Resilience of BST-2/Tetherin structure to single amino acid substitutions

Production of tetherin mutants

Tetherin mutants were generated from the membrane bound model of tetherin described by Ozcan & Berndsen (2017) using the list of human mutations found in Ensembl as of August 2018 (Aken et al., 2017). We wrote a YASARA macro to individually swap amino acids and minimally energy minimize the resulting structure. All models were equilibrated for 10 to 20 nanoseconds in explicit water solvent and a DOPC bilayer using an AMBER14 forcefield within YASARA structure (Case et al., 2014; Krieger & Vriend, 2015; Krieger et al., 2009). Simulation conditions were 298 K, pH 7.4, 0.9% mass fraction NaCl, at a density of 0.997 g/mL, with a 12 Angstrom non-bonded interaction cutoff, and a time step of 2.5 femtoseconds.

Analysis of models

Mutant simulations were analyzed using existing macros within YASARA to calculate simulation energies, root mean square fluctuation (RMSF), and the secondary structure over time. These macros were run on the complete data set. Data for each mutation were summarized using Rmarkdown. Root mean square deviation values were calculated using the Align function in YASARA structure. Global analysis of data was performed in R using the components within the tidyverse package and the cluster package (Ross, Wickham & Robinson, 2017; Wickham, 2016; Maechler et al., 2013). Ridgeline plots and some figures were made using the ggridges and cowplot packages of Claus Wilke (Wilke, 2018a; Wilke, 2018b).

Human missense mutations are absent from beginning of coiled-coil

The structure of tetherin has been determined however the connection between the crucial functions of tetherin and this structure is less clear (Swiecki et al., 2011; Hinz et al., 2010; Schubert et al., 2010; Yang et al., 2010). The primary sequence of tetherin is not well-conserved across species, although the general organization appears to be (Heusinger et al., 2015; Blanco-Melo, Venkatesh & Bieniasz, 2016). Areas of structural or functional importance generally have a conserved protein sequence. Given the lack of apparent sequence conservation between species, it is difficult to predict the important structural features of tetherin. We decided to compare sequences within the human species to identify areas of amino acid stability and therefore relative importance. We retrieved variation data from the Ensembl database containing the missense mutations within the ectodomain of tetherin (Table S1). We focused on the mutations within the helical portion of the ectodomain because there is better, high quality structural information on this region. Mapping the amino acid locations of SNPs showed that there was a lower density of mutations between residues 93 and 117 (Fig. 2A). We note that synonymous mutations, which do not affect the amino acid sequence, still occur in this region, but amino acid changes appear to be lacking (Fig. 2A). Although the data set is limited in scope, the information suggests that amino acids in this region, which includes one disulfide and the beginning of the coiled-coil region of the ectodomain, are important for tetherin structure or function. In addition to a general position search, we mapped the mutations on the 3MQB structure and observed that 66 of 78 mutations appear to occur on the exterior of the protein (“outside”) rather than at the interface of the tetherin dimer (“inside”) (Fig. 2B) (Yang et al., 2010). Moreover, the frequency of many of these outside mutations is higher than for inside mutations (Table S1). The simplest interpretation of this distribution of mutations is that the solvent and membrane facing surfaces of the protein are more tolerant of single amino acid changes. This distribution supports the previously demonstrated role of dimerization in viral tethering and stability, although dimerization is not entirely essential for tethering of all viruses (Andrew et al., 2009; Sakuma, Sakurai & Yasuda, 2009; Schubert et al., 2010; Welbourn et al., 2015; Du Pont et al., 2016). The 78 mutations occur at 59 unique sites within the protein and some positions have several polymorphisms (summarized in Table S1). We also observed that the areas of highest mutation density were in the N-terminal portion of the ectodomain, which has been suggested to be more flexible and is stabilized by the inter-molecular disulfides (Welbourn et al., 2015; Du Pont et al., 2016), or at the C-terminal end of the coiled-coil region, which in structural simulations unwinds relatively easily as it contains a non-canonical heptad (Andrew et al., 2012; Du Pont et al., 2016; Ozcan & Berndsen, 2017). Therefore the locations of SNPs align well with previous functional and structural data on tetherin. However, the potential effects of these mutations on tetherin structure are unclear.

Missense mutations do not alter overall structure

To determine the association between structure and function in tetherin mutants, we computationally made point mutations in the previously described full-length model of tetherin at positions between 49 and 163, where the structure of tetherin appears mostly helical (Ozcan & Berndsen, 2017). All models equilibrated within 10 to 20 ns of simulation. We then compared the root mean square deviation, root mean square fluctuation, and secondary structure for all mutations to a similar simulation on a model lacking mutations to identify structural and dynamic changes due to the mutation. The root mean square deviation (RMSD) is a measure of the differences in the 3-D position of one or more elements of the ectodomain. We calculated the RMSD of the average structure from each mutant simulation from alignment of the ectodomain alone with a wild-type tetherin that had also been equilibrated. The data were then plotted as RMSD vs. amino acid number as shown in Fig. 3. Any large RMSD values would suggest a strong effect of the mutation on the overall tetherin structure. We observed that there were distinctly large RMSD values across the ectodomain mutations and no single mutation disrupted the entire coiled-coil (Fig. 3). These data suggest that mutations in tetherin do not cause major changes to the helical structure.

Mutations alter local tetherin structure

Single changes to the amino acid sequence may not cause drastic changes to protein structure. Therefore, we next analyzed our simulations for local changes in helical structure. Based on the distance between hydrogen bond donors and acceptors, amino acid positions were binned into types of secondary structure at 400 steps throughout the simulation. Representative plots of secondary structure across the simulation are shown in Fig. 4. In addition, we simulated a point mutation known to affect the structure of tetherin, Gly118Cys, as a positive control for structural disruptions (Welbourn et al., 2015). The wild-type simulation shows a stable helical structure which is matched in behavior by the Asp129Glu and Asp103Asn mutations. The postive control, Gly118Cys, shows a chunk of 4 amino acids which are in a coiled structure, but the effects are limited to this region. Similarly, mutations Arg64Pro, Ala100Pro and Ile120Phe showed distinct regions where the secondary structure is altered compared to the wild-type as shown in Fig. 4A. Proline is well-known to be disruptive to alpha helices because it is unable to form the hydrogen bonds characteristic of this secondary structure (Pace & Scholtz, 1998). Position 64 within tetherin is the most diversely mutated position in the data that we analyzed with 5 SNPs but only the Arg64Pro had any effects on structure (Table S1). Phenylalanine is aromatic and occupies a larger volume than isoleucine suggesting the local secondary structure changes are due to the alterations in the interface between the two helices. We note that none of the changes in secondary structure spread along the coil. Instead, cracks in the helices stayed localized to a single area within the ectodomain, supporting the biochemical and computational findings of Du Pont et al. (2016). We next analyzed where the changes in structure were localized and plotted the number of times an amino acid adopted a different environment versus position (Fig. 4B). The effects on secondary structure appear primarily at the ends of the coil, which are known to be unstructured, rather than proximal to the site of substitution except for a few mutations (Fig. 4B) (Swiecki et al., 2011). These data suggest that the potential disruptive effects of most mutations are minimized by the coiled-coil structure of tetherin.

While most mutations had minimal effect on the helical structure of tetherin, three mutations Arg64Pro, Ala100Pro, and Ile120Phe did disrupt the helices. Arg64Pro and Ala100Pro induced alterations in secondary structure near to the site of the polymorphism. In neither of these instances do the alterations in structure include more than two consecutive amino acids. Ile120Phe however induces a disruption in secondary structure that is not directly adjacent to the mutation site and is maintained throughout the simulation. For Ala100Pro mutation this disruption in secondary structure was reproduced while for Ile120Phe, the disruption in secondary structure was inconsistent (Fig. S1). Visualization of the energy minimum structure from the simulation for both the wild-type and Ile120Phe structure shows that Ile120 is an interior amino acid and the side chains could form a van der Waals contact (Fig. 5A). However, in the structure of the Phe substitution, the side chain is displaced from the interface (Fig. 5A). Further comparison of the distance between the tetherin monomers and at sites along the helix shows that the Phe substitution noticeably changes the distribution of measurements on either side of the mutation over a region extending at least 56 amino acids in length (Fig. 5B). The alteration in the amino acid distances appeared consistently in replicate simulations of this mutation despite the lack of effect on secondary structure in some simulations (Fig. S2). In comparison, the Ala100Pro mutation does not alter distribution of distances except for slight differences adjacent to the mutation site (Fig. 5C). Observation of the individual measurement sites over time show that amino acids in the region of 105 to 123 quickly adjust to a new distance followed by slower adjustments at positions 81, 84 and 148 (Fig. S3). This behavior is not observed in the wild-type simulations (Fig. S3). These data suggest that the structure of tetherin can tolerate many changes in the sequence and that for most changes that disrupt the helical structure there is limited spreading of breaks.

Effect of mutations on ectodomain dynamics

Protein structure is not just based on the static arrangement of amino acids, but the ensemble of structures that the protein can adopt, which is sometimes referred to as protein dynamics. To quantify dynamics, we analyzed the root mean square fluctuation (RMSF), which describes the amount of movement each amino acid undergoes during the simulation. We then correlated the per residue RMSF values between each mutation simulation and the wild-type simulation. We observed moderate correlations at most positions but wanted additional methods to analyze the shape of the cluster. Using the central point in the cluster would allow for easier visualization of how each mutation affects the global dynamics of the system and allow for easier comparison of the effects of mutations. Therefore, we next calculated the mean and medoid centers of the RMSF clusters. The data for the medoid and mean cluster centers in the mutants correlated well suggesting that our data are normally distributed (Fig. 6). We plotted all the data determined by the k-means method at each amino acid position to determine if there were any trends or clustering. As seen in the right pane of Fig. 6, mutations at the ends of the ectodomain tended to have larger k-mean values suggesting larger RMSF values and more dynamics in the simulation. Mutations on the interface of the two helices did not increase dynamics relative to the outside mutations. These data are consistent with our previous finding that disruptions to secondary structure occurred preferentially toward the termini of the ectodomain and that the ectodomain can tolerate some changes to the structure.

Comparison with Tetherin mutations with known functional defects

Given the general lack of effect of point mutations on the structure of tetherin, we next looked at known non-functional mutations described previously in the literature (Table S2). We have shown that cysteine mutations or truncations that alter the orientation of the helices resulting in a poor interface are non-functional (Welbourn et al., 2015; Andrew et al., 2012). However, others have shown multiple alanine or charge swap mutations can also reduce viral tethering (Yang et al., 2010; Hammonds et al., 2012). Yang and coworkers proposed that swapping the charge of amino acids in the N-terminus altered the interaction of tetherin with the lipid membrane. In Fig. 7A, we suggest that this is unlikely because the mutation sites are not on a single face of the coiled-coil and not all sites interact with the membrane in our model. However, we cannot rule out a mechanism where the orientation of the helices relative to the membrane are maintained via interactions between these amino acids and the lipid heads. Simulations show a relatively stable distance between the coils with few disruptions outside of the sites of mutation (Figs. 7B–7C).

We next analyzed simulations containing the tetra-alanine and tetra-serine mutations of tetherin by Hammonds et al. (2012). As with the charge swap mutations, the tetra-alanine mutations show essentially no structural changes (Fig. 8). In contrast, the tetra-serine substitution showed large but localized changes in structure as assessed by distance measurements (Fig. 8C). Given that the tetra-serine mutation sites are all at the interface of the coiled-coil, these data further link alterations of the coiled-coil interface to defects in function (Welbourn et al., 2015; Andrew et al., 2012; Hammonds et al., 2012; Yang et al., 2010; Du Pont et al., 2016). Additionally, these results match up to our results with the single point mutations showing limited structural effects induced by mutation.