Granuloma-Specific Expression of Mycobacterium Virulence Proteins from the Glycine-Rich PE-PGRS Family

See allHide authors and affiliations

Science  26 May 2000:
Vol. 288, Issue 5470, pp. 1436-1439
DOI: 10.1126/science.288.5470.1436


Pathogenic mycobacteria, including the agent of tuberculosis, Mycobacterium tuberculosis, must replicate in macrophages for long-term persistence within their niche during chronic infection: organized collections of macrophages and lymphocytes called granulomas. We identified several genes preferentially expressed whenMycobacterium marinum, the cause of fish and amphibian tuberculosis, resides in host granulomas and/or macrophages. Two were homologs of M. tuberculosis PE/PE-PGRS genes, a family encoding numerous repetitive glycine-rich proteins of unknown function. Mutation of two PE-PGRS genes produced M. marinum strains incapable of replication in macrophages and with decreased persistence in granulomas. Our results establish a direct role in virulence for some PE-PGRS proteins.

Pathogenic mycobacteria initiate long-term infection by entering host macrophages where they cause extensive remodeling of their vacuolar environment to prevent vacuolar acidification and lysosomal fusion (1,2). Replication in macrophages plays a crucial role in persistence in vivo, distinguishing pathogenic from nonpathogenic mycobacteria (3, 4).

Because mycobacteria are facultative intracellular pathogens, we hypothesized that they selectively express specific genes in host macrophages to facilitate replication and persistence in vivo. Consequently, we used differential fluorescence induction (DFI), a fluorescence-activated cell sorter (FACS)–based method (5–7), to search for M. marinum genes activated in macrophages. M. marinum, which grows relatively rapidly at 25° to 35°C, causes a systemic tuberculosis-like disease in fish, frogs, and other cold-blooded animals, which are its natural hosts, and a peripheral disease in warm-blooded animals (8–11), including a peripheral granulomatous disease in immunocompetent humans (12). The clinically silent phase of infection in leopard frogs (Rana pipiens) is characterized by granulomas consisting largely of modified (epithelioid) macrophages and containing only a few M. marinum (10). The general pattern of infection is similar to that of clinically silent (latent) human tuberculosis.M. marinum can be adapted easily to rapid growth at 37°C and can cause systemic infection of lungs and spleens of mice that protects against a subsequent challenge with virulent M. tuberculosis (13). M. marinum is closely related to M. tuberculosis genomically (14). Both organisms also occupy identical vacuolar niches in cultured mammalian macrophages (1, 2). Yet unlike M. tuberculosis, M. marinum is neither an aerosol nor serious (biosafety level 3) human pathogen.

DFI was adapted to enrich for M. marinum sequences with enhanced expression in macrophages (Fig. 1A) (7, 15,16). Eight macrophage-activated promoters (maps) were identified from three promoter-trap libraries of M. marinum DNA fragments that represented about one-fourth of the M. marinum genome (Table 1 and Fig. 1A). We found very highly activated macrophage-induced mycobacterial promoters compared with earlier studies (5,6), perhaps as a result of our modification of DFI, where bacteria were subjected to both positive and negative cell sorting. All the identified mags (macrophage-activated genes) are homologs of M. tuberculosis genes (Table 1) (17). Intracellular activation of themaps was confirmed by confocal microscopy, showing that many of them were expressed weakly or not at all outside macrophages (Fig. 1B). All six maps tested were also active in granulomas of frogs for several months (Fig. 1C and Table 1), suggesting that the bacteria are sequestered in an intracellular niche during persistent infection.

Figure 1

Macrophage-dependent bacterial gene expression. (A) Independent maps exhibited varied levels of ex vivo and macrophage-dependent gene expression. J774 macrophages were infected for 24 hours with M. marinumbearing various promoter fusions (23). The geometric mean fluorescence of bacteria released from the macrophages (x-2) and bacteria grown ex vivo (x-1) (15) was analyzed and quantitated with a FACStar cytometer and CellQuest Software, respectively (Becton Dickinson). Histograms show relative fluorescence intensity of bacteria grown ex vivo (shaded) and those released from macrophages (solid line). The magnitude of macrophage-activated fluorescence (Table 1) was calculated by dividing x-2 by x-1. For M. marinum bearing the constitutively expressed promoter fusion hsp60::gfp (28, 29), x-1 and x-2 are identical as expected. GFP, green fluorescent protein. (B) J774 cells were grown on glass cover slips and infected with M. marinum bearing hsp60::gfp or map24::gfp for 24 hours and stained with polyclonal antisera to M. marinum followed by rhodamine-conjugated antibody to rabbit immunogloblin G (red), without permeabilizing the macrophage monolayer, so as to stain extracellular bacteria only. The cover slips were examined by laser confocal microscope with a dual [rhodamine and fluorescein isothiocyanate (for GFP)] filter. Intracellular bacteria are designated by an arrowhead, and extracellular bacteria by an asterisk. For M. marinum hsp60::gfp, the intracellular bacteria are green fluorescent and the extracellular bacteria appear yellow, being both green and red fluorescent owing to GFP expression and antibody staining. For M. marinum map 24::gfp, promoter activation causes the intracellular bacteria to fluoresce green, but the extracellular bacteria are pure red, indicating no GFP expression outside the macrophage. (C) A fluorescentM. marinum map 24::gfp within a liver granuloma. Cryosections (10 μm) were prepared from livers and spleens of frogs infected with M. marinum bearing the differentmap gfp fusions so as to produce chronic granulomas (10). The frog cell nuclei were stained with propidium iodide. Laser confocal microscopy was used to capture images of serial 1-μm sections each under red and green filters, which were combined to build a single image showing the red nuclei of the granuloma cells and the green bacteria. Bar, 4 μm (B and C).

Table 1

Characteristics ofmaps and mags. Promoters with intracellular activation of fluorescence (Fig. 1A), trapped by differential fluorescence induction (15, 16), were sequenced with a primer derived from the pFPV27 vector (5) and their sequences compared with the National Center for Biotechnology Information (NCBI) sequence databases. Putative functions were assigned based on those determined or postulated for close homologs (>50% amino acid identity). Percent amino acid identities are derived from the mag sequences contained in the promoter insert spanning lengths of 64 to 130 amino acids. ND, not determined.

View this table:

To determine if genes in addition to the mags were activated in host granulomas, we used DFI to screen directly for genes induced in chronically infected frogs (10, 15,16). Seven independent promoter sequences activated within granulomas were identified from one promoter library. map25, map 85, and map 86 were reisolated, confirming that these genes are expressed in granulomas of infected animals, as well as in cultured macrophages. In addition, we identified a second class of promoters activated specifically in host granulomas but not in cultured macrophages. Their downstream genes are homologs of M. tuberculosis Rv0321 (dcd), encoding a deoxycytidine triphosphate deaminase; Rv1106c, encoding a probable cholesterol dehydrogenase; Rv0631c (recC), encoding RecC; and Rv1205, encoding a protein of unknown function (17).

Two genes selectively expressed in both macrophages and granulomas,mag 24 and mag 85, are homologous to members of the large M. tuberculosis PE/PE-PGRS protein family, which comprises roughly 5% of the coding DNA of M. tuberculosis(Table 1 and Fig. 1) (17). The genes of this family are scattered throughout the genome of M. tuberculosis and other closely related mycobacteria (14, 17, 18). This family, characterized by a relatively conserved ∼110–amino acid NH2-terminus, is subdivided into the PE and PE-PGRS subfamilies (17). In the PE-PGRS proteins, the conserved NH2-terminus is followed by a region of irregularly spaced glycine dipeptide repeats spanning 100 to >500 residues. Most terminate just after the glycine-rich region, while a few have unique COOH-terminal extensions beyond. The function of these proteins is unknown and has been the subject of considerable speculation (17, 19). The PE-PGRS genes vary among clinical isolates of M. tuberculosis(18), leading to the hypothesis that they represent a source of antigenic diversity or that their glycine repeats inhibit host major histocompatibility complex class I processing, akin to the glycine repeats of the Epstein-Barr virus EBNA-1 protein (17).

Sequence analysis of the M. marinum genome flankingmap 24 revealed a PE-PGRS gene (mag 24-1) (Fig. 2 A) (17, 20). However, we found two additional open reading frames (mag24-2 and mag 24-3) downstream of mag 24-1, both encoding PE-PGRS proteins homologous to MAG 24-1 (Fig. 2, A and C). All three MAG 24 proteins have the unique COOH-terminal extensions homologous only to those of the M. tuberculosis Rv1651c and Rv3812 proteins, which are, in turn, homologous to each other (Fig. 2C) (17).

Figure 2

Identification, homologies, and disruption of the mag 24 genes. (A) Overlapping Sac I and Bgl II fragments flanking map 24 of the M. marinum genome were isolated by genomic cloning and sequenced. The 440–base pair (bp) map 24 insert included 186 bp of thepheT gene and 47 bp of mag 24-1. In M. tuberculosis also Rv1651c and pheT are adjacent but in the opposite orientation to each other. The kanamycin insertion mutations L1, P59, and L2 were made in the three mag 24 genes, respectively, in either the Sac I or Bgl II fragment (21). The resultant suicide plasmids were used to create theM. marinum gene disruption mutants L1D (mag24-1), P59D (mag 24-2), and L2D (mag 24-3) by sucrose counterselection (21). The complemented mutant L1D-T was created by reintroducing the Bgl II fragment into L1D with selection for the apramycin resistance gene in the plasmid (21, 30). (B) Southern blot analysis of Bgl II–digested chromosomal DNA from the wild-type (lanes W), L1D mutant (lane M), and the complemented mutant L1D-T (lane C) probed with the map 24 insert probe (Fig. 1A), showing the 1.2-kb upward shift of the L1D Bgl II fragment and the presence of both wild-type and mutant alleles in L1D-T. (C) Protein similarity profile for MAG24-1, MAG24-2, MAG24-3, Rv1651c, and Rv3812. Their predicted protein sequences were aligned with the GCG programpileup with a zero gap extension penalty, and the average similarity of the alignment (dotted line) calculated in the GCGplotsimilarity program with the blosum62 scoring matrix. The five proteins are similar at their NH2- (N) and COOH- (C) termini (50% identity or greater) but not in their intervening glycine-rich region (G), whereas the other M. tuberculosisPE-PGRS proteins have homology only to the NH2-terminus of these proteins.

To dissect the roles of the three MAG 24 proteins, we constructedM. marinum strains with insertion mutations in each of the three mag 24 genes by homologous recombination (Fig. 2, A and B) (21). All three mutants grew at rates similar to those of wild-type M. marinum in minimal medium (22, 23), indicating that they are not auxotrophic mutants and the proteins do not play a role in bacterial replication outside host cells.

The mag 24-1 mutant, L1D, and the mag 24-3 mutant, L2D, did not replicate in macrophages (Fig. 3A). The mag 24-2 mutant, P59D, was not attenuated for macrophage replication (Fig. 3A), consistent with the finding that a transcriptional fusion of the region upstream of mag 24-2 (map 24-2 in Fig. 2A) togfp was not expressed either constitutively or in macrophages (22). The intergenic distances between the threemag 24 genes (Fig. 2A), coupled with the results of the mutational analysis, indicated that each of the mag 24 genes was transcribed independently. MAG 24-1 and MAG 24-3 may be functionally redundant because macrophage persistence was not completely abrogated in either of these mutants (Fig. 3A). To confirm the role of MAG 24-1 in macrophage replication, we introduced the wild-type mag 24 genes into L1D by homologous recombination (Fig. 2, A and B). The resulting partial merodiploid strain, L1D-T, was restored for macrophage replication (Fig. 3A).

Figure 3

Effect of the mag 24 mutations on replication of M. marinum in J774 macrophages and frogs. (A) J774 macrophages were infected with the wild-type, L1D, L1D-T, P59D, and L2D M. marinum strains (23) with the substitution of streptomycin for amikacin. Intracellular counts obtained periodically (23) are represented as the means of triplicate samples. Error bars represent standard errors of the means. The 0 time point is 5 hours after infection. The infecting inocula were virtually the same for all strains except that of the wild-type strain, which was about sevenfold lower. The plateau in growth of P59D and L1D-T at day 11 is due to a cytopathic effect on the monolayer, leading to the release of the intracellular bacteria, which are not counted in the assay (23). (B) Frogs were injected intraperitoneally with wild-type (3.8 × 105 bacteria per frog) or L1D mutant (7.0 × 105 bacteria per frog) for the individual assay, and with a mixture of 3.8 × 105wild-type and 7.0 × 105 L1D M. marinum for the competition assay. Bacterial counts were obtained from spleens and livers at 8 weeks after infection (10). For the competition assay, the infected tissue homogenates were plated on solid bacterial medium in the presence and absence of kanamycin to distinguish wild-type (kanamycin-sensitive) from mutant (kanamycin-resistant) bacteria. The mutant bacteria were out-competed by the wild-type bacteria by 10- to 24-fold.

The L1D mutant bacteria also had diminished persistence in chronic granulomas of leopard frogs (10), persisting at an ∼25-fold lower level than wild-type bacteria in spleen (Fig. 3B) with similar results being obtained in liver (22). Histopathological examination of livers from the L1D-infected frogs revealed either a marked decrease in the granulomatous response or a qualitative change in the host response to the infection (15, 22). Whereas frogs infected with wild-typeM. marinum have defined granulomas composed mainly of epithelioid macrophages, the L1D-infected animals had more diffuse lesions consisting primarily of lymphocytes (10,15).

The mag 24-1 gene first induced at 7 to 20 hours after entry into cultured macrophages (22) was expressed indefinitely in granulomas of infected frogs (Fig. 1C). Consistently, disruption ofmag 24-1 diminished M. marinum replication in macrophages as well as persistence in granulomas. Our results show an immediate role in pathogenesis for some PE-PGRS proteins by facilitating macrophage replication, similar to the effect of the secreted Erp protein of M. tuberculosis(24). Their contribution to bacterial replication in cultured macrophages, where antigen presentation does not play a role in bacterial survival, indicates that they affect persistence in granulomas independent of, or in addition to, an effect on antigen presentation. A recent screen of macrophage-inducedM. tuberculosis promoters expressed in the attenuatedMycobacterium bovis BCG (bacillus Calmette-Guerin) strain has also revealed a PE-PGRS gene (6).

There is precedence for virulence determinants to be interspersed in highly variable regions in other bacteria, e.g., the Neisseria gonorrhea pili and the streptococcal M proteins (25). Although the subcellular localization of the PE-PGRS proteins is not known, their amino acid composition predicts them to be surface proteins like the M proteins and pili (17, 22).

M. marinum and Mycobacterium ulcerans are phylogenetically the most closely related mycobacteria to the M. tuberculosis complex (14). This is not to say that M. marinum is a surrogate for M. tuberculosis, but that these two pathogenic mycobacteria likely share a common ancestor (14) and employ indistinguishable strategies to replicate in macrophages and persist in granulomas. Perhaps the most telling similarity is that at the histological level, M. marinum infection of the human dermis is virtually indistinguishable from human dermal tuberculosis (12, 22).

The hallmark of chronic mycobacterioses, whether in the mammalian or amphibian host, is the maintenance of lifelong asymptomatic infection. The metabolic and replicative state of the mycobacteria in granulomas is largely unknown owing to the experimental intractability of investigating such lesions (26). Our ability now to identify mycobacterium gene classes specifically induced in different phases of chronic granulomas provides new avenues to better understand central questions of mycobacterial infection and disease.

  • * To whom correspondence should be addressed. E-mail: lalitar{at}


View Abstract

Stay Connected to Science

Navigate This Article