Coevolution-based inference of amino acid interactions underlying protein function
- UT Southwestern Medical Center, United States
- The University of Chicago, United States
Figures
Figure 1 with 2 supplements
A deep coupling scan (DCS) for the PDZ binding pocket.
(A), The thermodynamic double mutant cycle (TDMC), a formalism for studying the energetic coupling of pairs of mutations in a protein. Given two mutations ( at postion and at position ), …
Figure 1—figure supplement 1
The bacterial two-hybrid assay for PDZ ligand binding.
(A) The bacterial two-hybrid system consists of three parts, (1) a PDZ domain fused to the C-terminus of a -cI DNA binding domain, (2) a ligand peptide as a C-terminal fusion to the -subunit of …
Figure 1—figure supplement 2
Reproducibility and quality of the bacterial two-hybrid assay.
The panels show 2D histograms of all single and double mutant binding free energies for four independent experimental trials of the bacterial two-hybrid assay for PSD95pdz3. The data show that the …
Figure 2 with 5 supplements
Distributions of pairwise thermodynamic couplings in a single PDZ homolog (PSD95pdz3).
Each subplot shows the distribution of coupling free energies (, see Equation 1, main text) for all measured mutants at one pair of positions in the α2-helix (numbering per Figure 1C) in PSD95pdz3.…
Figure 2—figure supplement 1
Distributions of double mutant cycles for each α2 helix position pair (numbering as in Figure 1C) for PSD95pdz2.
As in Figures 2–3, distributions are fit to single or double Gaussians, using the Bayes Information Criterion to choose the model. The position of zero coupling is indicated by the solid line and …
Figure 2—figure supplement 2
Distributions of double mutant cycles for each α2 helix position pair (numbering as in Figure 1C) for Shank3pdz.
As in Figure 3. 2–3, distributions are fit to single or double Gaussians, using the Bayes Information Criterion to choose the model. The position of zero coupling is indicated by the solid line and …
Figure 2—figure supplement 3
Distributions of double mutant cycles for each α2 helix position pair (numbering as in Figure 1C) for Syntrophinpdz.
As in Figures 2–3, distributions are fit to single or double Gaussians, using the Bayes Information Criterion to choose the model. The position of zero coupling is indicated by the solid line and …
Figure 2—figure supplement 4
Distributions of double mutant cycles for each α2 helix position pair (numbering as in Figure 1C) for ZO1pdz.
As in Figures 2–3, distributions are fit to single or double Gaussians, using the Bayes Information Criterion to choose the model. The position of zero coupling is indicated by the solid line and …
Figure 2—figure supplement 5
Distributions of coupling in relative enrichment () for a sampling of position pairs in the GB1 domain of the immunoglobulin-binding protein G.
The data are from (Olson et al., 2014), and position numbering is as given in that work; note that coupling is calculated here directly from enrichment scores. As in Figures 1 and 2, distributions …
Figure 3 with 1 supplement
Homolog-averaged pairwise thermodynamic couplings in the PDZ domain.
Each subplot shows the distribution of coupling free energies (, see Equation 1, main text) for all measured mutants at one pair of positions in the α2-helix, but here averaged over the five …
Figure 3—figure supplement 1
Amino acid contributions to distributions of double mutant cycle coupling energies for the PDZ α2 helix.
(A), Boxplots showing the couplings for each amino acid substitution with every other amino acid substitution in the homolog-averaged dataset. The data show that every amino acid substitution is …
Figure 4
A basic model for observed distributions of double mutant cycle coupling energies.
(A), A schematic representation of two coupled equilibria in a protein molecule – a reaction with equilibrium constant corresponding to function (here, binding), an internal two-state …
Figure 5
Conservation and idiosyncrasy in the pattern of energetic couplings over PDZ homologs.
(A–E), Matrices of mutation averaged pairwise thermodynamic couplings for the α2-helix in each PDZ homolog. The color scale is chosen to represent the full range of measured energetic couplings. The …
Figure 6 with 2 supplements
Coevolution-based inference of energetic couplings - SCA.
(A), Coevolution of sequence positions corresponding to the top eigenmode of the SCA matrix, derived from an alignment of 1656 eukaryotic PDZ domains (the 'Poole' alignment). The data show that a …
Figure 6—figure supplement 1
Robustness of the SCA to alignment size.
(A–F) The graphs show scatterplots of homolog-averaged experimental couplings in the α2 helix against the first eigenmode of the SCA coevolution matrix for six independent trials of randomly …
Figure 6—figure supplement 2
The relationship between mutation-averaged couplings and predictions from SCA for individual domains.
(A–E), Scatterplots of experimental couplings in the α2 helix of each individual PDZ homolog against the first eigenmode of the SCA coevolution matrix. Linear fits are shown in red lines with …
Figure 7
Homolog-averaged thermodynamic couplings and protein sectors.
Analysis of the top eigenmodes of the SCA coevolution matrix exposes groups of coevolving amino acids that empirically are found to form physically contiguous networks in the tertiary structure, …
Figure 8 with 2 supplements
Coevolution-based inference of energetic couplings - DCA.
(A), The matrix of direct couplings () from the DCA method, with tertiary contacts in the PDZ structure (1BE9) indicated by white or black circles. By convention (Morcos et al., 2011), trivial …
Figure 8—figure supplement 1
Robustness of DCA to alignment size.
(A–F), The graphs show scatterplots of homolog-averaged experimental couplings in the α2 helix against computed DCA model coupling energies for six independent trials of randomly sub-sampling the …
Figure 8—figure supplement 2
The relationship between mutation-averaged couplings and predictions from DCA for individual domains.
(A–E), scatterplots of experimental couplings for individual PDZ homologs against the coupling values computed from the DCA model. Linear fits are shown in red lines with adjusted Pearson …
Tables
Table 1
Summary of data collection.
For each PDZ homolog, we indicate the target ligand, the wild-type affinity, the top-hit sequence identity within the ensemble of homologs, and the assay/sequencing statistics.
| PDZ homolog | Ligand | Ligand sequence | Affinity | Top ID% (PDZ) | No. of single mutants (out of 171) | No. of double mutants (out of 12,996) | Mean cycles/position pair (out of 361) | Sequence readsunsel/sel |
|---|---|---|---|---|---|---|---|---|
| PSD95pdz3 | CRIPT | TKNYKQTSV | 0.8 μM (McLaughlin et al., 2012) | 41.0 (Syntrophinpdz) | 171 | 11,531 | 320 | 12,190,079/14,358,962 |
| PSD95pdz2 | NMDAR2A | KMPSIESDV | 3.6 μM (Stiffler et al., 2006) | 40.5 (PSD95pdz3) | 171 | 12,072 | 335 | 9,402,209/14,965,473 |
| Shank3pdz | Dlgap1/2/3 | YIPEAQTRL | 0.2 μM (Stiffler et al., 2006) | 25.0 (PSD95pdz2) | 171 | 10,454 | 290 | 17,232,329/6,429,999 |
| Syntrophinpdz | Scn5a (Nav1.5) | PDRDRESIV | 1.6 μM (Stiffler et al., 2006) | 41.0 (PSD95pdz3) | 171 | 10,757 | 298 | 8,227,200/15,248,680 |
| ZO-1pdz | Claudin8 | SIYSKSQYV | 4.6 μM (Zhang et al., 2006) | 37.5 (PSD95pdz3) | 171 | 11880 | 330 | 5,365,041/11,523,044 |
Additional files
-
Supplementary file 1
Salinas_Ranganathan_data.
- https://doi.org/10.7554/eLife.34300.023
-
Transparent reporting form
- https://doi.org/10.7554/eLife.34300.024
