Differences and commonalities in plasma membrane recruitment of the two morphogenetically distinct retroviruses HIV-1 and MMTV

Retroviral Gag polyproteins are targeted to the inner leaflet of the plasma membrane through their N-terminal matrix (MA) domain. Because retroviruses of different morphogenetic types assemble their immature particles in distinct regions of the host cell, the mechanism of MA-mediated plasma membrane targeting differs among distinct retroviral morphogenetic types. Here, we focused on possible mechanistic differences of the MA-mediated plasma membrane targeting of the B-type mouse mammary tumor virus (MMTV) and C-type HIV-1, which assemble in the cytoplasm and at the plasma membrane, respectively. Molecular dynamics simulations, together with surface mapping, indicated that, similarly to HIV-1, MMTV uses a myristic switch to anchor the MA to the membrane and electrostatically interacts with phosphatidylinositol 4,5-bisphosphate to stabilize MA orientation. We observed that the affinity of MMTV MA to the membrane is lower than that of HIV-1 MA, possibly related to their different topologies and the number of basic residues in the highly basic MA region. The latter probably reflects the requirement of C-type retroviruses for tighter membrane binding, essential for assembly, unlike for D/B-type retroviruses, which assemble in the cytoplasm. A comparison of the membrane topology of the HIV-1 MA, using the surface-mapping method and molecular dynamics simulations, revealed that the residues at the HIV-1 MA C terminus help stabilize protein–protein interactions within the HIV-1 MA lattice at the plasma membrane. In summary, HIV-1 and MMTV share common features such as membrane binding of the MA via hydrophobic interactions and exhibit several differences, including lower membrane affinity of MMTV MA.

Retroviral Gag polyproteins are targeted to the inner leaflet of the plasma membrane through their N-terminal matrix (MA) domain. Because retroviruses of different morphogenetic types assemble their immature particles in distinct regions of the host cell, the mechanism of MA-mediated plasma membrane targeting differs among distinct retroviral morphogenetic types. Here, we focused on possible mechanistic differences of the MAmediated plasma membrane targeting of the B-type mouse mammary tumor virus (MMTV) and C-type HIV-1, which assemble in the cytoplasm and at the plasma membrane, respectively. Molecular dynamics simulations, together with surface mapping, indicated that, similarly to HIV-1, MMTV uses a myristic switch to anchor the MA to the membrane and electrostatically interacts with phosphatidylinositol 4,5-bisphosphate to stabilize MA orientation. We observed that the affinity of MMTV MA to the membrane is lower than that of HIV-1 MA, possibly related to their different topologies and the number of basic residues in the highly basic MA region. The latter probably reflects the requirement of C-type retroviruses for tighter membrane binding, essential for assembly, unlike for D/B-type retroviruses, which assemble in the cytoplasm. A comparison of the membrane topology of the HIV-1 MA, using the surface-mapping method and molecular dynamics simulations, revealed that the residues at the HIV-1 MA C terminus help stabilize protein-protein interactions within the HIV-1 MA lattice at the plasma membrane. In summary, HIV-1 and MMTV share common features such as membrane binding of the MA via hydrophobic interactions and exhibit several differences, including lower membrane affinity of MMTV MA.
Retroviruses are divided into several groups according to their morphogenesis (1)(2)(3). HIV, together with the other representatives of lentiviruses but also alpharetroviruses and gammaretroviruses, assemble their immature particles from individual Gag polyprotein precursors underneath the plasma membrane (PM) of the host cell immediately before or simultaneously with its budding (4,5). This morphogenesis is typical for so-called C-type retroviruses. By contrast, B-and D-type retroviruses, such as betaretroviruses, e.g. mouse mammary tu-mor virus (MMTV) and Mason-Pfizer monkey virus (M-PMV), respectively, assemble their immature particles in the pericentriolar region of an infected cell; the resulting particle is then transported to the PM, where it buds (6).
In betaretroviruses, the targeting of retroviral structural Gag polyproteins to the periplasmic region for the assembly of immature particles is mediated by the conserved cytoplasmic targeting/retention signal (CTRS) located in the sequence of the N-terminal matrix (MA) domain of Gag (7)(8)(9), which has been shown to interact with dynein motor machinery (10). Interestingly, a single amino acid substitution in CTRS (R55W in the case of M-PMV MA and D56A in the case of MMTV MA) is able to redirect the morphogenetic pathway of Gag from B/D-type to C-type (7)(8)(9)11). The requirement of CTRS for targeting to the intracytoplasmic site of assembly has been confirmed by introducing this sequence into the MA of a Ctype retrovirus, murine leukemia virus. As a result, the C-type murine leukemia virus converted to the B/D-type virus (12).
The PM binding of individual Gag molecules (C-type retroviruses) or preassembled immature particles (B/D-type retroviruses) is mediated by several factors. One of them is a bipartite signal in the MA comprising the N-terminal myristoylation of the MA and a highly basic region (HBR) located at the surface of the MA (6,(13)(14)(15). For the HIV-1 MA, it has been observed that the myristoylated MA can adopt either myristoyl-exposed or myristoyl-sequestered states and that the myristoyl exposure is coupled with MA trimerization (16). This so-called myristic switch mechanism has been proposed to be utilized by HIV-1 in the reversible interaction between the MA and the PM (16). The mechanism of the MA-PM interaction was further elucidated when phosphatidylinositol 4,5-bisphosphate (PI4,5P 2 ), the negatively charged phospholipid characteristic for the PM, was determined as another factor essential for the targeting of HIV-1 Gag to the PM (17). This suggests that PI4,5P 2 may serve as a specific motif recognized by a HBR. Furthermore, it has been detected that even soluble PI4,5P 2 (with truncated acyl chains) interacts directly with the HIV-1 MA, and this specific interaction induces the exposure of myristoyl, which can subsequently anchor the MA to the membrane and stabilize the orientation of MA at the membrane (18). This experiment has also shown the possibility of strengthening the interaction between PM and MA by the 29-acyl chain of PI4,5P 2 , buried in the MA (18). Further, it has been shown that the cells enriched with PI4,5P 2 endosomes retarget Gag molecules to the vesicles in which assembly and intravesicular budding have been observed (17).
Subsequent studies have confirmed the important role of PI4,5P 2 in the membrane targeting of HIV-1 Gag, but they have also reported the positive effect of other phospholipids on the affinity of HIV-1 Gag to the membrane (19)(20)(21)(22). Specifically, it was shown that increasing amounts of phosphatidylserine (PS) enhance the Gag affinity to liposomes containing PI4,5P 2 and that the HIV-1 Gag or MA can also bind to liposomes containing phosphatidylinositol 3,5-bisphosphate (19,20). Additionally, Vlach and Saad (22) have reported that the HIV-1 MA directly binds other soluble phospholipids such as PS, phosphatidylethanolamine (PE), and phosphatidylcholine (PC), and more importantly, Vlach and Saad (22) have also shown that the MA interaction with these phospholipids at the critical micellar concentration triggers the exposure of myristoyl. Two distinct binding sites for PI4,5P 2 and PC/PE/PS have been determined in the structure of HIV-1 MA. Based on this, the so-called "trio engagement" model has been proposed. In this model, the polar heads of both phospholipids interacting with the MA contribute to the electrostatic interactions, which are supported by hydrophobic interactions mediated by their 2'acyl chains buried in MA hydrophobic pockets together with the myristoyl buried into the membrane (22). The interaction between HIV-1 MA and an artificial membrane containing palmitoyloleoyl-PC, palmitoyloleoyl-PE, stearoyloleoyl-PS, and stearoylarachidonoyl-PI4,5P 2 has further been studied by coarse-grained molecular dynamics (CG-MD) simulations, showing the exposure of myristoyl and its anchoring to the membrane (21). Nevertheless, during the simulations, both acyl chains of all the phospholipids remained buried within the membrane, which makes the trio engagement model questionable (22).
Simultaneously, the influence of the hydrophobic part of the membrane bilayer on the affinity of Gag was analyzed in the study using phospholipid liposomes with different compositions of saturated or unsaturated acyl chains and variable amounts of cholesterol (23). The results clearly show that Gag prefers lipids with both acyl chains unsaturated and that cholesterol enhances the binding of the HIV-1 MA additively to PI4,5P 2 . This is in good agreement with other studies showing that protein and lipid components of membrane nanodomains are enriched in HIV-1 viral envelope compared with the PM (24)(25)(26) and that the HIV-1 MA has a higher affinity to raft-like liposomes than those with nonraft composition (20). All of these findings support the theory that immature HIV-1 particles bud in the regions of membrane nanodomains. Moreover, data indicating that HIV-1 Gag supports the formation of PI4,5P 2 /cholesterol nanodomains at the inner leaflet of the PM at the initial stages of assembly rather than targeting to preexistent PI4,5P 2 -enriched domains have recently been published (27).
However, studies of other retroviruses have shown that the mechanisms occurring in HIV-1 are not common to all retroviruses. To arrange the interaction of Gag with PM, HIV-1 uses a bipartite signal in the MA including the electrostatic interac-tions of the MA HBR with the membrane component PI4,5P 2 and the hydrophobic interactions mediated by the anchoring of myristoyl to the membrane. However, the MAs of some retroviruses such as Rous sarcoma virus, equine infectious anemia virus, and avian sarcoma virus are not myristoylated. In these retroviruses, the Gag-membrane binding is mediated solely by the electrostatic interactions between the basic regions of the MA and acidic phosphoinositides (28)(29)(30)(31)(32). Affirmatively, recent results have shown that the release of Rous sarcoma virus and equine infectious anemia virus VLPs (virus like particles) from the host cells is PI4,5P 2 -dependent (33)(34)(35). On the contrary, PI4,5P 2 is not required for the association of the human T-lymphotropic virus type 1 Gag, the murine leukemia virus Gag, and the human endogenous retrovirus K Gag with PM (35,36). Although the release of virus-like HIV-2 particles is PI4,5P 2 -dependent as in HIV-1, the trimerization of the HIV-2 MA and its interaction with PI4,5P 2 does not trigger the myristoyl exposure (37). Similarly, the myristoyl exposure was not observed in the case of the M-PMV MA when it interacted with truncated PI4,5P 2 (38) or during the CG-MD simulations of the M-PMV MA interaction with the lipid bilayer (39). Nevertheless, the tendency of the nonmyristoylated M-PMV MA to form trimers in contrast to its myristoylated form suggests that the exposure of myristoyl does occur and that it induces a conformational change of M-PMV MA, which favors trimerization (40). All of this evidence leads to the conclusion that the mechanisms of retroviral PM targeting can significantly differ in various retroviruses and cannot be generalized.
Because the interaction of the MMTV MA with the PM has not been described, we have thoroughly investigated the interaction between the MMTV MA and membranes of different compositions. We have used a combination of the MS-based protein surface mapping method and CG-MD simulations. This approach has already been utilized to describe M-PMV MA-membrane interaction (39). Additionally, we have applied surface mapping to broaden our knowledge of the HIV-1 MA interaction with the membrane.

Liposome binding of HIV-1 and MMTV MAs
Because the HIV-1 MA has a different affinity to various phospholipids, we have mapped the interactions between the HIV-1 MA and two differently composed liposomes. The first type of liposomes mimicked the phospholipid composition of the inner leaflet of the PM and contained PC, PE, PS, and PI4,5P 2 (designated as complex liposomes or PC/PE/PS/PI4,5P 2 liposomes) in the molar ratio 45:45:5:5. The second type of liposomes contained only PC and PS (labeled as PC/PS liposomes) in the molar ratio 66:34 and was used to determine the influence of negatively charged phospholipids other than PI4,5P 2 on the topology of the HIV-1 MA on the membrane. The same experiment was performed with MMTV MA to determine the differences between the MA binding of two morphogenetically distinct retroviruses, HIV-1 and MMTV.
We have found that the complex PC/PE/PS/PI4,5P 2 liposomes are bound the HIV-1 MA more effectively than the PC/PS liposomes, because ;60% of the total HIV-1 MA was associated with complex liposomes, in contrast to ;40% of the MA detected as bound to the PC/PS liposomes (Fig. 1). Our results are in good agreement with the study performed by Kroupa et al. (41), in which the membrane interaction and the affinity of M-PMV and HIV-1 MAs to differently composed liposomes was also inspected. Via a liposome pelleting assay, Kroupa et al. (41) found that HIV-1 MA efficiently interacted with PC/PS liposomes (;20% of the HIV-1 MA was bound to PC/PS liposomes) with almost half-effectivity as compared with PI4,5P 2 -containing liposomes (more than 40% of the HIV-1 MA was bound to PI4,5P 2 liposomes). The more significant difference between the HIV-1 MA binding to PC/PS and PI4,5P 2 in their data is probably caused by the higher proportion of PS in the PI4,5P 2 -containing liposomes used by them (30 mol % in contrast to our 5 mol %). Indeed, it has already been shown that the increasing amount of PS in PI4,5P 2 -containing liposomes enhances their affinity to HIV-1 Gag (19).
Liposome-binding assay (LBA) with the MMTV MA has shown significantly lower affinity of this protein to both types of liposomes than that of the HIV-1 MA (Fig. 1). Interestingly, the affinity of the MMTV MA to PI4,5P 2 -containing liposomes was even lower than the affinity of the HIV-1 MA to PC/PS liposomes, because only 40% of the initial MMTV MA remained bound to PC/PE/PS/PI4,5P 2 liposomes. Additionally, only 26% of the MMTV MA remained bound to PC/PS liposomes. The lower affinity of the M-PMV MA, a D-type retrovirus, to liposomes has previously been observed by Kroupa et al. (41), where only 5% of the M-PMV MA was bound to PI4,5P 2 -containing liposomes in contrast to more than 40% of the HIV-1 MA bound to the same liposomes. In our previous study, we found that ;50% of the M-PMV MA was bound to PI4,5P 2 -containing liposomes (39). However, the concentration of the phospholipids used in the mixture of M-PMV MA and liposomes was almost five times higher than that used by Kroupa et al. (41) and twice as high as in our current study to obtain a sufficient amount of protein bound to the liposomes for subsequent mapping experiments.

The surface mapping of HIV-1 and MMTV MAs
Because the targeting of the HIV-1 MA to the PM enables effective release of infectious viral particles from the host cell, it has been studied by various techniques. Although the MMTV MA also controls the PM targeting, the structural aspects of the interaction have not been studied yet. In the present study, we have used a surface mapping method (also termed protein covalent labeling or protein footprinting) to extend the current knowledge of the topology of HIV-1 MA molecules at the PM and to obtain the initial information on the topology of the MMTV MA at the PM. The mapping is based on the determination of surface accessibility and the modification of the amino acid residues of the studied protein by suitable monofunctional agents. Only surface-accessible residues can be modified, whereas the residues buried in the protein core or mediating protein interactions are not accessible and thus remain unmodified. The positions of specifically modified residues are subsequently identified by MS.
In our study, we have mapped the surface accessibility of the lysine, tyrosine, and tryptophan residues of HIV-1 and MMTV MAs in solution (free state) and bound to artificial liposomes (bound state). The lysines and tyrosines were acetylated by Nacetylimidazol (NAI), which resulted in the mass shift of 42.01 Da, whereas tryptophans were modified with the 2-hydroxy-5nitrobenzyl group, which originated from dimethyl(2-hydroxy-  When the surface accessibility of lysines, tyrosines, and tryptophans in the free HIV-1 MA protein was mapped, it was observed that the residues Lys 15 , Trp 16 Fig. 2A). The modification of all of these residues was detected at least in two of three repetitions (see Table S1 for detailed results). This implies that these residues are surface-accessible when the protein does not interact with the membrane. When mapping the free HIV-1 MA, only four of all the mapped residues (Lys 26 , Tyr 29 , Tyr 79 , and Lys 95 ) remained unmodified. One of the possible explanations is that some of these residues are located inside the protein molecule, which makes them inaccessible to the modifying agent. This can be indeed the case of Tyr 29 , which is buried in the protein core as seen from the published structure of the monomeric HIV-1 MA protein (the structure 1UPH in the Protein Data Bank (16); Fig. S1). In addition, some residues that are expected to be surface-accessible according to the known three-dimensional structure cannot be detected as modified because of other factors that can affect the modification reaction. One of the factors could be the unsuitable microenvironment of the residue, preventing the modification (for example the hydrophobic environment of the residue modified by the hydrophilic group and vice versa) (42). Another possibility is that the modification can affect the protease cleavage site and thus the spectrum of the digested peptides. In our study, we have used trypsin to generate peptide digest. Because this protease cleaves at the C-side of lysines and arginines, it is possible to expect its reduced cleavage activity after modified lysines. This complicates especially the analysis of the N- terminal HBR of the HIV-1 MA, which is enriched mainly in lysines. Indeed, we have failed to cover the HBR sequence by the detected peptides in the nontreated control HIV-1 MA (see Table S1 for detailed results). This was apparently caused by the occurrence of too-small peptides in the HBR tryptic digest. By contrast, we have covered the HIV-1 MA HBR sequence when some of the lysines (Lys 27 , Lys 30 , and Lys 32 ) of the free soluble MA were modified (see Table S1 for detailed results). To ensure the detection of peptides containing a HBR, we also tested the usage of other proteases such as chymotrypsin, Asp-N, and Arg-C, but none of them provided a better result than trypsin (data not shown).
The surface mapping of the HIV-1 MA bound to complex liposomes has provided substantial findings on the inaccessibility of the residues Lys 27 , Lys 30 , and Lys 32 compared with the free HIV-1 MA, where all of these residues were modified by NAI ( Fig. 2A). This result has confirmed the involvement of the residues of the HBR in the interaction of liposomes. In the case of PC/PS liposomes, the residues of the HBR have not been detected ( Fig. 2A). However, we also consider these residues as not modified because, as mentioned above, the residues located in the HBR are difficult to detect unless they are modified. Actually, in the treated free HIV-1 MA, the HBR was covered because of the modifications of lysines, whereas in the nontreated sample, this region was usually not detected (it was only detected in one peptide of the six peptides analyzed; see Table  S1). Additionally, although the unmodified peptide containing HBR residues was detected when MA was bound to complex liposomes, it was only detected in one of three repetitions, which also indicates how difficult it is to detect this sequence.
Interestingly, the residues Tyr 86 , Lys 98 , Lys 110 , Lys 112 , and Lys 113 located in the C-terminal part of the HIV-1 MA sequence were also inaccessible in the MA bound to liposomes. Nevertheless, the involvement of these residues in the interaction with membrane phospholipids is unlikely, because they are located on the opposite side of the molecule from the HBR and the HIV-1 MA membrane interface (18,21,43). We thus hypothesize that these residues are inaccessible because of their involvement in the stabilization of protein-protein interactions within the HIV-1 MA lattice at the membrane (see the chapter: The mechanism of MMTV MA-membrane interaction, in the Discussion).
The surface mapping of HIV-1 MA has also shown that the same residues were involved in the binding of the MA to complex liposomes and PC/PS liposomes. This similar topology of the HIV-1 MA bound to these two types of liposomes suggests that the interaction of HIV-1 with a PS-containing membrane as the only negatively charged phospholipid is mediated by the HBR.
The surface mapping of the MMTV MA has been performed in the same way as that of the HIV-1 MA. We have mapped the surface accessibility of the lysine, tyrosine, and tryptophan residues of the free MMTV and MMTV MAs bound to PC/PE/PS/ PI4,5P 2 and PC/PS liposomes.
Within the surface mapping of free MMTV MA residues, Lys 7 , Lys 10 , Lys 28 , Trp 45 , Trp 57 , Lys 58 , Tyr 67 , Lys 78 , Tyr 81 , and Trp 86 were modified (Fig. 2B). According to the published structure of the MMTV myristoylated MA monomer (chain in the 4ZV5 structure in the Protein Data Bank (44)), almost all of these residues are surface-accessible (Fig. S2). Only residue Trp 57 is oriented toward the protein core, which corresponds to the weak modification detected for this residue. In fact, the modification of this residue was detected only once in three independent experiments (see Table S2 for detailed results). On the other hand, residues Tyr 36 and Lys 41 were not modified in the free state of the MMTV MA, and residue Lys 65 was not detected at all. Residue Tyr 36 is buried in the protein core, where it is oriented toward the cavity at the N terminus of the MMTV MA, which is enriched in hydrophobic residues and expected to mediate hydrophobic interactions with myristoyl. Additionally, we suppose that the reasons why residues Lys 41 and Lys 65 were not modified or even detected are related to the primary structure of the MMTV MA. Residue Lys 41 is located in the part of the sequence lacking other lysines or arginines. Therefore, its modification impeding the action of trypsin results in the generation of a too-large, undetectable peptide. By contrast, two arginines in a close proximity to residue Lys 65 (Arg 62 and Arg 66 ) can result in the creation of too-small and thus also undetectable peptides.
The surface mapping of the MMTV MA bound to liposomes has shown a significant difference between the amino acid accessibility of the MA bound to complex liposomes and PC/ PS liposomes (Fig. 2B). When compared with the free MMTV MA, five residues (Lys 7 , Trp 45 , Trp 57 , Lys 58 , Lys 78 , and Trp 84 ) lost their accessibility and became resistant to modification when the MA was bound to complex liposomes, which indicated their involvement in the interaction. In contrast, only three residues (Lys 28 , Trp 45 , and Trp 57 ) became inaccessible when the MMTV MA was bound to PC/PS liposomes. Moreover, the residue Lys 28 has a less informative value, because it was detected only once in three samples of the MMTV MA bound to PC/PS liposomes, and it was not detected in the complex liposomes. Overall, the higher number of detected inaccessible residues of the MMTV MA upon its interaction with complex liposomes shows that the MMTV MA interacts with complex liposomes more tightly than with PC/PS liposomes. This also corresponds to the results of LBA, in which the lower affinity of the MMTV MA to PC/PS liposomes was observed.

CG-MD simulations of HIV-1 and MMTV MA-membrane interaction
To interpret the results from surface mapping and to deepen the knowledge of the interaction of retroviral MAs with a membrane, we performed a CG-MD simulation of the interaction between a membrane and the HIV-1 MA or MMTV MA. Although the interaction of the HIV-1 MA and the plasma membrane has already been studied by CG-MD (21), the authors of the study used a shortened version of the HIV-1 MA (2-114 amino acids), which did not contain the C-terminal region, in which the changes in the surface accessibility of several lysine residues have been detected by our surface mapping. Therefore, we have performed our CG-MD simulation with the full-length myristoylated HIV-1 MA (2-136 amino acids, a chain in the 2H3I structure in the Protein Data Bank). In the case of the MMTV MA, the CG-MD simulations have been performed with the myristoylated MA structure (a chain in the 4ZV5 structure in the Protein Data Bank), published by Doležal et al. (44). The simulation of both proteins has been performed with two types of membranes with the same composition as the liposomes used in surface mapping or LBA.
During HIV-1 MA-membrane interaction simulations, the membrane insertion of myristoyl and the subsequent anchoring of the MA was observed in both PC/PS and PC/PE/PS/ PI4,5P 2 membranes. However, when PC/PS membranes were used in the simulation, the insertion of HIV-1 MA myristoyl occurred in all three performed simulations (individual simulations were performed for 1-2 ms), whereas in the case of the complex membrane (PC/PE/PS/PI4,5P 2 ), the HIV-1 MA myristoyl insertion occurred in four of nine simulations (each simulation was performed for 2 ms). However, also in the simulations where myristoyl was not inserted into the membrane, the HIV-1 MA was able to bind the membrane with a similar orientation to the HIV-1 MA with myristoyl inserted into the membrane. Additionally, our simulations have shown that myristoyl has to be exposed from the HIV-1 MA structure prior to HIV-1 MA-membrane binding; otherwise, it remains sequestered in the MA already bound to the membrane. This observation is in agreement with the results of previous CG-MD simulations performed by Charlier et al. (21), in which the myristoyl was inserted into the membrane in all five simulations performed but had always been exposed before the HIV-1 MA reached the membrane. Additionally, the authors have suggested an equilibrium between sequestrated and exposed myristoyl in water, even for the monomer, because the cost required for myristoyl release out of its hydrophobic pocket was only ;5 kcal·mol 21 . They have also observed that after its release, the myristoyl group remained most of the time at the protein surface to prevent too many undesirable contacts with the surrounding water molecules. Moreover, other NMR studies with bicelles or micelles have also suggested that myristoyl is more readily exposed in the proximity of a lipid bilayer (16,22,45).
A representative snapshot of HIV-1 MA simulations, in which the myristoyl was anchored to the membrane, is shown in Fig. 3A. When the myristoyl was anchored to the membrane, the orientation of HIV-1 MA was stable and equal in both types of membranes. This is well-documented in Fig. 4, in which the mean number of contacts with phosphate groups in the membranes is almost identical for the residues of the HIV-1 MA bound to PC/PS and complex liposomes (Fig. 4A) (86). The interaction of the HIV-1 MA with both membranes is driven mainly by residues Lys 26 , Lys 27 , Lys 30 , Lys 32 , and His 33 , forming the HBR of the HIV-1 MA, together with the residues Trp 36 and Thr 43 . The Trp 36 residue has already been suggested to play a role in the PI4,5P 2 binding by Saad et al. (18) as well as Charlier et al. (21).
The projection of contacts with the membrane onto the structure of the HIV-1 MA (Fig. 4D) shows that the region that is in contact with the membrane corresponds to the area of the positively charged residues of the HIV-1 MA, forming its HBR located close to the protein N terminus (Fig. 4B). Further, the significant number of the contacts of N-terminal residues can be explained by their proximity to the N-terminally attached myristoyl that is anchored to the membrane.
The course of myristoyl-membrane anchoring in MMTV MA simulations was similar to that observed in the simulations with the HIV-1 MA. The myristoyl of the MMTV MA was anchored to the membrane four times of five simulations (individual simulations were performed for 1-2 ms) when the PC/PS membrane was used and in four of eight simulations (with each simulation being performed for 2 ms) when the complex membranes were used. Nevertheless, myristoyl was released from the protein core in all the replicas performed with the MMTV MA. Therefore, the possible reason for the lower frequency of the membrane insertion of the MMTV MA is different from the one in the case of the HIV-1 MA. The simulations have shown that the MMTV MA protein initially interacts with the membrane bilayer via electrostatics interactions, but these interactions have to be arranged by a specific region of protein to ensure the proper orientation of the protein on the membrane, which is needed for the myristoyl insertion into the lipid bilayer. However, in the case of our complex lipid bilayer, the protein can be transiently bound to the membrane through some electrostatic interactions "nonspecifically," e.g. with other regions of the protein that are likely to prevent its proper orientation. Indeed, when we plotted the minimal distance between the center of mass of the MMTV MA and the membrane (represented by the phosphate groups) as a function of time, we could observe that the protein was closely associated with the lipid bilayer and thus had reduced rotational freedom, leading to a lower chance of membrane insertion (Fig. S3). This finding is also supported by the mean number of contacts between individual amino acid residues and the membrane, which are distributed nonspecifically along the whole protein, when the MMTA MA is bound to the membrane but myristoyl is not inserted into the membrane (Fig. S4).
A representative snapshot of MMTV MA simulations, in which the myristoyl was anchored to the membrane, is shown in Fig. 3B. Although a detailed analysis showed that the PC/PS membrane was sufficient for the interaction with the MMTV MA and myristoyl anchoring, we observed that the orientation of the protein through the simulations was rather variable. The four simulations with the successful myristoyl insertion resulted in significantly different orientations of the protein toward the lipid bilayer measured as the angle between the vector formed by the first helix and the lipid bilayer normal (Fig. S3, C  and D). Additionally, the mean number of contacts of the individual residues of the MMTV MA with the phosphate groups in the membrane shows that the interaction between the MMTV MA and the PC/PS membrane is predominantly arranged by the myristoyl and the residues occurring at the N terminus of the MA (Fig. 4E). The engagement of other residues than those at the N terminus is minor, and the contacts of lower values are distributed among almost all residues in the MMTV MA sequence.
In the case of simulations with the complex membrane, it was evident that the orientation of the MMTV MA was stabilized after the insertion of myristoyl into the membrane (Fig.  S3C), and the resulting angle between the first helix and the lipid bilayer normal did not differ between different replicas (Fig. S3D). In addition to myristoyl and the N-terminal sequence of MA, two more distinct regions have been observed in close contact with membrane phosphate groups (Fig. 4E). These regions involve the residues located in helix II (Gln 37 -Glu 49 ) and helix III (Arg 59 -Ser 75 ). The residues with the highest number of contacts (more than three) in these two regions were the hydrophobic residue Trp 45 and the positively charged residues Lys 41 , Arg 59 , Arg 66 , and His 71 . Regarding the N-terminal part of the MMTV sequence, the residues closest to the membranes were shown to be Gly 2 , Ser 4 , and mainly Lys 7 . However, these residues were observed to have a high value of contacts also in the MMTV MA bound to the PC/PS membranes.
When the electrostatic potential is projected onto the MMTV MA structure (Fig. 4F), it is possible to observe a less extensive basic region on the surface of the MMTV MA than in the case of the HIV-1 MA (Fig. 4, B and F). Actually, in the structure of the MMTV MA, the basic residues are clustered only at the N terminus (Lys 7 and Lys 10 occurring close to the myristoyl) and in the region located close to the C-terminal part of the MMTV MA, which we have therefore labeled as the C-terminal HBR. This region comprises the residues Lys 58 , Arg 59 , Arg 62 , Lys 65 , Arg 66 , and His 71 , which have been identified to interact with the MMTV MA bound to complex liposomes, e.g. the liposomes containing PI4,5P 2 . We suppose that the interactions between the residues of the C-terminal HBR of the MMTV MA and PI4,5P 2 molecules stabilize the orientation of the MMTV MA on the membrane. This stabilization is clearly visible when the contacts of individual MMTV MA residues with the membrane are visualized onto the protein structure (Fig. 4H). When the MMTV MA is in contact with the complex membrane, the interacting residues are clustered in one specific region of the MMTV MA structure, whereas at the PC/PS membranes, the interacting residues are distributed in several distinct parts of the protein structure. Because the Cterminal HBR is located on the edge of the membrane-interacting region, we suppose that the interaction with the C-terminal HBR can be mediated by the molecules of PI4,5P 2 because of the extended range of their polar heads as compared with PS polar heads (for more details, see Fig. S5).

Discussion
The mechanism of HIV-1 MA-membrane interaction The interaction between the HIV-1 MA and the phospholipid membrane has already been intensively studied by various methods. Our results of CG-MD simulations support the results of previously performed MD simulations of the HIV-1 MA-membrane interaction, in which the shortened version of the MA was used (21). Similarly, our simulation has shown the importance of the electrostatic interaction between the HBR of the HIV-1 MA and the negative polar heads of phospholipids in the membrane. Additionally, we have also observed the exposure of the myristoyl from the structure of the HIV-1 MA and its subsequent anchoring to the membrane. However, we have not observed the exposure of any fatty acid from the membrane for the stabilization of the MA-membrane interaction described as the trio engagement model (22).
The role of the HBR for the HIV-1 MA-membrane interaction has also been confirmed by surface mapping. This is in good agreement with the previous studies in which the importance of HBR residues was shown and thoroughly inspected (18-21, 46, 47). Lys 30 and Lys 32 in particular were shown to interact directly with PI4,5P 2 (19,20,46,47) and to play a significant role in the PM targeting of the HIV-1 Gag, because mutations of these residues retargeted the Gag from the PM to intracellular vesicles (46,48,49). It has recently been shown that residues Lys 18 , Lys 26 , Lys 27 , Lys 30 , and Lys 32 lose NMR signal intensity upon binding to nanodiscs composed of palmitoyloleoyl-PC:POPS:PI(4,5)P 2 (43). Interestingly, it has also been revealed that the residues within the HBR may serve to regulate hydrophobic interactions mediated by the myristoyl because the HIV-1 MA double mutant K26T/K27T lost its specificity of binding to PI4,5P 2 -containing liposomes and bound efficiently to liposomes consisting of PC/PS or exclusively of PC (50). The authors have concluded that the mutant more readily exposes the myristoyl than the WT MA, in which the lysines 26 and 27 suppress nonspecific membrane binding.
Additionally, the electrostatic interactions between the HIV-1 MA HBR and the membrane have been shown to be regulated by the virus in the host cells. Experiments from several laboratories have documented well that the negative charge of the HIV-1 HBR may be neutralized by RNAs (50)(51)(52)(53)(54)(55). It has been shown that the same MA N-terminal residues that mediate the HIV-1 Gag-PM interaction are occupied by RNAs, which prevents the soluble HIV-1 Gag from binding to intracellular membranes (5,47,50). Recent data indicate that tRNAs are predominantly involved in this process (56). Current understanding is that the RNA-mediated neutralization of the basic character of the HIV-1 HBR prevents premature intracellular membrane-induced assembly at membranes not containing PI4,5P 2 , as shown both in vitro and in cells (50,(57)(58)(59). Moreover, it may prevent the intracellular aggregation of the HIV-1 Gag, possibly induced by an inappropriate interaction with long RNAs, including the genomic RNA of the virus. This is consistent with the results of the CLIP assay, which excluded such a direct interaction of the viral RNA with the HIV-1 MA in the cytoplasm of infected cells (56). The assay has also suggested that the myristoylation of MA is not required for the binding of tRNAs. However, both the exposure of myristoyl from the hydrophobic pocket of the MA and the release of the tRNA from the HBR must collectively mediate the binding of the HIV-1 Gag to the PM. The myristic switch may be also regulated by RNA interactions with the HBR as concluded from the fact that the membrane interaction of the 25KT/ 26KT HIV-1 MA was enhanced upon RNase treatment (50).
In our study, we are reporting two additional significant findings regarding the binding of the HIV-1 MA with the membrane. First, by the use of surface mapping and CG-MD, we show that the amino acids involved in the interaction with both studied types of liposomes are the same. This is also in agreement with the results of LBA, showing that the HIV-1 MA reliably binds PC/PS liposomes, albeit with lower intensity than it binds PI4,5P 2 -containing liposomes. Because the HIV-1 MA binds both PS and PI4,5P 2 with the same region, the different strength of the interaction seems to be caused by the more suitable orientation of PI4,5P 2 polar heads toward the basic residues. Vlach and Saad (22) have used NMR to determine whether the amino acids responsible for binding PS and PC are distinct from those binding PI4,5P 2 . They have suggested that residue Arg 43 is responsible for the interaction with the PS polar head, whereas residues Leu 41 , Phe 44 , Val 46 , Ile 60 , Leu 64 , and Leu 75 , forming the hydrophobic pocket on the MA surface, stabilize the interaction with the PS 2?-acyl chain. Although residue Arg 43 has been shown to be involved in the interactions with both membranes, in our CG-MD simulations, the exposure of the PS 2?-acyl chain from the membrane and its binding to the MA hydrophobic pocket have not been observed.
Another significant finding is related to the changes detected in the surface accessibility of the C-terminal lysine residues upon the HIV-1 MA-membrane interaction. Because CG-MD simulations have clearly shown that the C-terminal region is not in contact with the membrane when the HIV-1 MA is in the proper orientation with the membrane-anchored myristoyl, we suppose that the C-terminal residues play a role in the multimerization of the HIV-1 MA on the membrane. Nevertheless, it has already been reported that the HIV-1 MA forms trimers even in vitro, and the trimeric structure has been described in several studies (43, 60, 61). The interface of MA monomers in the trimer consists of residues predominantly located in the loops between helices II, III, and IV (residues 42-47, 59, 63, and 69-74), indicating that the identified C-terminal residues located in the helix IV are not involved in the trimerization of the HIV-1 MA. However, it has also been observed that the HIV-1 MA organizes its trimers into a hexameric lattice upon its interaction with the membrane containing PI4,5P 2 (62), and several models of HIV-1 organization on the membrane have been proposed (60-63). The arrangement of HIV-1 MA mole-cules on the membrane is frequently discussed in the context of Env recruitment, because it has already been confirmed that the HIV-1 MA interacts with the cytoplasmic tail (CT) of Env. The HIV-1 MA residues Leu 13 , Glu 17 , Leu 31 , Val 35 , and Glu 99 , which are responsible for CT coordination (64)(65)(66), are positioned in the close proximity to the C-terminal helix. The proposed models indicate that all these residues are oriented toward the hexameric centers (61)(62)(63) at membrane-proximal positions. Therefore, we presume that the C-terminal lysines Lys 98 , Lys 103 , Lys 110 , and Lys 112 , which are not accessible to the modification agent when the HIV-1 MA is bound to the membrane, stabilize the MA hexameric lattice on the membrane. Because the C-terminal helix is also rich in acidic amino acid, this stabilization may be mediated by the electrostatic interactions between the C-terminal helixes of HIV-1 MA molecules. Moreover, there are several publications describing the importance of the C-terminal residues of the MA for both the early and late phases of the HIV-1 cycle. The mutation K98E has caused an early post-entry defect (67), the mutation A100E has blocked virus production in CEM cells (14), and the mutation E99V has prevented Env incorporation (68). Additionally, various deletions in this region have prevented HIV-1 Env incorporation into virus particles (69) or impaired virus release from the host cells (70). HIV-1 mutants with deleted 96-99 or 100-107 MA amino acid residues were assembly-competent and only partially defective in the virus release, whereas the mutant D96-120 was retarded at the PM, although it was still able to multimerize and associate with the membrane (70).

The mechanism of MMTV MA-membrane interaction
The mechanism of MMTV MA interaction with the membrane has not been studied yet. To elucidate the aspects of this interaction, we have performed LBA, the surface mapping of the MMTV MA, and a CG-MD simulation of this interaction. As a result, we have found that MMTV binds both PI4,5P 2 -containing membranes and PC/PS membranes. The affinity of the MMTV MA to PC/PS membranes was lower than to PI4,5P 2containing membranes. Nevertheless, the affinity of the MMTV MA to both types of membranes is lower than the affinity of the HIV-1 MA to the same membranes. The aforementioned differences of the MMTV MA binding to different types of membranes were detected by the use of both surface mapping and CG-MD simulations. By means of surface mapping, more residues were detected as inaccessible to the modification agent when the MMTV MA was bound to PI4,5P 2 -containing liposomes than to those comprised only of PC and PS. This is in accord with the results of CG-MD simulations showing that the MMTV MA binds to both types of membranes but that it is stably oriented only at PI4,5P 2 -containing membranes. Therefore, we suppose that PI4,5P 2 stabilizes the orientation of the MMTV MA on the membrane via electrostatic interactions between the C-terminal basic region of the MMTV MA and PI4,5P 2 polar heads. In the case of the PC/PS membrane, the interaction with the MMTV MA is mainly mediated by myristoyl the anchoring to the membrane, but the orientation of the protein on the membrane is rather flexible. Therefore, only two residues have been detected to be stably involved in the interaction by surface Retroviral recruitment of the plasma membrane mapping, and the CG-MD simulations have shown that contacts with the membrane are of lower value and distributed among almost all residues in the MMTV MA sequence.
To rationalize the inaccessibility of MMTV MA residues detected by surface mapping, we compared the surface mapping results with the results of the CG-MD simulations. Among the residues inaccessible upon the MMTV MA interaction with the complex liposomes, only Lys 7 and Trp 45 were shown to have a significantly high number of contacts with the membrane. Therefore, the positions of individual inaccessible residues in the CG-MD simulations were inspected in further detail. The simulations of the MMTV MA interaction with the complex membrane (Fig. S5) suggest that Lys 7 and Trp 45 are completely buried in the membrane. Other residues do not occur in such a close proximity to the membrane, but it can be seen that Trp 57 is oriented toward the protein core, which most likely makes it inaccessible to the modification agent. Interestingly, the residues Lys 58 and Trp 84 are located close to each other in the structure of the MMTV MA, and polar heads of two PI4,5P 2 molecules occur in close proximity. Moreover, the residues Arg 59 and Arg 66 , which were shown to have a high number of contacts with the membrane, are located in the same region, with a high concentration of PI4,5P 2 molecules around it. Therefore, Lys 78 appears to be the only inaccessible residue to have no contact or only transient contacts with the membrane phospholipids, with its modification not being sterically obstructed by the polar heads of PI4,5P 2 molecules. Because our CG-MD simulations were performed with the monomeric structures of proteins, the aspects of the multimerization of individual MA molecules could not been studied. Nevertheless, the multimerization of the MA on the membrane is expected, and we suppose that the Lys 78 residue can be somehow involved in this process.
In the case of the MMTV MA bound to the PC/PS membrane, only Trp 45 and Trp 57 were observed to be inaccessible. A closer look at this interaction (Fig. S5) has revealed that Trp 45 is buried in the membrane and Trp 57 is buried in the protein core, as with the interaction of the MMTV MA with complex liposomes. Lys 58 , Lys 78 , and Lys 84 , which were shown to be differently accessible upon the interaction of the MMTV MA with different types of liposomes, are in the simulations of the MMTV MA interaction with the PC/PS membrane rather far from the membrane. Therefore, small PS headgroups cannot obstruct their modification. The only discrepancy between the surface mapping and CG-MD simulations concerns the residue Lys 7 , which was shown to interact strongly with phosphate groups of the PC/PS membrane. However, because we have observed that the orientation of the MMTV MA on the PC/PS membrane is rather variable, we suggest that Lys 7 can be transiently released from the membrane and thus modified by NAI.

Plasma membrane recruitment of morphogenetically different retroviruses
A comparison of our findings on the targeting of the MMTV MA to the plasma membrane with those on the targeting of the HIV-1 MA has revealed several common features as well as several differences. The important common feature is the use of myristoyl for anchoring to the membrane. Additionally, the myristoyl, which is initially buried in the MA hydrophobic pocket, was exposed spontaneously before the MA proteins reached the membrane. Several factors triggering the exposure of myristoyl have already been proposed for the HIV-1 MA (16,18,19,50,71,72), one of which was the electrostatic interaction between the HBR in the MA and the PI4,5P 2 present in the membrane. Although the electrostatic interactions play a significant role in the interaction, they are not a prerequisite for myristoyl exposure. The same was previously concluded for the HIV-1 MA, because the spontaneous exposure of myristoyl was observed before the MA interacted with the membrane in CG-MD and all-atom simulations performed by Charlier et al. (21). Although it is widely accepted that the switch is triggered by the binding of PI4,5P2 to the HIV-1 HBR (18), Tang et al. (16) have shown that myristoyl exposure can be triggered in solution without any phospholipids, only by the trimerization of the HIV-1 MA. Their model of the dynamic equilibrium between sequestered and exposed myristoyl in solution can explain myristoyl exposure both in our simulations and in the simulations performed by Charlier et al. (21). Another possible explanation favoring the myristoyl exposure in host cells is the proposed model based on the interaction of the HIV-1 MA with calmodulin (72,73).
In the case of the MMTV MA, the exposure of myristoyl has already been observed in the solution without the presence of membrane phospholipids (44). The same applies for the effect of multimerization on myristoyl exposure in our simulation, because the myristoyl moieties of both proteins were exposed without any influence of protein multimerization. Additionally, the experiments with the PC/PS membrane have also shown that the anchoring of the myristoyl of both MA proteins is independent of the membrane composition, because the anchoring of myristoyl occurred for both types of the studied membranes.
However, the hydrophobicity of the pockets harboring the myristoyl differs in the MAs of various retroviruses. This can significantly affect their tendency of myristoyl release. In our previous study, we used multiscale MD simulations to describe the MA-membrane interaction of M-PMV, which is morphogenetically similar to MMTV, and the release of myristoyl was not observed during the simulations, although the M-PMV MA stably interacted with the membrane by its HBR bound to PI4,5P 2 molecules (39). Additionally, the exposure of myristoyl was not observed in the case of HIV-2 MA under any of the conditions ensuring the exposure of the myristoyl of the HIV-1 MA (37). Therefore, it can be concluded that regarding the exposure of myristoyl and its role in membrane binding, the MMTV MA is more similar to the HIV-1 MA than to M-PMV or the HIV-2 MA and that the tendency to myristoyl exposure is independent of the morphogenetic type of retrovirus.
The significant difference between HIV-1 and MMTV MAs detected in our study lies in the location and proportions of their HBRs. The HBR of the HIV-1 MA is located in the loop between the helices I and II, including nine basic residues. On the other hand, the basic patch of the MMTV MA attracting PI4,5P 2 molecules has been found in its helix III, where only six basic residues are localized. Although both the anchoring of myristoyl to the membrane and the electrostatic interactions between the HBR and negatively charged phospholipids are necessary for the stable orientation of both HIV-1 and MMTV MA proteins on the membrane, the electrostatic interactions of the C-terminal HBR ensuring a stable orientation of the MMTV MA are PI4,5P 2 -dependent, whereas in the case of the HIV-1 MA, they are PI4,5P 2 -independent. Additionally, the comparison of HIV-1 and MMTV MA LBA results shows that the HIV-1 MA HBR is able to mediate stronger electrostatic interactions than the HBR of the MMTV MA because HIV-1 exhibits a higher affinity to the liposomes than the MMTV MA under the same experimental conditions. Moreover, the higher affinity of the HIV-1 MA to PI4,5P 2 -containing liposomes as compared with another B/D-type retrovirus, M-PMV, has been observed previously (39,41), although its HBR is located at the N terminus of the MA and comprises eight basic residues. Therefore, we suppose that the affinity of different MAs to PI4,5P 2 is related to the morphogenesis of particular retroviruses. In HIV-1, monomers or trimers of Gag initially interact with the PM; on the other hand in D-type retroviruses, preassembled immature particles come into contact with the PM. Therefore, the surface of B/D-type particles is formed by multiple MA molecules, promoting the interactions. In addition, it is necessary to consider the interactions of individual monomers and trimers of the MA within the MA lattice.
On the other hand, a significant role in this process may also be played by other factors, such as the interaction with Env molecules spanning the host-cell membranes or the interaction of retroviral MA proteins with RNA. Actually, there is an indication that MA RNA interactions could be distinct for different types of retroviruses (35); in addition, differences in the length Env cytoplasmic tail have been described, showing that in the case of HIV-1 and other lentiviruses, CT is significantly longer than other retroviruses (74, 75).
In summary, our in vitro and in silico study of two morphogenetically distinct retroviruses, HIV-1 and MMTV (assembling at the plasma membrane and in the cytoplasm, respectively), has shown some common features. These include the binding of the MA to the membrane by the hydrophobic interaction mediated by the myristoyl anchored to the membrane and the electrostatic interactions of MA basic residues with PI 4 ,5P 2 , both stabilizing MA orientation. Despite this similarity, we have observed lower affinity of MMTV MA to the membrane, likely a result of the different topology and the lower total charge of the HBR. This probably reflects the requirement of the monomeric Gag molecules of C-type retroviruses for stronger membrane binding, essential for their accumulation and assembly. In contrast, the MA molecules of B/D-type retroviruses assembling in the cytoplasm interact with the membrane in the context of a preassembled organized lattice. Additionally, a comparison of the membrane topology of MMTV and HIV-1 MAs by surface mapping and MD simulations has indicated that the residues located at the C terminus of HIV-1 MA could be responsible for the stabilization of proteinprotein interactions within the HIV-1 MA lattice at the plasma membrane in contrast to MMTV, where the C terminus contributes to the polar heads of PI 4 ,5P 2 within the membrane.

Experimental procedures
Materials PC, PE, PS, and PI4,5P 2 , all isolated from the porcine brain, were obtained from Avanti Lipids Inc. NAI and Pierce MS-grade trypsin protease wre from Thermo Fisher Scientific, and DHNBS was from MP Biomedicals. Other chemicals were purchased from Sigma-Aldrich.

Sample preparation
Myristoylated HIV-1 and MMTV MAs were prepared as previously described (16,76). Briefly, the MAs were expressed with a C-terminal histidine tag in the Escherichia coli BL21 (DE3) strain together with eukaryotic N-myristoyltransferase. The amino acid sequences of both proteins are shown in Fig. 2. After the production of the MA, the bacterial cells were disrupted with a One-Shot Cell Disruptor (Constant Systems) at a pressure of 2 kPa. Histidine-tagged MA proteins were bound onto a nickel-nitrilotriacetic acid (Qiagen) column and released by means of 300 mM imidazole in 50 mM phosphate buffer, pH 8. For the further purification of the protein, gel permeation chromatography in a HiLoad 26/60 Superdex 75PG chromatography column (Amersham Biosciences) with isocratic elution by phosphate buffer, pH 6.5 (20 mM sodium dihydrogen phosphate, 100 mM NaCl and 0.01% (v/v) mercaptoethanol), was used. Finally, the sample was concentrated using Amicon Ultra-15 Ultracel 5k (Millipore) to a final concentration of 0.1 mM.

Liposome-binding assay
Two types of liposomes have been prepared in this study: PCPS liposomes containing PC and PS in the molar ratio 66:34 and PCPEPSPIP liposomes containing PC, PE, PS, and PI4,5P 2 in the molar ratio 45:45:5:5. The mixtures of phospholipids were prepared by dissolving individual phospholipids in chloroform or in the mixture of chloroform/methanol/water (20:9:1) in the case of PI4,5P 2 to mix phospholipids properly. Afterward, the chloroform solution was evaporated, and the phospholipid mixture was resuspended in phosphate buffer, pH 6.5, to obtain a final concentration of 20 mg·ml 21 . Liposomes were formed using a mini-extruder (Avanti Polar Lipids, Inc.) with a 100-nm polycarbonate filter. 20 ml of HIV-1 or MMTV MA solutions were added to the same volume of solution with liposomes and incubated at room temperature for 45 min. 75 ml of 67% (w/w) sucrose in the phosphate buffer were then added, and the solution was placed in a centrifuge tube, overlaid first with 120 ml of 40% (w/w) sucrose and then with 40 ml of 4% (w/w) sucrose. The prepared gradient was centrifuged at 540,000 3 g and 4°C for 45 min, divided into three fractions, and the distribution of protein in the fractions was examined by SDS-PAGE. The gel was stained by Imperial protein stain (Thermo Scientific). The amount of the bound MA was quantified by ImageJ software (National Institute of Health).

Covalent labeling of proteins
The covalent labeling of proteins was performed like in an already-published study (39). Here we labeled free HIV-1 and MMTV MAs in solution, and both proteins bound to liposomes. The labeling of bound proteins was performed in the top fraction of the gradient obtained by LBA, containing only liposome-bound proteins. The labeling was performed directly after the LBA, and the concentration of the MA in the top fraction was determined by the SDS-PAGE analysis of the LBAs previously performed under the same conditions. In the HIV-1 MA, tyrosine together with lysine residues were modified by the 600-fold molar excess of NAI in phosphate buffer, pH 6.5, at 25°C for 1 h, whereas tryptophan residues were modified by the 120-fold molar excess of DHNBS in phosphate buffer at 25°C for 15 min. In the MMTV MA, tyrosines and lysines were modified by the 600-fold molar excess of NAI, and tryptophans were modified by the 360-fold molar excess of DHNBS. The reaction conditions were the same as in the case of the HIV-1 MA. All reactions were terminated by the agent removal using Zeba Microspin desalting columns (Pierce).

Protein digestion
The modified proteins were digested by Pierce MS-grade trypsin in the reaction buffer with trypsin to the substrate ratio 1:20 (w/w) at 37°C for 2 h. The digestion was terminated by the addition of TFA to a final concentration of 0.5% (v/v).

Nano-HPLC/nano-ESI-MS/MS
The LC-MS/MS analysis was performed by means of UHPLC Dionex Ultimate 3000 RSLC nano (Dionex) coupled with an ESI-Q-TOF Maxis Impact (Bruker ZipTip Daltonics) mass spectrometer. Prior to the analysis, the peptides were purified and concentrated by pipette tips (Millipore) containing C 18 reversed phase. The peptides were eluted by 50% acetonitrile in 0.1% TFA, air-dried, and dissolved in the solution of 0.1% formic acid and 3% acetonitrile in water. Subsequently, they were loaded into the trap column, an Acclaim PepMap 100 C18 (100 mm 3 2 cm; particle size, 5 mm; Dionex), with a mobile-phase flow rate of 5 ml min 21 of A (0.1% formic acid and 3% acetonitrile in water) for 5 min. The peptides were then separated in the Acclaim PepMap RSLC C18 analytical column (75 mm 3 150 mm; particle size, 2 mm; Dionex) and eluted with the mobile-phase B (0.1% formic acid in acetonitrile) using the following gradient: 0-5 min with 3% B, 5-35 min with 3-35% B, 35-37 min with 35-90% B, 37-50 min with 90% B, 50-52 min with 90-3% B, and 52-60 min with 3% B. The flow rate during the gradient separation was set to 0.3 ml min 21 . The peptides were eluted directly to the ESI source-captive spray (Bruker Daltonics), where the capillary voltage was set to 1400 V, the flow of drying gas was set to 3 liters min 21 , and the temperature was set to 150°C. All the measurements were performed in the positive ion mode using data-dependent analysis with precursor-ion selection in the range of 400-1400 Da; up to five precursor ions from each MS spectrum were selected for fragmentation by collision-induced dissociation. The preferred charge state of the precursors was 2-4, the dynamic exclusion was set to 30 s, and the MS/MS spectra were recorded in the range of 50-2200 m/z.
Peak lists were extracted from the raw data into an mgf file with the Data Analysis 4.1 software (Bruker Daltonics). The identification of the peptides of HIV-1 and MMTV MAs was performed by an in-house Mascot server 2.4.1, using a database containing sequences of the HIV-1 MA, the MMTV MA, and the whole proteome of E. coli BL21(DE3) strain (downloaded from Uniprot (RRID:SCR_002380) in March 2017), where MAs were produced, together with a list of standard contaminants (downloaded from MaxQuant (RRID:SCR_014485) in May 2014). N-terminal myristoylation and the oxidation of methionines were set as a variable modification for all data sets, and the acetylation of lysines and tyrosines or the modification of tryptophans by the 2-hydroxy-5-nitrobenzyl group was set as an additional variable modification when the data sets of modified proteins were evaluated. Trypsin was selected as a protease, and three missed cleavage sites were set for the evaluation of the data sets of MMTV MA, whereas six missed cleavage sites were set for the evaluation of HIV-1 MA data sets. The tolerance of 10 ppm was used for MS spectra, and the tolerance of 0.05 Da was utilized for MS/MS spectra. The significance threshold was set to 0.05, and the evaluation only included MS/ MS spectra with an ion score greater than 15. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (77) partner repository with the data set identifiers PXD018604 and 10.6019/PXD018604.

Surface mapping evaluation
The surface mapping of HIV-1 and MMTV MAs was evaluated qualitatively by a comparison of modified residues in proteins covalently labeled in solution (free proteins) and in the top fraction of the LBA gradient (liposome-bound proteins). All the experiments were performed in three replicas. When the lysine, tyrosine or tryptophan residues were modified in all three repetitions, the residue was marked as stably modified. If the residue was modified only in one or two repetitions, it was marked as weakly modified. When a nonmodified residue was detected at least once within some identified peptide, it was marked as unmodified. If some residue was not detected within any of the identified peptides, it was marked as not detected.

Coarse-grained molecular dynamics simulations
The MMTV MA (Protein Data Bank code 4ZV5) and the HIV-1 MA (Protein Data Bank code 2H3I) were mapped into the MARTINI CG representation using the martinize.py script (78)(79)(80). The ELNEDYN representation with distance cutoff (rc) = 0.9 nm and spring force constant (fc) = 500 kJ·mol 21 ·nm 22 excluding the first nine residues (Gly 2 to Lys 10 in the case of the MMTV MA, and Gly 2 to Gly 10 in the case of the HIV-1 MA) was used to prevent any undesired large conformational changes during CG-MD simulations (81). The parameters for N-myristoylated glycine were taken from Charlier et al. (21). The MARTINI CG model for all lipid molecules used in this study was taken from Ingólfsson et al. (82). Lipid bilayers, in total composed of ;250 lipid molecules with the particular compositions corresponding to the ones used in experiments, were prepared using Charmm-GUI Martini Maker (83).
All CG-MD simulations were performed in GROMACS v5 (84). Lennard-Jones and electrostatic interactions were shifted to 0 between 0.9 and 1.2 nm and between 0 and 1.2 nm, respectively. A relative dielectric constant of 15 was used. Simulations were run in NPT ensemble. The temperatures of the protein, lipids, and solvent were coupled separately at 310 K using the Berendsen algorithm, with the coupling constant of 1.0 ps. The system pressure was coupled using the same algorithm with a coupling constant of 3.0 ps, a compressibility of 3.0, and a reference pressure of 1 bar. Simulations were performed using a 20fs integration time step. Initially, the protein was placed ;1.0 nm away from the membrane. Subsequently, the standard MARTINI water and Na 1 ions were added to ensure the electroneutrality of the system. The whole system was energy-minimized using the steepest descent method up to the maximum of 500 steps, and production runs were performed for up to 3 ms. The standard GROMACS tools and in-house codes were used for the analysis. The VMD program was utilized to prepare the figures (85).

Data availability
The MS proteomics data have been deposited to the Proteo-meXchange Consortium via the PRIDE (77) partner repository with the data set identifier PXD018604 and DOI 10.6019/ PXD018604. All other supporting data are to be shared upon request from the corresponding author.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.