Prediction of rho-independent transcriptional terminators in Escherichia coli

Abstract

A new algorithm called RNAMotif containing RNA structure and sequence constraints and a thermodynamic scoring system was used to search for intrinsic rho-independent terminators in the Escherichia coli K-12 genome. We identified all 135 reported terminators and 940 putative terminator sequences beginning no more than 60 nt away from the 3'-end of the annotated transcription units (TU). Putative and reported terminators with the scores above our chosen threshold were found for 37 of the 53 non-coding RNA TU and for almost 50% of the 2592 annotated protein-encoding TU, which correlates well with the number of TU expected to contain rho-independent terminators. We also identified 439 terminators that could function in a bi-directional fashion, servicing one gene on the positive strand and a different gene on the negative strand. Approximately 700 additional termination signals in non-coding regions (NCR) far away from the nearest annotated gene were predicted. This number correlates well with the excess number of predicted 'orphan' promoters in the NCR, and these promoters and terminators may be associated with as yet unidentified TU. The significant number of high scoring hits that occurred within the reading frame of annotated genes suggests that either an additional component of rho-independent terminators exists or that a suppressive mechanism to prevent unwanted termination remains to be discovered.

Figure 1

Rho-independent intrinsic terminator construct. Regions include from 5′ to 3′: (i) A-region of 11 nt; (ii) a hairpin with a loop of 3–10 residues and with one of three types of stems: perfect stems of 4–18 base pairs, stems of 9–18 base pairs with internal loop no longer than 20% stem length, or stems of 7–18 base pairs with 1–5 nt bulges in either 5′ or 3′-side of stems; (iii) a spacer of 0–2 nt (any bases except T); (iv) T-region divided into three parts: proximal part of 5 nt, distal part of 4 nt, and extra part of 3 nt. (We did not number spacer nucleotides since no more than 5% putative terminators contained 2 nt spacers and no more than 15% putative terminators contained 1 nt spacers.)

Figure 2

Distance distribution between the start of the hits and the end of nearest annotated genes (A) and score distribution (B) for the RNAMotif hits of Set A containing 135 reported terminators (16). A translation termination codon of an ORF was considered the end of a protein-encoding gene and the processed end was considered the end of a non-coding RNA gene.

Figure 3

Distance distribution between the start of the hits and the end of nearest annotated genes (A) and score distribution (B) for the population of RNAMotif hits with scores under threshold –4.0 (ends of genes were designated as coordinate 0 on the x-axis). The data were fit by Spotfire software.

Figure 4

The score distribution histogram (A) and the Gaussian curve fitting it (B) for RNAMotif hits starting between –10 and +60 nt from the end of annotated genes (Set B). The lowest value of score bins were plotted on the x-axis. Each bin is 1 kcal/mol wide. The data were fit by Graphpad Prizm software.

Figure 5

Comparison of hairpin structure parameters for hits in Set A (light bars) and Set B (dark bars): (A) stem length distribution; (B) loop length distribution.

Figure 6

An example of bi-directional putative terminators starting 4 nt (score –11.28) and 25 nt (score –11.95) downstream of annotated genes in (+) and (–) strand, respectively.

Figure 7

The score distribution histogram (A) and the Gaussian curves fitting it (B) for RNAMotif hits beginning in NCRs further than 60 nt downstream from the end of annotated genes. The lowest value of score bins were plotted on the x-axis. Each bin is 1 kcal/mol wide. The data were fit by Graphpad Prizm software.

Figure 8

True terminator prediction probability obtained from equation 1 scores (triangles) versus free energy of hairpin formation [ΔG⁰_{37 (hairpin)}] (dimonds).