PredictRegulon: a web server for the prediction of the regulatory protein binding sites and operons in prokaryote genomes

Abstract

An interactive web server is developed for predicting the potential binding sites and its target operons for a given regulatory protein in prokaryotic genomes. The program allows users to submit known or experimentally determined binding sites of a regulatory protein as ungapped multiple sequence alignments. It analyses the upstream regions of all genes in a user-selected prokaryote genome and returns the potential binding sites along with the downstream co-regulated genes (operons). The known binding sites of a regulatory protein can also be used to identify its orthologue binding sites in phylogeneticaly related genomes where the trans-acting regulator protein and cognate cis-acting DNA sequences could be conserved. PredictRegulon can be freely accessed from a link on our world wide web server: http://www.cdfd.org.in/predictregulon/.

INTRODUCTION

With over 100 bacterial genomes sequenced, a key challenge of post-genomic research is to dissect the complex transcription regulatory network which controls the metabolic and physiological process of a cell. A first step towards this goal is to identify the genes within a genome that are controlled by a specific transcription regulatory protein. This paper describes a web server tool—PredictRegulon—for genome-wide prediction of potential binding sites and target operons of a regulatory protein for which few experimentally identified binding sites are known. This technique could utilize the available experimental data on binding sites of transcription regulatory proteins from various bacterial species (1–3) for identification of regulons in phylogenetically related species.

PREDICTREGULON METHOD

The program, PredictRegulon, first constructs the binding site recognition profile based on ungapped multiple sequence alignment of known binding sites. This profile is calculated using Shannon's positional relative entropy approach (4). The positional relative entropy Q_i at position i in a binding site is defined as

where b refers to each of the possible bases (A, T, G, C), f_b,i is observed frequency of each base at position i and q_b is the frequency of base b in the genome sequence. The contribution of each base to the positional Shannon relative entropy is calculated by multiplying each base frequency by positional relative entropy as follows:

where W_b,i refers to the weighted Shannon relative entropy of the base b (A, T, G, C) at position i. Finally, a 4 × L entropy matrix (L is the length of the binding site) is constructed representing the binding site recognition profile, where each matrix element is the weighted positional Shannon relative entropy of a base.

The profile, encoded as the matrix, is used to scan the upstream sequences of all the genes of the user-selected genome. The entropy score of each site is calculated as the sum of the respective positional nucleotide entropy (W_b,i). A maximally scoring site is selected from the upstream sequence of each gene. The score may represent the strength of interaction between regulatory protein and binding site (5). The lowest score among the input sites is considered as the cut-off score. The sites scoring higher than the the cut-off value are reported as potential binding sites conforming to the consensus profile.

Co-directionally transcribed genes downstream of the predicted binding site were selected as potential co-regulated genes (operons) according to one of the following criteria: (i) co-directionally transcribed orthologous gene pairs conserved in at least three genomes (6); (ii) genes belong to the same cluster of orthologous gene function category and the intergenic distance is <200 bp (7); (iii) the first three letters in gene names are identical (the gene names for all the bacterial species were assigned using the COG annotation); (iv) intergenic distance is <90 bp (8).

This method has two specific requirements: a few experimentally determined regulatory protein binding sites should be available for developing the binding site recognition profile, and the profile should be applicable to the genome where the regulator or its homologue is present. In the absence of any experimental information on the regulatory sites in a given genome one may look up the known regulatory motifs from other related species from one of the four online databases which host the information about known transcription regulatory protein binding sites in prokaryote genomes (1–3).

A limitation of this approach is that it may predict a few false positive sites as candidates. However, this limitation can be overcome by experimental validations, by either in vitro binding studies with double strand oligonucleotides containing the binding sites (designed based on prediction) and regulatory proteins or real-time PCR analysis of candidate co-regulated genes.

EXAMPLE: PREDICTION OF LEXA REGULON IN MYCOBACTERIUM TUBERCULOSIS

To demonstrate a typical usage of PredictRegulon, we predicted the LexA binding sites and LexA regulon of M.tuberculosis using the LexA binding sites of Bacillus subtilis. LexA regulators from B.subtilis and M.tuberculosis share a high sequence identity (45%) at protein level (data not shown). Table 1 lists the known LexA binding sites from B. subtilis given as input to the program (2) and Table 2 shows the output of predicted LexA binding sites in M.tuberculosis. The site column in Table 2 represents the predicted binding sites of LexA in M.tuberculosis. In a typical output the perfect match to the known binding sites and the downstream genes are highlighted with a yellow background, and the rest with score greater than cut-off is shown with a blue background (colours not shown in the table). Eighteen of these genes (indicated by ‘a’) belonging to the LexA regulon were also observed in data obtained by experimental means by others (9–12). The rest of the matches are potential novel regulatory sites which could be confirmed experimentaly.

Table 1. Known LexA binding sites of Bacillus subtilis from the PRODORIC database.

Binding site	Gene
AGAACAAGTGTTCG	dinC
AGAACTCATGTTCG	dinB
CGAACTTTAGTTCG	dinA
CGAATATGCGTTCG	recA
CGAACGTATGTTTG	dinC
CGAACCTATGTTTG	dinR
CGAACAAACGTTTC	dinR
GGAATGTTTGTTCG	dinR

Table 2. Output of PredictRegulon web server (predicted LexA binding sites).

Score	Position	Site	Gene	Synonym	COG	Product
5.37	−8	CGAACGTATGTTCG	—	Rv3776a	—	Hypothetical protein Rv3776
5.32	−100	CGAACATGTGTTCG	—	Rv3073ca	COG3189	Uncharacterized conserved protein
5.32	−144	CGAACATGTGTTCG	pyrR	Rv1379a	COG2065	Pyrimidine operon attenuation protein
5.22	−8	CGAACACATGTTCG	—	Rv3074a	—	Hypothetical protein Rv3074
5.2	−142	CGAACAATTGTTCG	—	Rv3371a	—	Hypothetical protein Rv3371
5.2	−64	CGAACAATTGTTCG	dnaE2	Rv3370ca	COG0587	DNA polymerase III
5.19	−36	CGAACGATTGTTCG	ruvC	Rv2594ca	COG0817	ruvC
5.14	−32	CGAAAGTATGTTCG	—	Rv0336a	—	Hypothetical protein Rv0336
5.14	−32	CGAAAGTATGTTCG	—	Rv0515a	—	Hypothetical protein Rv0515
5.14	−105	CGAACACATGTTTG	lexA	Rv2720a	COG1974	SOS-response transcriptional repressors
5.11	−122	CGAACAGGTGTTCG	recA	Rv2737ca	COG1372	recA
5.08	−87	CGAACAATCGTTCG	—	Rv2595a	COG2002	Hypothetical protein Rv2595
5.06	−44	CGAATATGCGTTCG	dnaB	Rv0058a	COG0305	Replicative DNA helicase
5.04	−263	GGAACTTGTGTTGG	ubiE	Rv3832c	COG2226	Methylase involved in ubiquinone biosynthesis
5.04	−23	AGAACGGTTGTTCG	splB	Rv2578ca	COG1533	DNA repair photolyase
5.02	−6	CGAATATGAGTTCG	—	Rv0071a	COG3344	Retron-type reverse transcriptase
5.01	−255	CGAACAAGTGTTGG	—	Rv1414	COG3616	Predicted amino acid aldolase or racemase
4.99	−181	GGAACGCGTGTTTG	—	Rv0750	—	Hypothetical protein Rv0750
4.98	−105	CGAACAACAGTTCG	baeS	Rv0600c	COG0642	Signal transduction histidine kinase
4.98	−186	CGAAGATGCGTTCG	rpsT	Rv2412	COG0268	Ribosomal protein S20
4.95	−242	TGAACGCAAGTTCG	fbpB	Rv1886c	COG0627	fbpB
4.95	−192	CGAACGGGAGTTCG	—	Rv1455	—	Hypothetical protein Rv1455
4.94	−270	AGAACCACCGTTCG	phd	Rv3181c	COG4118	Antitoxin of toxin–antitoxin stability system
4.94	−213	CGAACGACGGTTCG	pe	Rv2099ca	—	PE
4.92	−118	CGAACAGGTGTTGG	—	Rv0004	COG5512	Zn-ribbon-containing
4.92	−163	CGAACTTGCGTTCA	—	Rv1887	—	Hypothetical protein Rv1887
4.91	−239	GGAACGCGAGTTCG	fadB2	Rv0468	COG1250	3-hydroxyacyl-CoA dehydrogenase
4.91	−7	TGAACGAATGTTCC	—	Rv0039c	—	Hypothetical protein Rv0039c
4.9	−237	CGAAGCCTTGTTCG	dltE	Rv3174	COG0300	Short-chain dehydrogenase
4.89	−225	GGAAGGTGCGTTCG	frnE	Rv2466c	COG2761	Predicted dithiol-disulfide isomerase
4.88	−8	GGAAGCCATGTTCG	—	Rv0769	COG1028	Hypothetical protein Rv0769
4.88	−186	CGAAGAGGTGTTCG	coxS	Rv0374c	COG2080	Aerobic-type carbon monoxide dehydrogenase
4.88	−186	CGAACCGCAGTTCG	leuA	Rv3534c	COG0119	Isopropyl malate/citramalate synthases
4.85	−195	CGAACGGCTGTTGG	—	Rv2061c	COG3576	Hypothetical protein Rv2061c
4.85	−85	AGAACGGTTGTTGG	accA1	Rv2501c	COG4770	COG4770
4.84	−151	CGAAATTGTGTTCC	nuoB	Rv3146	COG0377	NADH:ubiquinone oxidoreductase
4.84	−217	CAAACATGTGTTCG	—	Rv2719ca	—	Hypothetical protein Rv2719c
4.84	−5	CGAACATGTATTCG	—	Rv1702ca	—	Hypothetical protein Rv1702c
4.84	−199	CGAAATCTTGTTTG	—	Rv1375	COG1944	Hypothetical protein Rv1375

Score: score of the binding sites, Position: position of the binding site relative to the translation start site, Site: binding site of a regulatory protein, Gene: gene downstream to the binding site, Synonym: synonym of the gene, COG: Cluster of Orthologous Gene code, Product: Gene product. ^arepresents the ORFs known to be regulated by the regulator. ‘a’ symbols are not part of the orginal output of the web server. Source of Genome: NCBI ftp site (ftp://ftp.ncbi.nih.gov/genomes/Bacteria/Mycobacterium_tuberculosis_H37Rv/), Accession no. NC_000962.

The web output of PredictRegulon also contains the hyperlinked gene-synonym and COG number. A click on the former shows the predicted operon context of the regulatory motif while a click on the latter opens a new page showing a description of this gene in the NCBI Conserved Domain Database, which is in turn linked to Pubmed for published information on this gene. These additional links provides users a simple way to browse and understand the functional/physiological implication of the genes that are part of predicted regulon.