecA2 appeared to control the extent rather than the quality of the host inflammatory response.

Fig 3.

SecA2 is needed for virulence and stable granulomas in mice. (A to C) C57BL/6 mice were infected via injection of different M. marinum strains into the tail vein in two independent experiments. (A) The lengths of all visible lesions on each tail were combined at each time point, and the averaged values represent the mean ± SEM (n = 18 per group). Significant differences between the ΔsecA2 and SA2^comp strains are noted (*, P < 0.05; **, P < 0.01). (B) Representative images of tails from infected mice at 4 and 7 weeks postinfection. (C) Bacterial burden in the whole tail was quantified by enumeration of CFU at multiple time points postinfection. Growth of the ΔsecA2 strain is significantly different (**, P < 0.003) from growth of SA2^comp at 21 days.

Fig 4.

Granulomas in mice are histopathologically similar. Mice were infected via injection of 10⁷ M. marinum wild-type, ΔsecA2, or SA2^comp bacteria into the tail vein. Cross sections of three infected mouse tails per experimental group were analyzed at 7 weeks postinfection by H&E and Ziehl-Neelsen (acid-fast) staining.

The growth of the mycobacterial strains in the mouse tail was measured by determination of the numbers of CFU over a 7-week period (Fig. 3C). Growth of the M. marinum strains peaked at 21 days. The bacterial numbers then declined, likely due to the onset of adaptive immunity (3). Although the ΔsecA2 strain's bacterial load was significantly less than that of both the wild type (P < 0.04) and the complemented strain (P = 0.0001) at the peak of infection (Fig. 3C), the difference was small. While there was significantly more sustained inflammation in the lesions caused by the wild-type and SA2^comp strains (Fig. 3B), the ΔsecA2 bacteria were not cleared significantly more than wild type after 4 more weeks of infection. These data suggest that in mice, M. marinum SecA2 is primarily required for maintenance of inflammatory lesions rather than for bacterial growth.

SecA2 is required for robust induction of TNF-α in both zebrafish and mice.

Studying the immune response in zebrafish in detail is challenging due to the limited number of reagents available to detect proteins important in immunity and inflammation. Fortunately, many of the key cytokines and immune cells found in mammalian systems are genetically conserved (25, 37), allowing us to analyze the expression of various target genes in response to infection. Using quantitative RT-PCR, expression of the mRNA of several cytokines was quantified from infected fish at 3, 7, and 11 days postinfection (Fig. 5). While there were no differences in cytokine induction at 3 days postinfection, the ΔsecA2 strain induced less TNF-α and interleukin-1β (IL-1β) at 7 days postinfection, even though the bacterial load was similar at this point. By 11 days postinfection, the wild-type and SA2^comp strains induced greater than 4-fold more TNF-α and IL-1β than the ΔsecA2 strain, while they induced equivalent amounts of IFN-γ and IL-12. There was a marked increase in IFN-γ between 7 and 11 days postinfection, consistent with the onset of adaptive immunity, but this was unaffected by the presence of SecA2. Thus, SecA2 makes an important contribution to the increase in inflammatory macrophage cytokines (TNF-α, IL-1β) during M. marinum infection in zebrafish that is not paralleled by an increase in IFN-γ or IL-12, cytokines primarily involved in adaptive immunity (6).

Fig 5.

SecA2 promotes induction of TNF-α and IL-1β in zebrafish. Cytokine mRNA expression from whole fish infected with different M. marinum strains or a buffer control was measured by quantitative RT-PCR using the gene-specific primers listed in Table 1. Values represent average fold induction relative to the PBS-treated control fish ± SEM (n = 8 per group per time point for all groups, except n = 6 for the PBS-treated control set). Data averaged from two independent experiments are shown. ANOVA analyses found statistically significant differences between groups for TNF-α and IL-1β (*, P < 0.05; **, P < 0.01).

At 3 and 6 weeks postinfection, cytokine protein profiles were obtained from the tails of uninfected naïve mice and mice infected with wild-type, ΔsecA2, or SA2^comp bacteria (Fig. 6). At 3 weeks postinfection, the ΔsecA2 strain induced less TNF-α, IFN-γ, and IL-17 and more IL-10 than wild-type or SA2^comp M. marinum. Thus, in both fish and mice, the infection with the ΔsecA2 strain induced less TNF-α at a point of infection where the bacterial burden was comparable to that of wild-type-infected animals. IFN-γ production was diminished in mice but not in fish, whereas IL-1β production was diminished in fish but not in mice, suggesting that these two hosts have both unique and overlapping cytokine responses to M. marinum infection. TNF-α secretion by mouse macrophages infected with ΔsecA2 or wild-type strains was similar (see Fig. S2 in the supplemental material), suggesting that the SecA2-dependent difference in TNF-α seen in vivo is not the result of differences in direct interaction with macrophages but reflects more complex immune cell interactions during infection.

Fig 6.

SecA2 is needed for early induction of TNF-α, IFN-γ, and IL-17 in mice. At 3 and 6 weeks postinfection, cytokine levels from the whole tail of each experimental or naïve uninfected animal were quantified by Luminex analysis. Values represent the average amount of cytokine detected (pg/ml) divided by the total amount of protein present in each tail suspension (mg/ml) ± SEM (n = 6 per experimental group per time point). Data averaged from two independent experiments are shown. Significant differences in TNF-α, IFN-γ, and IL-17 were found at 3 weeks between infection with the M. marinum ΔsecA2 strain and infection with either the wild-type or SA2^comp strain (*, P < 0.05; **, P < 0.01). No cytokines were detected in uninfected mice.

Rag2-knockout mice are susceptible to ΔsecA2 infection.

The kinetics of disease in both zebrafish (Fig. 1F) and mice (Fig. 3A) following infection by the wild-type or ΔsecA2 strain suggested that the role for secA2 is more prominent after the onset of adaptive immunity. To better examine only the innate immune response, Rag2-knockout mice (34), which lack mature B and T cells, were infected with the ΔsecA2 or the SA2^comp strain. Infection of these immunodeficient mice led to more severe disease, and the infected Rag2-knockout mice began to die at 3 weeks postinfection (see Fig. S3 in the supplemental material). However, there was no difference in mortality rate between mice infected with the ΔsecA2 and SA2^comp strains. Diagnostic necropsy with H&E and Ziehl-Neelsen stains of the limbs, tail, and nasal tissue of these mice revealed cellulitis, osteomyelitis, and numerous acid-fast-positive bacilli associated with severe inflammation (data not shown). As expected, no granulomatous structures were found. Therefore, adaptive immunity is required for control of M. marinum infection in mice, and the ΔsecA2 strain can be as pathogenic as wild type if mature B and T cells are absent.

The ΔsecA2 strain has an altered cell envelope.

SecA2 is needed for secretion of two lipoproteins to the cell envelope in M. smegmatis (14), so we subjected the ΔsecA2 strain to several stresses in vitro to test the functional integrity of its cell envelope. The ΔsecA2 strain was much more sensitive to SDS than the wild-type or SA2^comp strain (Fig. 7). Compared to wild-type strains, SDS decreased the growth of the ΔsecA2 strain to levels similar to those for the ΔkasB strain (Fig. 7), a mutant that has a defect in mycolic acid synthesis and is known to have a defective permeability barrier (13). However, unlike the ΔkasB strain, the ΔsecA2 strain did not show increased sensitivity to hydrogen peroxide or rifampin (data not shown), suggesting that the cell envelope defects of the two mutants are distinct. To further test a role for SecA2 in cell envelope structure, we visualized whole bacilli by cryo-electron tomography. Unlike traditional electron microscopy, this technique preserves cellular architecture, including the cell envelope, and allows three-dimensional reconstruction of the intact bacterial cell (26). Twenty-five percent of the ΔsecA2 bacilli had large protrusions from the cell envelope that were not present in wild type (see Fig. S4 in the supplemental material). These protrusions differed from the ribbon-like loops observed in the ΔkasB strain. While the contents of these protrusions are unknown, these data suggest a role for SecA2 in maintaining the cell wall. Thus, SecA2 is required for the functional and structural integrity of the cell envelope.

Fig 7.

The ΔsecA2 strain has abnormal cell envelope function. M. marinum wild-type, ΔsecA2, SA2^comp, and ΔkasB strains were incubated with various concentrations of SDS (0.01, 0.05, 0.1, and 0.25%, vol/vol) in 7H9 broth for 20 h at 30°C. Sensitivity was determined by enumerating CFU, and the assay detection limit is shown with a dashed line. Mean ± SD is shown, and is averaged from two independent experiments. Statistically significant differences between the SA2^comp strain and both the ΔsecA2 and ΔkasB strains are noted (*, P < 0.05; **, P < 0.01). For simplicity, lower error bars are not shown for the ΔsecA2 and ΔkasB strains.

DISCUSSION

These studies demonstrate that SecA2 is required for full expression of disease during infection of both zebrafish and mice with M. marinum. A careful analysis of both pathology and host cytokines revealed that SecA2 is required for both TNF-α mRNA production and the stability of mature granulomas in both species. We hypothesize that SecA2-dependent secretion of TNF-α is responsible for the effects on granuloma numbers, since TNF-α is known to be required for granuloma formation and maintenance during mycobacterial infection (16). TNF-α has a well-described host-protective role during mycobacterial infection in mice, since either absence of TNF-α (1) or neutralization of TNF-α (27) enhances susceptibility to infection. Similarly, in zebrafish larvae infected with M. marinum, knockdown of TNF receptor 1 facilitated the formation of destabilized granulomas that disintegrated and increased mycobacterial growth, killing the zebrafish host more rapidly (4). Thus, it is likely that early induction of TNF-α is associated with initial host protection, while secretion later in the disease may be exploited by the bacteria through granuloma usage as a niche for survival and possibly growth (9). In this context, it is interesting that the primary effects of SecA2 on infection are seen only after infection has been established, likely at the onset of adaptive immunity. SecA2-mediated secretion apparently enhances this late TNF-α synthesis. We suggest that this is required for bacterial survival and pathology in a natural host (zebrafish) and for pathology in mice, which are incidental hosts. This model is consistent with our observation that loss of M. marinum SecA2 does not affect TNF-α secretion in isolated macrophages, since its effects are not apparent until after recruitment of additional mechanisms of adaptive immunity. Consistent with this, the ΔsecA2 strain is as virulent as a complemented strain during infection of Rag2-knockout mice, which lack mature T and B cells. Thus, we predict that SecA2-secreted products are likely to affect host cells other than the macrophage. It is instructive to compare these findings with those from studies on the role of SecA2 in M. tuberculosis infection of macrophages and mice (2, 21, 38). In these studies, SecA2-dependent effects on growth in isolated macrophages were discovered and correlated with an increase in TNF-α secretion by these macrophages. The difference between the requirement for SecA2 in M. marinum and M. tuberculosis for growth in isolated macrophages may result from the increased growth rate of M. marinum in macrophages, which likely increases other bacterial products in sufficient quantities to lead to TNF-α production in this circumstance. The normal growth of the M. marinum ΔsecA2 strain in vitro allowed us to uncover additional effects of SecA2 on host-pathogen interactions that occur only in the in vivo environment.

How SecA2 affects TNF-α mRNA expression in vivo is unknown. Previous work has identified a number of secreted proteins that depend on SecA2, but none have been convincingly tied to its role in pathogenesis. Interestingly, SecA2 is required for secretion of two lipoproteins to the cell envelope in nonpathogenic M. smegmatis (14). Our findings that M. marinum bacteria lacking SecA2 have increased sensitivity to SDS and have morphological abnormalities of the cell envelope are consistent with a role for SecA2 in cell wall synthesis in M. marinum as well. This could suggest a role for the cell wall in SecA2-dependent manipulation of host immunity. While much remains to be done to explore this hypothesis, roles for bacterial lipoproteins in triggering host responses through Toll-like receptors and other mechanisms are well-known (16, 20), as is the ability of cell envelope components to disseminate widely and regulate immune responses during mycobacterial infection (33).

Given that fish and mice are evolutionarily distant in their relatedness, it is not surprising that our studies also identified differences between them for cytokine induction in response to infection with M. marinum ΔsecA2. Measurement of mRNA in fish and protein in mice may have further contributed to the observed difference. Understanding the basis for these differences would require a comparison of immune responses in fish and mice more detailed than is currently possible. However, we believe that by focusing on the common responses between the distant hosts, we have found an important mechanism through which SecA2 modulates host immunity. In both fish and mice, SecA2 enhances TNF-α during the adaptive immune response to M. marinum infection. We suggest that this is important for establishment of well-formed granulomas, which in turn can act as a niche for bacterial survival and proliferation.