The identification of Mycobacterium marinum genes differentially expressed in macrophage phagosomes using promoter fusions to green fluorescent protein


  • Lucia P. Barker,

    1. Microscopy Branch, Rocky Mountain Laboratories, National Institute of Allergy and Infectious Diseases, 903 South 4th Street, Hamilton, MT 59840, USA.
    Search for more papers by this author
  • Diane M. Brooks,

    1. Microscopy Branch, Rocky Mountain Laboratories, National Institute of Allergy and Infectious Diseases, 903 South 4th Street, Hamilton, MT 59840, USA.
    Search for more papers by this author
  • P. L. C. Small

    1. Microscopy Branch, Rocky Mountain Laboratories, National Institute of Allergy and Infectious Diseases, 903 South 4th Street, Hamilton, MT 59840, USA.
    Search for more papers by this author

Lucia P. Barker , E-mail LBARKER@NIH.GOV; Tel. (406) 363 9252; Fax (406) 363 9371.


Mycobacterium marinum, like Mycobacterium tuberculosis, is a slow-growing pathogenic mycobacteria that is able to survive and replicate in macrophages. Using the promoter-capture vector pFPV27, we have constructed a library of 200–1000 bp fragments of M. marinum genomic DNA inserted upstream of a promoterless green fluorescent protein (GFP) gene. Only those plasmids that contain an active promoter will express GFP. Macrophages were infected with this fusion library, and phagosomes containing fluorescent bacteria were isolated. Promoter constructs that were more active intracellularly were isolated with a fluorescence-activated cell sorter, and inserts were partially sequenced. The promoter fusions expressed intracellularly exhibited homology to mycobacterial genes encoding, among others, membrane proteins and biosynthetic enzymes. Intracellular expression of GFP was 2–20 times that of the same clones grown in media. Several promoter constructs were transformed into Mycobacterium smegmatis, Mycobacterium bovis BCG and Mycobacterium tuberculosis. These constructs were positive for GFP expression in all mycobacterial strains tested. Sorting fluorescent bacteria in phagosomes circumvents the problem of isolating a single clone from macrophages, which may contain a mixed bacterial population. This method has enabled us to isolate 12 M. marinum clones that contain promoter constructs differentially expressed in the macrophage.


Mycobacterium marinum causes chronic skin lesions known as ‘swimming pool granuloma’ on the extremities of human hosts (Mollohan and Romer, 1961; Huminer et al., 1986; Gluckman, 1995; Joe and Hall, 1995; Ramakrishnan, 1997). In immunocompromised human hosts, including AIDS patients, infections can lead to systemic disease and death (Tchornobay et al., 1992; Hanau et al., 1994; Parent et al., 1995). Consistent with the prevalence of human disease in the extremities, the optimal growth temperature of the organism is 25–35°C, although some strains grow well at 37°C (Aronson, 1926; Clark and Shepard, 1963).

M. marinum is classified taxonomically as a slow-growing member of the genus Mycobacterium (Rogall et al., 1990; Stahl and Urbance, 1990) and has been shown to be closely related to M. tuberculosis by DNA–DNA homology and 16S RNA sequence studies (Tonjum et al., 1998). There have also been recent advances in the ability to manipulate the organism genetically by performing gene disruption via homologous recombination (Ramakrishnan et al., 1997a). Because of the relatively rapid growth of M. marinum (4 h doubling time compared with 20 h for M. tuberculosis), the designation of the organism as a biosafety level two pathogen and the phenotypic similarity of M. marinum and M. tuberculosis during a macrophage infection (Barker et al., 1997; Ramakrishnan et al., 1997b), it is an ideal organism for the study of the virulence mechanisms of pathogenic mycobacteria.

M. marinum and other pathogenic mycobacterial species have been shown to survive and replicate in macrophages and HeLa cells (Shepard, 1956; 1957; 1958; McDonough et al., 1993). In addition, several animal models are available for the study of M. marinum pathogenesis that have shown a correlation of the ability of the organism to survive in macrophages and its virulence in animal models (Ramakrishnan and Falkow, 1994; Ramakrishnan et al., 1997b). Many groups have sought to characterize the mechanisms by which pathogenic mycobacteria survive the hostile environment of the host macrophage (Sturgill-Koszycki et al., 1994; Xu et al., 1994; Clemens and Horwitz, 1995). In these studies, it has been shown that phagosomes containing pathogenic mycobacteria do not fuse with lysosomes, and the vesicles are only mildly acidified. Studies in our laboratory have shown that the intracellular trafficking of M. marinum is analogous to that of M. tuberculosis (Barker et al., 1997). The molecular mechanisms by which mycobacteria circumvent the host cell endosomal network are, however, unclear.

Valdivia et al. (1996) have constructed a shuttle vector for the expression of a mutated green fluorescent protein (mGFP) in M. marinum (Valdivia et al., 1996). This vector was developed as a tool for fluorescence-activated cell sorter (FACS) isolation of individual or intracellular bacteria. The mGFP has been optimized for fluorescence emission, solubility and kinetics of chromophore formation in bacteria (Cormack et al., 1996). M. marinum organisms expressing mGFP were not adversely affected with respect to replication or intracellular survival (Valdivia et al., 1996). This led to the development of a promoter-trap vector with a multiple cloning site upstream of a promoterless mGFP gene. Recently, a similar construct was used in differential fluorescence induction (DFI) assays to sort Salmonella typhimurium clones differentially expressing mGFP in low pH environments (Valdivia and Falkow, 1996) and within host cells (Valdivia and Falkow, 1997). In the DFI strategy, a library of small fragments of bacterial DNA cloned upstream of the promoterless mGFP gene is constructed, and promoter activity is assayed under different environmental conditions by the level of fluorescence. Individual or intracellular bacteria that are expressing the mGFP are then isolated with the use of a FACS to separate and isolate fluorescent bacteria or host cells.

It is our objective to elucidate the molecular mechanisms by which pathogenic mycobacteria are able to survive in the host cell environment. In this study, we have used the DFI system to isolate M. marinum clones containing bacterial promoters, which are differentially expressed intracellularly, successfully. A unique aspect of this work is the isolation of bacteria directly from phagosomes rather than from intact macrophages. Mycobacteria are very hydrophobic and tend to clump in culture. Despite efforts to minimize clumping, in most experiments many bacteria are taken up by macrophages in clusters (McDonough et al., 1993). Even when individual bacteria are phagocytosed, most macrophages will contain numerous organisms. Therefore, the population of bacteria within a single macrophage is usually heterogeneous. However, within 2–3 days of infection, the bacteria replicate intracellularly, and the majority of mycobacteria are found distributed in vesicles containing a single bacterium (Xu et al., 1994). The isolation of vesicles from infected macrophages makes it possible to obtain a pure culture of an individual clone. In our studies, vesicle preparations from infected macrophages were sorted by FACS to isolate microorganisms that expressed mGFP intracellularly. Using this method, we identified 12 clones containing mGFP fusions differentially expressed intracellularly. We further tested the ability of representative clones to express mGFP in different mycobacterial species.


Vesicle preparation was optimized for gentle lysis that disrupted the cell membrane without compromising the nuclear membrane. Nucleic acid contamination is viscous and will tend to clump the preparation to the point at which vesicles cannot be retrieved. We therefore chose to Dounce homogenize vesicles and monitor cell and nuclear membrane lysis by light microscopy. Vesicles were isolated from RAW murine macrophages that were uninfected or that were infected with 1 μm fluorescent latex beads. 1Figure 1A shows a transmission electron micrograph of a latex bead enclosed in a macrophage vesicle. Very few nuclei or whole cells were seen in electron micrographs of these preparations. Preparations of latex beads in vesicles were analysed by FACS to determine the best parameters by which to sort fluorescent M. marinum of approximately 1–4 μm away from background vesicles and/or cell debris. By calibrating the FACS instrument with latex beads of known size (1–6 μm), we could gate the FACS to isolate single organisms based on size and above-background fluorescence.

Figure 1.

. Electron micrographs of vesicle preparations. Transmission electron micrographs of vesicle preparations from macrophages infected with (A) 1 μm green fluorescent latex beads and (B) M. marinum. Bars represent 1 μm and large arrows point to (A) a 1 μm latex bead (lb) within a vesicle and (B) M. marinum organisms in vesicles. A variety of sizes of vesicles can be seen in both preparations (small arrows). As this is the post-nuclear fraction of the vesicle preparations, whole cells and nuclei have been removed by density gradient separation, and a variety of vesicles of different sizes remain. A majority of the organisms are in vesicles containing only one bacterium, and the phagosomal membrane surrounds the organism tightly.

After determining the optimal lysis and gradient spin steps for vesicle isolation (see Experimental procedures), macrophages were infected with stationary phase or presorted library PCL2M. This library contains 200–1000 bp fragments of M. marinum DNA cloned upstream of the promoterless mGFP gene. Approximately 6% of the PCL2M library clones were fluorescent on solid media. Infections of macrophages with PCL2M were allowed to proceed for 3 days, after which macrophages were lysed and a vesicle preparation performed to isolate individual M. marinum organisms (see Fig. 1B). As a majority of the organisms are in vesicles containing only one bacterium and the phagosomal membrane surrounds the organism tightly, the isolation of these phagosomes was relatively straightforward. Dilution plate counting of bacterial cfu from the samples obtained with the FACS indicated that a total of approximately 107 individual organisms in vesicles could be isolated and analysed with the FACS in each experiment.

Isolation of fluorescent clones by FACS

Two different preparations of the PCL2M promoter-capture library were analysed by FACS. Parameters were set to enrich for clones expressing mGFP intracellularly. In the first experiment, a stationary-phase culture of PCL2M was used to infect RAW murine macrophages. At 3 days after infection, the point at which most vesicles contain only one organism, the macrophages were lysed and the vesicles isolated. This vesicle preparation was analysed by FACS and was found to contain approximately 1.5% fluorescent events that corresponded to the size of single organisms in vesicles. Vesicles that were approximately 1–4 μm (i.e. that corresponded to a single bacterium within a vesicle) and were fluorescing above background levels were isolated away from the non-fluorescent population by gating the FACS to sort the desired population away from other organisms and vesicles (see Fig. 2A). A sample of the resulting suspension of organisms was then analysed and found to contain 58.6% organisms of high fluorescence and the proper size, an enrichment of 40-fold (Fig. 2B). This suspension was then diluted and plated for single cfus. Clones obtained by this isolation procedure were designated F-series (FACS-isolated) clones (e.g. F80).

Figure 2.

. FACS sorting of PCL2M. A. Presort. Macrophages infected for 3 days with the PCL2M library were lysed and the vesicle fraction isolated. After calibration of the FACS with 1 μm and 6 μm beads, the sorter was gated for events corresponding to 1–4 μm and above-background fluorescence (box). This sample contained 1.5% of the total events. Approximately 106 events corresponding to the proper size and fluorescence were collected. B. Post-sort. The sample gated as described above was reanalysed by FACS. Events of proper size and above-background fluorescence (box) were determined to be 58.6% of the total sample collected.

In order to reduce the number of FACS-selected clones that were constitutively expressing mGFP in tissue culture media, we presorted fluorescent organisms away from the PCL2M population. A culture of PCL2M was grown to stationary phase in DMEM media, and fluorescent clones (approximately 6% of the population) were FACS isolated away from the non-fluorescent population and discarded. The remaining non-fluorescent clones were used to infect RAW murine macrophages for 3 days, and vesicles were isolated for FACS analysis. In this experiment, 1.1% of the events corresponded to a single, fluorescent bacterium. A sample of this suspension of organisms was then analysed and found to contain 16.2% organisms of high fluorescence and the proper size, an enrichment of 15-fold. These organisms were also plated for single cfus. These clones were designated P-series (presorted) clones.

A total of 400 individual F-series clones and 300 individual P-series clones were tested for fluorescence in tissue culture media and intracellularly by infecting tissue culture wells with or without macrophages in parallel. Each well was examined visually and scored for fluorescence in the macrophage that exceeded fluorescence in media alone. In this way, four F-series clones (F80, F154, F278 and F320) and 12 P-series clones (P40, P99, P136, P140, P141, P144, P155, P163, P168, P170, P232 and P238) were chosen for further characterization. Although the enrichment of P-series clones was less successful than that for the F-series in the original FACS analyses, the percentage of P-series clones positively screened for differential expression within macrophages was higher [4/400 (1%) for F-series, 12/300 for P-series (4%)].

Fluorescence of clones in media and intracellularly by FACS analysis

In order to quantify promoter activity in the clones of interest, log-phase cultures of clones were seeded into 7H9 media, DMEM tissue culture media and RAW murine macrophage cultures. After 3–4 days, the macrophages were lysed with a mild detergent to release phagocytosed bacteria. Bacteria from macrophage lysates and the same bacteria grown in media were FACS analysed for fluorescence intensity. Strains GFP3R and G13 were included as controls. Strain GFP3R is the wild-type strain 1218R transformed with construct pMV262gfpmut3. This construct has the mGFP gene under the control of the mycobacterial heat shock promoter HSP60. Clone G13 was picked from solid media as a highly fluorescent constitutive promoter construct from the original PCL2M plating. Clone G13 contains a promoter insert that fluoresces at levels 10–20 times higher than the mycobacterial HSP60 promoter in GFP3R.

Five of the 16 original clones (F80, P40, P136, P140 and P170) exhibited no induction of fluorescence (fold induction ≤ 1.0) and were characterized no further. Table 1 shows the fold fluorescence induction of the remaining clones. Intracellular induction of M. marinum promoters ranged from 1.4-fold for clones F320 and F154 to an approximate 15-fold induction for clone P155. Most of the promoters were induced approximately two- to threefold in the macrophage with low standard error values. Construct P155, however, exhibits a great deal of variation in the amount of fluorescence induction observed depending on the experiment. Fluorescence intensity of P155 is, however, always higher in the macrophage than in tissue culture media. Preliminary work suggests that mGFP expression from the P155 construct may be dependent upon growth phase as well as environmental signals (data not shown), and kinetic studies are in progress.

Table 1. . Comparison of mGFP induction in macrophages and associated nucleotide sequences homologous to inserts. a. Fold induction in macrophages is the fluorescence value in macrophage lysates divided by the fluorescence values of the same clone grown in DMEM media supplemented with 10% fetal bovine serum. Background wild-type fluorescence has been subtracted from mean fluorescence values for each sample. Experiments were repeated at least twice, and values represent the mean fluorescence induction ± SEM.b. The genetic nomenclature assigned to each promoter fusion was based on homology to either known sequences or ORFs at the mGFP fusion junction. Those genes listed were determined as the most likely sequences reading into mGFP based on the polarity of transcription/translation, as determined by sequence homology. Those genes listed in bold have demonstrated homology with a P-value < 10−20.c. MTB, Mycobacterium tuberculosis.Thumbnail image of

Sequence analysis of selected clones

Partial sequences of the clones of interest were compared with known sequences with BLAST (Altschul et al., 1990) and MycDB (Bergh and Cole, 1994) database programs. Results of sequence comparisons are shown in Table 1. Of the inserts, G13, F320, P99, P141/144, P163, P232 and P238 are completely sequenced. None of these known sequences have coding regions that span the entire insert. This indicates that intergenic regions that may contain mycobacterial promoters are probably present rather than fortuitous promoters internal to open reading frames (ORFs). Further, all clones obtained exhibited homology to M. tuberculosis or M. leprae sequences, although, in some cases, the percentage identity was low. Sequence analysis identified clones P141 and P144 as sibling strains. Inserts G13, P99, P141/144, P168 and P238 exhibited homology to mycobacterial sequences or genes but did not yield information concerning specific gene products. Insert sequence from clone P163 showed high identity (< 10−24) to the Pseudomonas aeruginosa gene pvd D, a peptide synthetase-like gene that is located in an iron-starvation box in P. aeruginosa. A BLAST search of P163 ORFs yielded significant protein identity (< 10−17) between P163 and a peptide synthetase from Streptomyces pristinaespiralis.

The remaining clones, F154, F278, F320, P155 and P232, showed significant identity to known M. tuberculosis sequences. This is not surprising, as M. marinum has been shown to be closely related to M. tuberculosis (Tonjum et al., 1998). Clone P155 contained the insert sequence that exhibited the highest amount of induction in macrophages. The P155 insert sequence most probably reading into mGFP exhibits a high degree of identity (< 10−81) with an M. tuberculosis gene encoding a probable drug efflux pump. Clones F154 and P232 contain inserts with homology to M. tuberculosis housekeeping genes. F278 and F320 contain inserts homologous to the M. tuberculosis PPE protein family of unknown function and a putative M. tuberculosis membrane protein of unknown function respectively.

Transformation of other mycobacteria with control strains and representative clones

The expression of most mycobacterial genes identified thus far has not been restricted to a single species, e.g. many genes isolated from M. tuberculosis can be expressed in M. smegmatis (Jacobs et al., 1991). Nonetheless, we wanted to determine whether the clones isolated from the M. marinum PCL2M library would be expressed in other mycobacterial species. Strains M. bovis BCG, M. smegmatis and M. tuberculosis H37Ra were transformed with pMV262gfpmut3, G13, F154 and P155. Transformants were selected on kanamycin and colonies observed on a fluorescent microscope. Figure 3 shows representative transformant colonies compared with the same constructs in M. marinum. In all of the strains, mGFP fluoresced much more brightly with the G13 insert than with the HSP60 promoter construct. Additionally, fluorescence levels of F154 and P155 varied from strain to strain, but all of the insert constructs exhibited colony fluorescence to some degree, indicating that expression of mGFP from the promoters isolated from M. marinum would be expressed in other mycobacterial species as well.

Figure 3.

. Colonies of mycobacterial species transformed with promoter constructs. M. smegmatis, M. bovis BCG and M. tuberculosis H37Ra were transformed with pMV262gfpmut3, G13 and representative constructs from clones F154 and P155. The resulting colonies were then photographed on selective media and compared with M. marinum transformants (top row). Exposure times were identical for all strains (2 min), except for M. tuberculosis, for which the exposure time was 1 min in order to avoid overexposure. However, M. tuberculosis transformed with the P155 construct was exposed for 2 min.


Using an mGFP promoter-capture strategy outlined in Fig. 4, we have successfully isolated a series of promoter constructs that are differentially expressed within macrophages. By using a FACS to isolate macrophage vesicles rather than intact macrophages, we were able to circumvent the problems associated with isolating a mixed population of organisms from intact macrophages. We have identified 12 clones containing fusions that exhibit more fluorescence intensity when inside the macrophage than extracellularly. DNA isolated from these mGFP fusion constructs exhibited a high degree of identity with known M. tuberculosis DNA sequences. As M. marinum and M. tuberculosis are both able reside in similar phagocytic compartments (Barker et al., 1997) and are phylogenetically very closely related (Tonjum et al., 1998), it is probable that analogous molecular mechanisms are involved in the survival of these organisms in a hostile host cell environment.

Figure 4.

. FACS isolation of fluorescent GFP fusions expressed in macrophage phagosomes. Macrophages are infected with the M. marinum promoter-trap library PCL2M. After 3 days, the majority of phagosomes contain only one bacterium. Macrophages are lysed and vesicle preparations are sorted by size and fluorescence to enrich for clones specifically expressing GFP intracellularly.

This is the first report of the use of GFP as a tool for screening the promoter expression of a mycobacterial library intracellularly. However, several other groups have looked at the use of GFP as a reporter gene to quantify the amount of gene expression by mycobacterial species (Dhandayuthapani et al., 1995; Kremer et al., 1995; Parker and Bermudez, 1997). The addition of a procedure in which we isolate vesicles containing individual mycobacteria has enabled us to separate these fluorescent organisms from other bacteria, even in the same host cell, that express little or no mGFP. Further, by isolating vesicles rather than intact macrophages, we would be less likely to miss a weak promoter fusion that is differentially expressed within the macrophage in a background of non-fluorescent clones. The relatively low percentages of positive clones (1% of F-series and 4% of P-series) can be explained by the fact that this method is an enrichment rather than a selection. The clones isolated by FACS were contaminated with non-fluorescent clones and clones constitutively expressing mGFP. The sorting of organisms by FACS, however, enabled us to screen many fewer individual clones for differential expression in the macrophage.

Although the degree of induction of our promoter constructs appears to be relatively low in most cases, we are confident that our screening methods have identified several promoter constructs that are more active intracellularly. We believe that a reproducible two- or threefold intracellular induction of a particular gene could indicate a highly significant and physiologically relevant upregulation of a particular gene product. We found it necessary to use FACS analysis, rather than fluorometry, for quantification of promoter activity. The high concentration of organisms required for the measurement of population fluorescence confounded our data, as light scatter from the sample was too high to allow reproducible measurement of fluorescence intensity. The advantage of FACS analysis is that the measurement of fluorescence intensity for each individual bacterium is highly reproducible.

Although the level of differential expression of these M. marinum constructs is not as high as those found in enteric bacteria (Valdivia and Falkow, 1997), it is consistent with previous work done with mycobacterial systems. For example, Curcic et al. (1994) used a transcriptional fusion system with the xylE gene from Pseudomonas as a reporter gene and found a maximal eightfold reduction of mtrA gene expression when M. smegmatis was grown on various nitrogen sources. Dellagostin et al. (1995) found a maximum 6.3-fold intracellular induction of an M. leprae promoter sequence transformed into M. bovis BCG as measured by β-galactosidase activity.

In contrast to the induction of specific promoters in different environmental backgrounds, several groups have studied the range of promoter strength in mycobacterial species. Using the chloramphenicol acetyltransferase (CAT) reporter system and a LacZ reporter system, DasGupta et al. (1993) found that both M. tuberculosis and M. smegmatis exhibited an approximate 20-fold range of promoter activity, although M. tuberculosis promoters were less active than those from M. smegmatis (Jain et al., 1997). In a related study by Timm et al. (1994) also using a LacZ reporter system, an approximate 100-fold range of promoter activity was described for M. bovis BCG promoter constructs as well as M. smegmatis containing M. tuberculosis promoter constructs.

To summarize what is known about mycobacterial promoters thus far: (i) the induction of specific genes in different environmental backgrounds has not exceeded 10- to 20-fold using different reporter systems; (ii) the range of values for promoter strength within mycobacterial species is similar; and (iii) slow-growing M. tuberculosis appears to have many fewer strong promoters than does the fast-growing M. smegmatis. It is reasonable to suggest that the levels of induction that we have reported here represent biologically significant events.

Using the DFI strategy, we have identified 12 constructs that are upregulated in the macrophage. It is important to note that some of these fusions could contain chimeric inserts (i.e. non-contiguous fragments of insert DNA), and that this is most likely when the insert size is large. We also cannot rule out the possibility of cryptic promoters driving the expression of mGFP. In assigning gene identities to mGFP fusions, we have chosen the genes homologous to the insert sequences that are most likely to be reading into and driving the expression of mGFP. Further analysis is required before we know whether this preliminary assignment is correct.

All of the sequences we have characterized demonstrate homology to sequences from other mycobacterial strains, and some can be assigned non-ambiguous designation as specific genes. Not surprisingly, some of these sequences are homologous to housekeeping genes. We also found that the mGFP construct on the HSP60 promoter (pMV262gfpmut3) was induced in the macrophage. The upregulation of a heat shock/stress response promoter in the hostile environment of the macrophage was not unexpected. Dellagostin et al. (1995), however, did not measure any induction of the HSP60 promoter in macrophages when using a LacZ reporter system. This may be because of differences in sensitivity between the two assay systems. Another interesting result is the slight but reproducible induction of the constitutive and highly active G13 construct. The G13 promoter system expresses mGFP highly in all mycobacterial strains tested and, in M. marinum, is 10–20 times more active than the HSP60 promoter. As there are few strong mycobacterial promoters, this promoter could be extremely useful for obtaining a high yield of mycobacterial gene products. Work is in progress to characterize this promoter construct further.

The goal of this work is to identify mycobacterial genes required for macrophage survival and/or virulence. Whether the genes that we have identified are required for the survival of pathogenic mycobacteria in the macrophage can only be addressed by deleting the gene of interest and determining the effects of this deletion on the virulence of the organism. We can then explore the effects of deleting or interrupting these sequences in the wild-type strain in a macrophage tissue culture system with regard to bacterial survival and trafficking through the host cell.

Experimental procedures

Bacterial strains and growth conditions

M. marinum 1218R was originally obtained from the American Type Culture Collection (ATCC 927), as were M. tuberculosis H37Ra (ATCC 25177) and M. bovis BCG (ATCC 35743). M. smegmatis mc2155 was provided by William R. Jacobs, Albert Einstein College of Medicine, NY, USA. The M. marinum 1218R isolate has been shown to be virulent in the frog model (Ramakrishnan et al., 1997b) and in the guinea pig and fish models (unpublished data). Individual isolates of M. marinum 1218R were streaked for single colonies on 7H10 media (Difco) supplemented with 10% Middlebrook oleic acid, albumin, dextrose and catalase enrichment (OADC; Difco) and grown at 32°C for 5–7 days. A single colony was inoculated into 7H9 media (Difco) supplemented with 10% OADC and 0.2% Tween 80 and allowed to incubate at 32°C for 10–14 days. The cultures were then frozen in 0.5 ml aliquots at −70°C in 7H9 media with 10% OADC. This culture was designated passage 2. Aliquots were inoculated as needed into 30 ml of liquid growth media and grown without shaking for 7–10 days. These cultures were designated passage 3. Passage 3 M. marinum used for experiments were in stationary phase, having been in media for more than 10 but less than 30 days, unless otherwise specified. M. tuberculosis strain H37Ra, M. smegmatis mc2155 and M. bovis BCG were cultured under similar conditions, but grown and maintained at 37°C. Liquid and solid media were supplemented with 30 μg ml−1 kanamycin as needed. Escherichia coli strain DH5α was maintained in Luria broth or on LB agar with or without kanamycin.

Macrophage cell line

The mouse macrophage cell line RAW 264.7 (ATCC TIB71) was maintained at 37°C in 5.0% CO2 in Dulbecco's modified Eagle medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (Gibco). RAW macrophages are fully capable of phagocytosing bacteria and latex beads. Further, RAW cells are able to kill avirulent mycobacteria in vitro (unpublished data). RAW macrophages were not used beyond passage 10 from frozen stocks.

Library construction

Genomic DNA was isolated from M. marinum 1218R and digested for various time intervals with Sau3A. The digests were resolved on a 1.0% agarose gel, and bands corresponding to DNA fragments between 200 and 1000 bp, were excised from the gel. Size-fractionated DNA was phosphatase treated and ligated into the BamHI site of promoter-trap plasmid pFPV27 (constructed and provided by Raphael Valdivia, Stanford University). This construct is derived from vector pFPV2 (Valdivia et al., 1996) and contains the ori sequences of both mycobacterial and E. coli strains, a kanamycin resistance selectable marker and a promoterless mutant green fluorescent protein (mGFP) gene. Constructs were transformed into E. coli DH5α. A total of 2.8 × 105 individual kanamycin-resistant clones were generated in E. coli. These colonies were suspended in Luria broth with 15% glycerol and frozen in 1 ml aliquots. This E. coli library was designated PCL2 (for promoter capture library 2). Promoter capture library 1 was discarded as not being representative of the M. marinum genome. An aliquot of PCL2 was grown at 37°C overnight, and the plasmid library was isolated and transformed into M. marinum 1218R. A total of 3.5 × 104 individual kanamycin-resistant clones was generated in the 1218R transformation. Colonies were scraped off plates into 7H9 media supplemented with 10% OADC and frozen in 1 ml aliquots. This library was designated PCL2M (for promoter capture library 2 in M. marinum). Aliquots of PCL2M were grown for 3–4 days in 7H9 media to ensure that the bacteria were in log-phase growth. The log-phase PCL2M bacteria were then diluted 1:10 in 7H9 media or DMEM supplemented with kanamycin and grown for 10–14 days before use in experiments, unless otherwise specified.

Macrophage infections with bacterial strains

Macrophages were seeded into tissue culture dishes or microchambers at a concentration of 2 × 105 cells ml−1 total volume. Macrophages were grown to semi-confluence before bacteria were added at a multiplicity of infection (MOI) of 1–5:1. After 4 h of growth at 37°C and 5% CO2, the medium was removed, and fresh DMEM supplemented with 100 μg ml−1 amikacin was added to kill extracellular organisms. Monolayers infected with M. marinum were then transferred to 32°C. At 24 h after infection, the concentration of amikacin was reduced to 20 μg ml−1.

Isolation and FACS sorting of vesicles

To isolate vesicles, 20 tissue culture dishes of infected or control macrophages were placed on ice and washed three times with ice-cold phosphate-buffered saline (PBS), pH 7.2, and monolayers were scraped into 2.5 ml of PBS per plate. Cells were pooled into two tubes and spun down at 500 r.p.m. (100 × g) in a Beckman J-6M centrifuge for 5 min at 4°C. The supernatant was removed, and the cell pellets were gently resuspended in ice-cold homogenization buffer (HB). HB is 3 mM imidazole (Calbiochem) in 250 mM sucrose at pH 7.2, to which was added one tablet of complete protease inhibitor (Boehringer Mannheim) per 50 ml. This solution was immediately centrifuged at 1200 r.p.m. (200 × g) for 10 min at 4°C. The supernatant was removed, and 0.5 ml of ice-cold HB was added to each tube. Samples were combined and placed in a 1.0 ml glass Dounce homogenizer (Wheaton Scientific) and homogenized 35 strokes. A volume of 20 μl of this homogenized suspension was observed under light microscopy to ascertain that at least 50% of the macrophages were lysed without compromising the nuclear membranes. The homogenate was centrifuged at 2000 r.p.m. (500 × g) for 2 min at 4°C. The milky supernatant containing the vesicle suspension was then gently separated from the whole-cell and nuclear pellet for FACS analysis. Vesicle preparations of macrophages infected with latex beads or M. marinum were fixed with periodate–lysine–paraformaldehyde (Robertson et al., 1963; McLean and Nakane, 1974), subjected to 0.5% reduced OsO4 (McDonald, 1984) and visualized on a Hitachi model HU11-E1 transmission electron microscope at 75 kV. FACS analysis was performed on a Facstar instrument modified for five-parameter operation (Becton-Dickinson Immunocytometry Systems). Before each analysis, 1 μm and 6 μm green latex beads (Polysciences) were used to calibrate and scale the machine. For PCL2M library sorts, forward scatter (size) was on the x-axis and the fluorescence intensity on the y-axis. Fluorescence was measured using logarithmic amplification.

Fluorescence screening of selected clones

Fluorescent bacterial clones that were sorted away from non-fluorescent vesicles and vesicles containing non-fluorescent organisms were screened further by assaying fluorescence visually on a Bio-Rad MRC 1000 scanning confocal microscope. Control strains were GFP3R (pMV262gfpmut3 in 1218R) and G13, a highly fluorescent constitutive promoter fusion isolated from the original PCL2M library. Those clones that appeared to fluoresce intracellularly but not on 7H10 agar or in media were inoculated into DMEM media or into culture dishes containing RAW macrophages at a MOI of 1–5:1. Growth rates of all clones measured by optical density did not vary from that of the wild-type M. marinum 1218R (data not shown). After 3–4 days of growth in either media or macrophages, bacteria were harvested from macrophage culture by lysing macrophages with 0.1% Triton-X in PBS for 5 min, after which the detergent was diluted with 9 volumes of PBS. A total of 5000 bacteria from media or macrophage lysates was assayed for fluorescence by FACS. Bacteria were detected by side scatter as described previously (Valdivia and Falkow, 1996), and fluorescence and side-scatter data were collected using logarithmic amplification. Quantitative measurements and distribution of fluorescence from the different fusions were determined with the Consort 30 software program (Becton-Dickinson). To calculate the amount of fluorescence induction, the fluorescence value of a given clone from the macrophage lysate was divided by the value of fluorescence of the same clone grown in DMEM.

Expression of mGFP in other mycobacterial species

Plasmid preparations from control and representative clones were transformed into M. smegmatis, M. bovis and M. tuberculosis strains as described previously (Jacobs et al., 1991). Transformants were observed directly on selective media for the expression of mGFP. A Leitz Orthoplan epifluorescent microscope with a 2.5 × objective lens was used to observe and photograph colonies.

Sequencing of selected clones

Plasmid mini-preps were performed on clones of interest as described previously (Ramakrishnan et al., 1997a). A total of 0.1 μg of DNA was transformed into library efficiency DH5α competent cells (Gibco BRL) according to the manufacturer's instructions. Plasmid mini-preps were performed on at least five single-colony transformants. Each plasmid preparation was digested with EcoRI and resolved on an agarose gel to ascertain that inserts were stable. Polymerase chain reaction (PCR) amplification of insert sequences followed previously described protocols (Maniatis et al., 1989). Inserts were sequenced using PCR products with flanking vector sequences or the native plasmid as templates. Primers that annealed 15 bp away from the BamHI insert site were chosen to flank the insert DNA and read from both directions into the fragment. All DNA sequencing was performed using a model 370A stretch automated DNA sequencer according to the manufacturer's instructions (Applied Biosystems). Homologies to known sequences were determined with MycDB (Bergh and Cole, 1994) and BLAST software (National Center for Biotechnology Information at the National Library of Medicine) (Altschul et al., 1990).


We thank Olivia Steele-Mortimer for advice/expertise on vesicle preparations, Raphael Valdivia for the provision of and helpful discussion of vector pFPV27, Lalita Ramakrishnan and Benjamin Chang for vector pMV262gfpmut3, Stanley F. Hayes and Elizabeth Fischer for electron microscopy preparations, and Gary Hettrick and Bob Evans for help with graphics. We also thank Patti Rosa, Jos van Putten, Katie George and Jay Carroll for critical review of this manuscript. Thanks also to K. and H. Fischer for helpful discussion. Our sincerest thanks to Brian Stevenson, Steve Porcella and Stanley Falkow for advice, expertise and encouragement.