A complete map of specificity encoding for a partially fuzzy protein interaction

This article has been Reviewed by the following groups

Read the full article See related articles


Thousands of human proteins function by binding short linear motifs embedded in intrinsically disordered regions. How affinity and specificity are encoded in these binding domains and the motifs themselves is not well understood. The evolvability of binding specificity - how rapidly and extensively it can change upon mutation - is also largely unexplored, as is the contribution of ‘fuzzy’ dynamic residues to affinity and specificity in protein-protein interactions. Here we report the first complete map of specificity encoding for a globular protein domain. Quantifying >200,000 energetic interactions between a PDZ domain and its ligand identifies 20 major energetically coupled pairs of sites that control specificity. These are organized into six modules, with most mutations in each module reprogramming specificity for a single position in the ligand. Nine of the major energetic couplings controlling specificity are between structural contacts and 11 have an allosteric mechanism of action. The dynamic tail of the ligand is more robust to mutation than the structured residues but contributes additively to binding affinity and communicates with structured residues to enable changes in specificity. Our results quantify the binding specificities of >1,800 globular proteins to reveal how specificity is encoded and provide a direct comparison of the encoding of affinity and specificity in structured and dynamic molecular recognition.

Article activity feed

  1. This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/12053815.


    Quantifying the mutational tolerance between short linear motifs (SLiMs) and peptide binding proteins has classically been a difficult process due to the large combinatorial space. PDZ domains have served as a model system for studying such interactions - as they are important for their roles in cellular signaling towards neuronal arborization (Bustos et al., 2014; Omelchenko et al., 2020). These domains bind C-terminal linear motifs and often serve a scaffolding role. The PDZ3 domain of PSD-95 performs "fuzzy" interactions with dynamic residues, such as in the case of PDZ3-CRIPT interaction. "Fuzzy" interactions describe proteins that bind with each other in an ensemble of heterogeneous binding conformations, each of which can be characterized by a measure of entropy, which is experimentally described by high crystallographic b-factor. Some prior work studying the PSD-95 elucidated the structure of PDZ3 (Doyle et al., 1996) and further research described the "fuzzy" interaction between PSD-95 and CRIPT (Niethammer et al., 1998). Additionally, mutational approaches were performed by the Ranganathan lab, uncovering individual residues that result in specificity changes. Despite prior data, there is still an incomplete understanding of how the PDZ3-CRIPT and similar systems address the ensemble nature of "fuzzy" interactions.

    In this paper, the authors utilize deep mutational scanning (DMS) and a protein fragment complementation assay (PCA) to query protein-protein interactions (PPI) in the PDZ3-CRIPT system. The authors designed four library strategies to map PDZ3-CRIPT interactions comprehensively. Then, they used DMS data combined with MoCHI, an analytical tool for interpreting deep mutational scan data using neural network machine learning. This method is well-suited to understanding the mutational tolerance of the PDZ-CRIPT system as complete mutational coverage is necessary to piece apart the effects of individual residues and positions in the system and a computational tool is necessary to provide interpretable energetic couplings to describe epistasis.

    The paper presents a comprehensive analysis specifically detailing energetic coupling of interactions using MoCHI, as identification of higher-order interactions represents the application of the MoCHI method to a model biological system for protein-protein interactions. The high-throughput results presented encapsulate previously known interactions, including H372A (Raman et al., 2016), and a series of interactions involved in epistatic ligand specificity change (McLaughlin et al., 2012). Their complete map is limited to only complete combinatorial coverage (20^n, where n is number of residues) of four residues at the N and C termini of CRIPT, with single and double mutation libraries in cis and trans comprising the rest of the generated libraries. While this does limit the interpretation of mutational effects, it is sufficient to understand the interaction between CRIPT and PDZ3, specifically the differences between the N and C termini of the CRIPT; with the N terminus representing a more "fuzzy" interaction comparatively to the fully structured C terminus. The authors find that the fuzzy N terminus is more robust to mutations. The analysis performed resulted in 20 major coupling pairs, recovering known interaction pairs as well as specificity changes.

    This paper contributes to the protein-protein interaction field by providing a map of interactions in the PDZ-CRIPT model system in a high throughput manner. Firstly, the authors are able to recapitulate PDZ-CRIPT system interactions with complete combinatorial coverage for 4 residues in the N-terminus and 4 residues in the C-terminus of CRIPT and discovered 20 major coupling pairs, recovering the known interaction pairs. Secondly, utilizing MoCHI allows a description of epistatic interactions, one of which was previously described to change the specificity of PDZ at the T-2 position to an aromatic residue. Overall, this DMS approach combined with MoCHI neural network provides interpretable biophysical relationships in this model system.

    Major comments:

    1. The authors describe PDZ-CRIPT as an IDR system. Prior papers describe PDZ-CRIPT as a model system for allosteric protein-protein interactions and framing it as such would be illuminating. The interactions described are more well-attributed to a protein-protein interaction model instead, as the binding energy contributions are described as additive residue-specific contributions that impart PDZ with the ability to bind to multiple conformations of CRIPT with some mutational tolerance. We suggest the authors explain their interpretation of the system for clarity.

      1. A definition for quantifying the degree of the "fuzziness" of CRIPT would be greatly appreciated, our suggestion would be to define the "fuzzy" N-terminus in contrast to the C-terminus as well as known IDRs.

    2. The authors show electrostatic interactions drive the affinity for dynamic residues. However, mutations in the PDZ domain do not alter the effects of mutations in the dynamic residues. The authors suggest that multiple charge changes may be required to alter the specificity. The authors can reframe this in the discussion as a limitation of their design strategies as they do not test multiple combinatorial mutations relating to charge and/or other biophysical properties, such as disorder tendency/flexibility. Discussing and comparing it to other designed or hypothesis-driven mutational studies could be beneficial in illustrating the broad applicability and limitations of the presented method.

    1. The authors used an unbiased approach and have confirmed previously found interactions for H372A in PSD95 which the authors mention. Other previous couplings and interactions have been observed for N326 in PSD95 and Q3 in CRIPT, as well as S1 in CRIPT (Niethammer et al., 1998). For this same system, (Nedrud et al., 2021) using a DMS approach showed similar coupling, and (Crean et al., 2023) showed similar coupling pairs using molecular dynamics. We suggest the authors include previous papers identifying similar results and incorporate these results into their discussion.

    Minor comments:

    1. To provide a rough visualization of variance, we suggest including the raw read distributions in the input and outputs similar to the Figure S4e division in order to convey the distribution of reads per mutant.

    2. We suggest adding S1a and S1b into the main text as they contribute greatly to understanding the experimental methods performed. A description of the experimental setup would be useful for those who are not familiar with the yeast display. Additionally, the visualization of the experimental setup and mutational strategies employed would help with the framing and flow of the paper.

    3. In Figure 2a, an inset equation,ŷ = h(g(fb(x))), is unexplained. We believe this to be an equation explaining how the weights are updated, where ŷ is the output variable, h and g are equations describing the two-state linear model, fb is a function describing binding fitness scores, and x is the amino acid position. Adding a description in the text will be useful for readers who are unfamiliar with MoCHI.

    4. In Figure 3b, Hamming distance distribution is used to visualize the quality of the library, we suggest moving this figure as well as Figure 6c to the supplementary since these break the flow of the paper and don't provide new insights or results.

    5. In Figure 3d the borders are a bit hard to discern, this can be improved by darkening/thickening the outline and changing the range of the heatmap from [-1,1] to [-1,0] for visual clarity.

    6. The text on the x-axis is crowded and difficult to read. We recommend changing the x-axis labels in Figure 3e for clarity. One idea is to keep only the amino acid letters on the label while adding an additional axis label that demarcates the numerical position of every 5th amino acid.

    7. Figure 3g: there is a lack of data for many mutants in a specific area of the PDZ3 (P346-E352), we feel it could be beneficial to mention this in the text (referring readers to S4e) and/or highlight it in Figure 3g.

    8. In Figure 4a the number of specificity-changing mutations is visualized by color with a threshold on the number of specificity-changing mutations (10+). We think this might obfuscate the total number of specificity-changing mutations per position on CRIPT. We suggest adding numbers within the boxes to demarcate the high-specificity changing positions beyond 10.

    9. In Figure 7, Q-3 is annotated as Q-5. We also found it difficult to understand due to the low contrast between the green superimposed on yellow. We suggest changing the color scheme for clarity. 

    This review was written by two first-year PhD students in the UCSF BMI and Biophysics programs as part of a graduate course in Peer Review.

    The authors of the review declare that they have no competing interests.

    Bustos, F. J., Varela-Nallar, L., Campos, M., Henriquez, B., Phillips, M., Opazo, C., Aguayo, L. G., Montecino, M., Constantine-Paton, M., Inestrosa, N. C., & Van Zundert, B. (2014). PSD95 suppresses dendritic arbor development in mature hippocampal neurons by occluding the clustering of NR2B-NMDA receptors. PLoS ONE, 9(4). https://doi.org/10.1371/journal.pone.0094037

    Crean, R. M., Slusky, J. S. G., Kasson, P. M., & Kamerlin, S. C. L. (2023). KIF - Key Interactions Finder: A program to identify the key molecular interactions that regulate protein conformational changes. Journal of Chemical Physics, 158(14). https://doi.org/10.1063/5.0140882

    Doyle, D. A., Lee, A., & Lewis, J. (1996). Crystal Structures of a Complexed and Peptide-Free Membrane Protein-Binding Domain: Molecular Basis of Peptide Recognition by PDZ tures of the domain in complex with peptide and in the. In Cell (Vol. 85).

    McLaughlin, R. N., Poelwijk, F. J., Raman, A., Gosal, W. S., & Ranganathan, R. (2012). The spatial architecture of protein function and adaptation. Nature, 491(7422), 138–142. https://doi.org/10.1038/nature11500

    Nedrud, D., Coyote-Maestas, W., & Schmidt, D. (2021). A large-scale survey of pairwise epistasis reveals a mechanism for evolutionary expansion and specialization of PDZ domains. Proteins: Structure, Function and Bioinformatics, 89(8), 899–914. https://doi.org/10.1002/prot.26067

    Niethammer, M., Valtschanoff, J. G., Kapoor, T. M., Allison, D. W., Weinberg, R. J., Craig, A. M., & Sheng, M. (1998). CRIPT, a Novel Postsynaptic Protein that Binds to the Third PDZ Domain of PSD-95/SAP90. Neuron, 20(4), 693–707. https://doi.org/10.1016/S0896-6273(00)81009-0

    Omelchenko, A., Menon, H., Donofrio, S. G., Kumar, G., Chapman, H. M., Roshal, J., Martinez-Montes, E. R., Wang, T. L., Spaller, M. R., & Firestein, B. L. (2020). Interaction Between CRIPT and PSD-95 Is Required for Proper Dendritic Arborization in Hippocampal Neurons. Molecular Neurobiology, 57(5), 2479–2493. https://doi.org/10.1007/s12035-020-01895-5

    Raman, A. S., White, K. I., & Ranganathan, R. (2016). Origins of Allostery and Evolvability in Proteins: A Case Study. Cell, 166(2), 468–480. https://doi.org/10.1016/j.cell.2016.05.047

    Competing interests

    The authors declare that they have no competing interests.