|Home | About | Journals | Submit | Contact Us | Français|
There are many species of environmental mycobacteria (EM) that infect animals that are important to the economy and research and that also have zoonotic potential. The genomes of very few of these bacterial species have been sequenced, and little is known about the molecular mechanisms by which most of these opportunistic pathogens cause disease. In this study, 18 isolates of EM isolated from fish and humans (including strains of Mycobacterium avium, Mycobacterium peregrinum, Mycobacterium chelonae, and Mycobacterium salmoniphilum) were examined for their abilities to grow in macrophage lines from humans, mice, and carp. Genomic DNA from 14 of these isolates was then hybridized against DNA from an M. avium reference strain, with a custom microarray containing virulence genes of mycobacteria and a selection of representative genes from metabolic pathways. The strains of EM had different abilities to grow within the three types of cell lines, which grouped largely according to the host from which they were isolated. Genes identified as being putatively absent in some of the strains included those with response regulatory functions, cell wall compositions, and fatty acid metabolisms as well as a recently identified pathogenicity island important to macrophage uptake. Further understanding of the role these genes play in host specificity and pathogenicity will be important to gain insight into the zoonotic potential of certain EM as well as their mechanisms of virulence.
Environmental mycobacteria (EM) have long been recognized for their abilities to cause opportunistic diseases in humans as well as in animals of economic and research significance, including cattle and fish. The influences that all members of this group have on human and animal health are just emerging, however. Unlike the obligate pathogens Mycobacterium tuberculosis and Mycobacterium leprae, EM are cultured from the soil and water and are extremely hardy under conditions that would not allow many other bacteria to survive. They are routinely isolated from many sources of water that have been treated for human consumption (11, 18), and once in the host, the great majority of these organisms live within cells.
EM can cause a wide variety of disease pathology in humans and animals. In humans, the Mycobacterium avium-intracellulare complex (MAC) is the group of EM that most commonly affects patients infected with human immunodeficiency virus (HIV), while Mycobacterium ulcerans is most common in non-HIV-positive individuals (3). Infection with MAC in otherwise healthy individuals is usually associated with pulmonary disease in adults and lymph node disease in children (17, 28), while MAC causes disseminated disease in immune-suppressed individuals. Another member of the MAC, Mycobacterium avium subsp. paratuberculosis, is the cause of Johne's disease in ruminants, which is of particular economic concern to the cattle industry. Mycobacterium marinum, Mycobacterium fortuitum, and Mycobacterium chelonae cause dermal infections in humans. M. chelonae and M. fortuitum are associated with the infection of injection sites or wound sites at hospitals (9), whereas infection with M. marinum is most commonly linked to contact with swimming pools and fish tanks (15).
These three species of mycobacteria also cause disease in wild and cultured fishes (12). In addition, many other species that cause infection in humans, such as Mycobacterium triplex (36, 53), Mycobacterium peregrinum (27), and Mycobacterium haemophilum (27, 52), have also been shown to lead to serious infections in fishes. Mycobacteriosis has a significant impact on scientific groups using zebrafish (2, 50, 52) or other fish, such as medaka (45), for research models, as it is necessary to obtain and maintain mycobacterium-free fish. Zebrafish have been recognized for over 30 years as an excellent model for the study of early vertebrate development (49). Many fish species, including zebrafish (40, 41) and goldfish (Carassius auratus) (33, 44, 48), are now also being utilized as models for infectious diseases, including diseases caused by mycobacterial infection. For example, studies by Pozos and Ramakrishnan (40, 42) have proposed M. marinum as a surrogate bacterial model for M. tuberculosis because the two bacteria are genetically closely related.
Despite the importance of EM to the health of humans and many animals, very few species have been studied in depth in the laboratory to determine the molecular and cellular bases for their abilities to cause disease. Of the EM species mentioned, only M. avium, M. avium subsp. paratuberculosis, M. marinum, and M. ulcerans have been genomically sequenced and annotated, and there are a limited number of published studies that have identified putative factors involved in virulence in these strains.
Microarray technology is currently being utilized for comparative studies of bacteria at the genomic DNA level, both within serovars of the same species and between species of the same genus. Porwollik et al. (37-39) have done numerous experiments comparing the genomic DNA of serovars of Salmonella enterica with different virulence genes and hosts and have shown that there are often hundreds of genomic differences within serovars. Hamelin et al. (22) used a microarray containing all known Escherichia coli virulence and antimicrobial resistance genes to survey the population of environmental E. coli isolates in the Great Lakes, specifically in areas with known pollution (23). Microarrays have also been used to compare the genomic DNA of strains of sequenced mycobacteria. Behr et al. (5) first used a whole-genome array to compare the Bacille Calmette-Guérin (BCG) vaccine strain to M. tuberculosis. Their findings indicated at least nine deletions in the BCG strains that may have played a role in the differences in virulence between the two species. Similarly, microarrays have been used to compare the genomic DNA of M. tuberculosis with that of Mycobacterium microti, a member of the same clade that has very low virulence in humans (20). For species of the MAC, including M. avium subsp. avium, M. avium subsp. paratuberculosis, M. avium subsp. silvaticum, and Mycobacterium intracellulare, multiple groups have used a whole-genome microarray to analyze regions of deletions and polymorphisms that may explain some of the differences in pathogenicity and host specificity within this closely related group (30, 34, 47).
For this study, the virulence genes and genomes of some of the unstudied and unsequenced EM were compared to that of M. avium, a species that we know more about. Several strains of M. avium, M. peregrinum, M. salmoniphilum, and M. chelonae were examined for their abilities to invade and grow in human, mouse, and carp macrophage cell lines. Genomic DNA from a subset of these strains was then hybridized to a selective microarray containing probes for numerous genes thought to be involved in the pathogenicity of M. avium and M. tuberculosis. By comparing a number of EM species and incorporating data from similar experiments with other EM species, such as M. marinum, our knowledge about the molecular and cellular mechanisms of virulence used by this group of opportunistic pathogens will be expanded. Because there is a wide range of hosts for EM, it is interesting to explore whether there are genetic differences among these species that lead to a different pathogenic potential.
Eighteen strains of EM were analyzed with macrophage invasion assays, and 14 of these strains, as well as M. tuberculosis H37Rv, were hybridized against the DNA microarray. The strains were obtained from different sources and hosts and have been characterized previously, as listed in Table Table1.1. The four serovars of M. avium and M. tuberculosis H37Rv were grown at 37°C on Middlebrook 7H10 agar (Difco, Detroit, MI) supplemented with oxalate acid-albumin-dextrose-catalase (OADC; Difco, Detroit, MI) for 7 to 10 days. Pure colonies were inoculated into 7H9 broth (Difco, Detroit, MI) supplemented with 0.1% Tween 80 (Sigma) and OADC and grown to log phase in a shaking incubator prior to use in the assays. The MP ATCC and MP101 strains of M. peregrinum were grown on similar media for 5 to 7 days prior to the inoculation of broth, and the MP102 strain was grown at 30°C. All M. chelonae and M. salmoniphilum strains were grown at 30°C for 3 to 5 days prior to the inoculation of broth. For macrophage assays, mycobacteria in log phase growth were washed and resuspended in Hanks' balanced salt solution. Inocula of approximately 3 × 107 cells were established using McFarland's concentration standards, followed by serial dilutions and plating on 7H10 to confirm the number of bacteria.
Three cell lines were used in the macrophage assays. The human monocyte (U-937) (ATCC CRL-1593.2) and mouse peritoneal macrophage (RAW264.7) (ATCC TIB-71) cell lines were obtained from the ATCC, and the carp leukocyte cell (CLC) line was a gift from Jeffery Cirillo (University of Nebraska, Lincoln, NE). U-937 cells were cultured in RPMI 1640 medium (GIBCO) supplemented with 10% fetal bovine serum (FBS) (Sigma, St. Louis, MO). RAW264.7 cells were cultured in Dulbecco's modified Eagle's medium (GIBCO) supplemented with 10% FBS. CLCs were cultured in Eagle's minimal essential medium (ATCC) supplemented with 10% FBS and 10 mM nonessential amino acids (Sigma, St. Louis, MO) on collagen IV precoated flasks and plates. Cells were maintained in 25-cm2 or 75-cm2 flasks and transferred to 24-well culture plates, such that the monolayers were 80% confluent for the assays. The U-937 cells were treated with 500 ng/ml of phorbol 12-myristat 13-acetate (Sigma) overnight to promote maturation and adherence. U-937 and RAW264.7 cells were grown at 37°C with 5% CO2, while CLCs were grown at 28.5°C with 5% CO2, until the time of the assays.
Assays were performed as described previously by Bermudez and Petrofsky (6), except that infected 24-well plates were incubated at 37°C and/or 28.5°C, depending on the cell type and the experiment. Briefly, the inoculum was added into duplicate wells for each time point to 24-well culture plates containing U-937, RAW264.7, or CLCs at a multiplicity of infection of 10. After a 1-h infection time, the supernatant was removed, and the wells were washed three times with Hanks' balanced salt solution to remove extracellular bacteria. Sterile water containing a 1:5 dilution of 0.025% sodium dodecyl sulfate (SDS) was added to two of the monolayers to lyse the cells. The lysate was diluted serially and plated onto 7H10 agar to determine the number of CFU per ml. Media were replenished in the remaining wells. After 2 and 4 days, the duplicate cell monolayers were lysed and plated as described above. Assays were performed in triplicate, and the resulting CFU from all assays were analyzed to determine the uptake and growth of all indicated mycobacterial strains in the three cell types.
Oligonucleotide probe sequences were generated from the sequences of 181 predicted open reading frames (ORFs) of M. avium serovar 104 (MAC104) as described below. These 181 genes encode proteins with predicted or characterized roles in the virulence of mycobacteria, along with representative genes of various metabolic pathways. Additionally, the array probes included 15 sequences from a recently identified pathogenicity island (PI) in M. avium (14). MAC104 gene sequences were compared to those of M. tuberculosis and M. smegmatis and other mycobacterial sequences when homologues were present. Regions that were most similar across multiple mycobacterial sequences for each gene were analyzed by ProbeSelect (29) against the entire M. avium genome. The parameters for the probe design were specificity to the MAC104 gene, bias toward the 3′ end of the selected gene sequences, minimal secondary structures, and a length of 49 to 51 nucleotides. As controls for establishing a score for the absence of genes (see below), probe sequences were selected from 20 M. tuberculosis genes that could not be found in MAC104. Additionally, as negative controls, oligonucleotide probe sequences were designed for five ORFs of Escherichia coli K-12. The probes designed for the array represent approximately 4% of the ORFs in the MAC104 genome. A complete list of the 224 oligonucleotide sequences can be found in Table S1 in the supplemental material. Oligonucleotide probes were synthesized by QIAGEN/Operon (Alameda, CO). Lyophilized probes were resuspended in a buffer containing 3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and 1.5 M betaine and printed in quadruplicate on Corning UltraGap slides, using a BioRobotics Microgrid II TAS robotic spotter. The assembled array had a total of 1,120 features.
The strains indicated in Table Table11 were grown on Middlebrook 7H9 supplemented with 10% OADC. After sufficient growth, the resulting bacterial pellet was resuspended in 10 mM Tris and 0.5 mM EDTA. Bacterial lysis was performed by overnight shaking with 20 mg/ml lysozyme (Sigma) at 37°C, followed by the addition of 1% SDS and 1 μg/μl proteinase K and incubation for 3 h. After the addition of 5 M NaCl and cetyl trimethyl ammonium bromide buffer, DNA was precipitated from the lysate twice by chloroform-isoamyl alcohol (24:1) (Sigma), twice by phenol-chloroform-isoamyl alcohol (25:24:1) (Sigma), and then with cold isopropanol and finally washed with 80% ethanol. The resulting DNA pellet was dried and resuspended in DNase-free water and treated with RNase (Roche). The quantity and purity of each DNA sample was evaluated by spectrophotometric analysis at 230, 260, and 280 nm.
DNA of the bacterial strains was fluorescently labeled by the direct incorporation of fluorescent nucleotide analogs in a Klenow polymerase fill-in reaction mixture. Each 60-μl reaction was carried out with 4 μg DNA, partially digested with Sau3A1 (Roche) random primers (10 μg hexamers) (Invitrogen, Alameda CA), 6 μl of 10× Klenow buffer (Roche), 1 mM Cy3- or Cy5-dCTP (Amersham), 10 μg of each deoxynucleoside triphosphate-dCTP, and 1,000 U of Klenow polymerase (Roche). After incubation at 37°C for 5 h, labeled DNA was purified using a Qiaquick PCR purification kit (QIAGEN, Alameda, CA).
Labeled samples of DNA (with MAC104 as the reference and another strain as the test) were combined in a hybridization mixture containing the Cy3- and Cy5-labeled DNAs, a formamide-based hybridization buffer, and COT-1 DNA. The solution was applied to array slides that had been preincubated with the same hybridization mixture without DNAs, covered with a glass coverslip, and placed into a Boekel Scientific hybridization chamber for overnight incubation at 42°C. The slides were then washed with 2× SSC (diluted from 20× SSC; Ambion), 0.2% SDS, 2× SSC, and finally 0.2× SSC. Arrays were scanned after 10 min in the dark, using a ScanArray 4000 model scanner (Packard BioScience Company, Meriden, CT).
Fluorescence intensities and background signals from each channel were determined using a ScanArray 4000 model scanner. Data were uploaded to QuantArray to quantify the intensities generated by the spots. Background fluorescence was subtracted from the fluorescence for each spot, and the data set was normalized using a Lowess transformation. To reduce the analysis of uninformative genes, spots for which the average signal intensity was less than the average intensity of the E. coli negative controls for that array were removed from the analysis.
The methods of Porwollik et al. (39) were used to determine whether a gene was present or absent. Briefly, the ratio of the number of genes in the test strain to that in MAC104 was determined for each spot (Cy3/Cy5 or Cy5/Cy3, depending on the experiment); this ratio was log2 transformed and then normalized with respect to all ratios (such that the mean log2 is 0). The median value of the spots for each gene was calculated, and genes with a median log2 of less than −1.05, based on 2 to 4 experiments, were scored as being absent. This cutoff was determined by calculating the average log2 value for all spots that should be present in both MAC104 and M. tuberculosis and subtracting two standard deviations.
For a subset of genes with a median ratio of less than −1.05 in one or more strains tested, primers were designed to amplify regions encompassing the oligonucleotide sequence on the array. Primers were designed for each gene from both the MAC104 and the M. smegmatis sequences. A range of annealing temperatures from 55°C to 65°C was used in situations where there was questionable or no amplification with an annealing temperature of 60°C with either set of primers. Table Table22 lists the genes and primers used to verify the presence or absence of these genes.
For the macrophage assays, the results shown are the averages and standard deviations of three replicates. The significance of differences in growth was determined by Student's t test (P < 0.05).
Human monocytes (U-937), mouse macrophages (RAW264.7), and carp leukocytes (CLCs) were exposed to one strain of M. avium, two strains of M. peregrinum, three strains of M. chelonae, and six strains of M. salmoniphilum at 37°C and/or 28.5°C. The uptake after a 1-h exposure and growth within the cells after 2 and 4 days were assayed by counting the CFU collected from the lysed cells, as described previously (6). MAC101 grew by 732% (a 7.3-fold increase) from the baseline to day 2 and by 1,259% (a 12.6-fold increase) from day 2 to day 4 in U-937 cells incubated at 37°C. This strain increased by 1,788% (a 17.88-fold increase) over 2 days from the baseline and by 324% (a 3.24-fold increase) from day 2 to day 4 in RAW264.7 cells grown at 37°C. However, MAC101 growth decreased by 155% (a 1.55-fold decrease) after 2 days from the baseline and decreased by an additional 265% (a 2.65-fold decrease) between days 2 and 4 in CLCs grown at 28.5°C (Table (Table3).3). MAC101 exhibited a pattern similar to that of MAC104, based on data from previous work (24). MAC101 was also assayed in U-937 and RAW264.7 cells at 28.5°C and CLCs at 37°C. In human and mouse cells treated at these lower temperatures, an increase in bacterial CFU was observed for human cells (by 394% over baseline, a 3.94-fold increase by day 2; and by a 228% additional increase, a 2.28-fold increase between days 2 and 4), but growth was observed only after 2 days for mouse cells (increased by 324%; a 3.24-fold increase between baseline and day 2, with a subsequent decrease of 1,778%, a 17.78-fold decrease between days 2 and 4). In CLCs treated at 37°C, MAC101 did not exhibit growth after 2 days (a decrease of −211%, a 2.11-fold decrease) but did grow between days 2 and 4 (by 437%, a 4.37-fold increase) (Table (Table33).
Two strains of M. peregrinum (MP101 and MP102) were tested with all three strains of cells at both temperatures. MP101 was able to infect and exhibited growth over 4 days in U-937 cells, RAW264.7 cells, and CLCs at 37°C and 28.5°C (Table (Table3).3). MP102 was also able to replicate in different species of macrophages at 37°C but was not able to grow in the mouse macrophages at 28.5°C between the baseline and day 2 (−1,000%, a 10.00-fold decrease), although it did show subsequent growth after 4 days (424%, a 4.24-fold increase) (Table (Table33).
Of the three M. chelonae strains tested, two were isolated from outbreaks in zebrafish colonies and the other (MCH) was isolated from incidental infections of zebrafish. CFU of the MCH strain of M. chelonae increased compared to those at the baseline after 2 and 4 days in CLCs at 28.5°C but did not grow in human or mouse cells at 37°C or 28.5°C (Table (Table3).3). The other strains isolated from zebrafish, MCJA and MCJ78, grew in all three cell types, regardless of the incubation temperature (Table (Table33).
Six isolates of M. salmoniphilum from salmonid fish grew in U-937 cells at 37°C after 2 days but exhibited little additional growth between days 2 and 4, with the exception of the MCER strain (Table (Table3).3). In RAW264.7 cells incubated at 37°C, there was again growth between baseline and day 2 but little additional growth over the subsequent 2 days (Table (Table3).3). All six strains showed an increased number of CFU in CLCs at 28.5°C over the course of the 4-day assay, in which the trend toward less growth between days 2 and 4 was consistent with that of the other cell types (Table (Table33).
The genomic DNA of a subset of the EM strains analyzed in macrophage assays (Table (Table3)3) was compared against the genomic DNA from the reference strain, MAC104, by microarray analysis. Table Table44 (M. peregrinum strains), Table Table55 (M. chelonae strains), Table Table66 (M. salmoniphilum strains), and Fig. Fig.11 summarize the genes that were found to be absent by microarray analysis. A summary of all results is available in Table S1 in the supplemental material. The strains of M. avium are similar in their ability to grow in macrophages, and microarray analysis showed that there were only two genomic differences between MAC104 and the other three strains of M. avium tested. By the calculation described in Materials and Methods for determining the presence or absence of a gene, MAV3729 (mpt53) was scored as absent in MAC100, and MAV2536 (lprM) was scored as absent in MAC109. The mpt53 gene, however, was amplified from the genomic DNA of the MAC100 strain by PCR (Table (Table77).
The MP101 strain of M. peregrinum, isolated from zebrafish, replicated differently in macrophages than M. avium and had more genomic differences from other strains of M. avium than MAC104 (Table (Table4).4). Two genes were scored as absent from the MP101 strain (MAV2241 and MAV2479). No genes were absent from the MP102 strain isolated from tilapia. The probes designed for the putative invasion island indicated that parts of the region are absent in the MP101 strain but present in the MP102 strain. The MP ATCC strain, isolated from human sputum, was also missing segments of the invasion island as well as four other genes (MAV1540, MAV1705, MAV3389/carB, and MAV4209/mtrA) (Table (Table4).4). Four regions of the invasion island were further analyzed by PCR and could not be amplified in any of the three strains. The mtrA gene could be amplified by PCR from the DNA of the MP ATCC strain (Table (Table77).
Although the MCH, MCJA, and MCJ78 strains were all isolated from zebrafish, the MCH strain had a different profile of growth in macrophages. Similarly, the MCH strain differed genomically from the other two strains. The genomic DNA of all three strains also had a genomic profile more divergent from that of the strains of M. peregrinum than that of MAC104. The scores for the presence or absence of genes are summarized in Table Table5.5. There were five genes that were scored as absent in the MCH strain, including MAV4209/mtrA. Additionally, the invasion island is absent from this strain. None of the genes that were absent in the MCH strain is similar to those scored as absent in MCJA or MCJ78. The MCJA strain had five genes that were absent, while the MCJ78 strain was missing seven genes. Of these missing genes, MAV1169/kdpE was common to both strains. In the case of four genes that were scored as absent in only one of the strains (MAV0182/sodA, MAV2510/mmpL5, MAV3608, and MAV4679/pcaA), the score for the other strain was very near that of the cutoff of −1.05 (Table (Table5).5). The invasion island was present in both the MCJA and MCJ78 strains. PCR amplification from genomic DNA confirmed that MAV4209/mtrA and the invasion island were absent from MCH and that the MAV1169/kdpE gene was absent from the MCJA and MCJ78 strains (Table (Table77).
The strains of M. salmoniphilum were the most dissimilar group compared to those of MAC104. All strains tested by macrophage assays had comparable growth in the three types of macrophages. The profiles of genes that are present, absent, or putatively duplicated are very similar across the four strains tested by microarray. Table Table66 summarizes the genes putatively absent from each strain; MCBAND had 21 genes that scored as absent, MCMWF had 25 genes, MCSABLE had 29 genes, and MCTRASK had 21 genes absent. Nine genes were absent in all four strains. Many others were absent from two or three of the strains. In these cases, the score was typically very close to the −1.05 cutoff in the other strain(s) (Table (Table6).6). Table S1 in the supplemental material summarizes these values, and Fig. Fig.11 visually presents the presence and absence of all genes tested.
A subset of genes amplified by PCR had some correlation to the microarray results with these strains (Table (Table7).7). Although MAV0469 and MAV3729/mpt53 scored as absent in four and three of the strains, respectively, both genes could be amplified from all four strains by PCR with primers designed from an M. smegmatis sequence (but not with primers designed from an M. avium sequence). Although MAV0214/fbpA scored as absent in all four strains, it could be amplified by PCR using primers designed from M. smegmatis from every strain except MCTRASK. MAV0701/phoP was scored as absent from MCSABLE and MCTRASK and also did not amplify from MCTRASK DNA by PCR. The mtrA gene, absent as determined by microarray analysis, could not be amplified from DNA of any of the four strains by PCR (Table (Table77).
EM organisms have a significant ability to cause disease in humans and animals. Very little knowledge exists, however, about the mechanisms of disease for this group of pathogens, including those important to fish mycobacteriosis. To learn more about the virulence and genomics of some of these species with zoonotic potential, we used macrophage assays and a custom microarray analysis to compare strains of different EM species to that of M. avium.
Many of the EM strains of M. peregrinum and M. chelonae were isolated from zebrafish and have been characterized previously (27, 50). The strains isolated from salmonid fishes were first classified as M. salmoniphilum, but the species was eliminated and its members were assigned to M. chelonae (1, 43). However, the species has now been resurrected based on molecular analyses (51). Because all the strains of M. chelonae and M. salmoniphilum and the MP102 strain of M. peregrinum were unable to grow in culture at 37°C, it was notable that many of these isolates were able to grow in human and mouse macrophages at the same temperature. Previous work with M. marinum, another fish pathogen with zoonotic potential, indicates that strains that are not able to grow in culture at 37°C are capable of growing in macrophages at that temperature (26). All strains of mycobacteria isolated from fish replicated in the carp cells, and many of those strains also grew in the human and mouse cells. In contrast, neither the MAC101 nor the MAC104 strain (from humans) grew well in carp cells. This result does not seem to be a product of the optimum temperature for the cell type or bacterial strain, as the MAC101 strain was able to grow in human and mouse cells at 28.5°C.
The percentages of growth over 2 and 4 days for the three types of macrophages were not similar in mycobacteria isolated from the same hosts or even in the same species, although there were some trends. The M. salmoniphilum strains, isolated from salmonids, and the MCJA and MCJ78 strains, isolated from zebrafish, showed growth over 2 and 4 days in all three macrophages lines. The MCH strain of M. chelonae, isolated from zebrafish, however, was not able to grow in either mouse or human macrophages or in carp cells incubated at 37°C. When the human and mouse macrophages were incubated at a lower temperature, the MCH strain was able to grow. The MP101 strain of M. peregrinum, isolated from zebrafish, was able to grow in all cell types at either temperature, but the MP102 strain, isolated from tilapia, was not able to grow in mouse macrophages, unless they were incubated at 28.5°C. This strain was also unable to grow in carp cells incubated at 37°C. As indicated previously, the M. avium strains tested were unable to multiply well in carp cells at either temperature.
The macrophage growth profile correlates well with other virulence data. Watral and Kent (50), using intraperitoneal injection, showed that the MCH strain of M. chelonae did not cause any significant disease or mortality in zebrafish, nor could it be cultured from the zebrafish weeks after exposure. Moreover, this isolate was obtained from a few fish as an incidental infection in a large research colony (27). In a different study, in which zebrafish were exposed to mycobacteria by bath immersion or intubation, we showed that the MP101 strain of M. peregrinum (which did grow well in mammalian macrophages) was consistently able to infect and grow in zebrafish tissues, although signs of disease and mortality were rare (24). In data not included in that study, the MP102 strain of M. peregrinum could not be cultured from the zebrafish, nor did it cause any signs of disease. Similarly, we could not culture the M. avium MAC104 strain from zebrafish infected by bath immersion or by intubation, and the infected fish showed no signs of disease (M. J. Harriff, unpublished observations).
Genomic differences based on microarray analyses also correlated with phenotypic virulence data. A survey of the literature revealed over 120 genes predicted to play a role in the virulence of different species of mycobacteria. Many of the genes selected for the microarray were identified by a number of genome- or proteome-wide screens designed to detect M. avium or M. tuberculosis genes or proteins upregulated in human macrophages (7, 13, 16, 25). Over 50 of these gene products have no known or predicted function, including the ORFs in a 3-kb PI identified in our laboratory, that is important in virulence (14). Little is known about how the genes contained in the open reading frames in the PI may function in virulence, but it has been shown to allow for the infection of amoeba. This region is not present in M. tuberculosis but is present in other members of the MAC. As shown by microarray analysis, the island was completely absent in the MCH strain of M. chelonae, while portions of the island were also predicted to be absent in the ATCC and MP101 strains of M. peregrinum. Interestingly, the island is putatively duplicated in the M. salmoniphilum strains. This bacterium is a recognized pathogen of salmonids, based on both laboratory transmission studies (1) and field observations (8, 10). The in vivo and in vitro growth of these strains of mycobacteria suggest that the genes making up this PI should be studied further for their role in the virulence of EM.
In addition to genes predicted to play a role in virulence, sequences representing selected genes of many of the metabolic pathways in mycobacteria were included. It is not known if differences or absence of metabolic pathways can play a role in host specificity or levels of pathogenicity. Differences in metabolic pathways are also important to differences in the basic phenotypic characteristics of species, which may play roles in environmental survival, host specificity, and pathogenicity. For example, M. avium subsp. paratuberculosis is very closely related to M. avium subsp. avium at the genomic level (4); however, multiple studies have revealed that each subspecies has regions and genes that are not present in the other (34, 47). Among some of the important findings, microarray analysis revealed that M. avium subsp. paratuberculosis has a truncation of an early gene in the mycobactin synthesis operon (47). Further study of this truncation could reveal the reasons for the slow growth of M. avium subsp. paratuberculosis as well as for differences in hosts and pathogenicity levels. Some of the species of EM tested in this study were putatively missing genes from metabolic pathways that may play a role in the ability to persist in certain environments.
Although many virulence genes identified in screens and included on this array have no putative function, others have been studied for their role in virulence. The mtrA gene was scored as absent in the MCH strain of M. chelonae and three of the four strains of M. salmoniphilum (with the fourth strain score being very near to that of the cutoff). This gene encodes the response regulator protein of a two-component regulatory system that has been shown by Zahrt and Deretic to be essential to growth in broth and differentially expressed by virulent and avirulent mycobacteria (54). While the expression of this gene is constitutive and high in M. tuberculosis, it is induced to very high levels in the BCG vaccine strain upon entry into macrophages. A more recent study, by Fol et al. (19), showed that the regulation of the ratio of this protein product and its phosphorylated version are important to proliferation in macrophages. The protein product encoded by mtrB that phosphorylates MtrA is not essential (54), suggesting that there are other kinases that play a role in the modulation of this ratio. It is possible that in strains of EM without the MtrA protein, other response regulators have evolved to serve in the essential role for growth in the environment. When a host is encountered, however, the lack of this protein leads to an inability to replicate in the host cells. It is possible that the lack of the mtrA gene is important to these differences in both in vitro and in vivo growth. Although the strains of M. salmoniphilum were able to grow in the human and mouse macrophages, the change in growth from 2 to 4 days was much reduced from the initial growth over 2 days (with the exception of one strain in human cells), suggesting that these strains are not able to persist as well in macrophages as strains with the mtrA gene.
The mtrA/B system was not the only two-component regulatory system to be identified as being absent from certain strains of EM. The kdpE (MCJA and MCJ78) gene scored as absent in some of the strains, while the phoP and regX3 genes also scored as absent (or nearly absent) in one to three of the M. chelonae or M. salmoniphilum strains. These response regulatory genes have been shown by a number of groups to be important to the growth and virulence of M. tuberculosis in animal models (21, 31, 32, 35, 46), but there is still very little known about the mechanisms by which they do so.
Although microarrays have been a useful tool for comparing the genomes of different strains of sequenced mycobacteria, we determined that there are potential limitations to the use of this tool to compare unsequenced strains of EM. The cutoff score for absent genes was determined by hybridizing the DNA of M. avium and M. tuberculosis to oligonucleotides representing genes known to be present and absent in both strains. Even with a cutoff score of two standard deviations lower than the median absent score for M. tuberculosis, we were nevertheless able to amplify some genes by PCR that were identified as absent by the microarray. Other than 16S rRNA genes, ITS, and hsp60 DNA, there are very few sequence data in the database for M. peregrinum and M. chelonae. This lack of sequence data makes it difficult to know what the cause of the discrepancy is between the microarray and the PCR results. It is possible that, although they are present, the sequences of these genes are divergent enough from the MAC104 gene sequences that they do not hybridize to the oligonucleotide on the array. It is also possible that the oligonucleotide was selected from a region of the gene that is in fact absent, but the primers used for PCR are similar to regions of the genomic DNA that are still present. If this were the case, we would expect the amplified PCR fragments to have different sizes. As this was not the case, it is more likely that the former explanation accounts for the discrepancy. Further evidence for this explanation is that primers designed from M. smegmatis were required to amplify genes expected to be present from some strains. The use of PCR as a technique for verification has limitations as well, as even a single base difference could lead to a false-negative PCR result. The use of additional techniques, such as Southern blotting or reverse transcription-PCR, could be used to verify the presence or absence as well as the expression of specific genes or regions identified in this study. Overall, however, there was good correlation between the microarray and the PCR data, and there were many genes for which the test strain DNA so strongly hybridized to the array spot, compared to that of the MAC104 reference, to be considered putatively duplicated, indicating that these techniques together provide reliable results. Further optimization may improve this custom microarray as a tool for comparing the genomes and gene expression in unsequenced species of mycobacteria.
Given how little we know about the mechanisms of virulence for many of these species, this study has provided a useful set of data from which to base further investigation. Future study may involve an examination of the expression of these genes when the various strains of mycobacteria are exposed to cell types or an animal model, such as zebrafish or mice. Additionally, constructing recombinant strains that express genes or regions that are putatively absent and looking for differences in the ability to grow in cell types or cause disease could provide insight into the relative importance of those genes to virulence. As we isolate more species and strains of EM from hosts with which humans frequently come into contact, such as cattle, zebrafish, and other fishes, further understanding of how virulence is tied to the genomes of these organisms may provide a means by which we are able to determine the isolates with zoonotic potential. Finally, more knowledge about the mechanisms of virulence and pathogenic potential of these organisms may lead to advances in controlling disease in both humans and animals of economic and research importance, such as zebrafish.
We thank Scott Givan and Caprice Rosato of the Center for Gene Research and Biotechnology for support with microarray design, construction, and hybridizations. We also thank Denny Weber for editing the manuscript.
This work was supported by NIH grants NCRR 5R24RR017386-02 and AI-43199.
Published ahead of print on 2 November 2007.
†Supplemental material for this article may be found at http://aem.asm.org/.