Human pRb (UniProtKB identifier - P06400) and LTSV40 (UniProtKB identifier - P03070) are multi domain monomeric proteins. pRb contains 928 amino acids organized into three domains. The central or “pocket” domain (Pro374-Ala772) is made up of two subdomains known as the A and B boxes. These boxes, primarily composed of α -helices, contact each other through a large interface laying a binding site called the AB cleft. This site mediates the interaction between pRb and E2F (Chow and Dean, 1996). There is an alternative binding site, named the LXCXE binding cleft, that is entirely contained in box B. This site mediates the interaction between pRb and a wide range of associated proteins, all of which contain an LXCXE motif (Lee et al., 1998; Dick and Rubin, 2013).

LTSV40, made up of 708 amino acids, is composed of four folded domains: the J-domain (Met1-Asn102), the origin-binding domain (OBD) (Pro135-Pro249), the Zn-binding domain (Lys271-Gln354) and the ATPase domain (Arg357-Thr708) (An et al., 2012). The J-domain is connected to the OBD by an intrinsically disordered region (IDR) which spans from Leu103 to Asp134. Hereafter, this region will be referred to as Linker 1. Another IDR, Linker 2, comprises the sequence Gly250-Trp270 that joins the OBD with the Zn-binding domain. This last domain and the ATPase domain are linked by a short segment of just three aminoacids. The first five residues of Linker 1, Leu103, Phe104, Cys105, Ser106, and Glu107 bind to the LXCXE binding cleft of pRb. In the following, this segment will be named as the LFCSE motif of LTSV40.

The starting point to build the complex between pRb and LTSV40 was the structure reported in PDB entry 1GH6 (Kim et al., 2001). It contains the boxes A and B of the pRb pocket domain but lacks the flexible loop connecting them. This loop spans from residue Asp584 to Ser644 and is not part of any binding motif. Entry 1GH6 also contains a portion of LTSV40 ranging from Arg7 to Thr117. This segment accommodates the J-domain plus 15 of the 32 residues of Linker 1. The positions of residues in the loop between the boxes A and B of the pRb domain were predicted with Alphafold3 (Jumper et al., 2021; Varadi et al., 2024), using the known structure of pRb as a template. To complete the structure of LTSV40, we used information taken from PDB entries 1Q1S (Fontes et al., 2003) and 4GDF (Chang et al., 2013). Entry 1Q1S provides the conformation of the Asp119-Glu133 segment, almost completing the portion of Linker 1 that is absent in 1GH6. The only missing residue is Ala118. Accordingly, we used DS ViewerPro 5.0 to add Ala118 after Thr117 in structure 1GH6 and before Asp119 in structure 1Q1S, following the geometrical restrictions of the backbone atoms of Thr117 and Asp119, respectively. After that, we joined the two fragments by superimposing the backbone atoms of Ala118 of the two portions and then deleted one of the occurrences of this residue.

PDB entry 4GDF, on the other hand, contains the structure of the segment of LTSV40 that goes from Lys131 to Asp627. It encompasses part of the OBD, Linker 2, the Zn-binding domain and most of the ATPase domain. This sequence overlaps with the one in entry 1Q1S by the triad Lys131-Val132-Glu133. Therefore, to complete our model of LTSV40, we superimposed the overlapping sequences and then deleted one of them. In this way, our computational model of the pRb-LTSV40 complex comprised the pocket domain of pRb (Pro374-Ala772) and the four domains of LTSV40 (Arg7-Asp627) except for a small portion of the ATPase domain. Finally, to generate models of pRb and LTSV40 in their apo forms, we just removed pRb or LTSV40 from the PDB file of the pRb-LTSV40 complex. Figure 1 presents a graphical representation of the minimized conformations of pRb and LTSV40.

The models were prepared using the LEAP module of AMBER18 (Case et al., 2018), where they were neutralized by adding N a + ions. Water molecules were then added to fill an octahedral box extending 20 Å from the closest protein atom. To achieve a 0.15 M ionic concentration, additional N a + and C l − ions were randomly placed in the system.

Simulations of models containing LTSV40 were performed using two different force field combinations to assess the robustness of their results (Rauscher et al., 2015). In half of the simulations, we employed the ff19SB force field for proteins (Tian et al., 2019) along with the OPC water model (Izadi et al., 2014), which has been shown to improve the accuracy of MD simulations of intrinsically disordered regions compared to the widely used TIP3P model (Shabane et al., 2019). In the remaining simulations, we used the ff 99 SB disp force field for proteins along with the TIP 4 P disp water model (Piana et al., 2015; Robustelli et al., 2018), a combination whose accuracy has been recently validated by (Koder Hamid et al., 2022). Simulations of free pRb were only run with the ff19SB/OPC combination.

All simulations were conducted using the PMEMD module of AMBER18. Electrostatic interactions were computed with the Particle Mesh Ewald method, applying a 10.0 Å cutoff radius. Thus, direct-space calculations were performed for r < 10 Å, while reciprocal-space calculations handled longer-range interactions (Darden et al., 1993; Essmann et al., 1995). A 10.0 Å cutoff was also used for the rest of the non-bonded interactions. The models were initially minimized using 125,000 cycles of the steepest descent method, followed by 25,000 cycles of conjugate gradient minimization. During the simulations, the SHAKE algorithm was applied to constrain hydrogen-involving bond lengths, enabling an integration time step of 2.0 fs.

Molecular dynamics (MD) simulations were performed on the computational models of pRb, LTSV40, and the pRb-LTSV40 complex. Because the LTSV40 model includes intrinsically disordered regions and was assembled using data from various experimental sources, we anticipated its initial conformation could deviate significantly from any representative structure of the molecule in solution. Additionally, we observed that applying the heating protocol repeatedly to the same minimized structure resulted in conformations whose differences could be detected by the naked eye. With these considerations in mind, and aiming to explore the conformational space of LTSV40 as extensively as our computational resources allow, we ran 50 separate heating, equilibration and production phases for models including this protein. Instead, for simulations of isolated pRb, we run one heating and one equilibration stage, followed by 50 alternative production stages. The numerical details of the complete protocols are provided below.

The heating stages took the minimized models from 0 to 303 K. They were carried out at constant volume and lasted for 100 ns. The temperature was controlled with a Langevin thermostat using a collision frequency of 1.0 ps − 1 . The conditions were then changed from constant volume to a constant pressure of 1 atm, to allow the density to relax. This equilibration under NPT conditions lasted for 200 ns. The pressure was controlled with a Berendsen barostat with a coupling constant of 2.0 ps. Finally, the production stage for each equilibrated system was also run under NPT conditions and lasted for 150 ns. Each production stage was initiated from the last structure of the corresponding equilibration stage, but the initial velocities were randomly selected from a Maxwellian distribution at 303 K. Snapshots were taken every 0.25 ns. Consequently, for each model, we have 50 × 150 / 0.25 = 30000 samples.

We used Principal Component Analysis (PCA) (Kitao, 2022; Palma and Pierdominici-Sottile, 2023) to identify the most significant deformations of pRb in its apo form and when complexed with LTSV40. Furthermore, we used PCA as a dimensionality reduction tool to allow the subsequent application of clustering algorithms (Sittel and Stock, 2018) on the conformations Linker 1 of LTSV40, both in its apo form and when complexed with pRb. In the first case, we performed PCA on the Cartesian coordinates of the C α atoms of pRb boxes A and B but excluded the loop region connecting them, because this fragment is highly flexible and is not directly involved in any binding event. Including it in the analysis would have obscured the fluctuations we aimed to reveal. In the second case, we performed dihedral angle PCA (dPCA) on the backbone atoms from Leu103 to Asp134 (Sittel et al., 2014; Mu et al., 2005; Altis et al., 2008). In all cases, we applied the algorithm to a fictitious trajectory formed by concatenating the 50 independent simulations of each model. This strategy systematically improves the consistency of PCA results, as it mitigates the effect of spurious correlations (Cossio-Pérez et al., 2017; Palma and Pierdominici-Sottile, 2023).

We performed a cluster analysis to identify the most representative conformations of Linker 1 of LTSV40. The metric we employed to that end was the Euclidean distance between the samples, in a reduced-dimensionality (RD) space determined with dPCA. Using this metrics is slightly more elaborated than the most widely employed RMSD, but has significant advantages. RMSD requires fitting all the structures to a reference structure to eliminate the effect of global rotation and translation. Unfortunately, this procedure cannot be unambiguously performed and problems become particularly serious for molecules presenting large deformations, such as those analysed here (Palma and Pierdominici-Sottile, 2023). On top of that, transforming from the high dimensional space of Cartesian coordinates to the 1D space of RMSD can introduce projection errors that tend to put structures differing in important features into the same cluster (Sittel and Stock, 2018). Because of these considerations, we decided to also implement a metric based on distances in the RD space of dPCA. Since this algorithm employs internal coordinates, it does not require any adjustment to a reference structure. Besides, in principle, the dimensions of the space to be used can be set as high as desired, aiming to include all directions with significant eigenvalues. In practice, however, the sample size, N , limits the number of dimensions that can be effectively used, because the larger the dimension of the space, d , the poorer the statistics. If the number of subdivisions per dimension is n s , the average number of samples per bin is n ave = N / n s d . In our case, N = 30000 . Therefore, requiring n ave = 5 and n s between 5 and 10, we obtained that appropriate values of d are between 4 and 5. Thus, we set d = 5 aiming to include the maximum number of features as our dataset allows.

To perform the cluster analysis, we utilized the DBSCAN algorithm to estimate the relevance of the regions in the configurational space that were explored during the simulations. Unlike the more commonly used k-means algorithm, DBSCAN is deterministic and does not require prior knowledge of the number of clusters in the dataset. Instead, it requires the setting of two parameters: ϵ , which defines the maximum distance at which two points are considered part of the same cluster, and N min , the minimum number of points needed to establish a dense region (referred to as a core point).

To determine both parameters, we calculated the distances from each point to its k -th nearest neighbour for various values of k . These distances were then sorted in descending order. The appropriate value of k was identified by examining where the curves for k and k + 1 converge closely. The value of ϵ was chosen from the distance at which these curves begin to level off. Supplementary Figure S1 in the Supplementary Material section illustrates the curves derived from our data. Based on this analysis, we set k = 20 and ϵ = 0.8 .

Visual inspection of the LTSV40 configurations sampled from the MD simulations showed that an important parameter to characterize the state of this protein was the collinearity of the backbone of Linker 1. A quantitative assessment of this characteristic can be obtained by comparing two distances. One of them is the distance between the first and last C α of the segment under analysis, D n 0 , n 0 + m . The other is the sum of the distances between consecutive C α atoms in the segment, D sum = ∑ i = n 0 n 0 + m − 1 d i,i + 1 . (1)

In Equation 1, the segment under analysis starts at residue n 0 and ends at n 0 + m , while d i,i + 1 represents the distance between the C α atoms of residues i and i + 1 . For the analysis presented in the next section, the value of m was varied between 5 and 13. As the difference D iff = D n 0 , n 0 + m − D sum equals zero if all C α s are perfectly aligned, it is reasonable to consider a segment adopts a nearly collinear arrangement if D iff is below a pre-established threshold. However, the value of that threshold depends on the length of the segment, m . To quantify this dependence, we calculated D iff on a set of collinear structures of different lengths. The set of structures used for this purpose is available in the Supplementary Material section. A graph of D iff computed from this set, plotted as a function of m , showed that it can be fitted to the following straight line (see Supplementary Figure S2 at the SM): D iff m = 1.13 ∗ m − 2.37 . (2)

We applied Equation 2 to the residues of Linker 1 located on the C-terminal side of the LTSV40 LFCSE motif. This involved varying n 0 from 108 to 134. For each value of n 0 , we counted the number of snapshots in which the residue was part of a collinear segment formed by at least five residues (i.e., m ≥ 5 ). This allowed us to determine the probability of finding Linker 1 residues in extended segments containing five or more amino acids. It is important to note that we did not include the amino acids of the LFCSE motif in this analysis, as they necessarily adopt an almost collinear conformation to fit into the LXCXE binding cleft.

For the pRb-LTSV40 complex we measured the numbers of contacts between each residue of Linker 1, located on the C-terminal side of the LFCSE motif, and any residue of pRb pocket domain. To this end, we considered a contact existed when the intermolecular distance between any pair of atoms of the regions under consideration was shorter than 4.0 Å. This analysis was performed with module CPPTRAJ of AMBER18.

We used the Shortest Path Map (SPM) tool, available on the SPM server (Casadevall et al., 2024), to identify key residues influencing the structure and dynamics of Linker 1 in the pRb-LTSV40 complex. This analysis aimed to uncover critical residues involved in communication pathways between the molecular partners. To streamline computations, we focused on the C α atoms of residues Tyr68 to Lys131 in LTSV40 and the AB box of pRb, rather than the entire proteins. Distance and correlation matrices were computed using CPPTRAJ. An edge between nodes was established when their distance was below 6.0 Å, with a visualization/significance threshold set at 0.3.

To evaluate the stabilizing/destabilizing effects of pRb residues that bear the largest number of contacts with LTSV40 Linker 1 in the pRb-LTSV40 complex, we performed an alanine scanning calculation using the Molecular Mechanics-Generalized Born Surface Area (MM-GBSA) method (Genheden and Ryde, 2015). For each model of the pRb-LTSV40 complex, the analysis was conducted using snapshots taken at 1, 50, 100, and 150 ns of the production phase of each trajectory. Thus, with 50 trajectories per model, each MM-GBSA computation was based on 200 snapshots.

The protocol we followed computes the interaction energy and solvation free energy for both the complex and the isolated proteins, and the results are averaged to estimate the binding free energy. However, we did not include the entropic contribution in this estimation because the available procedure, which relies on normal mode analysis, is not suitable for proteins with disordered regions such as LTSV40. As a result, the computed Δ Δ G values provide a rough estimation of the stabilizing or destabilizing effects of the analysed residues; however, they cannot be directly compared with experimental data, as actual free energy values are required for such comparisons.

Our analysis of the pRb-LTSV40 complex mostly focused on the conformational ensemble of the viral protein, specifically examining the structures of Linker 1. As described in Section 2.4, we performed a dPCA analysis on this segment using the 30,000 structures obtained from the production phase of the MD simulations. The snapshots were then projected onto the first five dPCA eigenvectors, and the projections were subsequently used for DBSCAN cluster analysis. Employing the parameters outlined in Section 2.5, the algorithm identified 12 clusters that together account for 97.7% of the entire population of simulations done with the ff19SB/OPC force fields. The remaining 2.3% were classified as outliers. Using the same DBSCAN parameters, 9 clusters were identified in simulations employing the ff 99 SB disp / T I P 4 P disp force fields, representing 98.2% of the total population. The most representative structure of the main cluster, determined for each force field combination, is shown in Figure 2.

A simple visual inspection reveals a noticeable characteristic in the most representative conformations of Linker 1 in the pRb-LTSV40 complex (see Figure 2). The conserved SSDDE pentapeptide and its neighbouring regions are situated close to a corner of the pRb AB cleft. This corner features a pair of loops: one of them connects α 6 with α 7 in box A while the other connects α 15 with α 16 in box B. To quantify the interactions between these regions we determined the number of contacts between each residue of Linker 1 (excluding the LFCSE motif) and any residue of pRb, averaged over the whole set of snapshots collected from the production phase. The results presented in Figure 3 show that, for both force field combinations, residues within the Glu108–Glu115 sequence typically form two or more contacts with pRb. This outcome confirms that the characteristic observed in the representative structures by visual inspection also describes well what is found across the entire conformational ensemble. The average number of contacts per residue is largely consistent between the two force field combinations, except for Ser111. In simulations using ff19SB/OPC, Ser111 forms on average seven contacts, mostly with Lys740, but also with Arg741 and Asp750 of pRb. However, in ff 99 SB disp / T I P 4 P disp simulations, it establishes only two. This difference arises from variations in the preferred rotameric states of the Ser111 side chain. Importantly, it does not affect the key finding of the simulations, related to the predominant conformations adopted by Linker 1 in the complex, as detailed below.

The pRb-Linker 1 interactions discussed in the previous paragraph influence the conformations the linker can adopt within the complex, as evidenced by the analysis in Figure 4. The curves in this figure represent the probability of each residue being part of a linear segment of at least five residues. In the pRb-LTSV40 complex (red curve), residues Glu108 to Asp113 exhibit a high likelihood of forming such segments, with probabilities exceeding 0.50 in simulations performed with both force field combinations. Notably, in ff19SB/OPC simulations, this probability surpasses 0.85 for the first four residues, while in ff 99 SB disp / T I P 4 P disp simulations, it extends to the first five. In contrast, in the free form of LTSV40, these probabilities drop below 0.40 for the two sets of simulations, indicating that interaction with pRb is necessary to maintain the linear conformations. Due to the typical extended conformation of Linker 1 residues up to Asp113, the distance between Leu103, the first residue of the LFCSE motif, and Asp113, is larger than the distance between the LXCXE and AB clefts of pRb (which is nearly 30.0 Å). Supplementary Figure S10, in the SM section, shows the result of the SPM analysis for the pRb-LTSV40 complex. In reveals that residue-residue communication is strong within the structured pRb AB box and somewhat weaker along the extended LTSV40 Linker 1. Interestingly, the communication between the two fragments occurs between Asn757 of the pRb LXCXE binding cleft and Phe104 of the LTSV40 LFCSE binding motif.

Figure 4 also shows that, for both force field combinations, the probability of residues between Glu115 and Asn121 being part of a linear segment is low in both, isolated LTSV40 and LTSV40 in complex with pRb. This probability then increases, likely due to the presence of two consecutive prolines at positions 125 and 126. Additionally, the stretch of five consecutive basic residues from positions 127 to 131, which constitute the protein’s nuclear localization signal, favours extended conformations over linear ones. To complete the analysis, we also identified the pRb residues that stablish the larger number of contacts with Linker 1, excluding those that interact with the LFCSE motif. For this analysis we considered side chain heavy atoms. The residues we spotted are presented in Table 1 along with their average number of contacts with Linker 1 and their stabilizing or destabilizing effect for the binding, as measured by the Δ Δ G determined with the MM-GBSA alanine scanning algorithm. Supplementary Figure S11, at the SM section, shows an interaction diagram for pRb and LTSV40. It was created with the online version of PDBsum (Laskowski et al., 2018), using a representative structure of the complex sampled during the simulations.

As outlined in Section 2.1, Linker 1 and Linker 2 of LTSV40 are long IDRs that connect the J-domain with the OBD, and the OBD with the Zn-binding domain, respectively. Examining the whole set of conformations visited by free LTSV40 in our MD simulations, we observed that the two linkers typically adopt relatively curved conformations, positioning the J-domain and the Zn-binding domain close to the OBD. This feature is illustrated in Figure 5, which displays the representative structures of the most populated clusters in the conformational ensemble of free LTSV40, according to each of the force field combinations used in this work. These structures should be compared with those of Figure 2.

Another notable difference between the behaviour of free and complexed LTSV40 lies in the mobility of the LFCSE residues and those on its C-terminal side. This effect is illustrated in Supplementary Figure S12, which shows the difference in Root Mean Square Fluctuations (RMSF) between the free and complexed forms of the protein. As expected, fluctuations are significantly larger in the free form. The results obtained with both force field combinations are consistent.

Unlike the intrinsically disordered regions of LTSV40, the residues in the AB box of pRb exhibit minimal conformational changes when transitioning from the free to the complexed form. This can be observed in Supplementary Figure S12, which presents the differences in RMSF between the two states. Additionally, Figure 6 highlights the most notable deformations in both the apo form (panel A) and the complexed form (panel B). Specifically, the figure depicts the movements generated by a linear combination of the first two principal component analysis (PCA) eigenvectors, each scaled by its corresponding eigenvalue. Animations of these vectors are available in the Supplementary Material.

The analysis reveals two key dynamical and structural characteristics of the pRb pocket domain. First, the AB cleft, highlighted in yellow in Figure 6, is a rigid structure that retains its shape and rigidity upon LTSV40 binding. Second, in both the free form of pRb and the complex with LTSV40, the most significant deformations occur in the loop connecting α 6 and α 7 of the A box, which shifts toward the loop linking α 15 and α 16 of the B box. In the apo form of pRb, this second loop also exhibits high mobility.

Binding assays demonstrated that pRb can be complexed with a decapeptide of HPV E7 in its LXCXE cleft and a small portion of E2F in its AB cleft. This finding indicates that binding to the LXCXE and AB clefts can occur independently (Lee et al., 1998; Xiao et al., 2003). Additionally, experiments revealed that small peptides containing the LXCXE motif cannot inhibit E2F binding (Wu et al., 1993; Xiao et al., 2003). However, when full-length viral proteins or a significant portion of them are used in the assays, complexation with E2F does not occur (Wu et al., 1993; Xiao et al., 2003). This indicates that the inhibition of E2F binding is not directly due to the interaction of pRb with the LXCXE motif or its neighbouring segments but that more distant regions of the viral proteins are involved. Also, this raises the question of whether there is a mechanism common to proteins with a LXCXE motif and disordered regions towards their C-terminal side.

In this work, we present an analysis of the interaction between the pocket domain of pRb and the full viral protein LTSV40, to reveal information that could not be gained from experiments yet. To this end, we used MD to characterize the conformational ensembles of the isolated proteins as well as the complex between them. When bound to pRb, the LFCSE motif of LTSV40 adopts a nearly linear conformation to fit into the LXCXE binding cleft. Our simulations revealed that, in the complex, the adjacent residues Glu108–Asp113 also tend to adopt an extended conformation. This behaviour contrasts sharply with the free form of LTSV40, where the entire Leu103–Asp113 segment typically assumes a curved and more compact shape. We thus conclude that the linear conformations observed in the complex are caused by binding to pRb.

As a consequence of the linear shape, typical structures of the pRb-LTSV40 conformational ensemble have the C-terminal side of the LFCSE motif close to one corner of the AB cleft (see Figure 2). This corner should be occupied by E2F if the whole transcription factor was present Lee et al. (2002). In other words, due to the extended conformations adopted by Linker 1, the C-terminal side of the LFCSE motif blocks regions that would be occupied by E2F in the E2F-pRb complex. This findings would explain why small fragments of LTSV40 and E2F can simultaneously bind to the LXCXE and AB clefts of pRb, respectively, but simultaneous binding of longer constructs is not feasible. Supplementary Figure S13 of the SM section provides a graphical representation of the pRb-E2F-1 complex, as predicted by Alphafold3 using PDB structure 1O9K as a template. The template comprises the A and B boxes of pRB plus residues 409 to 426, which form the core of the binding region of the transcription factor. Although the conformation of the disordered region of E2F-1 in the figure is uncertain, according to Alphafold3 estimates, it is clear that the residues immediately adjacent to the fragment present in the PDB structure are placed in the same region as that occupied by part of LTSV40 Linker 1 in our simulations of the pRb-LTSV40 complex.

It is well established that the LXCXE motif is the minimum sequence required for binding the LXCXE cleft of pRb (Kim et al., 2001; Lee et al., 1998). However, viral oncoproteins have acidic residues on the C-terminal side of the LXCXE motif that also contribute to pRb binding (Lee et al., 1998). In LTSV40, this region contains the conserved pentapeptide SSDDE (Barbosa et al., 1990). According to our simulations, this motif mainly interacts with Lys722, Lys729, Gln736, Lys740 and Arg741 of the pRb B box. The two force field combinations employed in this study agree to indicate that Lys740 has a significant stabilizing effect in the complex, while the ff 99 SB disp / T I P 4 P disp also indicates an important involvement of Lys729 and Gln736. Site-directed mutagenesis studies have found that changing by alanine all Lysines of Table 1, along with Lys720 and Lys765, disrupted the pRb-LTSV40 complex (Brown and Gallie, 2002). The authors indicated that the interaction between the acidic portion of Linker 1 and the basic patch of pRb could play a role in the interaction between both proteins, but argued that the interaction would be labile due to the flexibility observed for this segment in the crystal structure of the complex (Brown and Gallie, 2002). Unfortunately, the study did not provide experimental data regarding the individual role of each lysine, and the crystal structure employed featured LTSV40 truncated at position Thr117. Our results indicate that, when the full LTSV40 sequence is considered, the interactions of these Lysines, along with those of Gln736 and Arg741, are persistent. These interactions may play a role in positioning the remainder of Linker 1 and the central domain of LTSV40 close to the pRb AB cleft, thereby preventing E2F binding. Thus, a simple competition between LTSV40 and E2F would explain the disruption of the pRb-E2F pathway due to the action of the viral oncoprotein.

Brown et. al. also refer to the possibility that the attachment of oncoproteins to the LXCXE binding cleft could lead to an allosteric conformational change in the AB cleft, potentially preventing E2F attachment. However, the authors suggested that this mechanism is unlikely due to the rigidity observed for these sites in the crystal structures. Supporting this suggestion, our examination of the primary deformations of pRb revealed that the two binding clefts remain relatively rigid and are not affected by the attachment of LTSV40. Instead, two flexible loops of pRb, one located between α 6- α 7 of the A box and the other linking α 15- α 16 of the B box, participate in a large concerted hinge-bending motion which is probably related to the E2F binding. motion probably related to the E2F binding.

We would like to conclude this section by clearly outlining the scope and limitations of our study. Molecular dynamics simulations have proven to be a highly valuable tool for studying the behaviour of proteins and other biomolecules, providing support and complementing experimental studies. However, they also come with significant limitations that must be taken into account when interpreting their results. Among these limitations are the shortcomings of additive force fields, which are used in most studies, including ours, and the challenges associated with adequately sampling the conformational space of most biomolecules. In this work, the issues arising from force field deficiencies were addressed by employing two alternative force field combinations, both of which have been reported to perform well for disordered proteins. To tackle the sampling problem, we conducted 50 alternating heating and equilibration stages for each model containing LTSV40. While these precautions increase the confidence in our results, they do not provide certainty, particularly because there are no alternative experimental data against which to validate the simulations predictions. Therefore, the results presented here should be regarded as a well-founded hypothesis rather than a definitive conclusion.

Having made this caveat, we must nevertheless point out that certain aspects of biomolecular behaviour are so robustly embedded in their sequence and structure that they consistently emerge in MD simulations, despite the limitations of the computational models employed. A notable example of this is the folded structure of Trp-cage, which was accurately predicted de novo over 20 years ago, even with the limitations of the force fields available at the time. The consistency of the ff19SB/OPC and ff 99 SB disp / T I P 4 P disp force field combinations in predicting an extended conformation for residues located on the C-terminal side of the LXCXE motif of LTSV40, in the pRb-LTSV40 complex, suggests that this could be a robust characteristic of these molecules. However, further validation by experimental means is required to confirm that.

HPV E7 is a viral oncoprotein consisting of 98 amino acids that exhibits significant similarities to LTSV40. Functionally, both proteins bind to pRb to disrupt the cell cycle regulation. Structurally, both have an LXCXE motif that interacts with the LXCXE binding cleft of pRb, as well as a conserved mostly acidic pentapeptide located on the C-terminal side of that motif (Chen and Paucha, 1990; Christensen and Imperiale, 1995). Additionally, in the two cases, the pentapeptide is located within an intrinsically disordered region (Pelka et al., 2008). These characteristics are also common to most polyomaviruses. Supplementary Figure S14, available at the SM section, presents an alignment of representative polyomavirus large antigens with the prediction of IDRs, highlighting the LXCXE motif position. These similarities suggest a common evolutionary background that has driven them to develop comparable strategies for manipulating the host cell machinery and promote tumorigenesis (Barbosa et al., 1990).

In HPV E7, the LXCXE motif comprises the sequence Leu22-Glu26, while the segment of E2F that binds pRb is Asp410-Asp427 (Lee et al., 2002). Similar to the LTSV40-E2F case, short sequences of HPV E7 and E2F (HPV E 7 20 − 29 and E 2 F 409−426 ), can attach simultaneously to pRb (Lee et al., 1998). Even if somewhat longer sequences, E 2 F 380−−437 and HPV E7 (HPV E 7 17 − − 98 ) are employed, the binding affinity to pRb of the E2F construct is the same as that of the shorter E 2 F 409−426 . However, if an even longer E2F construct is considered, E 2 F 243−−437 , the affinity for pRb is sensibly diminished (Xiao et al., 2003). From these observations, the authors of the study concluded that regions on the C-terminal side of the HPV E7 LXCXE motif interact with the AB pocket of pRb, interfering in the binding of E2F (Xiao et al., 2003). This is in line with Wu et. al. that also suggested that aminoacids in HPV E7 located at the carboxy terminus of the LXCXE motif are important for modulating E2F activity (Wu et al., 1993). In the same line, Huang et. al. determined that HPV E 7 20 − 29 can not inhibit the binding of the complete form of E2F to pRb, but when HPV E 7 20 − 98 is considered, the interaction pRb-E2F is completely inhibited (Huang et al., 1993). These authors suggested that the intact HPV E7 protein may occlude the E2F-binding site on pRb through steric hindrance or by eliciting a conformational change on pRb upon binding.

Our findings regarding the position of the LTSV40 portion on the C-terminal side of the LXCXE motif, as well as the rigidity of the pRb clefts, align perfectly with the above-mentioned experiments that explore the mechanism by which HPV E7 blocks E2F binding to pRb. These similarities strongly suggest that both viral oncoproteins may employ the same blocking strategy. While there are no experimental studies akin to those conducted with HPV E7 for other polyomaviruses, sequence similarities and predictions of their intrinsically disordered regions suggest that they might utilize related strategies.

Hypophosphorylated pRb binds E2Fs, thereby inhibiting cell proliferation. However, in response to growth signals, pRb becomes phosphorylated and loses its ability to bind E2Fs. The phosphorylation sites of pRb are located on the C-terminal side of the pocket domain (Zhou et al., 2022; Zarkowska and Mittnacht, 1997; Ewens et al., 2017) and the region connecting that domain with the phosphorylation sites contains IDRs. Additionally, it is well established that the phosphorylated region interacts with basic residues in the B-box of the pocket domain, thereby inhibiting the formation of the E2F-pRb complex (Brown and Gallie, 2002; Kim et al., 2001; Lee et al., 1998). On the other hand, it has been shown that phosphorylated pRb cannot bind LTSV40 (Knudsen and Wang, 1996), and studies of the interaction between pRb and LXCXE-containing peptides suggest that pRb phosphorylation induces the release of these peptides via a competitive mechanism (Lee et al., 1998).

Although E2Fs and LXCXE-containing peptides/proteins bind to distinct regions of pRb, their release may be governed by a shared mechanism triggered by pRb phosphorylation. Previous experimental studies on the E2F/pRb complex suggest that phosphates attached to phosphorylation sites of pRb interact with the B-box, disrupting E2F binding and/or promoting its release (Brown and Gallie, 2002). We recall that the phosphorylation site and the B-box are connected by a lengthy IDR. On the other hand, our MD simulations revealed that in the LTSV40/pRb complex, the IDR in Linker 1 of LTSV40 sterically occludes the E2F-binding cleft of pRb. Integrating these findings suggests to extend Brown and Gallie’s hypothesis by proposing a model in which phosphorylation-dependent structural rearrangements of pRb, involving the IDR connecting the phosphorylation site to the B-box, block access to both the E2F and LXCXE binding sites, thereby preventing complex formation in any of them.