Mol Biol Rep DOI 10.1007/s11033-014-3463-4
Genome-wide screening of pathogenicity islands in Mycobacterium tuberculosis based on the genomic barcode visualization Jiao Xie • Fengfeng Zhou • Guangyu Xu Guoqin Mai • Jie Hu • Guoqing Wang • Fan Li
•
Received: 17 March 2013 / Accepted: 13 June 2014 Ó Springer Science+Business Media Dordrecht 2014
Abstract Mycobacterium tuberculosis (M. tuberculosis) is one of the most widely spread human pathogenic bacteria, and it frequently exchanges pathogenesis genes among its strains or with other pathogenic microbes. The purpose of this study was to screen the pathogenicity islands (PAIs) in M. tuberculosis using the genomic barcode visualization technique and to characterize the functions of the detected PAIs. By visually screening the barcode image of the M. tuberculosis chromosomes, three candidate PAIs were detected as MPI-1, MPI-2 and MPI-3, among which MPI-2 and MPI-3 were known to harbor pathogenesis genes, and MPI-1 represents a novel candidate. Based on the functional annotations of Pfam domains and GO categories, both MPI-2 and MPI-3 carry genes encoding PE/PPE family proteins, MPI-2 encodes the type VII secretion system, and MPI-3 encodes genes for mycolic acid synthesis in the cell wall. Some of these genes were already widely used in early diagnosis or treatment of M. tuberculosis. The novel candidate PAI MPI-1 encodes CRISPR-C as family proteins, which are known to be associated with persistent infection of M. tuberculosis. Our data represents a molecular basis and protocol for Electronic supplementary material The online version of this article (doi:10.1007/s11033-014-3463-4) contains supplementary material, which is available to authorized users. J. Xie G. Xu J. Hu G. Wang (&) F. Li (&) Norman Bethune Medical College of Jilin University, Changchun 130021, Jilin, China e-mail: [email protected] F. Li e-mail: [email protected] F. Zhou G. Mai Shenzhen Institutes of Advanced Technology, and Key Lab for Health Informatics, Chinese Academy of Sciences, Shenzhen 518055, Guangdong, People’s Republic of China
comprehensive annotating the pathogenic systems of M. tuberculosis, and will also facilitate the development of diagnosis and vaccination techniques of M. tuberculosis. Keywords Mycobacterium tuberculosis Pathogenicity island Genomic barcode Diagnosis Vaccination
Introduction Tuberculosis is one of the most dreaded diseases caused by Mycobacterium tuberculosis (M. tuberculosis)of the last few centuries. Many aspects of M. tuberculosis physiology and pathogenesis still remain elusive. It has developed a complex matrix of virulence genes which attack the host organisms with various mechanisms, which makes its diagnosis and treatment extremely difficult [1, 2]. Point mutation, gene recombination and horizontal gene transfer (HGT) represent three major strategies in forming the pathogenic landscape of the modern M. tuberculosis strains [3]. In particular, HGT facilitates the dynamics of pathogenic functional modules of modern M. tuberculosis strains, by exchanging genetic material with the environmental microbes [4]. Pathogenicity island(PAI)is a genomic region with pathogenic genes that was acquired through HGT from another microbial genome [5–7]. Therefore, the acquisition of PAIs plays an essential role in the adaptive evolution of a microbe, by giving the PAI-receiving microbe novel virulent modules [6, 8–10]. At present, no experimental technique is available to perform the screening and functional analysis of PAIs from the pan-genome aspect. Large-scale PAI screening is mainly achieved based on either nucleotide statistics or phylogenetic analysis [11–13]. But all of these in silico methods lack a reasonable screening accuracy, due to that highly expressed genes also have altered compositions of single or di-
123
Mol Biol Rep
nucleotide sequences [14, 15]. Thus the integration and visual annotation of more information on the pathogen genome are necessary and it might provide a powerful tool for high-throughput and more accurate screening of PAIs. Our previous study found that the frequency spectrum for each K-mer nucleotide string (K, 1 \ K \ 6) within the region of equal length fragments in microbial genome has consistency. The barcode-like visual annotation (Barcode image) of a genome can be further obtained by making a digital and graphical process for the array matrix of the frequency spectrum. According to this theory, any microbial genome can be represented as a unique barcode image [15]. Genome barcode carries all the genetic information in the genome, and has a one-to-one correspondence with the genome sequence. This study applied the genomic visualization technique, genomic barcode, to the M. tuberculosis chromosomes, and screened the PAIs in these chromosomes based on the observation that PAIs are usually significantly different to the other genomic regions in the barcode feature. Furthermore, Pfam_Scan [16] and Blast2GO [17] were used to analyze the function of the screened PAIs and to describe the characteristics.
Results and discussion Genomic barcodes of M. tuberculosis We split the chromosome sequences of the five M. tuberculosis strains into a plurality of non-overlapping Fig. 1 Genomic barcodes of the five M. tuberculosis strains, i.e. a H37Rv, b H37Ra, c CDC1551, d F11, e KZN1435, f Mycobacterium leprae TN, g Escherichia coli K12 str MG1655, and h a random sequence with the same length as E. coli K12 str MG1655. The genomic barcodes were generated using the parameters W = 1,000 and K = 4
123
fragments of 1,000-bp lengths (W = 1,000). This study chose to use 4-mer strings for generating the genomic barcodes, as similar to [18]. So in this study, K = 4 and the total number of K-nucleotide pairs is N(K) = 128. Let the chromosome length be L, and the genomic barcode will be L/1,000 in height and N(K) in width, as shown in Fig. 1. Figure 1 shows that five strains of M. tuberculosis share a unique genomic barcode pattern, different strains have different barcode structure, of which the more closer sibship the more similar barcode graph, such as M. tuberculosis H37Rv, CDC 1551, F11 and KZN1435. In contrast, the farther sibship the more different barcode image, such as Mycobacterium leprae TN and Escherichia coli K12 str MG1655. This feature of genomic barcode was observed in all Prokaryotes’ DNA sequences, whereas the genomic barcode of a randomly generated DNA sequence (as in Fig. 1h) doesn’t even show any such patterns. So genomic barcode is an inherent characteristic of the genetic material of all life forms. The difference between two DNA sequences can be modeled by the Euclidean distance of the genomic barcodes of the two sequences. The two DNA sequences may come from two genomes or two genomic regions of the same genome. Although the majority of the sequence fragments of the same species share an almost identical barcode (Fig. 1a–e), some regions are significantly different in genomic barcode to the other regions. Such regions tend to have foreign origins or multiple repeats, and almost all the pathogenicity islands were exchanged through HGT events [19].
PE-PGRS family protein 11 3,922,471–3,950,263 MPI3
3,931,148–3,958,929
3,916,304–3,943,046
3,934,824–3,963,047
3,910,930–3,935,736
*28 K
151.6 ± 38.2
PPE family protein and PE-PGRS family protein 150.5 ± 28.6 10 3,727,488–3,767,102 MPI2
3,738,329–3,776,417
3,725,096–3,747,265
3,739,815–3,779,781
3,723,397–3,763,176
*39 K
147.1 ± 35.0 11 *14 K 3,130,179–3,144,352 3,112,942–3,126,002 3,130,192–3,143,764 3,118,224–3,131,773 MPI1
F11 CDC1551 H37Ra H37Rv
Location PAI
Table 1 The gene information of M. tuberculosis PAIs
A PAI is defined to be a large genomic region consisting of genes encoding important virulent regulatory factors, usually acquired through horizontal gene transferring events [23]. Three screening criteria were established for the PAIs in M. tuberculosis, (a) the abnormal genomic barcode fragment is longer than 10 K bps, (b) the abnormal barcode region exists in the chromosomes of all the five studied M. tuberculosis strains, (c) and the region harbors virulent genes. Three candidate PAIs were detected based on the above three criteria, and they were denoted as MPI1, MPI-2 and MPI-3, respectively. The locations and encoded genes of the three candidate PAIs in M. tuberculosis were shown in Table 1 and illustrated in Fig. 2. The entire three candidate PAIs were located in the region between 3,000,000 and 4,000,000 in the majority of M. tuberculosis chromosomes. This region contains a large amount of inserted fragment and is a hot region for gene recombination [24]. All the three candidate PAIs are flanked by tRNA genes, which are known to be the integration sites of PAIs [25]. MPI-2 and MPI-3 are close to each other, and may constitute one larger PAI, which was separated by new genomic recombination in between. Firstly, both MPI-2 and MPI-3 harbor genes from the same PE-PGRS protein family. The hypothesis is also supported by the barcode distance analysis. A recently acquired genomic island, e.g. a PAI, usually has a different genomic barcode pattern compared with the other regions of the current host genome, and will accumulate enough mutations over time to share a similar genomic barcode with the host genome [15]. Table 1 show that the barcode distances of MPI-2 and
KZN 1435
Pan-genome-scale screening of PAIs in M. tuberculosis
1,282,054–1,295,433
Number of genes Length (bps)
Barcode distance (X SD)
Function
According to the distribution frequency of barcode distances, we defined the genomic regions with Euclidean distance \90 or [150 as abnormal regions [18]. The barcode distances of all the 1,000-bp fragments in the five strains of M. tuberculosis were listed in the Supplementary Table 1.The abnormal barcode regions within the M. tuberculosis genome account for 18.0–18.8 % with an average value of 18.4 %, which is almost half of that in Escherichia coli (35.6 %). The data is consistent with the ratios of horizontal transferring events in the two bacteria [12, 20]. We have analyzed the abnormal fragments across the M. tuberculosis genomes, and found the following: *20 % of the abnormal fragments can be explained in terms of (a) horizontal gene transfers, (b) phage invasions and (c) highly expressed genes, based on PHX-PA [21] and Prophinder [22]. The continuous abnormal barcode regions of least 10 K bps long play an essential role in the exchange of genetic material between different pathogenic species and provide a molecular basis for M. tuberculosis pathogenicity.
Transposase and hypothetical protein
Mol Biol Rep
123
Mol Biol Rep
(a)
oriC MPI-3 MPI-2
Mycobacterium tuberculosis H37Rv 4411532bp MPI-1
(b) fadE26
240
fadE27
fadD17
PE_PGRS53 PE_PGRS53
PE_PGRS54
ilvX
MPI-1
Rv3510c
PE_PGRS55PE_PGRS56 PE_PGRS55 PE_PGRS56
MPI-2
PPE55 fadD18 PE_PGRS57 fadD18 PE_PGRS57
MPI-3
220 200 180 160 140 120 100 80 3000000
3200000
3400000
3600000
3800000
4000000
Fig. 2 Three PAI candidates in M. tuberculosis. a Barcode distance distribution of the genomic region between 3 and 4 M bps in the strain H37Rv of M. tuberculosis. Outer concentric circle: predicted coding regions on the plus strand, second concentric circle: predicted coding regions on the minus strand, third concentric circle: predicted
coding regions on the both strand, fourth concentric circle: buffer zone, fifth concentric circle: predicted PAI candidates, sixth concentric circle: GC content. b Barcode distance distributions and genes of three PAI candidates
MPI-3 were similar, suggesting that the times for these two PAIs entering the host genome were very close or the same. The alternative hypothesis is that the two PAIs were horizontally transferred into M. tuberculosis separately, and the two integration sites happened to be close to each other. Considering the random nature of HGT integration and other evidences, the alternative hypothesis is much less
likely than the original one. We could find same small peaks around location 3,500,000 in Fig. 2b, we didn’t consider this peak as a candidate PAI duo to its small size (less than 10 K). Phylogenetic analysis of these three PAIs showed that MPI-2 and MPI-3 are more conserved than MPI-1 in sequenced M. tuberculosis genome (http://www.ncbi.nlm.
123
Not found Not found Rv3506-Rv3521 3,925,000–3,959,000 3,922,471–3,950,263 MPI-3
Rv3504-Rv3514
Not found
Not found Rv3339c-Rv3355c
Rv2814-Rv2829c 3,120,566–3,137,012
3,724,615–3,769,807 Rv3343-Rv3355c
Not found
3,730,000–3,770,000 3,727,488–3,767,102 MPI-2
Rv2813-Rv2823c 3,118,224–3,131,733 MPI-1
Rv3341-Rv3350
Location Location Location
Genes
PMID: 11435108(206) PMID/Tool (citations)
One notorious feature of M. tuberculosis is that this pathogen can evade the attacks of the host immune system and remains a persistent infection in the host for months, years or even decades, without active replication. This is termed latent tuberculosis and M. tuberculosis under the latent status may regain activities whenever the host body’s immune system works abnormally. This represents the major problem for the tuberculosis control program. Functional annotation of genes in MPI-1 indicated that most of them encode transposases or hypothetical proteins. Although the gene function of the MPI-1 has not been reported previously, the PfamScan domain prediction suggested the functional gene Rv2820c of MPI-1 encoded a CRISPR-C as family protein. Previous report indicated that this protein family could activate the function of specific interval sequence encoding small RNA fragments (crRNA), which is closely related to persistent infection of M. tuberculosis [27]. Hence, we speculate that MPI-1 may be involved in the regulation of persistent infection of M. tuberculosis. Recently, some new reviews on biological functions of CRISPR system showed that this system was absence in Beijing sublineage [28, 29], which also consistent with our phylogenetic analysis of MPI-1. Further study of MPI-1 function may improve our understanding of the molecular mechanism of M. tuberculosis persistent
Table 2 Comparison with the known PAIs in M. tuberculosis H37Rv
The functional characterization of the M. tuberculosis PAIs
Genomic barcode
Genes
Island view (78)
Genes
PAIDB (28)
nih.gov/genome/browse/), which probably because MPI-1 may have resided in the M. tuberculosis genome for quite a long period of time led to the emergence of more mutations and recombination. This hypothesis also confirmed by genomic barcode distance as shown in Table 1. In order to analysis the advantage of genomic barcode visualization for screening of PAIs in M. tuberculosis,we compared our results to several published tools with high citation (http://scholar.google.com/), as shown in Table 2. We could found that these methods lack a reasonable screening accuracy or missed some PAIs, particularly so for PAIs that has been in the host genome for a long period of time. In addition, there are differences of the bountry of each PAIs. Actually, bountry analyses of PAIs are still elusive due to the limitations of the experimental techniques. Genomic barcode annotation could distinguish the bounry of each PAIs duo to the high-resolution base on and small windows sliding of 4-mer nucleatide and barcode distance calculate. For example, PAI candidates using genomic barcode tool do not break any operons in M. tuberculosis while others do [26]. However, this work is basically a proof-of-principle one, and more information of PAIs will be studied in the future work.
Location
Genes
Mol Biol Rep
123
Mol Biol Rep
infection, and facilitate the discovery of molecular drug targets and immune interventions. Genes within MPI-2 encode PPE family proteins, transposases, and the type VII secretion system. PPE55 protein showed an over-expression in the early stage of M. tuberculosis infection [30–32], and its N-termini has a signal peptide that may prompt a secreted trans-membrane protein involved in the body’s immune regulation during the early stage of M. tuberculosis infection [33–35]. So markers from MPI-2 can be used for the early diagnosis of M. tuberculosis infection. MPI-3 encodes PPE family proteins and enzymes responsible for fatty acid synthesis and degradation. The carboxylase encoded by the fad family genes in MPI-3 is a key protein responsible for providing the substrate for the last reaction of mycolic acid synthesis pathway [36]. Mycolic acids are important biological molecules in M. tuberculosis, usually synthesized through the type II fatty acid biosynthetic system. Thus MPI-3 has an important role in strengthening the latent persistence against the adverse growth conditions in the host body or in vitro environments. In addition, the PE-PRGS family proteins within MPI-3 are mainly involved in the interaction of M. tuberculosis with the host cells. Therefore, the genes within MPI-3 may also serve as the targets of anti-tuberculosis drugs. In summary, the genomic barcode provides an efficient and intuitive way to visualize a genome, from which hypotheses could be quickly formulated and tested using either computational or experimental techniques. Using this technique, we screened three candidate M. tuberculosis PAIs, including two with known functions and another one with unknown function. The further functional analysis of PAIs provides some novel insights for our understanding of the pathogenesis and evolution of M. tuberculosis and also opens a new avenue for choosing the diagnostic targets and vaccine targets.
Materials and methods Genome sequence data We focused on the PAIs of five firstly sequenced strains of M. tuberculosis, i.e. H37Rv, H37Ra, CDC1551, F11 and KZN1435, and collected their genome sequences and gene annotations from the NCBI Genome database [37] as of January 5, 2012. Calculation of genomic barcodes and their Euclidean distance The M. tuberculosis genomic sequences were partitioned into a plurality of non-overlapping fragments with equal
123
length W, and the frequency vector of the pairs of Knucleotide string and its reverse complement in each fragment was calculated. The appearance frequencies of k-mers are grouped into dozens of grayscale colors, e.g. L = 14 for k = 4 and WindowSize = 1,000. The reasonable prediction performances of this parameter combination for multiple applications suggest that WindowSize = 1,000-bp is good enough for k = 4 [15]. We mapped the frequencies in the frequency vectors to the gray scale integers (i.e. 0–255), and got an image for the given genome sequence. As in Fig. 1, the image of a genome sequence resembles a barcode used in supermarkets, and therefore called so. The more the color is close to white, the higher the frequency of the K-nucleotide string is in the fragment. For the array matrix of k-mers strings of the microbial genome, the barcode distance can be represented by the Euclidean distance of array. For example, for the array of any two matrices M1 and M2, the difference between them is defined as follow: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u L NðKÞ uX X t ðM1 ði; jÞ M2 ði; jÞÞ2 i¼1 j¼1
where L is the number of fragments, and N(K) is the number of all K-nucleotide sequences. The functional analysis of PAIs region using PfamScan and Blast2GO The genes in the candidate PAIs were consistently annotated using the programs PfamScan [38] and BLAST2GO [39]. The Pfam database (version 23.0) was downloaded from the website. The Pfam database is a large collection of protein domain families, including Pfam_ls and Pfam_fs, et al. We used the curated data set Pfam_ls, and annotated the genes in the candidate PAIs with Evalue e-4. BLAST2GO is a GO annotation and functional analysis tool based on sequence similarity search, and PAI genes were annotated using the online version of Blast2GO software with Evalue e-10. Acknowledgments This work was supported by National Natural Science Foundation of China (81101295 and 81071424), Specialized Research Fund for the Doctoral Program of Higher Education of China (20110061120093), China Postdoctoral Science Foundation (20110491311 and 2012T50304), Foundation of Jilin Provincial Health Department (2011Z049),Foundation of Jilin Provincial Science and Technology Department (20130522013JH), Norman Bethune Program of Jilin University (2012219). It was also supported in part by the Shenzhen Research Grant ZDSY 20120617113021359, China 973 program (2011CB512003 and 2010CB732606-6) and NSFC 31000447. Computing resources were partly provided by the Dawning supercomputing clusters at SIAT CAS.
Mol Biol Rep
References 1. Gazdik MA, McDonough KA (2005) Identification of cyclic AMPregulated genes in Mycobacterium tuberculosis complex bacteria under low-oxygen conditions. J Bacteriol 187:2681–2692 2. Sanz J, Navarro J, Arbue´s A, Martı´n C, Marijua´n PC, Moreno Y (2011) The transcriptional regulatory network of Mycobacterium tuberculosis. PLoS ONE 6:e22178 3. Gogarten JP, Townsend JP (2005) Horizontal gene transfer, genome innovation and evolution. Nat Rev Microbiol 3:679–687 4. Pallen MJ, Wren BW (2007) Bacterial pathogenomics. Nature 449:835–842 5. Ferna´ndez-Go´mez B, Ferna`ndez-Guerra A, Casamayor EO, Gonza´lez JM, Pedro´s-Alio´ C, Acinas SG (2012) Patterns and architecture of genomic islands in marine bacteria. BMC Genom 13:347 6. Hacker J, Blum-Oehler G, Mu¨hldorfer I, Tscha¨pe H (1997) Pathogenicity islands of virulent bacteria: structure, function and impact on microbial evolution. Mol Microbiol 23:1089–1097 7. Sui SJH, Fedynak A, Hsiao WWL, Langille MGI, Brinkman FSL (2009) The association of virulence factors with genomic islands. PLoS ONE 4:e8094 8. Becq J et al (2007) Contribution of horizontally acquired genomic islands to the evolution of the tubercle bacilli. Mol Biol Evol 24:1861–1871 9. Yu G et al (2011) Integrative analysis of transcriptome and genome indicates two potential genomic islands are associated with pathogenesis of Mycobacterium tuberculosis. Gene 489(1):21–29 10. Boyd EF, Almagro-Moreno S, Parent MA (2009) Genomic islands are dynamic, ancient integrative elements in bacterial evolution. Trends Microbiol 17:47–53 11. Karlin S (2001) Detecting anomalous gene clusters and pathogenicity islands in diverse bacterial genomes. Trends Microbiol 9:335–343 12. Veyrier F, Pletzer D, Turenne C, Behr MA (2009) Phylogenetic detection of horizontal gene transfer during the step-wise genesis of Mycobacterium tuberculosis. BMC Evol Biol 9:196 13. Juhas M, van der Meer JR, Gaillard M, Harding RM, Hood DW, Crook DW (2009) Genomic islands: tools of bacterial horizontal gene transfer and evolution. FEMS Microbiol Rev 33:376–393 14. Karlin S, Mrazek J, Ma J, Brocchieri L (2005) Predicted highly expressed genes in archaeal genomes. Proc Natl Acad Sci USA 102:7303–7308 15. Zhou F, Olman V, Xu Y (2008) Barcodes for genomes and applications. BMC Bioinform 9:546 16. Bateman A et al (2004) The Pfam protein families database. Nucleic Acids Res 32:D138–D141 17. Go¨tz S et al (2008) High-throughput functional annotation and data mining with the Blast2GO suite. Nucleic Acids Res 36:3420–3435 18. Wang G, Zhou F, Olman V, Li F, Xu Y (2010) Prediction of pathogenicity islands in enterohemorrhagic Escherichia coli O157:H7 using genomic barcodes. FEBS Lett 584:194–198 19. Gal-Mor O, Finlay BB (2006) Pathogenicity islands: a molecular toolbox for bacterial virulence. Cell Microbiol 8:1707–1719 20. Lawrence JG, Ochman H (1998) Molecular archaeology of the Escherichia coli genome. Proc Natl Acad Sci 95:9413
21. Mrazek J, Bhaya D, Grossman AR, Karlin S (2001) Highly expressed and alien genes of the Synechocystis genome. Nucleic Acids Res 29:1590–1601 22. Lima-Mendez G, Van Helden J, Toussaint A, Leplae R (2008) Prophinder: a computational tool for prophage prediction in prokaryotic genomes. Bioinformatics 24:863–865 23. Letek, M et al (2010) The genome of a pathogenic rhodococcus: cooptive virulence underpinned by key gene acquisitions. PLoS Genet 6(9):e1001145 24. Cole ST (1999) Learning from the genome sequence of Mycobacterium tuberculosis H37Rv. FEBS Lett 452:7–10 25. Freitas-Vieira A, Anes E, Moniz-Pereira J (1998) The site-specific recombination locus of mycobacteriophage Ms6 determines DNA integration at the tRNA(Ala) gene of Mycobacterium spp. Microbiology 144(Pt 12):3397–3406 26. Mao F, Dam P, Chou J, Olman V, Xu Y (2009) DOOR: a database for prokaryotic operons. Nucleic Acids Res 37:D459–D463 27. Marraffini LA, Sontheimer EJ (2010) CRISPR interference: RNA-directed adaptive immunity in bacteria and archaea. Nat Rev Genet 11:181–190 28. Arnvig K, Young D (2012) Non-coding RNA and its potential role in Mycobacterium tuberculosis pathogenesis. RNA Biol 9:427–436 29. Bryant J, Chewapreecha C, Bentley SD (2012) Developing insights into the mechanisms of evolution of bacterial pathogens from whole-genome sequences. Future Microbiol 7:1283–1296 30. Sampson SL (2011) Mycobacterial PE/PPE proteins at the hostpathogen interface. Clin Dev Immunol 2011:1–11 31. Delogu G et al (2006) PE_PGRS proteins are differentially expressed by Mycobacterium tuberculosis in host tissues. Microb Infect 8:2061–2067 32. Rachman H et al (2006) Unique transcriptome signature of Mycobacterium tuberculosis in pulmonary tuberculosis. Infect Immun 74:1233–1242 33. Singh KK, Dong Y, Patibandla SA, McMurray DN, Arora VK, Laal S (2005) Immunogenicity of the Mycobacterium tuberculosis PPE55 (Rv3347c) protein during incipient and clinical tuberculosis. Infect Immun 73:5004–5014 34. Koh KW, Soh SE, Seah GT (2009) Strong antibody responses to Mycobacterium tuberculosis PE-PGRS62 protein are associated with latent and active tuberculosis. Infect Immun 77:3337–3343 35. Singh PP, Parra M, Cadieux N, Brennan MJ (2008) A comparative study of host response to three Mycobacterium tuberculosis PE_PGRS proteins. Microbiology 154:3469–3479 36. Raman K, Yeturu K, Chandra N (2008) targetTB: a target identification pipeline for Mycobacterium tuberculosis through an interactome, reactome and genome-scale structural analysis. BMC Syst Biol 2:109 37. Barrett T et al (2012) BioProject and BioSample databases at NCBI: facilitating capture and organization of metadata. Nucleic Acids Res 40:D57–D63 38. Finn RD et al (2010) The Pfam protein families database. Nucleic Acids Res 38:D211–D222 39. Conesa A, Gotz S, Garcia-Gomez JM, Terol J, Talon M, Robles M (2005) Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research. Bioinformatics 21:3674–3676
123
Genome-wide screening of pathogenicity islands in Mycobacterium tuberculosis based on the genomic barcode visualization Jiao Xie • Fengfeng Zhou • Guangyu Xu Guoqin Mai • Jie Hu • Guoqing Wang • Fan Li
•
Received: 17 March 2013 / Accepted: 13 June 2014 Ó Springer Science+Business Media Dordrecht 2014
Abstract Mycobacterium tuberculosis (M. tuberculosis) is one of the most widely spread human pathogenic bacteria, and it frequently exchanges pathogenesis genes among its strains or with other pathogenic microbes. The purpose of this study was to screen the pathogenicity islands (PAIs) in M. tuberculosis using the genomic barcode visualization technique and to characterize the functions of the detected PAIs. By visually screening the barcode image of the M. tuberculosis chromosomes, three candidate PAIs were detected as MPI-1, MPI-2 and MPI-3, among which MPI-2 and MPI-3 were known to harbor pathogenesis genes, and MPI-1 represents a novel candidate. Based on the functional annotations of Pfam domains and GO categories, both MPI-2 and MPI-3 carry genes encoding PE/PPE family proteins, MPI-2 encodes the type VII secretion system, and MPI-3 encodes genes for mycolic acid synthesis in the cell wall. Some of these genes were already widely used in early diagnosis or treatment of M. tuberculosis. The novel candidate PAI MPI-1 encodes CRISPR-C as family proteins, which are known to be associated with persistent infection of M. tuberculosis. Our data represents a molecular basis and protocol for Electronic supplementary material The online version of this article (doi:10.1007/s11033-014-3463-4) contains supplementary material, which is available to authorized users. J. Xie G. Xu J. Hu G. Wang (&) F. Li (&) Norman Bethune Medical College of Jilin University, Changchun 130021, Jilin, China e-mail: [email protected] F. Li e-mail: [email protected] F. Zhou G. Mai Shenzhen Institutes of Advanced Technology, and Key Lab for Health Informatics, Chinese Academy of Sciences, Shenzhen 518055, Guangdong, People’s Republic of China
comprehensive annotating the pathogenic systems of M. tuberculosis, and will also facilitate the development of diagnosis and vaccination techniques of M. tuberculosis. Keywords Mycobacterium tuberculosis Pathogenicity island Genomic barcode Diagnosis Vaccination
Introduction Tuberculosis is one of the most dreaded diseases caused by Mycobacterium tuberculosis (M. tuberculosis)of the last few centuries. Many aspects of M. tuberculosis physiology and pathogenesis still remain elusive. It has developed a complex matrix of virulence genes which attack the host organisms with various mechanisms, which makes its diagnosis and treatment extremely difficult [1, 2]. Point mutation, gene recombination and horizontal gene transfer (HGT) represent three major strategies in forming the pathogenic landscape of the modern M. tuberculosis strains [3]. In particular, HGT facilitates the dynamics of pathogenic functional modules of modern M. tuberculosis strains, by exchanging genetic material with the environmental microbes [4]. Pathogenicity island(PAI)is a genomic region with pathogenic genes that was acquired through HGT from another microbial genome [5–7]. Therefore, the acquisition of PAIs plays an essential role in the adaptive evolution of a microbe, by giving the PAI-receiving microbe novel virulent modules [6, 8–10]. At present, no experimental technique is available to perform the screening and functional analysis of PAIs from the pan-genome aspect. Large-scale PAI screening is mainly achieved based on either nucleotide statistics or phylogenetic analysis [11–13]. But all of these in silico methods lack a reasonable screening accuracy, due to that highly expressed genes also have altered compositions of single or di-
123
Mol Biol Rep
nucleotide sequences [14, 15]. Thus the integration and visual annotation of more information on the pathogen genome are necessary and it might provide a powerful tool for high-throughput and more accurate screening of PAIs. Our previous study found that the frequency spectrum for each K-mer nucleotide string (K, 1 \ K \ 6) within the region of equal length fragments in microbial genome has consistency. The barcode-like visual annotation (Barcode image) of a genome can be further obtained by making a digital and graphical process for the array matrix of the frequency spectrum. According to this theory, any microbial genome can be represented as a unique barcode image [15]. Genome barcode carries all the genetic information in the genome, and has a one-to-one correspondence with the genome sequence. This study applied the genomic visualization technique, genomic barcode, to the M. tuberculosis chromosomes, and screened the PAIs in these chromosomes based on the observation that PAIs are usually significantly different to the other genomic regions in the barcode feature. Furthermore, Pfam_Scan [16] and Blast2GO [17] were used to analyze the function of the screened PAIs and to describe the characteristics.
Results and discussion Genomic barcodes of M. tuberculosis We split the chromosome sequences of the five M. tuberculosis strains into a plurality of non-overlapping Fig. 1 Genomic barcodes of the five M. tuberculosis strains, i.e. a H37Rv, b H37Ra, c CDC1551, d F11, e KZN1435, f Mycobacterium leprae TN, g Escherichia coli K12 str MG1655, and h a random sequence with the same length as E. coli K12 str MG1655. The genomic barcodes were generated using the parameters W = 1,000 and K = 4
123
fragments of 1,000-bp lengths (W = 1,000). This study chose to use 4-mer strings for generating the genomic barcodes, as similar to [18]. So in this study, K = 4 and the total number of K-nucleotide pairs is N(K) = 128. Let the chromosome length be L, and the genomic barcode will be L/1,000 in height and N(K) in width, as shown in Fig. 1. Figure 1 shows that five strains of M. tuberculosis share a unique genomic barcode pattern, different strains have different barcode structure, of which the more closer sibship the more similar barcode graph, such as M. tuberculosis H37Rv, CDC 1551, F11 and KZN1435. In contrast, the farther sibship the more different barcode image, such as Mycobacterium leprae TN and Escherichia coli K12 str MG1655. This feature of genomic barcode was observed in all Prokaryotes’ DNA sequences, whereas the genomic barcode of a randomly generated DNA sequence (as in Fig. 1h) doesn’t even show any such patterns. So genomic barcode is an inherent characteristic of the genetic material of all life forms. The difference between two DNA sequences can be modeled by the Euclidean distance of the genomic barcodes of the two sequences. The two DNA sequences may come from two genomes or two genomic regions of the same genome. Although the majority of the sequence fragments of the same species share an almost identical barcode (Fig. 1a–e), some regions are significantly different in genomic barcode to the other regions. Such regions tend to have foreign origins or multiple repeats, and almost all the pathogenicity islands were exchanged through HGT events [19].
PE-PGRS family protein 11 3,922,471–3,950,263 MPI3
3,931,148–3,958,929
3,916,304–3,943,046
3,934,824–3,963,047
3,910,930–3,935,736
*28 K
151.6 ± 38.2
PPE family protein and PE-PGRS family protein 150.5 ± 28.6 10 3,727,488–3,767,102 MPI2
3,738,329–3,776,417
3,725,096–3,747,265
3,739,815–3,779,781
3,723,397–3,763,176
*39 K
147.1 ± 35.0 11 *14 K 3,130,179–3,144,352 3,112,942–3,126,002 3,130,192–3,143,764 3,118,224–3,131,773 MPI1
F11 CDC1551 H37Ra H37Rv
Location PAI
Table 1 The gene information of M. tuberculosis PAIs
A PAI is defined to be a large genomic region consisting of genes encoding important virulent regulatory factors, usually acquired through horizontal gene transferring events [23]. Three screening criteria were established for the PAIs in M. tuberculosis, (a) the abnormal genomic barcode fragment is longer than 10 K bps, (b) the abnormal barcode region exists in the chromosomes of all the five studied M. tuberculosis strains, (c) and the region harbors virulent genes. Three candidate PAIs were detected based on the above three criteria, and they were denoted as MPI1, MPI-2 and MPI-3, respectively. The locations and encoded genes of the three candidate PAIs in M. tuberculosis were shown in Table 1 and illustrated in Fig. 2. The entire three candidate PAIs were located in the region between 3,000,000 and 4,000,000 in the majority of M. tuberculosis chromosomes. This region contains a large amount of inserted fragment and is a hot region for gene recombination [24]. All the three candidate PAIs are flanked by tRNA genes, which are known to be the integration sites of PAIs [25]. MPI-2 and MPI-3 are close to each other, and may constitute one larger PAI, which was separated by new genomic recombination in between. Firstly, both MPI-2 and MPI-3 harbor genes from the same PE-PGRS protein family. The hypothesis is also supported by the barcode distance analysis. A recently acquired genomic island, e.g. a PAI, usually has a different genomic barcode pattern compared with the other regions of the current host genome, and will accumulate enough mutations over time to share a similar genomic barcode with the host genome [15]. Table 1 show that the barcode distances of MPI-2 and
KZN 1435
Pan-genome-scale screening of PAIs in M. tuberculosis
1,282,054–1,295,433
Number of genes Length (bps)
Barcode distance (X SD)
Function
According to the distribution frequency of barcode distances, we defined the genomic regions with Euclidean distance \90 or [150 as abnormal regions [18]. The barcode distances of all the 1,000-bp fragments in the five strains of M. tuberculosis were listed in the Supplementary Table 1.The abnormal barcode regions within the M. tuberculosis genome account for 18.0–18.8 % with an average value of 18.4 %, which is almost half of that in Escherichia coli (35.6 %). The data is consistent with the ratios of horizontal transferring events in the two bacteria [12, 20]. We have analyzed the abnormal fragments across the M. tuberculosis genomes, and found the following: *20 % of the abnormal fragments can be explained in terms of (a) horizontal gene transfers, (b) phage invasions and (c) highly expressed genes, based on PHX-PA [21] and Prophinder [22]. The continuous abnormal barcode regions of least 10 K bps long play an essential role in the exchange of genetic material between different pathogenic species and provide a molecular basis for M. tuberculosis pathogenicity.
Transposase and hypothetical protein
Mol Biol Rep
123
Mol Biol Rep
(a)
oriC MPI-3 MPI-2
Mycobacterium tuberculosis H37Rv 4411532bp MPI-1
(b) fadE26
240
fadE27
fadD17
PE_PGRS53 PE_PGRS53
PE_PGRS54
ilvX
MPI-1
Rv3510c
PE_PGRS55PE_PGRS56 PE_PGRS55 PE_PGRS56
MPI-2
PPE55 fadD18 PE_PGRS57 fadD18 PE_PGRS57
MPI-3
220 200 180 160 140 120 100 80 3000000
3200000
3400000
3600000
3800000
4000000
Fig. 2 Three PAI candidates in M. tuberculosis. a Barcode distance distribution of the genomic region between 3 and 4 M bps in the strain H37Rv of M. tuberculosis. Outer concentric circle: predicted coding regions on the plus strand, second concentric circle: predicted coding regions on the minus strand, third concentric circle: predicted
coding regions on the both strand, fourth concentric circle: buffer zone, fifth concentric circle: predicted PAI candidates, sixth concentric circle: GC content. b Barcode distance distributions and genes of three PAI candidates
MPI-3 were similar, suggesting that the times for these two PAIs entering the host genome were very close or the same. The alternative hypothesis is that the two PAIs were horizontally transferred into M. tuberculosis separately, and the two integration sites happened to be close to each other. Considering the random nature of HGT integration and other evidences, the alternative hypothesis is much less
likely than the original one. We could find same small peaks around location 3,500,000 in Fig. 2b, we didn’t consider this peak as a candidate PAI duo to its small size (less than 10 K). Phylogenetic analysis of these three PAIs showed that MPI-2 and MPI-3 are more conserved than MPI-1 in sequenced M. tuberculosis genome (http://www.ncbi.nlm.
123
Not found Not found Rv3506-Rv3521 3,925,000–3,959,000 3,922,471–3,950,263 MPI-3
Rv3504-Rv3514
Not found
Not found Rv3339c-Rv3355c
Rv2814-Rv2829c 3,120,566–3,137,012
3,724,615–3,769,807 Rv3343-Rv3355c
Not found
3,730,000–3,770,000 3,727,488–3,767,102 MPI-2
Rv2813-Rv2823c 3,118,224–3,131,733 MPI-1
Rv3341-Rv3350
Location Location Location
Genes
PMID: 11435108(206) PMID/Tool (citations)
One notorious feature of M. tuberculosis is that this pathogen can evade the attacks of the host immune system and remains a persistent infection in the host for months, years or even decades, without active replication. This is termed latent tuberculosis and M. tuberculosis under the latent status may regain activities whenever the host body’s immune system works abnormally. This represents the major problem for the tuberculosis control program. Functional annotation of genes in MPI-1 indicated that most of them encode transposases or hypothetical proteins. Although the gene function of the MPI-1 has not been reported previously, the PfamScan domain prediction suggested the functional gene Rv2820c of MPI-1 encoded a CRISPR-C as family protein. Previous report indicated that this protein family could activate the function of specific interval sequence encoding small RNA fragments (crRNA), which is closely related to persistent infection of M. tuberculosis [27]. Hence, we speculate that MPI-1 may be involved in the regulation of persistent infection of M. tuberculosis. Recently, some new reviews on biological functions of CRISPR system showed that this system was absence in Beijing sublineage [28, 29], which also consistent with our phylogenetic analysis of MPI-1. Further study of MPI-1 function may improve our understanding of the molecular mechanism of M. tuberculosis persistent
Table 2 Comparison with the known PAIs in M. tuberculosis H37Rv
The functional characterization of the M. tuberculosis PAIs
Genomic barcode
Genes
Island view (78)
Genes
PAIDB (28)
nih.gov/genome/browse/), which probably because MPI-1 may have resided in the M. tuberculosis genome for quite a long period of time led to the emergence of more mutations and recombination. This hypothesis also confirmed by genomic barcode distance as shown in Table 1. In order to analysis the advantage of genomic barcode visualization for screening of PAIs in M. tuberculosis,we compared our results to several published tools with high citation (http://scholar.google.com/), as shown in Table 2. We could found that these methods lack a reasonable screening accuracy or missed some PAIs, particularly so for PAIs that has been in the host genome for a long period of time. In addition, there are differences of the bountry of each PAIs. Actually, bountry analyses of PAIs are still elusive due to the limitations of the experimental techniques. Genomic barcode annotation could distinguish the bounry of each PAIs duo to the high-resolution base on and small windows sliding of 4-mer nucleatide and barcode distance calculate. For example, PAI candidates using genomic barcode tool do not break any operons in M. tuberculosis while others do [26]. However, this work is basically a proof-of-principle one, and more information of PAIs will be studied in the future work.
Location
Genes
Mol Biol Rep
123
Mol Biol Rep
infection, and facilitate the discovery of molecular drug targets and immune interventions. Genes within MPI-2 encode PPE family proteins, transposases, and the type VII secretion system. PPE55 protein showed an over-expression in the early stage of M. tuberculosis infection [30–32], and its N-termini has a signal peptide that may prompt a secreted trans-membrane protein involved in the body’s immune regulation during the early stage of M. tuberculosis infection [33–35]. So markers from MPI-2 can be used for the early diagnosis of M. tuberculosis infection. MPI-3 encodes PPE family proteins and enzymes responsible for fatty acid synthesis and degradation. The carboxylase encoded by the fad family genes in MPI-3 is a key protein responsible for providing the substrate for the last reaction of mycolic acid synthesis pathway [36]. Mycolic acids are important biological molecules in M. tuberculosis, usually synthesized through the type II fatty acid biosynthetic system. Thus MPI-3 has an important role in strengthening the latent persistence against the adverse growth conditions in the host body or in vitro environments. In addition, the PE-PRGS family proteins within MPI-3 are mainly involved in the interaction of M. tuberculosis with the host cells. Therefore, the genes within MPI-3 may also serve as the targets of anti-tuberculosis drugs. In summary, the genomic barcode provides an efficient and intuitive way to visualize a genome, from which hypotheses could be quickly formulated and tested using either computational or experimental techniques. Using this technique, we screened three candidate M. tuberculosis PAIs, including two with known functions and another one with unknown function. The further functional analysis of PAIs provides some novel insights for our understanding of the pathogenesis and evolution of M. tuberculosis and also opens a new avenue for choosing the diagnostic targets and vaccine targets.
Materials and methods Genome sequence data We focused on the PAIs of five firstly sequenced strains of M. tuberculosis, i.e. H37Rv, H37Ra, CDC1551, F11 and KZN1435, and collected their genome sequences and gene annotations from the NCBI Genome database [37] as of January 5, 2012. Calculation of genomic barcodes and their Euclidean distance The M. tuberculosis genomic sequences were partitioned into a plurality of non-overlapping fragments with equal
123
length W, and the frequency vector of the pairs of Knucleotide string and its reverse complement in each fragment was calculated. The appearance frequencies of k-mers are grouped into dozens of grayscale colors, e.g. L = 14 for k = 4 and WindowSize = 1,000. The reasonable prediction performances of this parameter combination for multiple applications suggest that WindowSize = 1,000-bp is good enough for k = 4 [15]. We mapped the frequencies in the frequency vectors to the gray scale integers (i.e. 0–255), and got an image for the given genome sequence. As in Fig. 1, the image of a genome sequence resembles a barcode used in supermarkets, and therefore called so. The more the color is close to white, the higher the frequency of the K-nucleotide string is in the fragment. For the array matrix of k-mers strings of the microbial genome, the barcode distance can be represented by the Euclidean distance of array. For example, for the array of any two matrices M1 and M2, the difference between them is defined as follow: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u L NðKÞ uX X t ðM1 ði; jÞ M2 ði; jÞÞ2 i¼1 j¼1
where L is the number of fragments, and N(K) is the number of all K-nucleotide sequences. The functional analysis of PAIs region using PfamScan and Blast2GO The genes in the candidate PAIs were consistently annotated using the programs PfamScan [38] and BLAST2GO [39]. The Pfam database (version 23.0) was downloaded from the website. The Pfam database is a large collection of protein domain families, including Pfam_ls and Pfam_fs, et al. We used the curated data set Pfam_ls, and annotated the genes in the candidate PAIs with Evalue e-4. BLAST2GO is a GO annotation and functional analysis tool based on sequence similarity search, and PAI genes were annotated using the online version of Blast2GO software with Evalue e-10. Acknowledgments This work was supported by National Natural Science Foundation of China (81101295 and 81071424), Specialized Research Fund for the Doctoral Program of Higher Education of China (20110061120093), China Postdoctoral Science Foundation (20110491311 and 2012T50304), Foundation of Jilin Provincial Health Department (2011Z049),Foundation of Jilin Provincial Science and Technology Department (20130522013JH), Norman Bethune Program of Jilin University (2012219). It was also supported in part by the Shenzhen Research Grant ZDSY 20120617113021359, China 973 program (2011CB512003 and 2010CB732606-6) and NSFC 31000447. Computing resources were partly provided by the Dawning supercomputing clusters at SIAT CAS.
Mol Biol Rep
References 1. Gazdik MA, McDonough KA (2005) Identification of cyclic AMPregulated genes in Mycobacterium tuberculosis complex bacteria under low-oxygen conditions. J Bacteriol 187:2681–2692 2. Sanz J, Navarro J, Arbue´s A, Martı´n C, Marijua´n PC, Moreno Y (2011) The transcriptional regulatory network of Mycobacterium tuberculosis. PLoS ONE 6:e22178 3. Gogarten JP, Townsend JP (2005) Horizontal gene transfer, genome innovation and evolution. Nat Rev Microbiol 3:679–687 4. Pallen MJ, Wren BW (2007) Bacterial pathogenomics. Nature 449:835–842 5. Ferna´ndez-Go´mez B, Ferna`ndez-Guerra A, Casamayor EO, Gonza´lez JM, Pedro´s-Alio´ C, Acinas SG (2012) Patterns and architecture of genomic islands in marine bacteria. BMC Genom 13:347 6. Hacker J, Blum-Oehler G, Mu¨hldorfer I, Tscha¨pe H (1997) Pathogenicity islands of virulent bacteria: structure, function and impact on microbial evolution. Mol Microbiol 23:1089–1097 7. Sui SJH, Fedynak A, Hsiao WWL, Langille MGI, Brinkman FSL (2009) The association of virulence factors with genomic islands. PLoS ONE 4:e8094 8. Becq J et al (2007) Contribution of horizontally acquired genomic islands to the evolution of the tubercle bacilli. Mol Biol Evol 24:1861–1871 9. Yu G et al (2011) Integrative analysis of transcriptome and genome indicates two potential genomic islands are associated with pathogenesis of Mycobacterium tuberculosis. Gene 489(1):21–29 10. Boyd EF, Almagro-Moreno S, Parent MA (2009) Genomic islands are dynamic, ancient integrative elements in bacterial evolution. Trends Microbiol 17:47–53 11. Karlin S (2001) Detecting anomalous gene clusters and pathogenicity islands in diverse bacterial genomes. Trends Microbiol 9:335–343 12. Veyrier F, Pletzer D, Turenne C, Behr MA (2009) Phylogenetic detection of horizontal gene transfer during the step-wise genesis of Mycobacterium tuberculosis. BMC Evol Biol 9:196 13. Juhas M, van der Meer JR, Gaillard M, Harding RM, Hood DW, Crook DW (2009) Genomic islands: tools of bacterial horizontal gene transfer and evolution. FEMS Microbiol Rev 33:376–393 14. Karlin S, Mrazek J, Ma J, Brocchieri L (2005) Predicted highly expressed genes in archaeal genomes. Proc Natl Acad Sci USA 102:7303–7308 15. Zhou F, Olman V, Xu Y (2008) Barcodes for genomes and applications. BMC Bioinform 9:546 16. Bateman A et al (2004) The Pfam protein families database. Nucleic Acids Res 32:D138–D141 17. Go¨tz S et al (2008) High-throughput functional annotation and data mining with the Blast2GO suite. Nucleic Acids Res 36:3420–3435 18. Wang G, Zhou F, Olman V, Li F, Xu Y (2010) Prediction of pathogenicity islands in enterohemorrhagic Escherichia coli O157:H7 using genomic barcodes. FEBS Lett 584:194–198 19. Gal-Mor O, Finlay BB (2006) Pathogenicity islands: a molecular toolbox for bacterial virulence. Cell Microbiol 8:1707–1719 20. Lawrence JG, Ochman H (1998) Molecular archaeology of the Escherichia coli genome. Proc Natl Acad Sci 95:9413
21. Mrazek J, Bhaya D, Grossman AR, Karlin S (2001) Highly expressed and alien genes of the Synechocystis genome. Nucleic Acids Res 29:1590–1601 22. Lima-Mendez G, Van Helden J, Toussaint A, Leplae R (2008) Prophinder: a computational tool for prophage prediction in prokaryotic genomes. Bioinformatics 24:863–865 23. Letek, M et al (2010) The genome of a pathogenic rhodococcus: cooptive virulence underpinned by key gene acquisitions. PLoS Genet 6(9):e1001145 24. Cole ST (1999) Learning from the genome sequence of Mycobacterium tuberculosis H37Rv. FEBS Lett 452:7–10 25. Freitas-Vieira A, Anes E, Moniz-Pereira J (1998) The site-specific recombination locus of mycobacteriophage Ms6 determines DNA integration at the tRNA(Ala) gene of Mycobacterium spp. Microbiology 144(Pt 12):3397–3406 26. Mao F, Dam P, Chou J, Olman V, Xu Y (2009) DOOR: a database for prokaryotic operons. Nucleic Acids Res 37:D459–D463 27. Marraffini LA, Sontheimer EJ (2010) CRISPR interference: RNA-directed adaptive immunity in bacteria and archaea. Nat Rev Genet 11:181–190 28. Arnvig K, Young D (2012) Non-coding RNA and its potential role in Mycobacterium tuberculosis pathogenesis. RNA Biol 9:427–436 29. Bryant J, Chewapreecha C, Bentley SD (2012) Developing insights into the mechanisms of evolution of bacterial pathogens from whole-genome sequences. Future Microbiol 7:1283–1296 30. Sampson SL (2011) Mycobacterial PE/PPE proteins at the hostpathogen interface. Clin Dev Immunol 2011:1–11 31. Delogu G et al (2006) PE_PGRS proteins are differentially expressed by Mycobacterium tuberculosis in host tissues. Microb Infect 8:2061–2067 32. Rachman H et al (2006) Unique transcriptome signature of Mycobacterium tuberculosis in pulmonary tuberculosis. Infect Immun 74:1233–1242 33. Singh KK, Dong Y, Patibandla SA, McMurray DN, Arora VK, Laal S (2005) Immunogenicity of the Mycobacterium tuberculosis PPE55 (Rv3347c) protein during incipient and clinical tuberculosis. Infect Immun 73:5004–5014 34. Koh KW, Soh SE, Seah GT (2009) Strong antibody responses to Mycobacterium tuberculosis PE-PGRS62 protein are associated with latent and active tuberculosis. Infect Immun 77:3337–3343 35. Singh PP, Parra M, Cadieux N, Brennan MJ (2008) A comparative study of host response to three Mycobacterium tuberculosis PE_PGRS proteins. Microbiology 154:3469–3479 36. Raman K, Yeturu K, Chandra N (2008) targetTB: a target identification pipeline for Mycobacterium tuberculosis through an interactome, reactome and genome-scale structural analysis. BMC Syst Biol 2:109 37. Barrett T et al (2012) BioProject and BioSample databases at NCBI: facilitating capture and organization of metadata. Nucleic Acids Res 40:D57–D63 38. Finn RD et al (2010) The Pfam protein families database. Nucleic Acids Res 38:D211–D222 39. Conesa A, Gotz S, Garcia-Gomez JM, Terol J, Talon M, Robles M (2005) Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research. Bioinformatics 21:3674–3676
123
