Local gene regulation details a recognition code within the LacI transcriptional factor family

Abstract

The specific binding of regulatory proteins to DNA sequences exhibits no clear patterns of association between amino acids (AAs) and nucleotides (NTs). This complexity of protein-DNA interactions raises the question of whether a simple set of wide-coverage recognition rules can ever be identified. Here, we analyzed this issue using the extensive LacI family of transcriptional factors (TFs). We searched for recognition patterns by introducing a new approach to phylogenetic footprinting, based on the pervasive presence of local regulation in prokaryotic transcriptional networks. We identified a set of specificity correlations--determined by two AAs of the TFs and two NTs in the binding sites--that is conserved throughout a dominant subgroup within the family regardless of the evolutionary distance, and that act as a relatively consistent recognition code. The proposed rules are confirmed with data of previous experimental studies and by events of convergent evolution in the phylogenetic tree. The presence of a code emphasizes the stable structural context of the LacI family, while defining a precise blueprint to reprogram TF specificity with many practical applications.

Figure 1. HTH binding mode.

A) X-ray model for a LacI dimer bound to a palindromic BS (plotted with Jmol from the PDB structure 1lbg4). Only the binding domain of each monomer is shown (in light/dark purple, respectively). The hinge-helix and the recognition helix of each monomer are colored in yellow and red, respectively. B) Logo for the alignment of 2639 unredundant HTH-LacI domains. The AA coordinates of any particular domain will be referred by its position in this alignment –they match the numbering of the first 71 AAs of Escherichia coli's GalR and GalS regulators. Helix-1, helix-2 (or recognition helix) and the intermediate residues constitute the HTH motif itself. C) Logo for the alignment of the set of BSs associated to 370 LacI family members (BS sequences from RegTransBase [46]). In BS logos we avoided subscripts for left and right half sites coordinates.

Figure 2. Autoregulation and the search of conserved binding sequences.

A) Local regulation at the core of phylogenetic footprinting includes both autoregulation –which can be linked to the regulation of an upstream divergent operon– and downstream unidirectional adjacent regulation (BSs, white boxes). Red and green lines for the respective strict and extended regions of BS search. B–D) Examples of BS logos. Rest of cases in the Appendix of Text S1. Above each logo: the recognition sequence (AA-15, AA-16) and a triad of numbers (i/ii/iii) corresponding to i) the total number of TFs exhibiting the recognition sequence, ii) the number of TFs for which at least a BS was found, and iii) the total number of found BSs. E) Consensus-logo for the BSs associated to the TVSR group. The inserted position NT-2bis for the YQ-logo in C) has not been considered to build the consensus.

Figure 3. Degeneracies in TF binding.

A) Palindromic (P1 and P2) and non-palindromic BSs (M1 and M2). Nucleotides (NTs) in positions 4 and 5 in both half sites and strands were only considered. Colors distinguished different NTs pairs. Only the sense strand (black line) is included in the alignment of BSs. Half sites separated by dots. B) Scenarios for degeneracy. Spheres represent two different TFs sharing the same recognition amino acids. Arrows indicate what BSs they can bind. We considered both mixtures to have the same binding energy and termed them simply as M. See main text for details. C) Notation criterion for degeneracies uses different arrows between the corresponding left semisequences in the sense strand of the palindromic combinations.

Figure 4. Recognition code and experimental confirmations.

A) Sequence correlations between (AA-15, AA-16) and (NT-5, NT-4) extracted from correlations in Table S1. AAs sequences recognizing a same sequence of NTs were grouped. Here, we only considered significant palindromic NT sequences; for example, (NT-5, NT-4) = TG means . We included the case for (AA-15, AA-16) = YQ corresponding to the synthetic SymL site in C). Recognition degeneracies are represented as unidirectional arrows (asymmetrical intrinsic), bidirectional divergent arrows (symmetrical intrinsic), and bidirectional convergent arrows (extrinsic). Colors for polar (green), basic (blue), acidic (red) and hydrophobic (black) amino acids. B) Agreement between synthetic and natural data. Recognition of (NT-5, NT-4)-palindromes by different (AA-15, AA-16)-LacI mutants (YQ is the wild type). Data from –from which we only considered those sequences (AA-15, AA-16) with a natural correspondence in Table S1. Rest of BS positions as in SymL. The larger the TF/BS affinity, the stronger the repression of the -galactosidase activity. Experimental conditions limited repression to a factor of 200. Arrows indicated again degeneracy classes. Predictions for wild type YQ correspond to asymmetric natural BSs (see text). (NT-5, NT-4)-palindromes involved in the predicted correlations for PM (, see Table S1) lack an experimental test. Accordingly, PM do not exhibit a strong affinity for any of the tested palindromes (see Fig. S3), C) Natural and synthetic operators. A dot distinguishes the half sites. Flanking nucleotides separated by a space to help visualization of the highly conserved central region of the BSs. Colors identify different palindromic or mixed combinations in the specificity nucleotides (see Table S2 for more details).

Figure 5. Convergence of binding modes in the gene tree.

Gene tree involving all TFs with BSs in Table S1 (623 TFs) plus the 3 TFs with binding to natural SymL-like BSs (Fig. 4.C and Table S2). Only one BS per TF is shown. The external color code displays the specificity-associated positions –to help visualization of palindromic combinations right positions are read in the complementary (c) strand: . The color background in several branches corresponds to different recognition AAs (only a few recognition classes were enhanced). External color code in these branches shows darker colors to help visualization. Dots in branches denote bootstrap values larger than 80 (for 100 trees total, see Fig. S4 for more details).