Isolation of strong expression signals of Mycobacterium tuberculosis

Tuberculosis, caused by the intracellular pathogen Mycobacterium tuberculosis, remains a major public health problem of global importance. An understanding of the genetic elements controlling the expression of M. tuberculosis genes, especially those encoding virulence determinants, would allow the development of strategies to combat the disease. Many studies have attempted to identify promoters controlling the expression of genes within the host, in particular the macrophage, which is the preferred niche for M. tuberculosis during infection. These studies have relied on reporter genes such as the Aequoria victoria green fluorescent protein (GFP; Via et al., 1996 ; Triccas et al., 1999 ), or directly compared mRNA levels outside or within host cells (Graham & Clark-Curtiss, 1999 ). Alternatively, more generalized approaches have been used to identify a broad range of promoters, including those of a constitutive nature, through the use of reporter proteins and direct analysis of individual recombinant clones. These latter types of studies are hindered by the large number of recombinant colonies that need to be screened as the frequency of clones carrying true promoters is typically quite low (Timm et al., 1994a ; Das Gupta et al., 1993 ).

While the majority of studies analysing mycobacterial transcription regulation have focused on promoters whose activity varies in response to defined environmental signals, it is of interest to selectively define strong constitutive expression signals of M. tuberculosis. This would provide important information on those sequences required for optimal gene expression within mycobacteria. A limiting factor in the development of versatile expression systems for the mycobacteria is the lack of promoters permitting consistent, high-level expression of foreign genes in recombinant mycobacteria, analogous to the lac promoter of Escherichia coli (de Boer et al., 1983 ). Thus the isolation of strong mycobacterial promoters would permit expression of foreign genes in mycobacteria to aid protein purification and recombinant vaccine vector development. This is of particular importance as recombinant proteins obtained from fast-growing mycobacterial hosts, such as Mycobacterium smegmatis, have demonstrated superiority over the same proteins purified from E. coli expression systems, as assessed by structural and immunological analysis (Garbe et al., 1993 ; Roche et al., 1996 ; Triccas et al., 1996 ), an important consideration in the development of specific diagnostic reagents and improved vaccines.

In this report we have identified strong expression signals of M. tuberculosis. A random library of M. tuberculosis fragments was constructed in a vector containing a promoterless copy of the gene encoding the A. victoria GFP. After transformation into Mycobacterium bovis bacille CalmetteGuérin (BCG), recombinant bacteria displaying strong fluorescence were separated by fluorescence-activated cell sorting (FACS), resulting in the identification of M. tuberculosis sequences demonstrating strong promoter activity.

Bacterial strains and growth conditions.
E. coli DH5α was grown on liquid or solid LuriaBertani medium. Mycobacterial strains (M. tuberculosis strain 103 and M. bovis BCG Pasteur) were grown in liquid Middlebrook 7H9 medium (Difco) supplemented with ADC enrichment (Difco) or solid Middlebrook 7H10 medium (Difco) supplemented with OADC enrichment (Difco). When required, the antibiotic kanamycin was added at a concentration of 25 µg ml^-1 for both E. coli and mycobacteria.

M. tuberculosis genome library construction.
Construction of the M. tuberculosis promoter library in the E. coli/mycobacterial shuttle vector pJFX2 is described elsewhere (Triccas et al., 1999 ). The pJFX4 vector was constructed by placing the strong ß-lactamase promoter of Mycobacterium fortuitum (Timm et al., 1994b ) into pJFX2. Preparation of competent cells and electroporation of mycobacteria was carried out as described previously (Pelicic et al., 1997 ).

FACS analysis and sorting of fluorescent bacteria.
Bacteria were analysed with a FACScan (Becton-Dickinson Immunocytometry Systems) and sorted with an ELITE cell-sorting system (Beckman Coulter). After growth in 7H9 medium, BCG cells were resuspended in 1 ml PBS (approx. 1x10⁶ bacteria), analysed by FACS and bacteria exhibiting medium to high levels of fluorescence were collected. Bacteria were plated directly onto 7H10 solid medium to recover individual clones.

Macrophage preparation and infection.
Bone-marrow-derived macrophages were prepared as described previously (Triccas et al., 1999 ). Macrophage monolayers were infected with bacteria at an m.o.i. of 1:1. After 4 h infection, extracellular bacteria were removed by washing four times with PBS and incubation continued at 37 °C in 5% CO₂. After 6 d infection, BCG-infected macrophages were washed three times in PBS, centrifuged and lysed in water plus 0·1 % Tween 80. Recovered bacteria were analysed directly by flow cytometry.

Selection and analysis of clones harbouring strong M. tuberculosis expression signals.
Individual BCG colonies were analysed directly using an Axioscop fluorescence microscope (Zeiss) and those colonies exhibiting strong fluorescence inoculated into 7H9 medium in 24-well culture plates. After 7 d growth, the geometric mean of fluorescence for bacterial populations was determined by flow cytometry using the Lysis II program (Becton-Dickinson). The identity of the M. tuberculosis insert was determined by PCR, DNA sequencing and database analysis as described previously (Triccas et al., 1999 ).

Use of GFP to select M. bovis BCG clones harbouring strong expression signals of M. tuberculosis
In a first step to isolate strong promoter sequences of M. tuberculosis, a library of M. tuberculosis DNA (0·21·5 kb) was constructed in the vector pJFX2 (Triccas et al., 1999 ), which contains a promoterless copy of the gene encoding a strongly fluorescent version of GFP (Cormack et al., 1996 ). After transformation of this library into M. bovis BCG and subsequent growth in 7H9 medium, a pool of recombinant (r) BCG clones was obtained that displayed varying levels of GFP expression as assessed by flow cytometry (Fig. 1a). Bacteria exhibiting medium to high levels of GFP fluorescence were collected by cell sorting, resulting in a pool of rBCG enriched for fluorescent bacteria (Fig. 1b). After plating of the pool onto Middlebrook 7H10, visual inspection by fluorescence microscopy revealed that approximately 50% of the colonies were fluorescent. Forty-four colonies identified as strongly fluorescent were selected for further analysis.

(25K): Fig. 1. Selection by FACS of rBCG clones harbouring strong M. tuberculosis expression signals. A library of M. tuberculosis 103 DNA fragments was constructed in the pJFX2 vector and transformed into M. bovis BCG. The proportion of bacteria displaying GFP fluorescence after growth in 7H9 culture medium was determined by flow cytometry (a). Bacteria exhibiting medium to high levels of GFP fluorescence were collected resulting in a pool of rBCG enriched for fluorescent bacteria (b).

We next compared by flow cytometry the fluorescence of the 44 selected clones with that of BCG harbouring the pJFX4 plasmid (Triccas et al., 1999 ), which contains the gfp gene under the control of the ß-lactamase promoter (pblaF*) of M. fortuitum, one of the strongest mycobacterial promoters characterized to date (Timm et al., 1994b ). The strength of this promoter has been used to permit purification of important antigens of M. tuberculosis (Roche et al., 1996 ), Mycobacterium leprae (Triccas et al., 1996 ) and Mycobacterium avium (Triccas et al., 1998 ) from recombinant M. smegmatis. Systems based on pblaF* have also allowed overexpression of antigens in M. bovis BCG and subsequent detection of strong immune responses against the expressed products in animal models (Abdelhak et al., 1995 ). Of the 44 clones tested, 34 represented pure cultures showing fluorescence levels equal to or greater than that of BCG/pJFX4 (Fig. 2). The levels of fluorescence ranged from 0·91 to 3·50 times the level of gfp expression by virtue of pblaF*.

(26K): Fig. 2. Determination of relative M. tuberculosis promoter strength harboured by the isolated rBCG clones. Thirty-four colonies selected as strongly fluorescent by microscopy were analysed individually by flow cytometry. Results are represented as the fold difference in fluorescence of each clone compared to the fluorescence of BCG containing plasmid pJFX4 (gfp controlled by the M. fortuitum pblaF* promoter). The results are representative of two separate experiments.

Identification of genes under the control of the isolated M. tuberculosis expression signals
As the complete genome sequence of M. tuberculosis is available (Cole et al., 1998 ), we next determined the genomic location of the expression signals isolated. PCR amplification was performed on the 34 rBCG clones after DNA extraction by freeze-thawing. PCR products were obtained for 30 of the clones, which after sequencing were shown to represent nine different M. tuberculosis genomic fragments (Table 1). Clones within the same group most likely represent sibling strains, as all have identical inserts within the vector. By analysing the genomic location of the sequenced fragments and determining the direction of transcription with respect to the gfp gene, we were able to determine which sequences most likely corresponded to a predicted M. tuberculosis gene (Table 1). The majority of the promoters corresponded to genes encoding proteins of unknown function (clone groups A, B, D, E, F, G and H). The DNA fragment of clone group F was located within the Rv2850 ORF of M. tuberculosis and thus may represent a cryptic promoter with strong activity in M. bovis BCG. Clone group C corresponded to a gene encoding a possible calcium-binding protein, while clone group I corresponded to a protein predicted to be involved in magnesium transport. The importance of calcium and magnesium in cell physiology may indicate why these two genes appear to be strongly expressed in mycobacteria. We cannot exclude the possibility that activity of these promoters is different in M. tuberculosis itself compared to BCG, as genomic regions present in M. tuberculosis yet absent from BCG may encode products exerting enhancing or deleterious effects on promoter activity. Furthermore, while this plasmid-based system allows comparative analysis of promoter strength, it may not truly reflect the promoter activity in M. tuberculosis, where promoter/genes would most likely be present as a single copy.

Table 1. Analyses of clones harbouring strong M. tuberculosis expression signals

The cloned regions preceding the predicted translation start codon for the genes described in Table 1 were analysed using the BDGP Neural Network Promoter Prediction Program (Reese et al., 1996 ; www.fruitfly.org/seq-tools/promoter.html). Although we were able to identify -10 and -35 hexamers homologous to the consensus sequences of both E. coli (Harley & Reynolds, 1987 ) and mycobacterial promoters (Mulder et al., 1997 ), we did not identify patterns suggestive of a common sequence amongst promoters displaying elevated activity (see http://www.centenary.usyd.edu.au/research/tbres.html). The precise mapping of transcription start sites of these and other strong mycobacterial promoters would be required before a consensus pattern for strong promoters could be developed.

Activity of strong M. tuberculosis expression signals in M. smegmatis
We next determined if these M. tuberculosis expression signals that functioned strongly in M. bovis BCG also showed strong activity in M. smegmatis, the most widely used host for recombinant myobacterial gene expression. Plasmids were recovered from rBCG, electroporated into M. smegmatis and fluorescence levels of bacteria analysed by FACS. Restriction analysis of plasmids revealed no apparent deletions (data not shown). While essentially all promoters showed greater activity than the pblaF* promoter in BCG (Fig. 3a), only the promoter of Rv3108 showed an enhanced level of activity in M. smegmatis compared to pblaF* (1·73-fold greater fluorescence; Fig. 3b). All other clones showed fluorescence levels between 0·15 and 0·69 of that of M. smegmatis harbouring pJFX4. Re-isolation of the plasmids and introduction into BCG gave similar GFP levels as in Fig. 3(a) (data not shown), suggesting that the reduced promoter activity was not due to vector deletion or rearrangement events within M. smegmatis. The reason for the reduced activity of M. tuberculosis promoters within M. smegmatis is unclear, but may reflect differences in transcription machinery amongst mycobacterial species, or the existence of repressive elements within M. smegmatis.

(18K): Fig. 3. Comparison of the strength of M. tuberculosis promoters in M. bovis BCG (a) and M. smegmatis (b). Promoter strength was assessed by using flow cytometry to determine the fluorescence of each clone. Results are represented as the mean fold difference in fluorescence for the clones (±SEM; three experiments) compared to the fluorescence of either BCG (a) or M. smegmatis (b) containing plasmid pJFX4 (gfp controlled by the M. fortuitum pblaF* promoter).

Activity of the strong M. tuberculosis promoters within macrophages
Any strong mycobacterial promoter useful for recombinant vaccine construction would ideally maintain its activity within the in vivo environment. This was assessed by infecting murine bone-marrow-derived macrophages with rBCG clones of each promoter group and analysis of fluorescence after 6 d infection. All clones displayed fluorescence that was not significantly different from that achieved in normal culture medium (Fig. 4). This suggests that the activity of these expression signals is not affected after prolonged exposure to the in vivo environment.

(20K): Fig. 4. Activity of strong M. tuberculosis promoters within macrophages. The relative GFP fluorescence intensities of bacteria recovered from murine bone-marrow-derived macrophages after 6 d (solid bars) or grown in 7H9 medium for 7 d (white bars) were compared by flow cytometry. Results are expressed as the mean fluorescence level±SEM for three experiments.

In this report we have exploited the fluorescent properties of the A. victoria GFP to permit the isolation of strong expression signals of M. tuberculosis. By using FACS to isolate recombinant bacteria displaying strong fluorescence, we significantly limit the number of colonies required to be screened by initially eliminating the majority of bacteria not displaying promoter activity, and can also place a limit on the level of promoter strength preferred during the cell sorting process. Such advantages are not seen with systems based on the ß-galactosidase and chloramphenicol acetyltransferase reporter proteins which have previously been employed to isolate promoters of M. tuberculosis (Timm et al., 1994a ; Das Gupta et al., 1993 ). Apart from providing information on the genes of M. tuberculosis that are strongly expressed by the bacterium, the promoters identified in this study may prove useful in the overexpression of foreign genes in mycobacteria to aid protein purification and vaccine vector development. We thank Jean Rauzier for DNA sequencing, Dr Philippe Métézeau and Hélène Kiefer for FACS sorting, Dr Brendan Cormack for the pKEN-GFPmut2 vector and Daniel Roach for helpful discussions. J.A.T. was the recipient of an Institut Pasteur Cantarini Fellowship. This work was supported by the European Community (grant BMH4 CT92167) and a Pasteur Merieux Connaught Research and Development grant.