Abstract
Disordered proteins are conformationally flexible proteins that are biologically important and have been implicated in devastating diseases such as Alzheimer’s disease and cancer. Unlike stably folded structured proteins, disordered proteins sample a range of different conformations that needs to be accounted for. Here, we treat disordered proteins as polymer chains, and compute a dimensionless quantity called instantaneous shape ratio (Rs), as Rs = Ree2/Rg2, where Ree is end-to-end distance and Rg is radius of gyration. Extended protein conformations tend to have high Ree compared with Rg, and thus have high Rs values, whereas compact conformations have smaller Rs values. We use a scatter plot of Rs (representing shape) against Rg (representing size) as a simple map of conformational landscapes. We first examine the conformational landscape of simple polymer models such as Random Walk, Self-Avoiding Walk, and Gaussian Walk (GW), and we notice that all protein/polymer maps lie within the boundaries of the GW map. We thus use the GW map as a reference and, to assess conformational diversity, we compute the fraction of the GW conformations (fC) covered by each protein/polymer. Disordered proteins all have high fC scores, consistent with their disordered nature. Each disordered protein accesses a different region of the reference map, revealing differences in their conformational ensembles. We additionally examine the conformational maps of the nonviral gene delivery vector polyethyleneimine at various protonation states, and find that they resemble disordered proteins, with coverage of the reference map decreasing with increasing protonation state, indicating decreasing conformational diversity. We propose that our method of combining Rs and Rg in a scatter plot generates a simple, meaningful map of the conformational landscape of a disordered protein, which in turn can be used to assess conformational diversity of disordered proteins.
Significance
Intrinsically disordered proteins (IDPs) are proteins without stable three-dimensional conformations that represent a large fraction of the eukaryotic proteome and often nontrivially participate in serious diseases. Unlike structured proteins, they have diverse ensembles of conformations. Here, we report a framework to map those ensembles and introduce a metric, the fC score, that quantifies conformational diversity. We use dimensionless parameters of size and shape to quantify conformations and plot them on a graph, illustrating entire structural landscapes as simple maps, designed to be comparable for different proteins. Our maps show how IDPs, structured proteins, and polymers can have unique conformational landscapes and differing conformational diversities. We propose that this framework for analyzing conformations can provide new insights into protein disorder.
Introduction
Disordered proteins are proteins without stable three-dimensional structures. Unlike structured proteins, these proteins do not reach a single, stable conformation, but instead sample a range of different conformations. Proteins with disordered regions have been found to comprise substantial portions of the human proteome (1), and have been implicated in serious human diseases (2), such as cancer (3) and Alzheimer’s disease (4). Disordered regions have also been observed in regulatory and signaling proteins (5), suggesting that they have important biological functions (2,3,4).
Recent decades have seen remarkable innovations in the challenging (5,6,7,8) task of analyzing the structures of disordered proteins. For example, there are now resources available that predict disorder based on sequence (9,10). Structures generated by the well-known deep learning-based resource AlphaFold (11,12) often contain low-confidence regions that overlap with disordered regions (13), suggesting that AlphaFold can potentially be used to locate disordered regions in a structure. Because sufficient sampling of disordered protein conformation space is challenging in molecular dynamics (MD) simulations, there are now enhanced sampling techniques such as Hamiltonian replica-exchange MD that have been used to computationally generate accurate configurational ensembles of disordered proteins that are consistent with experimental data (14). On top of enhanced sampling techniques, force fields have been developed to improve modeling and simulation of disordered proteins (8,15,16).
The closer a protein structure is to the disorder end of the order-disorder continuum (17,18,19,20), the more challenging its structure can be expected to be for analysis. Highly disordered proteins tend to be highly flexible structures, which in turn likely means complicated energy landscapes (21), varying amounts of secondary structure, and likely varying timescales (21,22) at which they switch between different conformations—all of which are obstacles to reliably describing their conformations. Thus, describing the heterogeneous configurational ensembles of highly disordered proteins is a matter that requires careful thought.
Through this study, we provide a tool to help analyze disordered protein structures. We attempt to simplify descriptions of disordered protein conformational landscapes by drawing a map of conformational landscapes. We conduct simulations of proteins and polymer models and calculate two quantities, a measure of shape (instantaneous shape ratio, Rs) and a measure of size (radius of gyration, Rg), to define conformations at a specific instance. Shape ratio, calculated by dividing the square of end-to-end distance by the square of radius of gyration, is a dimensionless quantity that tells us how compact or extended a polymer’s conformation is. Lower shape ratio values (∼2 or lower) are expected for collapsed (compact) coils, whereas higher shape ratio values (10) tend to represent rod-like (extended) states. Shape ratio values close to 6 usually represent Gaussian coils. Our first step is to analyze the Gaussian Walk (GW), Random Walk (RW), and Self-Avoiding Walk (SAW) simple polymer models using these quantities. We generate a scatter plot of Rs against Rg to map the landscape of conformations of these polymer models. We then use the GW model landscape as a reference for the conformational landscapes of other proteins (disordered and structured) and polymers. We examine a variety of disordered and structured proteins, and the well-known gene delivery vector polyethyleneimine (PEI) at various protonation states. We find that all protein and polymer conformational landscape maps lie within the boundaries of the reference (GW) map. We notice that each protein/polymer has its own unique map—indicating differences in conformational landscapes and conformational diversity, even among disordered proteins. We quantify and rank the conformational diversity of proteins and polymers through their fC scores. On the scatter plot of GW and a given polymer, the fC score computes the fraction of GW conformations covered (i.e., close to at least one polymer conformation) by that polymer. Because the GW itself has high conformational diversity, any polymer that covers a high fraction of these conformations also has high conformational diversity. We compare the fC scores of different proteins and polymers to further demonstrate their differences in conformational diversity.
Methods
Monte Carlo simulations
We investigate the GW, RW, and SAW simple polymer chain models using Monte Carlo (MC) simulations. For GW chains, the chain conformation is generated such that the polymer beads are points connected in continuous space; the distance from one point to the next is randomly selected from a Gaussian distribution with mean 0 and standard deviation 1 using the gauss() method of the random module in Python 3.7. We simulate single chains with 25, 50, 75, 100, 125, and 150 monomers (and a few higher chain lengths in the supporting methods). For each GW chain length, we generate 30,000 snapshots, where each snapshot is a new randomly chosen chain configuration. Because we use a GW chain of length 100 as a reference landscape, we generate additional snapshots for this chain length to use a larger number of reference conformations.
For RW and SAW, we model the chains as connected beads sitting on a simple cubic lattice with periodic boundary conditions. SAW chains have excluded volume interactions and, therefore, multiple beads of a SAW chain cannot occupy the same lattice site, whereas multiple RW beads can occupy the same site. For RW and SAW models, we begin MC simulations by first generating a chain on a lattice. Then the chain is equilibrated using a combination of local movements and reptation moves whose acceptance is governed by the Metropolis rule (23,24). The unoccupied lattice sites are treated as being occupied by the solvent. The conformations of a chain with lengths (i.e., number of monomers) 25, 50, 75, 100, 125, and 150 in athermal solvent conditions (i.e., no solvent-bead interaction) are studied for RW and SAW. The lattice size we use for RW and SAW is 250 250 200—which is larger than the size of the chains. In addition to simulations in athermal solvent conditions, we simulate collapse behavior of the SAW chain using a solvent-bead interaction, εps, an interaction energy that is reduced by the thermal energy kT (k is the Boltzmann constant and T is the temperature). εps is applied between all pairs of neighboring polymer bead and solvent particles. The reduced solvent-bead interaction is a dimensionless number, which can be 0 (corresponding to athermal solvent) or have positive (corresponding to a poor solvent) or negative values (corresponding to an attractive interaction between the chain and solvent). The radius of gyration (Rg) and end-to-end distance (Ree) values of each chain are monitored during simulations to ensure equilibration. Of the equilibrated simulations, 30,000 snapshots (unless otherwise stated) of the RW and SAW chains (per chain length), with an interval of 25,000 MC steps between each snapshot, are collected. One such MC step consists of one attempted move per bead and reptation of entire chain. For analysis the latter ∼90% of these snapshots are used.
MD simulations
All MD simulations are performed using explicit solvent models. The disordered (except protamines) and structured protein simulation trajectories are from Song et al. (16). We obtained trajectories for ab40 (25) and tauF4 (26) (known disordered proteins), and ubiquitin (27), FKBP12 (28), and LUSH (29) (known structured proteins). These disordered proteins are multiple trajectories (5–200 ns trajectories, containing a total of 200,000 snapshots per protein), whereas the structured proteins are single trajectories (one 1000 ns trajectory for each, with 200,000 snapshots total). ab40 was simulated using the ff14SB (30) force field and the rest of the proteins listed so far were simulated using the ESFF1 (16) force field.
We re-examine 90 ns trajectories of the commonly studied nonviral gene delivery vector PEI that our group studied previously (31), using the generalized AMBER force field (32) in salt-free conditions (only counterions included). In brief, PEI is a 40-mer polymer chain with varying degrees of protonation of nitrogen atoms along its backbone (0p or 0% protonation up to 40p or 100% protonation; 90 ns trajectory for each protonation state).
We also re-examine protamine simulation trajectories that our group published previously (reference contains more simulation details) (33). In brief, the results we show here are the results of 13 parallel trajectories of each protamine (salmon, bull P1, and human P1), conducted using AMBER 18 (34). For each explicit solvent trajectory, we extract the latter 80 ns for analysis (which we consider to be the equilibrated region of each trajectory). For fC calculations, we reduce each protamine simulation’s snapshot count such that it represents ∼1040 ns over 208,000 snapshots (over 13 trajectories). We use the ff14SB (30) force field for protamines. In addition, we repeat our explicit solvent human P1 protamine simulation using the ESFF1 (16) force field, to compare its conformational landscape with that obtained using the ff14SB force field.
MD simulation trajectory analysis (especially Rg and Ree calculation) is done using CPPTRAJ (35) and conformations are visualized using VMD (36). All simulation details are summarized in Table S1.
Polymer physics quantities calculated
In polymer physics, the shape ratio (Rs) is a dimensionless quantity normally calculated as
| (1) |
where both < Ree2> and < Rg2> are mean values averaged over all the conformations and Ree and Rg are the end-to-end distance and radius of gyration, respectively (37). The ratio based on the two mean values is expected to be 6 for a flexible coil-like polymer chain (38). A higher shape ratio (∼12 or above) value is expected for rod-like or extended chain conformations, and a lower shape ratio value (∼2) is expected for collapsed or compact chain conformations.
In this study, we calculate an instantaneous shape ratio value (Rs) for individual simulation snapshots,
| (2) |
A fully straight chain is expected to have an Rs value of 12, although Rs values greater than 12 can be achieved by minimizing Rg while maximizing Ree. An example of a GW chain conformation with Rs = 25.2 is shown in Fig. S1, where the two ends of the chain are extended and far apart, but the inner beads are closer together.
Since we compare proteins/chains of different lengths, we normalize Rg values and make it dimensionless by dividing it by , where is calculated for a protein over its entire simulation. If there are multiple parallel trajectories for a protein, we compute by first calculating the mean Rg value of all trajectories and dividing all the Rg values of all trajectories by that mean. Fig. S2 presents the Rg2 and Ree2 distributions of each chain length for each of RW, SAW, and GW, and shows how Rg2 and Ree2 expectedly increase (Fig. S2, a and b, respectively) with chain length. GW and RW models have the expected exponent of 0.5 (Rg ∼ N0.5, N is the number of the beads/monomers in the chain), while SAW has the well-known Flory exponent, 0.59 (Fig. S3). On the other hand, Rs and do not depend on chain length for these three models (Fig. S2, c and d, respectively), which is not surprising, as both quantities are ratios in which the numerator and denominator scale with chain length in the same way.
In addition to examining instantaneous shape ratio values (sometimes referred to as shape ratio), we assess the impact of calculating the mean shape ratio value in two different ways. The two different definitions are:
| (3) |
and
| (4) |
We find small but systematic differences in the two definitions. Fig. S4 shows mean values of shape ratio of GW, RW, and SAW polymer chain models calculated using two different definitions (Rmsr1 and Rmsr2). For all three chain models, Rmsr1 remains greater than Rmsr2. In addition, values for GW and RW are within each others’ error margins, and SAW values are higher than those of GW/RW, with minimal variations (∼0.1) existing among mean values from chain length to chain length. We generally use the Rmsr2 definition for mean shape ratio unless otherwise stated.
Another quantity we investigate is the relative shape anisotropy (RSA). We calculate RSA using formulas previously used by Elder and Jayaraman (37). For GW polymer chain model, we calculate tensor such that:
| (5) |
where Rcom is the center of mass of the polymer and α and β represent Cartesian directions x, y, and z (37). For the proteins and PEI, tensors are calculated using CPPTRAJ (35). The gyration tensor terms are diagonalized to obtain the three eigenvalues (37,39) of : , , and such that:
| (6) |
and
| (7) |
These eigenvalues are the magnitudes of the eigenvectors pointing in the directions of the three principal axes of a given structure. Next, asphericity () is computed such that:
| (8) |
the minimum value of which is 0 representing a perfect sphere. Next, acylindricity () is computed:
| (9) |
the minimum value of which is 0 representing a cylindrically symmetric structure (37,40). Next, RSA is calculated using the formula:
| (10) |
RSA can range from 0 (representing a perfectly isotropic structure, such as a sphere) to 1 (representing a perfectly anisotropic structure, such as a rod) (37,40).
Two-dimensional scatter plots and fC score
Throughout this study, we generate scatter plots of against instantaneous shape ratio (Rs) for polymers and proteins to define and map their conformational landscapes. A two-dimensional scatter plot for GW and that for one or more polymers/proteins are plotted in the same plot. or Rs values for any protein or polymer fall within the boundary defined by GW on the two-dimensional scatter plot (i.e., they overlap with GW values), with GW serving as a reference. We then quantify the fraction of the reference (GW) map covered by a given protein or polymer through its fC score. Higher fC scores indicate more conformational diversity and vice versa. The fC calculation protocol and other technical details are provided in the supporting methods.
We use the GW (chain length, N = 100, with total 720,000 snapshots) as a reference map for other proteins/polymers. There are several reasons behind this choice. First, with simple chain models such as GW, RW, and SAW, the conformational space can be sampled very well within reasonable amounts of simulation, unlike with complex chains such as proteins. This can be seen in Fig. S5 where simulations with these simple chain models show excellent convergence as measured by the histogram distributions of Rg and Rs. With complex chains (e.g., protamine proteins), on the other hand, histogram distributions of these quantities from different trajectories differ from each other (see Fig. S6). Second, in two-dimensional scatter plots, the GW model provides the broadest map, with data from RW and SAW lying within the bounds of the GW model (see Figs. S7 and 2). This suggests that the GW model has a more diverse conformational landscape than the RW and SAW models, which is understandable since RW and SAW, unlike GW, are polymer chains constrained on a lattice that likely restricts their possible conformations. Third, the histogram distributions of Rg/Rgmean and Rs are nearly independent of GW chain length, as shown in Fig. S2, c and d, implying that the same GW chain length can be used as a reference for proteins/polymers of different lengths. As elaborated in the supporting methods, a longer GW chain would only add a few new outlier points around the edges of the two-dimensional scatter plot of a GW chain of length 100 (see Fig. SM8 b in the supporting methods). Hence, we use a GW chain of length 100 as a reference map. As we explain in the supporting methods, and demonstrate throughout this study, this GW chain effectively provides boundary values for all protein/polymer and Rs, which is one of the main expectations for the reference map. We further explore the implications of varying GW chain lengths in the supporting methods.
Figure 2.
Map of the conformational landscape of (a) GW polymer model, (b) GW and RW polymer models, and (c) GW and SAW polymer models. All y axes show instantaneous shape ratio (Rs). All x axes show relative radius of gyration (Rg/Rgmean). The RW and SAW plots show combined data for multiple chain lengths (i.e., N = 25, 50, 75, 100, 125, and 150, ∼27,000 snapshots per N). The reference (GW) plot shows data for a single chain length (N = 100) with 720,000 snapshots. To see this figure in color, go online.
There is some variation in the exact fC score with the number of GW or protein/polymer snapshots used (details in the supporting methods). We use the same number of reference GW snapshots in all cases (720,000). The number of protein/polymer snapshots are mostly kept close to one another, so the differences in fC scores are not due to the number of snapshots. The proteins tauF4, ab40, ubiquitin, FKBP12, and LUSH are each 200,000 snapshots (∼1000 ns) and each protamine is 208,000 snapshots (∼1040 ns), thus bringing it closer to 200,000. For RW/SAW, there are 162,054 snapshots each (27,009 snapshots for each of 6 chain lengths, at εps = 0, unless otherwise stated). For PEI simulations, each PEI protonation state contains 45,000 snapshots gathered over 90 ns.
Results and discussion
Examining shape ratios of GW, RW, and SAW and mapping their conformational landscapes
We use two main metrics to characterize the conformations of disordered proteins, the relative radius of gyration (), a measure of the size of the conformation, and shape ratio (Rs), a measure of the conformation’s shape. For GW, RW, and SAW chains, the values of these metrics have a wide range, with commonly taking values ranging from 0.5 to above 1.5 and Rs commonly ranging from 0 to ∼15 (see Figs. 1 and S2, c and d). This behavior reflects the disordered natures of these polymer chain models; they do not fold into stable conformations like proteins, but instead adopt a variety of conformations and have a wide range of and Rs values. Furthermore, RW and GW chain distributions are very similar, whereas that of SAW is different from these two, indicating that the presence of excluded volume interactions in SAW does have a small, but noticeable, impact on their sizes and shapes (Fig. 1). The Rs and distributions are also relatively constant for different chain lengths for all three chain models (Fig. S2, c and d), although neither Rs nor distributions are normal (probability plots for a GW chain are shown in Fig. S8).
Figure 1.
(a) Relative radius of gyration () distribution and (b) instantaneous shape ratio (Rs) distribution for a chain length (N = 150, where N is the number of monomers) in the GW (black), RW (red), and SAW (green) models. To see this figure in color, go online.
In this study, we map the conformational landscapes of polymers and proteins in conjunction with a reference (GW) model by generating a scatter plot of Rs against . For these scatter plots, we compute fC scores as described in the methods and supporting methods. The map of the conformational landscape of the GW polymer model alone is shown in Fig. 2 a. The fC of the RW is very close to 1, as the RW and GW maps are nearly identical (Figs. 2 b and S7 a), while SAW has a slightly smaller fC score at 0.989 (Figs. 2 c and S7 b). Our fC definition calculates the fraction of GW conformations that are covered by other chain models. Effectively, it represents the volume fraction of the GW distribution occupied by other chain models, not the area fraction, which explains why the fC score for SAW is high (Fig. S7 b), despite covering a smaller two-dimensional area than RW (Fig. S7 a). The GW scatter plot has a higher density of points near its center (where Rs is ∼6 and is ∼1), as many conformations are located there, which drops as one moves away from that center. The higher fC score of RW compared with the SAW is reflected in the individual and Rs distributions presented on the top and side, respectively, in Fig. S7, with the RW model having almost the exact same distribution as GW, while the distributions from SAW (Fig. S7 b) are slightly shifted from those of GW, leading to a slightly lower fC.
To examine how Rs and fC respond to known changes in chain conformation, we perform simulations of SAW chains in the presence of repulsive polymer-solvent interactions (i.e., εps > 0). As εps increases, repulsion between the chain and solvent increases, causing the chain to collapse and the mean Rs (Fig. 3 a) to decrease. The vertical line in Fig. 3 a identifies the θ-solvent threshold in polymer physics, determined from a plot of <Rg2>/N against εps (Fig. S9). The mean Rs values decrease noticeably after the solvent quality reaches the θ-solvent, from ∼6.0 in good solvent down to ∼3.5 or lower. We also note that shape ratio (Rs) values exhibit chain-length dependence in the collapsed regime (when εps > θ). The distributions of the shape ratio for different values of εps for a SAW chain of length 100 are shown in Fig. 3 b. As εps increases, the instantaneous shape ratio distributions shift to smaller values (Fig. 3 b). For highly collapsed chain (εps > 0.2), most conformations adopted by the SAW chain have Rs < 5. The chain is approaching a fully collapsed sphere for SAW when εps > 0.2, since the exponent in the <Rg2>∼N plot becomes less than 0.667, which is the value expected for a sphere model (see Table S2 for the exponents at different εps values).
Figure 3.
(a) The mean shape ratio <Rs>, calculated using the Rmsr2 formula, , of the SAW chain at different εps values. The gray vertical line indicates collapse transition threshold, identified as the crossover point in <Rg2>/N dependence on εps (as indicated in Fig. S9). (b) Histograms of instantaneous shape ratio (Rs) for a SAW chain with N = 100, at different εps values. To see this figure in color, go online.
We examine fC scores of a collapsed SAW chain of length 100 (Figs. 4 and S10) and find that the collapsed model has a lower fC than its uncollapsed counterpart. The fully collapsed model mostly has lower shape ratio values than the uncollapsed model (shown on the side histograms in Fig. 4), consistent with the trend demonstrated in Fig. 3 b, and narrower values (shown on the top histograms in Fig. 4), indicating smaller Rg range.
Figure 4.
Conformational landscape maps of collapsed SAW chain (εps = 0.3) and the uncollapsed SAW chain (εps = 0). Both the collapsed and uncollapsed chains have 162,054 snapshots from simulation of a single chain length (N = 100). The histograms on the top and side show distributions of and instantaneous shape ratio (Rs) respectively, with GW/SAW following the same color scheme as the scatter plots. The black dots are GW snapshots (N = 100, 720,000 snapshots). To see this figure in color, go online.
Fig. S10 presents fC scores at each εps investigated for SAW, with fC ranging from ∼1 to ∼0.8. The fC remains relatively high even for the fully collapsed chain, which is to be expected since the collapsed SAW chain does not resemble a folded structured protein. Instead, it represents a glassy disordered system.
Conformational landscape maps of known disordered and structured proteins
To map the conformational landscape of a protein, we generate scatter plots of instantaneous shape ratio (Rs) against , using fC scores to assess conformational diversity. Fig. 5 shows maps of two disordered proteins (ab40 (16,25) and tauF4 (16,26)). ab40 has a high fC at 0.757 (Fig. 5 a), showing enough conformational diversity that it occupies almost the entirety of the shape ratio distribution of GW (shown in the side histogram in Fig. 5 a). tauF4 (Fig. 5 b) also has a high fC at 0.595, although it is not as high as ab40. Each protein has a unique map, with noticeably different ranges of and Rs. Example conformation images from different parts of the tauF4 map are shown (Fig. 5 b), demonstrating that different regions of the map represent different conformational characteristics. For example, tauF4 conformations (i) and (iv) have very different shape ratio values at ∼12 and <1, respectively, but similar values (Fig. 5 b). This difference is visually detectable with conformation (i) being more extended (has higher end-to-end distance) than conformation (iv), which is more of a compact structure (Fig. 5 b). Conversely, tauF4 conformations (ii) and (iii), which have similar shape ratio values but very different values, are also noticeably different (Fig. 5 b), with (ii) appearing more compact than (iii), although both have similar end-to-end distances. Fig. 6 shows maps of two structured proteins (LUSH (16,29) and ubiquitin (16,27)). Unlike ab40 and tauF4, LUSH has a much smaller fC score of 0.011 (Fig. 6 a). The small fC score of LUSH is consistent with a stably formed three-dimensional conformation that does not change much. Maps of two additional structured proteins (ubiquitin in Fig. 6 b and FKBP12 in Fig. S11) have fC scores much closer to that of LUSH than the disordered proteins. However, ubiquitin has a slightly higher fC (0.114; Fig. 6 b) than the other structured proteins, but not high enough to be close to any of the disordered proteins. Ubiquitin and ubiquitin-like proteins are known to have a functionally relevant, flexible C-terminal tail (27,41,42,43,44). This flexibility is encapsulated in our map of ubiquitin’s conformational landscape; the flexibility of the tail causes variation in the end-to-end distance that is reflected in ubiquitin’s shape ratio distribution (specifically the two major peaks in the side histogram in Fig. 6 b), raising its fC score slightly above those of other structured proteins.
Figure 5.
Conformational landscape maps for the disordered proteins (a) ab40 and (b) tauF4, with select conformation images identified for (b). For each plot, the histograms on the top and side show distributions of and instantaneous shape ratio (Rs) respectively. The black dots are GW snapshots (N = 100, 720,000 snapshots) and magenta dots are protein snapshots from the respective protein simulations. To see this figure in color, go online.
Figure 6.
Conformational landscape map for the structured proteins (a) LUSH and (b) ubiquitin, with select conformation images identified for (b). For each plot, the histograms on the top and side show distributions of and instantaneous shape ratio (Rs), respectively. The black dots are GW snapshots (N = 100, 720,000 snapshots) and magenta dots are protein snapshots from the respective protein simulations. To see this figure in color, go online.
Finally, we examine maps of three known disordered proteins (salmon, bull P1, and human P1 protamines), whose structures we studied previously (33) (Fig. S12). Protamines are arginine-rich proteins that condense DNA in sperm cells, with biomedical applications as injectable biopolymers (45,46) and potentially as mRNA vaccine delivery agents (47,48,49). fC scores of all three protamines are expectedly close to those of other disordered proteins. Because reliable disordered protein MD simulations are dependent on choice of force field, we compare the effect of the ESFF1 (16) force field (intended to accurately simulate both disordered proteins and structured proteins) and the generic protein force field ff14SB on the human P1 conformational landscape. We notice that the ESFF1 force field facilitates sampling of more conformations than ff14SB with a higher fC score (0.797 vs. 0.714), although there is substantial overlap between the two (Figs. S12 b, S13, and S14).
By solely comparing fC, we demonstrate the stark difference in fC scores of disordered and structured proteins (Fig. 7). Among the proteins we use in our study, salmon protamine has the highest conformational diversity (and possibly highest disorder) and LUSH the lowest. However, we do observe some variation within the disordered proteins and within the structured proteins (Fig. 7).
Figure 7.
Bar plot of fC scores of all the disordered and structured proteins used in this study. To see this figure in color, go online.
Shape ratio and conformational landscape maps of PEI at different protonation states
Our group previously conducted MD simulations of the structure of PEI, which is known to be an effective nonviral gene delivery vector (23,31,50). Positive charges (protons) can be added to nitrogen atoms along the backbone of the PEI chain that affect its conformation. Every third atom along the PEI backbone is a nitrogen atom that can be protonated, which means there is potential for positive charges to be very close to one another. We studied a 40-mer charged polyelectrolyte polymer chain PEI, with protonation states ranging from a fully unprotonated chain (0p) to a fully protonated chain (40p). Average Rg values for PEI increase with increases in protonation, from 8.6 Å for the 0p chain to 31.8 Å at 40p (31). Here, we analyze those same PEI trajectories and calculate the mean shape ratio values at different protonation states (shown in Fig. 8, with corresponding Ree2 and Rg2 values shown in Fig. S15). We find that it trends upward with increasing protonation state (with Rmsr2 changing from ∼6 at 0p to ∼8 at 40p), suggesting that the PEI chain adopts a more extended (rod-like) conformation at the higher proton states, most likely due to increasing repulsion between positive charges. In addition, the error bars get smaller from 0p to 40p, suggesting lower variation in the higher protonation states than lower protonation states. Interestingly, we observe a slight drop in mean shape ratio after 20p (Fig. 8), despite steady increase up to and after that point, suggesting that, after 50% protonation, the forces acting on the chain are different and cause it to fold slightly. This is likely caused by repulsive forces between protonated nearest neighbor nitrogen atoms. At 50% protonation, every other nitrogen atom along the backbone is protonated (23,31), and additional protonation inevitably leads to nitrogen atoms that are very close together (i.e., nearest-neighbor nitrogen atoms) to become protonated. Upon comparing shape ratio distributions of PEI at every protonation state (Fig. S16), we notice that 0p and 5p behave differently from the remainder of the protonation states, with each showing a wider distribution than the higher protonation states.
Figure 8.
Variation of mean shape ratio of PEI as protonation state is increased, from 0% protonated (0p) to 100% protonated (40p), calculated two different ways. For the (Rmsr2) definition the error bars are standard deviation. For the (Rmsr1) definition propagated error (standard deviation) is shown in the form of error bars. To see this figure in color, go online.
We examine how PEI’s fC scores change with protonation state (Figs. 9 and S17), and notice that the fC increases immediately after 0p, and then trends downward with increasing protonation state. This suggests less conformational diversity, less flexibility, and (potentially) less disorder at the higher protonation states compared with the lower ones, likely caused by increased conformational restrictions imposed by positive charges along the PEI backbone. Consistent with the behavior described earlier with mean shape ratio values, there is a slight increase in fC between 20p and 25p, suggesting increase in flexibility immediately above 50% protonation.
Figure 9.
Conformational landscapes maps of PEI and corresponding fC at all protonation states. All y axes show instantaneous shape ratio (Rs). All x axes show relative radius of gyration (Rg/Rgmean). The black dots are GW snapshots (N = 100, 720,000 snapshots). To see this figure in color, go online.
Comparing instantaneous shape ratio with RSA
Thus far, we have mapped conformational landscapes by combining instantaneous shape ratio (a measure of the conformation shape) with relative radius of gyration (a measure of the size of a conformation). Since RSA is a known alternative measure of polymer shape, we compare RSA with instantaneous shape ratio (Fig. 10) for two of the same proteins as before: ab40 and LUSH. We notice similar behavior in that the disordered protein (ab40) has a noticeably higher fC than the structured protein (LUSH). Interestingly, there seems to be some correlation between RSA and (as suggested by the pattern of the GW dots in Fig. 10) that does not exist between shape ratio and (no discernible pattern in Fig. 2 a). We thus calculate Pearson correlation coefficients and find that for the GW model alone, the correlation coefficient between and RSA is much stronger at 0.696 than that between and instantaneous shape ratio (0.235).
Figure 10.
Maps of conformational landscapes (using RSA and ) for (a) ab40 and (b) LUSH, where the fC is calculated using RSA and . The black dots are GW snapshots (N = 100, 720,000 snapshots). To see this figure in color, go online.
We calculate fC scores using RSA for all the proteins in our study and, upon comparing them with those calculated using shape ratio (Table S3), we notice that ubiquitin has a noticeable difference (fC = 0.114 in Fig. 6 b and fC = 0.0021 in Fig. S18). This is a reflection of the difference in shape ratio distribution (i.e., the two major peaks of the shape ratio histogram in Fig. 6 b) and RSA distribution (single peak in RSA histogram in Fig. S18) For ubiquitin, we notice that shape ratio more clearly distinguishes two different ubiquitin conformations (with the N-terminal and C-terminal ends being either close together or further apart, as shown in Fig. 6 b) that RSA apparently fails to do (because the protein’s shape as defined by RSA does not change noticeably, Fig. S18).
Conclusions
In this study, we explore the possibility of using a two-dimensional scatter plot of instantaneous shape ratio (Rs) and relative radius of gyration (Rg/Rgmean) as a means to map the conformational landscapes of disordered proteins. We first generate such maps from simulations of simple polymer models such as GW, RW, and SAW. As we demonstrate using Rs and Rg, simulations with these simple polymer chain models are converged and thus their two-dimensional scatter plots are expected to have sampled the conformational space very well. We find that the conformational landscape of every protein and polymer in our study overlaps with, and is confined to the boundaries of, GW. Because of this and because our MC simulations are expected to have sampled the GW conformation space very well, we argue that the GW conformational landscape is a “universal” landscape that encompasses all (or most) protein/polymer conformational landscapes.
We next examine how the disordered (ab40, protamines, tauF4) and structured (ubiquitin, FKBP12, LUSH) proteins appear on our two-dimensional scatter plots. The two classes of proteins not only look different (compare Figs. 5 a and 6 a) but also have radically different fC scores, although differences are noticeable in the plots of proteins within each class as well (compare Fig. 5 a and b). Not surprisingly, the structured proteins have significantly lower fC scores (and thus lower conformational diversity) than disordered proteins (Fig. 7). In this study, we examine only a few structured and disordered proteins. It would be interesting to expand this list to include a broader spectrum of proteins (e.g., disordered proteins, folded proteins, mixed domain proteins) and map and compare their conformational landscapes and fC scores.
To further investigate if two-dimensional scatter plots could be used to reveal different conformational ensembles of disordered proteins, we map the well-known nonviral gene delivery vector PEI at various protonation states. When the protonation state changes from 0% (0p) to 100% (40p), the PEI chain changes from collapsed globule structure to extended rod-like structure (31). At both extremes of protonation states, the PEI chain is disordered. The fC varies from 0.39 (at 40p) to the highest value 0.93 (at 5p), both values representative of disordered proteins. The region of the reference map accessed by PEI generally narrows with increasing protonation state (Fig. 9, red regions). The collapsed SAW chain, which represents another class of disordered proteins, occupies regions of the two-dimensional map that differ from noncollapsed SAW chain (Fig. 4). These taken together suggest that such two-dimensional scatter plots can be used to reveal different conformational ensembles of disordered proteins.
There are challenges to investigating disordered protein structures, from ensuring adequate conformation sampling, to choosing appropriate force fields that accurately reproduce their conformations and visualizing different ensembles of disordered protein conformations. Multiple past studies have investigated metrics or models to describe conformational ensembles of disordered proteins (51,52,53). Recently, Zhao et al. investigated two intrinsically disordered proteins using simple physics-based hierarchical models (51). One challenge they noted is that one often cannot resolve different conformational ensembles sampled in different models through size distributions or single chain structure factors. To address the problem, they used Uniform Manifold Approximation and Projection to identify conformational subsets. It would be interesting to check if our proposed two-dimensional scatter plots can resolve conformational subsets in their study. Our previous study of protamine (33) reveals that multiple MD trajectories, with different starting seeds, tend to cluster on such two-dimensional maps in different regions.
In the supporting methods, we discuss the fC score, which can be used as a measure of the conformational diversity of a protein or polymer, in more detail. The exact fC score of a protein/polymer may depend on its number of snapshots and that of the reference (GW) chain (respectively) used for the scatter plot. Based on the proteins examined here, ∼1000 ns simulations for each protein, with snapshots taken every ∼5 ps appears sufficient, as they result in relative plateaus in fC scores, although longer simulations may be needed for some proteins. A prerequisite for a meaningful fC score is that its conformation landscape fully overlaps with our chosen reference landscape, which does happen for all the proteins/polymers in our study. Should a polymer landscape significantly exceed the boundaries of our reference landscape, we recognize that our fC score would require more cautious interpretation. In addition, we acknowledge that Rs and do not unambiguously define a conformation. Multiple conformations can exist for the same (, Rs) values, especially for Rs values around ∼6 and values around ∼1 (as the GW histograms show in Fig. S2, c and d). We further acknowledge that enhanced sampling techniques, especially if implemented on the disordered proteins from this study, are likely to increase coverage of their conformational landscapes, leading to increased fC scores. To that end, the framework presented in this paper would, arguably, be a valuable tool to confirm that enhanced coverage, as well as the specific conformations (in terms of size and shape) that were newly sampled.
Posttranslational modifications are changes to protein chemical structure (an example of which is phosphorylation) that have been known to modify its conformations and functions (54). Because a protein’s structure changes, it is likely that its conformational landscape changes as well with potential consequences in function. We think an interesting topic for future exploration using our landscape maps and fC scores is the change in conformational landscape of a protein before and after a posttranslational modification. Similarly, one could use our landscape maps and fC scores to explore how the conformational landscape changes for a protein after a point mutation. Furthermore, for proteins that are composed of both disordered and structured regions, such as Aggregatin (55), there is potential in using these maps and fC scores to study the interplay between these regions.
Data and code availability
A GitHub repository (https://github.com/hshadman/2d_conformational_landscape_map), with https://doi.org/10.5281/zenodo.10681432, is provided, containing a Python module (PyConforMap). This module automatically generates two-dimensional conformation plots and calculates fC for any given protein/polymer if a csv file containing its Rg2 and Ree2 data is provided. The repository includes a data file for a GW reference chain of length 100. The written code can be used to check whether fC reaches a plateau with respect to the number of GW or protein/polymer snapshots, respectively, as well as generate new GW reference chains (with any number of monomers and any number of snapshots).
Author contributions
H.S., Y.W., and J.D.Z. conceived ideas and designed the project. H.S. performed all protamine simulations and all RW and SAW simulations. H.S. wrote code for and performed all GW simulations. C.E.G. performed all PEI simulations. R.L., Z.L., and H.F.-C. provided all other disordered and structured protein simulation trajectories. H.S. wrote all Python code for the analysis, with input from Y.W. and J.D.Z. H.S. prepared the initial draft of the manuscript and H.S., Y.W., and J.D.Z. revised the manuscript. H.S. prepared all the figures.
Acknowledgments
R.L. gratefully acknowledges support from NIH (GM130367).
Declaration of interests
The authors declare no competing interests.
Editor: Diego Ferreiro.
Footnotes
Supporting material can be found online at https://doi.org/10.1016/j.bpj.2024.04.010.
Supporting material
References
- 1.Luong T.D.N., Nagpal S., et al. Muñoz V. A modular approach to map out the conformational landscapes of unbound intrinsically disordered proteins. Proc. Natl. Acad. Sci. USA. 2022;119 doi: 10.1073/pnas.2113572119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Uversky V.N., Oldfield C.J., Dunker A.K. Intrinsically Disordered Proteins in Human Diseases: Introducing the D 2 Concept. Annu. Rev. Biophys. 2008;37:215–246. doi: 10.1146/annurev.biophys.37.032807.125924. [DOI] [PubMed] [Google Scholar]
- 3.Iakoucheva L.M., Brown C.J., et al. Dunker A.K. Intrinsic Disorder in Cell-signaling and Cancer-associated Proteins. J. Mol. Biol. 2002;323:573–584. doi: 10.1016/s0022-2836(02)00969-5. [DOI] [PubMed] [Google Scholar]
- 4.Korsak M., Kozyreva T. In: Intrinsically Disordered Proteins Studied by NMR Spectroscopy. Felli I.C., Pierattelli R., editors. Springer International Publishing; Cham: 2015. Beta Amyloid Hallmarks: From Intrinsically Disordered Proteins to Alzheimer’s Disease; pp. 401–421. [DOI] [PubMed] [Google Scholar]
- 5.Stanley N., Esteban-Martín S., De Fabritiis G. Progress in studying intrinsically disordered proteins with atomistic simulations. Prog. Biophys. Mol. Biol. 2015;119:47–52. doi: 10.1016/j.pbiomolbio.2015.03.003. [DOI] [PubMed] [Google Scholar]
- 6.Pietrek L.M., Stelzl L.S., Hummer G. Structural ensembles of disordered proteins from hierarchical chain growth and simulation. Curr. Opin. Struct. Biol. 2023;78 doi: 10.1016/j.sbi.2022.102501. [DOI] [PubMed] [Google Scholar]
- 7.Best R.B. Computational and theoretical advances in studies of intrinsically disordered proteins. Curr. Opin. Struct. Biol. 2017;42:147–154. doi: 10.1016/j.sbi.2017.01.006. [DOI] [PubMed] [Google Scholar]
- 8.Gong X., Zhang Y., Chen J. Advanced Sampling Methods for Multiscale Simulation of Disordered Proteins and Dynamic Interactions. Biomolecules. 2021;11:1416. doi: 10.3390/biom11101416. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hu G., Katuwawala A., et al. Kurgan L. flDPnn: Accurate intrinsic disorder prediction with putative propensities of disorder functions. Nat. Commun. 2021;12:4438. doi: 10.1038/s41467-021-24773-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Atkins J.D., Boateng S.Y., et al. McGuffin L.J. Disorder Prediction Methods, Their Applicability to Different Protein Targets and Their Usefulness for Guiding Experimental Studies. Int. J. Mol. Sci. 2015;16:19040–19054. doi: 10.3390/ijms160819040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Jumper J., Evans R., et al. Hassabis D. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596:583–589. doi: 10.1038/s41586-021-03819-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Varadi M., Anyango S., et al. Velankar S. AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res. 2022;50:D439–D444. doi: 10.1093/nar/gkab1061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ruff K.M., Pappu R.V. AlphaFold and Implications for Intrinsically Disordered Proteins. J. Mol. Biol. 2021;433 doi: 10.1016/j.jmb.2021.167208. [DOI] [PubMed] [Google Scholar]
- 14.Shrestha U.R., Smith J.C., Petridis L. Full structural ensembles of intrinsically disordered proteins from unbiased molecular dynamics simulations. Commun. Biol. 2021;4:243. doi: 10.1038/s42003-021-01759-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Song D., Luo R., Chen H.-F. The IDP-Specific Force Field ff14IDPSFF Improves the Conformer Sampling of Intrinsically Disordered Proteins. J. Chem. Inf. Model. 2017;57:1166–1178. doi: 10.1021/acs.jcim.7b00135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Song D., Liu H., et al. Chen H.F. Environment-Specific Force Field for Intrinsically Disordered and Ordered Proteins. J. Chem. Inf. Model. 2020;60:2257–2267. doi: 10.1021/acs.jcim.0c00059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hsu C.C., Buehler M.J., Tarakanova A. The Order-Disorder Continuum: Linking Predictions of Protein Structure and Disorder through Molecular Simulation. Sci. Rep. 2020;10:2068. doi: 10.1038/s41598-020-58868-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Sormanni P., Piovesan D., et al. Vendruscolo M. Simultaneous quantification of protein order and disorder. Nat. Chem. Biol. 2017;13:339–342. doi: 10.1038/nchembio.2331. [DOI] [PubMed] [Google Scholar]
- 19.Dyson H.J., Wright P.E. Intrinsically unstructured proteins and their functions. Nat. Rev. Mol. Cell Biol. 2005;6:197–208. doi: 10.1038/nrm1589. [DOI] [PubMed] [Google Scholar]
- 20.Toth-Petroczy A., Palmedo P., et al. Marks D.S. Structured States of Disordered Proteins from Genomic Sequences. Cell. 2016;167:158–170.e12. doi: 10.1016/j.cell.2016.09.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Strodel B. Energy Landscapes of Protein Aggregation and Conformation Switching in Intrinsically Disordered Proteins. J. Mol. Biol. 2021;433 doi: 10.1016/j.jmb.2021.167182. [DOI] [PubMed] [Google Scholar]
- 22.Choi U.B., Sanabria H., et al. Weninger K.R. Spontaneous Switching among Conformational Ensembles in Intrinsically Disordered Proteins. Biomolecules. 2019;9:114. doi: 10.3390/biom9030114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ziebarth J.D., Wang Y. Understanding the Protonation Behavior of Linear Polyethylenimine in Solutions through Monte Carlo Simulations. Biomacromolecules. 2010;11:29–38. doi: 10.1021/bm900842d. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Frenkel D., Smit B. 2nd ed. Academic Press; San Diego: 2002. Understanding Molecular Simulation: From Algorithms to Applications. [Google Scholar]
- 25.Hou L., Shao H., et al. Zagorski M.G. Solution NMR Studies of the Aβ(1−40) and Aβ(1−42) Peptides Establish that the Met35 Oxidation State Affects the Mechanism of Amyloid Formation. J. Am. Chem. Soc. 2004;126:1992–2005. doi: 10.1021/ja036813f. [DOI] [PubMed] [Google Scholar]
- 26.Sibille N., Huvent I., et al. Landrieu I. Structural characterization by nuclear magnetic resonance of the impact of phosphorylation in the proline-rich region of the disordered Tau protein: Phospho-Tau Structure. Proteins. 2012;80:454–462. doi: 10.1002/prot.23210. [DOI] [PubMed] [Google Scholar]
- 27.Vijay-Kumar S., Bugg C.E., Cook W.J. Structure of ubiquitin refined at 1.8Åresolution. J. Mol. Biol. 1987;194:531–544. doi: 10.1016/0022-2836(87)90679-6. [DOI] [PubMed] [Google Scholar]
- 28.Szep S., Park S., et al. Saven J.G. Structural coupling between FKBP12 and buried water. Proteins. 2009;74:603–611. doi: 10.1002/prot.22176. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Kruse S.W., Zhao R., et al. Jones D.N.M. Structure of a specific alcohol-binding site defined by the odorant binding protein LUSH from Drosophila melanogaster. Nat. Struct. Biol. 2003;10:694–700. doi: 10.1038/nsb960. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Maier J.A., Martinez C., et al. Simmerling C. ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB. J. Chem. Theor. Comput. 2015;11:3696–3713. doi: 10.1021/acs.jctc.5b00255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Gallops C.E., Yu C., et al. Wang Y. Effect of the Protonation Level and Ionic Strength on the Structure of Linear Polyethyleneimine. ACS Omega. 2019;4:7255–7264. [Google Scholar]
- 32.Joung I.S., Cheatham T.E. Determination of Alkali and Halide Monovalent Ion Parameters for Use in Explicitly Solvated Biomolecular Simulations. J. Phys. Chem. B. 2008;112:9020–9041. doi: 10.1021/jp8001614. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Shadman H., Gallops C.E., et al. Wang Y. Exploring Structures and Dynamics of Protamine Molecules through Molecular Dynamics Simulations. ACS Omega. 2022;7:42083–42095. doi: 10.1021/acsomega.2c04227. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Case D.A., Ben-Shalom I.Y., et al. Kollman P.A. Univ. Calif. San Franc; 2018. Amber 18. [Google Scholar]
- 35.Roe D.R., Cheatham T.E. PTRAJ and CPPTRAJ: Software for Processing and Analysis of Molecular Dynamics Trajectory Data. J. Chem. Theor. Comput. 2013;9:3084–3095. doi: 10.1021/ct400341p. [DOI] [PubMed] [Google Scholar]
- 36.Humphrey W., Dalke A., Schulten K. VMD: Visual molecular dynamics. J. Mol. Graph. 1996;14:33. doi: 10.1016/0263-7855(96)00018-5. [DOI] [PubMed] [Google Scholar]
- 37.Elder R.M., Jayaraman A. Coarse-Grained Simulation Studies of Effects of Polycation Architecture on Structure of the Polycation and Polycation–Polyanion Complexes. Macromolecules. 2012;45:8083–8096. [Google Scholar]
- 38.Rubinstein M., Colby R.H. Oxford University Press; Oxford ; New York: 2003. Polymer Physics. [Google Scholar]
- 39.Noguchi H., Yoshikawa K. Morphological variation in a collapsed single homopolymer chain. J. Chem. Phys. 1998;109:5070–5077. [Google Scholar]
- 40.Mattice W.L., Suter U. Wiley; New York: 1994. Conformational Theory of Large Molecules: The Rotational Isomeric State Model in Macromolecular Systems. [Google Scholar]
- 41.Cappadocia L., Lima C.D. Ubiquitin-like Protein Conjugation: Structures, Chemistry, and Mechanism. Chem. Rev. 2018;118:889–918. doi: 10.1021/acs.chemrev.6b00737. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Monia B.P., Ecker D.J., Crooke S.T. New Perspectives on the Structure and Function of Ubiquitin. Nat. Biotechnol. 1990;8:209–215. [Google Scholar]
- 43.Sloper-Mould K.E., Jemc J.C., et al. Hicke L. Distinct Functional Surface Regions on Ubiquitin. J. Biol. Chem. 2001;276:30483–30489. doi: 10.1074/jbc.M103248200. [DOI] [PubMed] [Google Scholar]
- 44.Vijay-Kumar S., Bugg C.E., et al. Cook W.J. Three-dimensional structure of ubiquitin at 2.8 A resolution. Proc. Natl. Acad. Sci. USA. 1985;82:3582–3585. doi: 10.1073/pnas.82.11.3582. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Wu J., Jones N., et al. Li S.-D. Protamine-mediated efficient transcellular and transmucosal delivery of proteins. J. Contr. Release. 2023;356:373–385. doi: 10.1016/j.jconrel.2023.03.002. [DOI] [PubMed] [Google Scholar]
- 46.Hong Y., Najafi S., et al. Hwang D.S. Hydrophobicity of arginine leads to reentrant liquid-liquid phase separation behaviors of arginine-rich proteins. Nat. Commun. 2022;13:7326. doi: 10.1038/s41467-022-35001-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Alberer M., Gnad-Vogt U., et al. Von Sonnenburg F. Safety and immunogenicity of a mRNA rabies vaccine in healthy adults: an open-label, non-randomised, prospective, first-in-human phase 1 clinical trial. Lancet. 2017;390:1511–1520. doi: 10.1016/S0140-6736(17)31665-3. [DOI] [PubMed] [Google Scholar]
- 48.Stitz L., Vogel A., et al. Petsch B. A thermostable messenger RNA based vaccine against rabies. PLoS Neglected Trop. Dis. 2017;11 doi: 10.1371/journal.pntd.0006108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Mai Y., Guo J., et al. Yang J. Intranasal delivery of cationic liposome-protamine complex mRNA vaccine elicits effective anti-tumor immunity. Cell. Immunol. 2020;354 doi: 10.1016/j.cellimm.2020.104143. [DOI] [PubMed] [Google Scholar]
- 50.Boussif O., Lezoualc’h F., et al. Behr J.P. A versatile vector for gene and oligonucleotide transfer into cells in culture and in vivo: polyethylenimine. Proc. Natl. Acad. Sci. USA. 1995;92:7297–7301. doi: 10.1073/pnas.92.16.7297. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Zhao Y., Cortes-Huerto R., et al. Rudzinski J.F. Investigating the Conformational Ensembles of Intrinsically Disordered Proteins with a Simple Physics-Based Model. J. Phys. Chem. B. 2020;124:4097–4113. doi: 10.1021/acs.jpcb.0c01949. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Förster D., Idier J., et al. Malliavin T.E. Low-resolution description of the conformational space for intrinsically disordered proteins. Sci. Rep. 2022;12 doi: 10.1038/s41598-022-21648-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Alston J.J., Ginell G.M., et al. Holehouse A.S. The Analytical Flory Random Coil Is a Simple-to-Use Reference Model for Unfolded and Disordered Proteins. J. Phys. Chem. B. 2023;127:4746–4760. doi: 10.1021/acs.jpcb.3c01619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Kumar A., Narayanan V., Sekhar A. Characterizing Post-Translational Modifications and Their Effects on Protein Conformation Using NMR Spectroscopy. Biochemistry. 2020;59:57–73. doi: 10.1021/acs.biochem.9b00827. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Yan T., Liang J., et al. Wang X. FAM222A encodes a protein which accumulates in plaques in Alzheimer’s disease. Nat. Commun. 2020;11:411. doi: 10.1038/s41467-019-13962-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
A GitHub repository (https://github.com/hshadman/2d_conformational_landscape_map), with https://doi.org/10.5281/zenodo.10681432, is provided, containing a Python module (PyConforMap). This module automatically generates two-dimensional conformation plots and calculates fC for any given protein/polymer if a csv file containing its Rg2 and Ree2 data is provided. The repository includes a data file for a GW reference chain of length 100. The written code can be used to check whether fC reaches a plateau with respect to the number of GW or protein/polymer snapshots, respectively, as well as generate new GW reference chains (with any number of monomers and any number of snapshots).










