Multiple phenotypic protein expressions arising from one genome represent variations in the protein relative abundance and their stoichiometry. A lack of definite compositional parts challenges the modeling of protein megacomplexes and cellular architectures. Despite the advances in protein structural predictions with AI, the mechanism of protein interactions and the emergence of megacomplexes they assemble remains unclear. Here, we present a statistical physics framework of grand canonical ensemble to explore the protein interactions that drive the emergent assembly of a megacomplex using the observational mass spectrometry datasets including protein relative abundance and the cross linked connections. Using chromatin remodeler megacomplex, INO80, as an example, we discovered a class of divergent protein that plays a critical role in orchestrating the assembly beyond nearest neighbors, dependent on the excluded volumes exerted by others. With the constraints of the excluded volumes by varying crowding contents, these divergent subunits orchestrate and form clusters with selective components growing into configurationally distinct architectures. We propose a machinery view for the INO80 chromatin remodeler complex where each loosely associated subunits can be occasionally recruited for parts as attachment into a core assembly driven by excluded volumes. Our computational framework provides a mechanistic insight into taking the macromolecular crowding as necessary physicochemical variables representing cell states to remodel the configurations of protein megacomplexes with structurally loose modules.
There has been a rapidly growing interest in understanding why many cellular phenotypes arise from one genome in the content of environmental gradients enabled by high-throughput experimentations, imaging, and advance analytic tools guided by AI [1]. Given that genomic information and derived knowledge is abundantly available, we gained insight about biological networks from gene expression data [2][3][4], single-cell RNA sequencing [5], and protein sequences and structures [6]. However, the multiple phenotypic expressions at the proteomic level involving variations in protein interactions and their abundance representing a cell state [7] remain poorly understood (Figure 1A) [8,9]. Protein assemblies that dynamically adapt to a cell state [10,11] by associating into biomolecular condensates [12][13][14][15][16] or megacomplexes [17][18][19][20] are outstanding examples that manifest a growing complexity from genotype to phenotype expressions shaped by environmental influences. Despite the advance in protein structural predictions with AI, the mechanism of protein interactions and the emergence of megacomplexes they assemble remains unclear.
To address this gap, several computational and theoretical frameworks have been developed [12][13][14][15][16][21][22][23] to investigate the protein interactions that drive macromolecular assemblies involving heterotypical constituents. One approach is the scaffold-client framework, in which proteins interact with more partners (i.e. higher valency) are considered scaffolds to recruit the clients interacting with few partners (i.e. lower valency) [24,25]. While there are extended models to include more comprehensive interactions between scaffolds and clients, this framework is still under the assumption of a fixed protein level [26,27], representing a static cell state.
To incorporate the consideration of dynamic cell states that influence the constituents as well as the configurations of protein assemblies, we deployed an open system based on a grand canonical ensemble framework [28] to test a hypothesis that the excluded volumes of subunits limit the configurational search of protein interactions in a reservoir-like cytoplasm environment, and that drives protein assemblies. We used yeast chromatin remodeler INO80 complex [29][30][31][32][33][34][35] as an example for the assembly of a megacomplex with many structurally loose parts in a cell-like environment and they remodels [29,30] upon environmental cues.
We represented a dynamic cell state where a system continuously exchanges components and moves toward equilibrium with the surrounding particle reservoir. We generated structural models by inferring interaction energies and chemical potentials between loose parts from experimental data for INO80. Through thermodynamic perturbations such as abundance variation and crowding [36][37][38][39], we teased out the critical components in INO80 which orchestrates interactions beyond nearest neighbors. The excluded volumes contributed to the depleted interactions [40] between modules at high crowding content, giving rise to distinct configurational state of structural machinery for INO80 under various crowding conditions (Figure 1B).
We model the cellular impact of relative protein abundance into chemical potential [41] and their interactions from cross-linking mass spectrometry for the INO80 complex (see Method section). For each fifteen subunits of the INO80 complex, we iteratively adjust the chemical potential until the simulated particle number converges to its experimental abundance while keeping the other fourteen subunits’ abundance fixed. For most subunits, their convergence is reached within a few iterations (see Supplement Figure S4). However, for a small subset of subunits, the particle number oscillates around the experimental abundance and are unable to converge toward experimental estimates, regardless of hyperparameters. We call this specific category of subunits “divergent”. Others are by comparison “convergent”.
To better understand the underlying principles of the newly discovered divergent property of subunits in INO80, we plotted the phase diagram of density vs. chemical potential of the subunits that display divergent behavior at a reduced temperature unit, T=1, and the volume fractions of subunits, 𝜙 equals 0.05, 0.1, and 0.2 in Figure 2. This range of 𝜙 represents the volume fractions of crowding contents in a cell [38,42]. Density represents normalized abundance and 𝜙 is defined as the proportion of a system’s total volume that is occupied by the INO80 subunits. In Figure 2A, the profiles for the divergent subunits are plotted in thick lines. They exhibit a discontinuity in density around a critical chemical potential 𝜇 * . For example, the profile of Ies4 (yellow curve) at 𝜙 = 0.1 shows a density discontinuity spanning from roughly 0 to 0.014 at chemical potential 𝜇 * = -9.8, within which the empirical value (indicated by the yellow arrow) falls
This content is AI-processed based on open access ArXiv data.