Identification of the nature of reading frame transitions observed in prokaryotic genomes

Abstract

Our goal was to identify evolutionary conserved frame transitions in protein coding regions and to uncover an underlying functional role of these structural aberrations. We used the ab initio frameshift prediction program, GeneTack, to detect reading frame transitions in 206 991 genes (fs-genes) from 1106 complete prokaryotic genomes. We grouped 102 731 fs-genes into 19 430 clusters based on sequence similarity between protein products (fs-proteins) as well as conservation of predicted position of the frameshift and its direction. We identified 4010 pseudogene clusters and 146 clusters of fs-genes apparently using recoding (local deviation from using standard genetic code) due to possessing specific sequence motifs near frameshift positions. Particularly interesting was finding of a novel type of organization of the dnaX gene, where recoding is required for synthesis of the longer subunit, τ. We selected 20 clusters of predicted recoding candidates and designed a series of genetic constructs with a reporter gene or affinity tag whose expression would require a frameshift event. Expression of the constructs in Escherichia coli demonstrated enrichment of the set of candidates with sequences that trigger genuine programmed ribosomal frameshifting; we have experimentally confirmed four new families of programmed frameshifts.

Figure 1.

Examples of patterns facilitating programmed frameshifting. (A) ‘−1’ programmed frameshift is used in dnaX gene to express two subunits of DNA polymerase III. The Logo for (A) was derived from aligned sequences from 9 genera (Escherichia, Salmonella, Neisseria, Vibrio, Shigella, Citrobacter, Enterobacter, Yersinia, Serratia). The frameshift signal consists of conserved frameshift pattern AAA_AAA_G (‘slippery sequence’) and two stimulators. The upstream stimulator is a Shine-Dalgarno–like sequence that interacts with ribosome, while the downstream stimulator makes a hairpin secondary structure (7). (B) ‘+1’ programmed frameshift is used in prfB gene to auto regulate expression of Release Factor 2. The Logo for (B) was derived from 413 sequences (138 genera). The frameshift signal consists of conserved frameshift motif with consensus CTT_TGA_C and upstream Shine-Dalgarno–type sequence stimulator.

Figure 2.

An example of a ‘frameshift box’. Predicted frameshift position appears in between two stop codons situated in different frames (TAG stop codon upstream and the TGA stop codon downstream). The true frameshift position is always located inside the ‘frameshift box’, the region between two stop codons.

Figure 3.

(a) List of GeneTack clusters corresponding to known cases of programmed frameshifting. #, row index; Cluster ID, unique identifier of a cluster (can be used for a search in GeneTack database); Cluster name, designated protein function; Size, number of fs-genes in the cluster (number of different genera is specified in parenthesis); Type, frameshift direction (possible mechanism: PTR, programmed transcriptional realignment, PRF, programmed ribosomal frameshifting, II, internal initiaion); Heptamer, overrepresented heptamer (the fraction of the cluster’s fs-genes that contain the heptamer); Frameshift site Logo, logo of the frameshift site (see text for details); Sharma et al. ID, ID of the corresponding Sharma et al cluster(s) (2). (b) Summary of predicted programmed frameshifts, selected from GeneTack clusters for experimental verification. First seven column headers are the same as in Figure 3a. Experimental results (X/Y)—X programmed frameshift candidates out of selected Y candidates from a given cluster have shown detectable level of frameshifting; numbers in parentheses give frameshifting efficiency (in percentage points) for the X candidates.

Figure 4.

Experimental validation of predicted programmed frameshifting. The frameshifting efficiency in each experiment was estimated as the ratio of the product translated with the frameshift to the total amount of products translated with and without frameshift. The fs-gene ID’s are listed below the graph along with the names of clusters. Note that in the last two clusters, frameshifting was observed for only one of the constructs.

Figure 5.

Classification of predicted frameshifts was done by using features specified in Table 1. One of the most important properties of a predicted fs-gene was its membership in a cluster. Singleton fs-genes (not orphan genes) are likely to be a result of indel mutation or sequencing error, while clustered fs-genes could represent programmed frameshifts, phase variation and translational coupling, as well as clusters of pseudogenes or genes with indel mutations.