Percolation of the phd repressor-operator interface

Abstract

Transcription of the P1 plasmid addiction operon, a prototypical toxin-antitoxin system, is negatively autoregulated by the products of the operon. The Phd repressor-antitoxin protein binds to 8-bp palindromic Phd-binding sites in the promoter region and thereby represses transcription. The toxin, Doc, mediates cooperative interactions between adjacent Phd-binding sites and thereby enhances repression. Here, we describe a homologous operon from Salmonella enterica serovar Typhimurium which has the same pattern of regulation but an altered repressor-operator specificity. This difference in specificity maps to the seventh amino acid of the repressor and to the symmetric first and eighth positions of the corresponding palindromic repressor-binding sites. Thus, the repressor-operator interface has coevolved so as to retain the interaction while altering the specificity. Within an alignment of homologous repressors, the seventh amino acid of the repressor is highly variable, indicating that evolutionary changes in repressor specificity may be common in this protein family. We suggest that the robust properties of the negative feedback loop, the fuzzy recognition in the operator-repressor interface, and the duplication and divergence of the repressor-binding sites have facilitated the speciation of this repressor-operator interface. These three features may allow the repressor-operator system to percolate within a nearly neutral network of single-step mutations without the necessity of invoking simultaneous mutations, low-fitness intermediates, or other improbable or rate-limiting mechanisms.

FIG. 1.

Alignments of nucleotide and protein sequences from the P1 and Salmonella addiction operons. A. Alignment of P1 and Salmonella operator regions. In order to identify possible Salmonella operators, regions upstream of P1 phd and Salmonella phd were compared. The P1 sequence contains two palindromic Phd-binding sites spaced 13 bp apart, from center to center. The Salmonella DNA contains three very similar palindromic sites with the same 13-bp spacing between the centers of adjacent sites. The consensus P1 and Salmonella operators are eight nucleotides long and differ only in the first and last positions. Thus, it appears that the 6-bp core of the operator is conserved and that the promoter specificity may be regulated by the two flanking base pairs. B. Alignment of P1 and Salmonella Phd proteins. The two proteins have approximately 41% amino acid identity. An alternative start for the Salmonella Phd protein would add two amino acids to the N terminus of the protein (see Fig. 3). C. Alignment of P1 and Salmonella nucleotide sequences at the phd/doc boundary. Relative to the P1 sequence, the Salmonella sequence has a 3-bp insertion into the stop codon of phd. The insertion generates a new stop codon at the same position, destroys the original start codon for Doc, but simultaneously generates a new start codon in a slightly different position. Thus, the stop codon for P1 Phd and the start codon for P1 Doc overlap by 1 bp (TAATG), whereas the stop codon for S. enterica serovar Typhimurium Phd and the start codon for S. enterica serovar Typhimurium Doc overlap by 4 bp (ATGA). The overlapping start and stop codons are consistent with the hypothesis that toxin and antitoxin synthesis may be translationally coupled. D. Alignment of P1 and Salmonella Doc proteins. The two proteins have approximately 45% amino acid identity. Relative to the P1 Doc protein, the Salmonella Doc protein has two additional amino acids at the N terminus, due to the insertion discussed above, and six fewer amino acids at the C terminus.

FIG. 2.

β-sheet model for DNA binding by Phd. A. N-terminal alignment. The alignment of P1 Phd and S. enterica serovar Typhimurium Phd with two proteins (Arc and Mnt) that are known to contact DNA through an N-terminal antiparallel β-sheet structure is shown. The amino acids that are known (Arc and Mnt) (41, 63) or hypothesized (P1 Phd and S. enterica serovar Typhimurium Phd) to contact the DNA are in bold. Residues in these positions that differ from those found in wild-type P1 Phd are underlined. B. Model. We propose that in a Phd dimer, the N termini form an antiparallel β-sheet, as indicated. Positions 3, 5, and 7, shown in bold, are proposed to contact bases in the major groove of the DNA site. The model is reasonably consistent with general observations and models regarding β-sheet DNA-binding proteins (76, 79). C. Repressors. If the model is correct, then differences in the DNA-binding specificity in this protein family are likely to map to residues 3, 5, and 7. In the case of P1 and S. enterica serovar Typhimurium Phd proteins, positions 3 and 7 differ, but the amino acid at position 5 is conserved. In order to test the model, single and double mutations at positions 3 and 7, in both P1 Phd and S. enterica serovar Typhimurium Phd, were constructed, as shown.

FIG. 3.

Percolation of position 7. Although the amino acid identity at position 7 of Phd is an important determinant of DNA-binding specificity, this position is not well conserved. In an alignment of probable Phd homologs, positively charged (RKH), polar (NQST), hydrophobic (AGLF), and negatively charged amino acids (D) are all observed at position 7 (indicated in bold). In protein-DNA complexes, positive, polar, hydrophobic, and negative amino acids most commonly contact G, A, T and C bases, respectively (50). The position-specific iterated BLAST program (3) was used to identify list of Phd homologs, not necessarily associated with a Doc homolog. The sequence of P1 Phd was used to initiate this program. A low limit (expectation < 0.1) was used for inclusion. At each iteration, sequences were manually selected for the presence of the N-terminal region, and proteins lacking homology in this region were excluded from the analysis. Although the inclusion limit was low, the program converged upon a group of sequences for which the expectation of a chance match was less than 1 E-6. These sequences were then downloaded in their entirety and realigned in Clustal W (80) with double the default gap penalties. Sequences with large loops, extensions, or deletions were removedfrom the alignment. Sequences were then grouped according to the chemical nature of the residue corresponding to the seventh amino acid of P1 Phd. For each amino acid observed at position 7 (relative to P1 Phd), two representative sequences are shown, if available. Alignments were shaded on the BoxShade server (http://www.ch.embnet.org/software/BOX_form.html) running BoxShade version 3.21, written by Kay Hofmann and Michael D. Baron. Identical (black) and conserved (grey) amino acids were shaded if at least half of the amino acids at that position were identical or similar. The output was saved in “new RTF” format and edited in Microsoft Word.

FIG. 4.

Nearly neutral network. We suggest that operator-repressor space can be traversed by a nearly neutral network of single step mutations. Such a network requires a degree of fuzzy recognition between amino acid side chains and DNA bases. The degeneracy of the protein-DNA recognition code is well established (11). In the case of the P1 addiction operon and its closer homologs, the development of a nearly neutral network might be further facilitated by the robust properties of the negative feedback loop, by the presence of oligomeric interactions and by the duplication and divergence of the repressor-binding sites. To the first approximation, amino acid-base contacts appear to be fairly independent (9, 10, 12, 28, 50), however, both quantitative and qualitative exceptions to this generalization (reflecting a high degree of interdependence) have been well documented (14, 49, 62). In this figure, a repressor-operator family with a distinctly different or extensively remodeling repressor-operator interface would generate a parallel or intersecting ridgeline. Although biochemically acceptable, the transition between two such ridges might be evolutionarily difficult. The general tendency toward approximately independent contacts (75) in repressor-operator interactions (and other protein-ligand interactions) may reflect the constraints on the evolutionary process by which the system was generated rather than the structural requirements of the protein-DNA interaction itself.