From Corynebacterium glutamicum to Mycobacterium tuberculosis—towards transfers of gene regulatory networks and integrated data analyses with MycoRegNet

Abstract

Year by year, approximately two million people die from tuberculosis, a disease caused by the bacterium Mycobacterium tuberculosis. There is a tremendous need for new anti-tuberculosis therapies (antituberculotica) and drugs to cope with the spread of tuberculosis. Despite many efforts to obtain a better understanding of M. tuberculosis' pathogenicity and its survival strategy in humans, many questions are still unresolved. Among other cellular processes in bacteria, pathogenicity is controlled by transcriptional regulation. Thus, various studies on M. tuberculosis concentrate on the analysis of transcriptional regulation in order to gain new insights on pathogenicity and other essential processes ensuring mycobacterial survival. We designed a bioinformatics pipeline for the reliable transfer of gene regulations between taxonomically closely related organisms that incorporates (i) a prediction of orthologous genes and (ii) the prediction of transcription factor binding sites. In total, 460 regulatory interactions were identified for M. tuberculosis using our comparative approach. Based on that, we designed a publicly available platform that aims to data integration, analysis, visualization and finally the reconstruction of mycobacterial transcriptional gene regulatory networks: MycoRegNet. It is a comprehensive database system and analysis platform that offers several methods for data exploration and the generation of novel hypotheses. MycoRegNet is publicly available at http://mycoregnet.cebitec.uni-bielefeld.de.

INTRODUCTION

Year by year, approximately two million people die worldwide from tuberculosis (1) and one-third of the world's total population suffer from this communicable disease (http://www.who.int) caused by the bacterium Mycobacterium tuberculosis. Tuberculosis is the leading cause of death to people living with HIV and claims on average 200 000 lives every year, most of them in Africa. Persons infected with tuberculosis will not directly develop the characteristic full-blown clinical picture, but in most cases the latent form, which can progress to an active condition after years. About 10–15 people can be infected by a person with active tuberculosis a year, if she or he is left untreated (http://www.who.int). Although there is effective treatment to cure patients with tuberculosis, and new strategies have been developed to stop its further dissemination, its containment is still a serious problem (2). The number of multi-resistant strains not responding to standard drug treatments is increasing constantly worldwide (3,4). Consequently, there is a tremendous need for new anti-tuberculosis therapies (antituberculotica) and drugs to cope with the spread of tuberculosis. Despite many efforts to obtain a better understanding of the pathogenicity of M. tuberculosis and its survival strategy in humans, many questions are still unresolved. The molecular mechanisms responsible for resisting the human immune system and their activation are not perceived sufficiently so far; most notably, its ability to remain within the human host for years in a clinically latent state (5). Among other cellular processes in bacteria, pathogenicity is controlled by transcriptional regulation. Thus, various studies on M. tuberculosis concentrate in the analysis of transcriptional regulation in order to gain new insights on pathogenicity and other essential processes ensuring mycobacterial survival. The identification and characterization of transcriptional regulation on a genome-wide level will enable a better understanding of drug metabolism in M. tuberculosis and facilitate the development of new antibiotics, which are urgently needed. At present, studies focus mainly on the analysis of single regulons, or distinct subunits of the complex transcriptional regulatory network of M. tuberculosis [see e.g. (3,5,6)]. Bioinformatics platforms for data storage and public access of transcriptional regulation exist for M. tuberculosis, similar to other organisms such as Escherichia coli [RegulonDB (7)] or Corynebacterium glutamicum [CoryneRegNet (8)]. MtbRegList (9) and MTBreg (http://www.doe-mbi.ucla.edu/services/mtbreg) offer information relevant to regulatory interactions in M. tuberculosis H37Rv (MT) accumulated from literature or attained from computational predictions. While MtbRegList contains predicted and characterized regulatory DNA motifs cross-referenced with transcription factors (TFs), MTBreg combines a collection of conditionally regulated proteins together with information about selected TFs. However, both systems are designed as data repositories and only provide nonsatisfying bioinformatics support necessary for transcriptional gene regulatory network visualization, analysis, and reconstruction. Recently, the TB database has become available. This integrated online platform for tuberculosis research combines the annotated genome and expression data with a suite of bioinformatic tools for data analysis (10). The scope of TB database is placed on investigating and providing expression data, while little support is given for the reconstruction of regulatory networks based on these findings. Hence, there is currently no online platform or database system available, which aims to an appropriate data handling and analysis of transcriptional regulation in M. tuberculosis on a genome-wide level.

Here, we introduce MycoRegNet, an online accessible, user-friendly platform dedicated to the biomedical researcher, who is interested in the regulation of gene expression in the human pathogen M. tuberculosis. MycoRegNet is online available at http://mycoregnet.cebitec.uni-bielefeld.de.

The first idea of our approach is based on the assumption that orthologous TFs tend to regulate the expression of orthologous target genes for taxonomically closely related species (11–13). Corynebacterium glutamicum and M. tuberculosis are taxonomically classified into the suborder Corynebacterineae of the Actinobacteria phylum and are thus taxonomically closely related (14). Hence, the industrially important amino acid producer C. glutamicum has been successfully applied as model organism, e.g. for investigating cell envelope synthesis of M. tuberculosis (15–17). We therefore started with the well-examined regulatory network of C. glutamicum ATCC 13032 (CG) (18), which is stored in the corynebacterial reference database CoryneRegNet (8). Our comparative genomics approach aims for a reliable transfer of known regulatory interactions from CG to MT. Instead of relying exclusively on the detection of orthologous genes, we consider further evidence by means of an integrated TF binding site (TFBS) prediction. The resulting data were subsequently stored in an online platform designed for the visualization and analysis of the deduced transcriptional regulatory network, which enables the execution of bioinformatics tools for further hypotheses generation: MycoRegNet. The remainder of this article is structured as follows: we first describe the workflow used for the transfer of C. glutamicum data to M. tuberculosis in detail. The design of MycoRegNet is briefly introduced afterwards. It aims to overcome typical data integration problems and to supply online visualization and hypotheses generation tools. In the last section, we illustrate and discuss these functionalities. We finally conclude that MycoRegNet is an appropriate reference database and platform for gene regulatory network analysis of M. tuberculosis.

MATERIALS AND METHODS

The network reconstruction pipeline mainly consists of the detection of (i) conserved genes between CG and MT and (ii) binding sites upstream the conserved genes in MT. Based on the corresponding results, a list of putative gene regulatory interactions in MT is generated and imported into the MycoRegNet database back-end (see Figure 1 for a graphical overview of the workflow).

Figure 1.

Diagram of the prediction pipeline. The diagram shows the main steps performed during transfer of gene regulations from C. glutamicum to M. tuberculosis. Starting with an orthology detection, the next step was a prediction of conserved regulations. Based on that, a TFBSs prediction provided further evidence. Finally, the results can be exported as TAB-delimited files and imported into the MycoRegNet data repository.

Detection of orthologous genes

Generally, the detection of orthologous genes is not straightforward, since analysis can be perturbed by factors like paralogs or sequence divergences in the genomes of interest. To reduce such effects, we searched for orthologous genes by performing bidirectional BLASTP (19) searches on the corresponding protein sequences. Therefore, we scanned the CG genome for sequence similarities with the MT genome and vice versa, performing BLASTP with an E-value cut-off of 10⁻⁴ in both directions. As a result, we obtained amino acid sequence pairs, so called bidirectional best hits (BBHs), representing the reciprocal best alignments of respective protein sequences. Thus, identified BBHs were considered to be putative orthologous proteins in CG and MT, which in turn indicates the respective genes to be regulated in both bacteria by orthologous TFs.

Transfer of regulatory interactions

Based on the previously identified BBHs, regulatory interactions characterized in CG were transferred to MT. We utilized the comprehensive data on transcriptional regulation in CG collected in the corynebacterial reference database CoryneRegNet (8), which contains 806 regulatory interactions of 72 TFs and 544 regulated target genes on CG (status: January 2009). For each regulatory interaction taken from CG, both the gene encoding the TF and the target gene were compared to the list of predicted orthologs in the MT genome. Only if both, the TF as well as its target gene, were identified as BBH, the regulatory interaction was transferred from CG to the orthologous counterparts in MT and was considered as a candidate transcriptional regulation in MT. Furthermore, we assume the regulatory role of the TF (activation or repression) to be conserved as well, including known autoregulations.

Further evidence through conserved TFBSs

In the last step of our regulatory network prediction pipeline, we add further evidence to the orthology-based approach introduced above by combining the preliminary results with the prediction of TFBSs. Therefore, all known binding sites of characterized TFs of CG with potential orthologs in MT were utilized to create appropriate motif profiles. TF binding motifs were modeled as so called position weight matrices (PWMs), the most widely used model for that purpose. However, we applied only PWMs of corynebacterial TFs deduced from more than 20 binding sites, i.e. of the TFs GlxR, RamB, AmtR, DtxR and LexA. To detect instances of the respective motifs in MT, we employed the TFBS matching tool PoSSuMsearch (20) and scanned 580-bp long, noncoding DNA sequences upstream all genes and operons, which have been detected as potential orthologs to target genes of the respective TF. The upstream sequences ranged from +20 bp relative to the transcription start. In our initial approach, we performed a restrictive search by setting the P-value threshold to 10⁻⁵. Due to the low number of detected binding sites in the first PoSSuMsearch runs, we decided to decrease the P-value threshold since the set P-value might be chosen too restrictive for our TFBS predictions. To determine a new P-value, we considered P-values of binding site matches of the PWM for GlxR upstream 26 target genes as marking value, where the binding of the GlxR ortholog Rv3676 in MT was experimentally verified (21–24). For P-value definition, we chose the binding site that match upstream one of these genes with the worst P-value. Thus, we finally set the P-value to <10⁻² and defined for each target gene/operon in MT the TFBS match with the lowest P-value as prediction for the respective binding site.

Taken together, the outcome of the above introduced workflow is a list of transcriptional regulations for MT where (i) the TF is conserved, and (ii) the target gene is conserved between CG and MT as well, and additionally (iii) a binding site is predicted, if the target gene/operon is controlled by one of the five TFs where a TFBS search was performed for. Hence, the resulting predictions present most likely regulatory interactions in MT due to the taxonomically close relation between CG and MT. This is the data we aim to integrate into the MycoRegNet platform together with validated knowledge we have from (5,25–34).

Data integration with the MycoRegNet platform

Based on our experiences with CoryneRegNet, we designed MycoRegNet in a very similar way: as an ontology-based data warehouse for mycobacterial TFs and regulatory networks. We set it up as a sister project of CoryneRegNet to store, analyze and visualize the regulatory interactions in M. tuberculosis that are derived from the above introduced prediction pipeline. MycoRegNet is composed of two main parts: (i) A web front-end running on an Apache HTTP web server that manages user-database interactions as well as the execution of further online bioinformatics computations. (ii) The back-end consists of data preprocessing tools and a MySQL database that stores all data corresponding to the deduced and ontologically restructured mycobacterial gene regulatory interactions. This process comprises the integration of transcriptional regulations, the complete genome sequence of MT along with the genome annotation as stored in the GenBank database (NCBI) (35), operon predictions available from the Virtual Institute of Microbial Stress and Survival (VIMSS) (36), precalculated PWMs and other preprocessed data necessary for subsequent online TFBS detections, and stimulons derived from literature (25–31). The import and conversion software is implemented in Java, while the web pages generated at front-end level are developed in PHP. An embedded Java applet realizes the visualization of gene regulatory networks from the included data. A SOAP-based web Service (37) client/ server system implemented by means of NuSOAP enables a bidirectional interconnection with GenDB (38) and EMMA (39). The server is open access and provides well-structured data access via the SOAP interface to any other bioinformatics client. GenDB is an open source system for the annotation of prokaryotic genomes, while EMMA is a web-based application for the storage and analysis of transcriptomics data from microarrays. By means of the clients for GenDB and EMMA, data integrated in MycoRegNet is supplemented with up-to-date information on the genome annotation of MT (GenDB) and gene expression data preanalyzed with EMMA. To give one example, the Web Service client for GenDB facilitates the mapping of all genes controlled by a certain regulator to KEGG pathways (40) in order to provide an overview on the general nature of a TF of interest. Furthermore, the automatic annotation pipeline of GenDB can be used to regularly update gene function assignments.

RESULTS AND DISCUSSION

Here, we first summarize the database content. Subsequently, we present and discuss the benefits of MycoRegNet from the end-user perspective. We first describe the web interface with special attention to the TFBS prediction feature and the network visualization and analysis capability. We briefly describe the Web Service access afterwards and finally demonstrate the platforms' visualization functionality by means of an application example.

The database content

By using the above described transfer pipeline for regulatory interactions, we identified 1012 of 3991 proteins from MT as putative orthologs to proteins from CG. Based on the respective set of genes coding for the orthologous proteins, we detected 226 of 806 regulatory interactions from CG as likely conserved in MT (Table 1). Our initial findings reveal 24 partial conserved regulons affecting processes of the carbohydrate metabolism, cellular program, macroelement and metal homeostasis, SOS and stress response, specific biosynthesis as well as processes governed by sigma factors. By setting the P-value threshold to 102, we could put further evidence to 129 target genes 40 by predicting binding sites upstream the respective target genes/operons regulated by the TF orthologs of GlxR, RamB, AmtR, DtxR and LexA in MT (Table 2). All in all, we obtained a set of regulatory interactions which is based on good evidence. The database content comprises 618 regulatory interactions for 515 target genes regulated by 26 TFs. Several gene expression experiments are also directly stored within MycoRegNet's database back-end (data not shown). We also integrated genome annotation data of M. tuberculosis CDC1551 for future investigations concerning transcriptional regulation in another ecotype of M. tuberculosis.

Table 1.

Putative gene regulations of CG in MT

TF	Target genes
Carbohydrate metabolism
Rv0465c	Carbohydrate metabolism
	Rv0211 (pckA), Rv0247c (-), Rv0363c (fba), Rv0408 (pta)
	Rv0409 (ackA), Rv0465c (-), Rv0467 (icl), Rv0896 (gltA2)
	Rv0904c (accD3), Rv0951 (sucC), Rv0952 (sucD), Rv1475c (acn)
	Rv1837c (glcB), Rv1862 (adhA), Rv2193 (ctaE), Rv2241 (aceE)
	Rv2332 (mez), Rv2967c (pca), Rv3318 (sdhA)
	Cell division and septation
	Rv1009 (rpfB)
	Specific biosynthesis pathways
	Rv0884c (serC), Rv1010 (ksgA), Rv1011 (ispE), Rv1379 (pyrR)
	Rv1380 (pyrB), Rv1381 (pyrC)
Rv0792c	Carbohydrate metabolism
	Rv0753c (mmsA)
Rv1719	Carbohydrate metabolism
	Rv0554 (bpoC), Rv1074c (fadA3), Rv1719 (-), Rv2503c (scoB), Rv2504c (scoA)
Rv3676	Carbohydrate metabolism
	Rv0211 (pckA), Rv0247c (-), Rv0400c (fadE7), Rv0465c (-)
	Rv0467 (icl), Rv0896 (gltA2), Rv0904c (accD3), Rv0951 (sucC)
	Rv0952 (sucD), Rv1098c (fumC), Rv1130 (-), Rv1161 (narG)
	Rv1162 (narH), Rv1163 (narJ), Rv1436 (gap), Rv1437 (pgk)
	Rv1438 (tpi), Rv1475c (acn), Rv1837c (glcB), Rv1854c (ndh)
	Rv1862 (adhA), Rv1872c (lldD2), Rv2029c (pfkB), Rv2193 (ctaE)
	Rv2194 (qcrC), Rv2195 (qcrA), Rv2196 (qcrB), Rv2200c (ctaC)
	Rv2524c (fas), Rv2967c (pca), Rv3010c (pfkA), Rv3043c (ctaD)
	Rv3279c (birA), Rv3280 (accD5), Rv3318 (sdhA), Rv3548c (-), Rv3676 (-)
	Cell division and septation
	Rv1009 (rpfB), Rv2145c (wag31), Rv2201 (asnB)
	Macroelement and metal homeostasis
	Rv0820 (phoT), Rv0928 (pstS3), Rv0929 (pstC2), Rv0930 (pstA1), Rv2220 (glnA1)
	Rv2832c (ugpC), Rv2833c (ugpB), Rv2834c (ugpE), Rv2835c (ugpA), Rv2918c (glnD)
	Rv2919c (glnB), Rv2920c (amt), Rv3859c (gltB)
	SOS and stress response
	Rv0867c (rpfA), Rv3048c (nrdF2), Rv3217c(-), Rv3219 (whiB1), Rv3681c (whiB4)
	Specific biosynthesis pathways
	Rv0884c (serC), Rv1010 (ksgA), Rv1011 (ispE), Rv1092c (coaA)
	Rv3001c (ilvC)
	Rv3002c (ilvN), Rv3003c (ilvB1)
Cellular Program
RelA	Sigma factor module
	Rv1221 (sigE), Rv2710 (sigB), Rv3221A (-), Rv3911 (sigM)
	SOS and stress response
	Rv2720 (lexA)
Macroelement and metal homeostasis
Rv0485	Macroelement and metal homeostasis
	Rv0132c (fgd2)
PhoP	Macroelement and metal homeostasis
	Rv0545c (pitA), Rv0757 (phoP), Rv0758 (phoR), Rv0820 (phoT)
	Rv0928 (pstS3), Rv0929 (pstC2), Rv0930 (pstA1), Rv1095 (phoH2)
	Rv2832c (ugpC), Rv2833c (ugpB), Rv2834c (ugpE), Rv2835c (ugpA)
Rv0827c	Macroelement and metal homeostasis
	Rv0827c (-)
Rv1994c	Macroelement and metal homeostasis
	Rv1994c (-)
IdeR	Carbohydrate metabolism
	Rv0247c (-), Rv3318 (sdhA)
	Macroelement and metal homeostasis
	Rv0827c (-), Rv0844c (narL), Rv1285 (cysD), Rv1286 (cysN)
	Rv2391 (nirA), Rv2392 (cysH), Rv2393 (-), Rv2895c (viuB)
	Rv3044 (fecB), Rv3841 (bfrB)
Rv3160c	Macroelement and metal homeostasis
	Rv1848 (ureA), Rv1849 (ureB), Rv1850 (ureC), Rv1852 (ureG)
	Rv2220 (glnA1), Rv2918c (glnD), Rv2919c (glnB), Rv2920c (amt)
	Rv3664c (dppC), Rv3665c (dppB), Rv3666c (dppA), Rv3859c (gltB)
Rv3173c	Macroelement and metal homeostasis
	Rv0132c (fgd2), Rv0485 (-), Rv1079 (metB), Rv1133c (metE)
	Rv1175c (fadH), Rv1285 (cysD), Rv1286 (cysN), Rv1294 (thrA)
	Rv1296 (thrB), Rv1392 (metK), Rv2124c (metH), Rv2334 (cysK1)
	Rv2391 (nirA), Rv2392 (cysH), Rv2393 (-), Rv3025c (iscS)
	Rv3028c (fixB), Rv3029c (fixA), Rv3173c (-), Rv3340 (metC)
	Rv3341 (metA)
Sigma factor module
SigB	Carbohydrate metabolism
	Rv0363c (fba), Rv1023 (eno), Rv1098c (fumC), Rv1436 (gap)
	Rv1437 (pgk), Rv1438 (tpi), Rv3010c (pfkA)
	SOS and stress response
	Rv3132c (devS)
	Specific biosynthesis pathways
	Rv2210c (ilvE)
SigM	SOS and stress response
	Rv0384c (clpB), Rv1464 (csd), Rv1465, Rv1471 (trxB1)
	Rv3418c (groES), Rv3913 (trxB2), Rv3914 (trxC)
SOS and stress response
HspR	SOS and stress response
	Rv0350 (dnaK), Rv0351 (grpE), Rv0352 (dnaJ1), Rv0353 (hspR)
	Rv0384c (clpB), Rv0440 (groEL), Rv2745c
HrcA	SOS and stress response
	Rv0440 (groEL), Rv3418c (groES)
LexA	Cell division and septation
	Rv2748c (ftsK)
	SOS and stress response
	Rv1235 (lpqY), Rv1638 (uvrA), Rv1696 (recN), Rv2592c (ruvB)
	Rv2593c (ruvA), Rv2594c (ruvC), Rv2720 (lexA), Rv2736c (recX)
	Rv2737c (recA), Rv3370c (dnaE2), Rv3395c, Rv3585 (radA)
Rv2745c	SOS and stress response
	Rv0782 (ptrBb), Rv2460c (clpP2), Rv2461c (clpP1), Rv2725c (hflX)
	Rv3596c (clpC1), Rv3715c (recR), Rv3716c
WhiB1	SOS and stress response
	Rv3913 (trxB2), Rv3914 (trxC)
MtrA	SOS and stress response
	Rv0917 (betP), Rv3476c (kgtP)
CspA	Carbohydrate metabolism
	Rv1837c (glcB)
Specific biosynthesis pathways
PyrR	Specific biosynthesis pathways
	Rv1379 (pyrR), Rv1380 (pyrB), Rv1381 (pyrC), Rv2883c (pyrH)
ArgR	Specific biosynthesis pathways
	Rv1383 (carA), Rv1384 (carB), Rv1652 (argC), Rv1653 (argJ)
	Rv1654 (argB), Rv1655 (argD), Rv1656 (argF), Rv1657 (argR)
	Rv0488

Putative gene regulations of CG in MT, predicted in silico by using the introduced MycoRegNet pipeline

Table 2.

Detected binding sites upstream transferred target genes of CG in MT

TF	Gene ID	Gene name	Operon	Binding motif
Rv0465c	Rv0211a	pckA		ATAACTACGCAGG
	Rv0249c	–	Rv0249c-Rv0248c-Rv0247ca	AGTAGTTCGCGAT
	Rv0363ca	fba	–	CGTACTTCTCAAA
	Rv0407	pta	Rv0407-Rv0408a-Rv0409a	CGTGCTGTGCTCA
	Rv0465c	–	Rv0465ca-Rv0464c	CTAACTCTGCGAA
	Rv0467a	icl	–	CAAAATTTGCAAA
	Rv0884ca	serC	–	ATGGCATGGCCGA
	Rv0896a	gltA2	–	TGAGCAGATCACT
	Rv0904ca	accD3	–	ATTGCATGGCAAG
	Rv0951	sucC	Rv0951a-Rv0952a	AGTGCTAAGCCGT
	Rv1009	rpfB	Rv1009a-Rv1010a-Rv1011a	TCTACTTACCAAA
	Rv1379	pyrR	Rv1379a-Rv1380a-Rv1381a-Rv1382-Rv1383-Rv1384-Rv1385	AGTGCTACGCTGC
	Rv1475c	acn	Rv1475ca-Rv1474c	ACTGCTAGGCTGA
	Rv1837ca	glcB	–	TAGGCTGAGCAAT
	Rv1862a	adhA	–	TGTGCTGGGCTAA
	Rv2193	ctaE	Rv2193a-Rv2194-Rv2195-Rv2196	ACTACAAAGCGTC
	Rv2241	aceE	Rv2241a-Rv2242	CAAACAGCGCAAG
	Rv2332a	mez	–	TGCGCTCTGCGAA
	Rv2967ca	pca	–	CATGCAATGTCAA
	Rv3316	sdhA	Rv3316-Rv3317-Rv3318a-Rv3319	GTTGCATTGCCCC
IdeR	Rv0249c	–	Rv0249c-Rv0248c-Rv0247ca	TTAGATGAGCGCACCCACG
	Rv0827ca	–	–	CTATGGATCGCTGTACTAC
	Rv0844ca	narL	–	CGACGAGCAGCTAAACTCA
	Rv1285	cysD	Rv1285a-Rv1286a	GAGGGCGAGGCACACGTCA
	Rv2391	nirA	Rv2391a-Rv2392a-Rv2393a	TCAGGTGCGCGTCTCCCAG
	Rv2895ca	viuB	–	TAAGCGAAGCCGAACGCCA
	Rv3044a	fecB	–	GTAGACCAGGCTCCCCTTG
	Rv3316	sdhA	Rv3316-Rv3317-Rv3318a-Rv3319	CTAAGAAAAGCCAGCCTAA
	Rv3841a	bfrB	–	CTAGGAAAGCCTTTCCTGA
LexA	Rv1235	lpqY	Rv1235a-Rv1236-Rv1237-Rv1238	TCGACTATCTATCCGA
	Rv1638a	uvrA	–	TCGAATGTCAGCTCGC
	Rv1696a	recN	–
	Rv2594c	ruvC	Rv2594ca-Rv2593ca-Rv2592ca	TCGAACGATTGTTCGG
	Rv2720a	lexA	–	TCGAACACATGTTTGA
	Rv2737c	recA	Rv2737ca-Rv2736ca	TCGAACAGGTGTTCGG
	Rv2748ca	ftsK	–	CCGACCAGGTGCTCGC
	Rv3370ca	dnaE2	–	TCGAACAATTGTTCGA
	Rv3395c	–	Rv3395ca-Rv3394c	TCGAACATATTTTCGA
Rv3160c	Rv1848	ureA	Rv1848a-Rv1849-Rv1850a-Rv1851-Rv1852a-Rv1853	GTGTCTACTGCGCGATGATCGAGAGCAT
	Rv2220a	glnA1	–	CAACACGGGGTTGACTGACGGGCAATAT
	Rv2920c	amt	Rv2920ca-Rv2919ca-Rv2918ca	AAGTTTTACGTTAATCCTGATGAAACAT
	Rv3666c	dppA	Rv3666ca-Rv3665c-aRv3664ca-Rv3663c-Rv3662c	GTGGTAGCTAACGGTCACCGGCGAGTGT
	Rv3859c	gltB	Rv3859ca-Rv3858c	CGCTTGACGGACAGCCTATCGACAAGAC
Rv3676	Rv0211a	pckA		TGTGAGCAGGCTTATA
	Rv0249c	–	Rv0249c-Rv0248c-Rv0247ca	TGTGATCTGTAACACC
	Rv0400ca	fadE7	–	AGTGATGAGCACCCCG
	Rv0465c	–	Rv0465ca-Rv0464c	TTTGTCGAGGCTCACG
	Rv0467a	ivl	–	TGTTACAACGCTCACA
	Rv0820a	phoT		GGTGGTGATCCGCACC
	Rv0867ca	rpfA		TGTGACATTACCCACA
	Rv0884ca	serC		TGTGAGCTTGTTCACA
	Rv0896a	gltA2	–	GGCGTTGAACATCACC
	Rv0904ca	accD3	–	CGTGAGTCGTATCACG
	Rv0928	pstS3	Rv0928a-Rv0929a-Rv0930a	ACTGAATTGAAACTCA
	Rv0951	sucC	Rv0951a-Rv0952a	TGTGAGTTGGATCACG
	Rv1009	rpfB	Rv1009a-Rv1010a-Rv1011a	GGTGGCGCTCATCACC
	Rv1092ca	coaA		TGCCACGTAGGTCACG
	Rv1099c	fumC	Rv1099c-Rv1098ca-Rv1097c
	Rv1130	–	Rv1130a-Rv1131	TGTGGATAAGTCCAGG
	Rv1161	narG	Rv1161a-Rv1162a-Rv1163a-Rv1164-Rv1165-Rv1166	TGCGTTGAACGGCACG
	Rv1436	gap	Rv1436a-Rv1437a-Rv1438a	GGTTGTTTAGCCAACA
	Rv1475c	acn	Rv1475ca-Rv1474c	TGTAACTGCCGACATA
	Rv1837ca	glcB	–	AGGGATGCACTACACA
	Rv1854ca	ndh	–	TGTGGCTGATGACACA
	Rv1862a	adhA	–	CGTGGGGCGCCACACA
	Rv1872ca	lldD2	–	GATGCCGTAGCGCACT
	Rv2029c	pfkB	Rv2029ca-Rv2028c-Rv2027c-Rv2026c	GGTGACGAGTCGCGCA
	Rv2145c	wag31		CGTGACTGGCGTCCCA
	Rv2193	ctaE	Rv2193a-Rv2194a-Rv2195a-Rv2196a	GGTGGATAGGTTCACC
	Rv2200c	ctaC	Rv2200ca-Rv2199c	TGTGATACAGGAGGCG
	Rv2201	asnB		GCTGTCGAAGACCACG
	Rv2220a	glnA1		TGTGACGGAAAAGACG
	Rv2524ca	fas	–	CGTTACCCACGACACG
	Rv2835c	ugpA	Rv2835ca-Rv2834ca-Rv2833ca-Rv2832ca	GGTGATGCCGGGCACG
	Rv2920c	am	Rv2920ca-Rv2919ca-Rv2918ca	AGTGGACCAATTCCCC
	Rv2967ca	pca	–	CGTGGTGGTGGTCACC
	Rv3003ca	ilvB1	Rv3003ca-Rv3002ca-Rv3001ca	TGTGGTGGCCACCCCA
	Rv3010ca	pfkA	–	GGTGATGGCGATGACC
	Rv3043c	ctaD	Rv3043ca-Rv3042c	AGTGGATCGCATCCCG
	Rv3048ca	nrdF2		GGTGACTGGAAACGCA
	Rv3217ca	–		TGTGGTGGCGGTCGCA
	Rv3219a	whiB1		AGTGAGATAGCCCACG
	Rv3279c	birA	Rv3279ca-Rv3278c	TATCGGCTGCCGCACA
	Rv3280	accD5	Rv3280-aRv3281-Rv3282	CGGGACGTCGACCACA
	Rv3316	sdhA	Rv3316-Rv3317-Rv3318a-Rv3319	CGAGACGTTTTCCACG
	Rv3549c	–	Rv3549c-Rv3548ca	GGTGATCGGCATTGCA
	Rv3676a	–		TGTCACCTACGACAGA
	Rv3681ca	whiB4		TGAGATACAGGTAACA
	Rv3859c	gltB	Rv3859ca-Rv3858c	TGCTCCGGATTTCACA

Detected binding sites of GlxR (ortholog in MT: Rv3676/Crp), RamB (ortholog in MT: Rv0465c), AmtR (ortholog in MT: Rv3160c), DtxR (ortholog in MT: IdeR/Rv3173c) and LexA (ortholog in MT: Rv2720/LexA) orthologs of CG in MT. Code:

^aTransferred target gene of CG in MT.

The user interface

As for other online databases, MycoRegNet's web interface provides the three major capabilities: browsing the database content, searching by specifying filter criteria and basic visualization possibilities. Furthermore, the front-end offers the execution of computational features. At the main page (Figure 2), one has the option to search or to browse the database content. The user may browse the data repository by clicking on an ecotype name of interest and is provided with an overview on the selected organism. Alternatively, using one of the provided options within the search form, the database can be searched for specific gene/protein identifiers, gene/protein names, regulator types or functional modules. The search results are presented in tabular form, listing all relevant information for subsequent investigation. Furthermore, the following built-in features can be accessed from the main page, directly: TFBScan [for TFBS predictions; see below) and COMA (to check for contradictions within microarray gene expression studies, given the regulatory network stored in the database; refer to (8) for more details]. Detailed information on the results can be obtained via respective links at the result page. By selecting a particular gene, the corresponding gene details page is invoked. It presents a detailed overview of all available data attached to the gene of interest. Besides general information about the gene/protein (position in the genome, nucleotide sequence, etc.), it comprises a graphical representation of the genomic context, regulated target genes (if encoding a TF) including the TFBSs, etc., and stimulons that initiate a differential gene expression level. The integrated Web Service client for GenDB maintains the representation of up-to-date gene annotation data. General information (description, comments, an assigned function, etc.) is listed as well as the EC numbers for enzymes, and links to COG (41) and GO (42). Additionally, all target genes of a TF of interest, are linked to KEGG pathways and a list of regulated pathways is displayed.

Figure 2.

MycoRegNet main page. The main page includes a typical search mask, a statistical overview of the database content, an entry point to browse the integrated organisms, and links to more specific statistics, the system documentation and a tutorial on how to use the MycoRegNet Web Service.

TFBS prediction

With the integrated PoSSuMsearch software, MycoRegNet provides a statistically sound tool for the prediction of TFBSs based on PWMs, which have been precalculated during data import. To our knowledge, PoSSuMsearch is the only TFBSs profiling tool that offers exact P-value calculations and at the same time provides reasonable response times on genome-wide runs. There are three ways to access this feature through the MycoRegNet web site: (1) The TFBScan button at the main page offers the possibility to upload user-defined sequences in FASTA format. (2) At any gene details page the user can predict TFBSs in the upstream sequence of the selected gene. (3) If the gene of interest encodes a TF, the PWM learned from the known TFBSs of the TF may be used to scan for further TFBSs in the upstream sequences of all other mycobacterial genes. The predicted results may further be visualized as graphs. The interface is easy to use: one just has to choose a background model (nucleotide distribution) and a P-value threshold. For further details regarding the prediction of prokaryotic TFBSs by utilizing PoSSuMsearch, the reader is referred to (20,43).

Gene regulatory network visualization

As mentioned earlier, MycoRegNet also provides a network visualization toolkit: GraphVis. It is a Java applet, which graphically reconstructs regulatory networks as graphs based on selected genes and a user-defined graph depth cutoff. It traverses all regulatory interactions from the starting point until the graph depth cutoff has been reached. Finally, a Java applet window appears showing the regulatory network as graph, where nodes represent genes and edges regulatory interactions. GraphVis allows the user to zoom into the graph, apply different layout styles, remove selected elements or retrieve detailed information on selected genes. Furthermore, it is possible to extend the graph dynamically with more genes/regulations from the database and to display the operon grouping of presented genes. Visualized networks may also be graphically compared between two species or between a predicted and an evidenced network by utilizing special comparative graph layout algorithms. In addition, GraphVis features the projection of gene expression data onto the genes of a visualized network. The user can choose to apply gene expression data from the stimulon repository of MycoRegNet or from own tab-delimited text or MS-Excel files, which can be uploaded to GraphVis directly. It is also possible to use expression data extracted from EMMA by means of the integrated Web Service interconnection. According to the differential expression level of the genes, the concerned nodes are resized within the graph. Thus, the user can achieve a comprehensive overview of the transcriptional response of M. tuberculosis to a certain stimulus.

Well-structured data access by using Web Services

Although no real standard in bioinformatics, a growing number of platforms offer SOAP-based Web Service access to their data repositories [refer to some EBI resources (44) or to the BRENDA database (45), just to name two of them]. Many databases still provide flat files for exchange with other data processing systems. Thus, the developers of novel tools and platforms have to perform updates in regular time intervals and to adjust the downloaded data for their special purpose. On that account, gene regulatory data stored in MycoRegNet can also be accessed via the integrated Web Service server. The data can be integrated directly into corresponding projects without further time-consuming data processing. Detailed information on how to use the MycoRegNet Web Services is available from the main page via the Web Service button.

Application example—the regulatory network of the GlxR ortholog Rv3676 (Crp_MT) in MT

Both GlxR (Cg0350) of C. glutamicum and CRP_MT (Rv3676) of M. tuberculosis belong to the Crp-Fnr family of TFs (46) and have been characterized as cAMP sensing homologs of E. coli Crp (23,47,48). Crp-cAMP-dependent gene regulation is commonly involved in carbon catabolite repression and forms one of the possible connections between carbon metabolism and virulence (49,50).

In mycobacteria, cAMP signalling is the subject of intensive research, as it may be related to virulence of these strains (51,52). It is noteworthy that M. tuberculosis contains 16 putative adenylate cyclases, as well as 10 putative cyclic nucleotide binding proteins (53,54), hinting at a crucial and diverse role for cAMP signalling in mycobacteria.

GlxR of C. glutamicum has been in the focus of interest in the last years (48,55–59), and available data indicates GlxR as global regulator with about 150 target genes in functional diverse network modules, such as carbohydrate metabolism, aerobic and anaerobic respiration, fatty acid metabolism, aromatic compound degradation, glutamate uptake and nitrogen assimilation, the cellular stress response and resuscitation.

Previous studies suggested a similar vital role for Crp_MT in M. tuberculosis. Published data implicate Crp_MT in virulence, hypoxia and nutrient starvation (21,23,24,60). Deletion of Crp_MT altered the expression of 16 genes, and caused an impaired growth phenotype in bone marrow-derived macrophages as well as in tuberculosis mouse models (24). Several suggestions for a putative Crp_MT regulon have been made, although these studies relied solely on the detection of putative binding sites (23,24,60).

As part of our pipeline, the known regulatory interactions of GlxR collected in CoryneRegNet have been used to reconstruct the regulon of the orthologous TF Crp_MT. Due to the apparent vital role of these regulators in their respective organisms, and the available data on putative target genes and characterized binding sites, we chose them as application case for our analysis. Employing our pipeline, regulatory interactions with 64 target genes could be transferred from C. glutamicum GlxR to M. tuberculosis Crp_MT. Furthermore, we considered 26 genes with experimental evidence of regulation by Crp_MT as potential target genes (21–24).

Based on experimentally verified binding sites of Crp_MT (21–23) together with binding sites predicted by the TFBS search of our pipeline, we complemented the suggested regulon with the prediction of Crp_MT binding sites in the upstream regions of putative target genes.

In contrast to the TFBS searched within our pipeline, we created an adopted and optimized PWM for Crp_MT from experimentally verified and predicted binding sites, and applied it for TFBS search. To detect the novel binding sites, we set the P-value threshold to 10⁻⁵ and performed a restrictive search on sequences upstream genes/operons concering the whole genome of MT. Again, we used PoSSuMsearch for binding site prediction scanning 580-bp long upstream sequence, ranging from +20 bp relative to the transcription start site. Using Weblogo (61), we generated a sequence logo from the detected binding sites of Crp_MT and from the appropriate binding sites of GlxR that were used for PWM creation (see Methods section). The resulting sequence logos are shown in Figure 3.

Figure 3.

Sequence logo of the predicted Crp_MT binding sites (A) in comparison to the sequence logo of GlxR (B). The sequence logo models the binding site motif of Crp_MT. It was deduced from the predicted binding sites in Table 3. The height of each letter within an individual stack represents the nucleotide's frequency relative to the particular motif position; thus, the degree of a nucleotide's conservation is indicated by the stack according to the respective position.

In total, we identified 207 putative target genes of Crp_MT, organized in 121 transcription units (see Table 3 and Figure 4). Of this set, 17 genes belong to the mycobacterial core regulon (62) and 41 were reported as essential for M. tuberculosis (63,64). Furthermore, at least 17 genes of the suggested regulon are connected to antibiotic resistance and virulence of M. tuberculosis (65–69). Based on annotation information for M. tuberculosis (69), knowledge about orthologous C. glutamicum genes and operon structures, we attributed individual target genes to distinct functional modules.

Table 3.

Predicted Crp_MT binding sites

Gene ID	Gene	Motif positiond	Motif sequence	Operon
Carbohydarate metabolism
Rv0211a	pckA	−166	TGTGAGCAGGCTTATA	–
Rv0249c	sdhCD	−104	TGTGATCTGTAACACC	Rv0249c-Rv0248c-Rv0247ca
Rv0249c	sdhCD	−410	GGTGTCGGAGGTCACA	Rv0249c-Rv0248c-Rv0247ca
Rv0458	adhA	−41	TGTGAGCTGTATTACA	Rv0458-Rv0459
Rv0465c	–	−167	TTTGTCGAGGCTCACG	Rv0465ca–Rv0464cg
Rv0467a,g	icl	−341	TGTTACAACGCTCACA	–
Rv0896a	gltA2	−356	GGCGTTGAACATCACC	–
Rv0951	sucC	−173	TGTGAGTTGGATCACG	Rv0951a,f–Rv0952a,f
Rv1099c	–	−515	GCTGATGAATCCCACG	Rv1099c–Rv1098ca,f–Rv1097c
Rv1130	prpD2	−152	TGTGGATAAGTCCAGG	Rv1130a–Rv1131
Rv1436	gap	−48	GGTTGTTTAGCCAACA	Rv1436a,f–Rv1437a,f–Rv1438a,f
Rv1475c	acn	−462	TGTAACTGCCGACATA	Rv1475ca,f–Rv1474c
Rv1552	frdA	−284	TGTGATCTAGGTCACGb	Rv1552–Rv1553–Rv1554–Rv1555
Rv1837ca	glcB	−381	AGGGATGCACTACACA	–
Rv1862a	adhA	−227	CGTGGGGCGCCACACA	–
Rv1872ca	lldD2	−200	GATGCCGTAGCGCACT	–
Rv2029ca	pfkB	−410	GGTGACGAGTCGCGCA	Rv2029c–Rv2028c–Rv2027c–Rv2026cf
Rv2967ca,f	pca	−389	CGTGGTGGTGGTCACC	–
Rv3010ca	pfkA	−532	GGTGATGGCGATGACC	–
Rv3316	sdhC	−386	CGAGACGTTTTCCACG	Rv3316–Rv3317–Rv3318a–Rv3319
Rv3676a	CRP	−538	TGTCACCTACGACAGA	–
Fatty acid metabolism
Rv0097	–	−526	TGTCACGCCGGCCACG	Rv0097–Rv0098e–Rv0099c–Rv0100e–Rv0101
Rv0166	fadD5	−84	TGTGACCCAGACAACA	–
Rv0400ca,f	fadE7	−5	AGTGATGAGCACCCCG	–
Rv1185c	fadD21	−168	CGTGACGCCCCTCACG	–
Rv1714	–	−405	GGTGACGGCGGCCACA	Rv1714f–Rv1715f–Rv1716–Rv1717–Rv1718
Rv2485cc	lipQ	−91	TGTGATCCTCGACACA	–
Rv2486	echA14	−287	TGTGTCGAGGATCACA	–
Rv2524ca,f	fas	−259	CGTTACCCACGACACG	–
Rv2930c	fadD26	−498	TGTTAATCTCGTCACA	Rv2930–Rv2931–Rv2932f–Rv2933–Rv2934–Rv2935–Rv2936g–Rv2937g–Rv2938–Rv2939
Rv3279c	birA	−38	TATCGGCTGCCGCACA	Rv3279ca–Rv3278ce
Rv3280	accD5	−331	CGGGACGTCGACCACA	Rv3280a–Rv3281e,f–Rv3282
Rv3549c	–	−67	GGTGATCGGCATTGCA	Rv3549c–Rv3548ca
Nitrogen assimilation
Rv1538c	ansA	−187	TGTGAGCACCACCACA	–
Rv2220a,f,g	glnA1	−1	TGTGACGGAAAAGACG	–
Rv2920c	amt	−2	AGTGGACCAATTCCCC	Rv2920ca–Rv2919ca,g–Rv2918ca
Rv3859c	gltB	−398	TGCTCCGGATTTCACA	Rv3859ca,f–Rv3858cf
PGRS
Rv0453	PPE11	−269	GGTGACCAAACTCACG	–
Rv1386	PE15	−133	TGTGACCAAACTCACCb	Rv1386e–Rv1387c,e
Rv2408	PE24	−213	GGTGATCGGCGTCACG	–
Rv2591	P_PGRS44	−38	CGTGACATGTGTCACA	–
Rv3136c	PPE51	−16	AAGGAGCTGAGACACA	–
Rv3650	PE33	−83	TGTGATGCACTTGACA	–
Respiration
Rv1161	narG	−512	TGCGTTGAACGGCACG	Rv1161a–Rv1162a–Rv1163a–Rv1164–Rv1165–Rv1166f
Rv1623cc	cydA	−181	CGTGGTGATCGGCACA	–
Rv1854ca	ndh	−109	TGTGGCTGATGACACA	–
Rv2193	ctaE	−517	GGTGGATAGGTTCACC	Rv2193a,f–Rv2194a,f–Rv2195a,f–Rv2196a,f
Rv2200c	ctaC	−23	TGTGATACAGGAGGCG	Rv2200ca,f–Rv2199c
Rv3043c	ctaD	−227	AGTGGATCGCATCCCG	Rv3043ca,f–Rv3042cf
Other cellular processes
Rv0019cg	fhaB	−69	CGTGACTTTGCTGACGb	–
Rv0079	–	−110	GGTGACACAGCCCACA	Rv0079–Rv0080
Rv0103c	ctpB	−159	TGTGACGGGCGTCACA	–
Rv0104	–	−1	TGTGACGCCCGTCACA	–
Rv0145	–	−59	AGTGATGTGCCACACAb	Rv0145–Rv0146
Rv0188c	–	−356	AGAGAACAACGTCGCA	–
Rv0194	–	−517	TGTCATCTAGATCACG	–
Rv0232	–	−53	CGTGATGCAGCGCACA	Rv0232–Rv0233
Rv0250ce	–	−37	TGTGATCTGTAACACC	–
Rv0360c	–	−2	CGTGACCAAGCGCACA	–
Rv0457c	–	−43	TGTAATACAGCTCACA	–
Rv0470A	–	−212	TGTGGTGGGAATCACA	–
Rv0483	lprQ	−116	TGTGTTTGGTATCACA	–
Rv0793	–	−375	TGTGATGGTGCGCACG	–
Rv0820a,g	phoT	−538	GGTGGTGATCCGCACC	–
Rv0867ca	rpfA	−443	TGTGACATTACCCACAb	–
Rv0884ca,f	serC	−91	TGTGAGCTTGTTCACAb	–
Rv0885	–	−133	TGTGAACAAGCTCACA	Rv0885–Rv0886
Rv0904ca	accD3	−2	CGTGAGTCGTATCACG	–
Rv0928	pstS3	−6	ACTGAATTGAAACTCA	Rv0928a,g–Rv0929a,g–Rv0930a,g
Other cellular processes
Rv0993	galU	−8	TGTGAACGATGTCACG	Rv0993f–Rv0994g–Rv0995g
Rv0950c	–	−153	CGTGATCCAACTCACAb	–
Rv0992c	–	−109	CGTGACATCGTTCACA	Rv0992–Rv0991–Rv0990
Rv1009	rpfB	−271	GGTGGCGCTCATCACC	Rv1009a–Rv1010a,g–Rv1011a,f
Rv1057	–	−248	CGTGACCTAGGTAACA	–
Rv1092ca,f	coaA	−242	TGCCACGTAGGTCACG	–
Rv1111ce	–	−411	GGTGACATGAGTCACG	–
Rv1158c	–	−69	TGTCACTTGAGTCACAb	Rv1158ce–Rv1157ce
Rv1159	pimE	−77	TGTGACTCAAGTGACA	–
Rv1230c	–	−79	GGTGATCTAGTTCACGb	–
Rv1291c	–	−323	TGTGATCGGCGCCACC	–
Rv1314c	–	−294	GGTGATCCGGGCCACG	–
Rv1324e	–	−104	TGTGATCTTGGTCATA	–
Rv1482c	–	−23	TGTGACTCAGCACACT	–
Rv1566c	–	−235	CGTGACTGAAATCACA	–
Rv1568	bioA	−553	TGTGATTTCAGTCACG	Rv1568–Rv1569g–Rv1570–Rv1571
Rv1592cc	–	−215	TGTGATAGGCGCCACG	–
Rv1757c	–	−351	TGTGACGGCGGCCACG	–
Rv1779c	–	−89	TGTGAACAACACCACA	–
Rv1780	–	−147	TGTGGTGTTGTTCACA	–
Rv1890c	–	−7	TGTGTCGTGGCCCACA	–
Rv1891e,g	–	−63	TGTGGGCCACGACACA	Rv1891–Rv1892–Rv1893e
Rv2145ca,f	wag31	−463	CGTGACTGGCGTCCCA	–
Rv2172ce	–	−2	TGTGACCCTCAACACG	–
Rv2180c	–	−304	TGTGTGGAACAACACA	–
Rv2201a,f	asnB	−336	GCTGTCGAAGACCACG	–
Rv2258c	–	−459	GGTGACGTCGACCACG	–
Rv2362c	recO	−224	TGTGGGCTGGCTCACA	Rv2362c–Rv2361cf–Rv2360c
Rv2377c	mbtH	−268	TGTGGTTCACCTCACT	–
Rv2406c	–	−34	TGTGAACCAGCTCACC	–
Rv2407	–	−242	GGTGAGCTGGTTCACA	–
Rv2428c	ahpC	−93	GGTGTGATATATCACC	–
Rv2450c	rpfE	−509	TGTGGCGCAGGTCACC	–
Rv2450c	rpfE	−422	CGTGATTCGGCTCACG	–
Rv2455c	–	−237	AGTGACCAATACCACA	Rv2455c–Rv2454c–Rv2453c
Rv2650c	–	−305	CGTGAGGAGCCTCACG	–
Rv2699c	–	−116	TGTGATGTAAATCACA	–
Rv2700e,f	–	−138	TGTGATTTACATCACA	–
Rv2712c	–	−296	GGTGAGGTAGAGCACA	–
Rv2835c	ugpA	−513	GGTGATGCCGGGCACG	Rv2835c–Rv2834ca–Rv2833ca,f–Rv2832ca,f
Other cellular processes
Rv2874	dipZ	−351	TGTGGCGGAGTTCACA	–
Rv3003c	ilvB1	−335	TGTGGTGGCCACCCCA	Rv3003ca,f–Rv3002ca,f–Rv3001ca,f
Rv3048ca,f	nrdF2	−2	GGTGACTGGAAACGCA	–
Rv3053c	nrdH	−347	GGTGATCTGCGACACG	Rv3053c–Rv3052c–Rv3051cf
Rv3217ca,e	–	−278	TGTGGTGGCGGTCGCA	–
Rv3219 a,c	whiB1	−176	AGTGAGATAGCCCACGb	–
Rv3613cc	–	−458	CGTGACGAATCCCCCA	–
Rv3617	ephA	−315	TGTGACCGGTGTCACT	Rv3617–Rv3618
Rv3645	–	−179	TGTGAGCCGAATCACG	–
Rv3681ca	whiB4	−106	TGAGATACAGGTAACA	–
Rv3729	–	−190	TGTGACCACGGCCACG	–
Rv3843c	–	−505	GGTGAGGTAAGTCACA	Rv3843ce–Rv3842cg
Rv3856c	–	−547	TGTGGGCTTCGTCACA	–
Rv3857c	–	−341	TGTGGGCTTCGTCACAb	–
Consensus			TGTGANNNNNNTCACA

Crp_MT binding sites detected by the TFBS search of the introduced pipeline and by the additional TFBS search with adopted and optimized PWMs. Bold letters indicate conserved pentamers of the motif. Codes:

^aTransferred target gene from CG.

^bExperimentally verified binding site by EMSA/CHiP/RT-PCR (21–23).

^cGene showed altered expression in microarray studies of ΔRv3676 versus WT (24).

^dMotif position relative to the translation start site.

^eCore gene.

^fEssential gene.

^gGene involved in virulence processes

Figure 4.

Reconstructed network of the GlxR ortholog Crp_MT. The network reconstruction of the Crp_MT regulon is based on the 121 transcription units presented in Table 3. It was generated by the integrated network reconstruction tool GraphVis of MycoRegNet. Transcription units relying on binding site predictions/experimental verifications that were reported previously in (22–24,60) and correspond with our findings are colored according to the appropriate publication. Arrows and gene IDs (node labels) coloured in red indicate a repressive regulation of Crp_MT, green arrows correspond to an activating regulation.

Similar to present knowledge on GlxR, results implicate Crp_MT in the regulation of several functional modules such as carbohydrate metabolism (40 target genes), fatty acid metabolism (33 target genes), respiration (16 target genes) and nitrogen assimilation (7 target genes). Therefore, the position of the GlxR homolog Crp_MT as global regulator in the transcriptional regulatory network seems to be conserved in M. tuberculosis. It is interesting to note that the suggested regulon comprises genes involved in essential functional modules, e.g. the citrate cycle, as well as genes involved in the synthesis of the cellular envelope which plays an important role in the virulence of M. tuberculosis. Together with the supposed regulation of further virulence−associated genes this might explain why a functional Crp_MT is required for virulence in model systems (24).

CONCLUSIONS

With MycoRegNet, we have set up a system that allows researchers of the tuberculosis community to perform comprehensive analysis and visualizations of the gene regulatory network of MT. With its TFBS prediction it further provides easy access to a method that helps to generate new hypotheses in silico. As the sister project to CoryneRegNet, the MycoRegNet database content was generated through our comparative genomics pipeline, which provided us with reliable transfers of gene regulatory interactions from the reference organism C. glutamicum to M. tuberculosis. With MycoRegNet, the corresponding data are publicly available and can be accessed easily through the web interface, or in a well-structured manner by using the MycoRegNet Web Service to maintain the reconstruction, visualization, and validation of mycobacterial regulatory networks at different hierarchical levels. Taken together, MycoRegNet is a reference resource for the tuberculosis community to gain a better understanding of the complex coherences of transcriptional gene control. It has the potential to assist researchers at the development of new vaccines and drugs to treat and prevent tuberculosis. Although MycoRegNet has been initially designed for MT, it may also serve for other mycobacterial strains in future, such as the already integrated M. tuberculosis CDC1551.

FUNDING

GenoMik-Plus initiative of the German Federal Ministry of Education and Research (grant 0313805A); ERA-NET PathoGenoMics SPATELIS project (to J.K.); German Academic Exchange Service (to J.B. for his work at ICSI). Funding for open access charge: The work was financed in part by the GenoMik-Plus initiative of the German Federal Ministry of Education and Research, grant 0313805A.

Conflict of interest statement. None declared.