Molecular mechanism of HIV-1 resistance to sifuvirtide, a clinical trial–approved membrane fusion inhibitor

Host cell infection with HIV-1 requires fusion of viral and cell membranes. Sifuvirtide (SFT) is a peptide-based HIV-1 fusion inhibitor approved for phase III clinical trials in China. Here, we focused on characterizing HIV-1 variants highly resistant to SFT to gain insight into the molecular resistance mechanism. Three primary substitutions (V38A, A47I, and Q52R) located at the inhibitor-binding site of HIV-1's envelope protein (Env) and one secondary substitution (N126K) located at the C-terminal heptad repeat region of the viral protein gp41, which is part of the envelope, conferred high SFT resistance and cross-resistance to the anti-HIV-1 drug T20 and the template peptide C34. Interestingly, SFT's resistance profile could be dramatically improved with an M–T hook structure–modified SFT (MTSFT) and with short-peptide inhibitors that mainly target the gp41 pocket (2P23 and its lipid derivative LP-19). We found that the V38A and Q52R substitutions reduce the binding stabilities of SFT, C34, and MTSFT, but they had no effect on the binding of 2P23 and LP-19; in sharp contrast, the A47I substitution enhanced fusion inhibitor binding. Furthermore, the primary resistance substitutions impaired Env-mediated membrane fusion and cell entry and changed the conformation of the gp41 core structure. Importantly, whereas the V38A and Q52R substitutions disrupted the N-terminal helix of gp41, a single A47I substitution greatly enhanced its thermostability. Taken together, our results provide crucial structural insights into the mechanism of HIV-1 resistance to gp41-dependent fusion inhibitors, which may inform the development of additional anti-HIV drugs.

showed greatly increased anti-HIV activity and genetic barrier to inducing drug resistance (25).
Among the newly designed fusion inhibitor peptides, SFT was engineered on the basis of a C34-based gp41 core structure, in which the charged glutamic acid and lysine residues were introduced into the solvent-accessible site while the residues responsible for the NHR binding were maintained unchanged (Fig. 1). Further, a serine residue was added to the N terminus of SFT to increase its helical stability, and Glu 119 was substituted by threonine to enhance its binding with the NHR pocket site. We previously determined the crystal structure of SFT bound to a target mimic peptide, which verified an electronically constrained ␣-helical peptide with significantly improved binding stability (26). As anticipated, SFT showed greatly increased activities in inhibiting both diverse subtypes of WT and T20resistant strains (19,26). In the phase I clinical trials, SFT demonstrated its good safety and pharmacokinetic profiles, especially with a dramatically prolonged half-life (19). In an advanced stage, SFT has been evaluated in clinical phase II trials and approved for clinical phase II trials in China, and thus it will hopefully become the next fusion inhibitor for clinical application. In this study, we focused on characterizing SFT-induced HIV-1 variants to gain new insights into the genetic pathways and mechanisms of HIV-1 resistance to newly developed fusion inhibitors. Several new mutations were identified as conferring high resistance to SFT and cross-resistance to T20 and C34. The mechanisms underlying the resistance were further characterized from multiple aspects, including the binding stability of inhibitors, the functionality of viral Env glycoproteins, the conformation of gp41 core structure, and the structural changes of the NHR helices. The data have provided critical information for the structure-function relationship of HIV-1 Env and gp41-dependent fusion inhibitors and will definitely aid in the development of new anti-HIV drugs.

Identification of SFT-induced resistance mutations
To characterize the structure-activity relationship of newly designed HIV-1 fusion inhibitors, we previously performed the The negatively and positively charged residues introduced for potential formation of salt bridges in SFT are indicated with solid black lines. B, the interaction between the NHR and CHR of gp41. In the current model, the CHR sequence of gp41 packs back to the NHR sequence to form a hairpin-like structure. Three hairpins associate with each other to form a 6-HB conformation. The dashed lines between the NHR and CHR indicate the interaction between the residues located at the e, g and a, d positions in the NHR and CHR, respectively. The NHR-and CHR-derived peptides are depicted as lines to express their sequences and binding sites.

HIV-1 resistance to sifuvirtide
in vitro selection of escape HIV-1 variants to SFT, in which the drug concentration was raised from 0.2 to 9,600 nM after 38 generations of virus passage over 9 months (25). To identify the genetic pathway of SFT-induced resistance, the entire env genes of resistant viruses were amplified by PCR and cloned for DNA sequencing. As shown in Fig. 2, two Env clones (M1 and M2) carry a single V38A mutation; one Env clone (M3) carries a single A47I mutation; 12 Env clones (M4 -M15) carry V38A/ A47I/N126K mutations, and three Env clones (M16 -M18) carry V38A/A47I/Q52R/N126K quaternary mutations. Therefore, four mutant viruses were finally identified from the pool of SFT-resistant HIV-1 variants. The V38A, A47I, and Q52R located within the inhibitor-binding site of gp41 might determine the resistance as primary mutations, whereas the N126K in the CHR served as a secondary mutation. After analyzing all of the cloned Env sequences, no consistent substitutions that apparently caused the resistance were identified in the other sites of gp41 or in gp120 sequence (Fig. S1), and we thus focused on the mutations on the NHR and CHR sequences of gp41.

Resistance profile of HIV-1 mutants to SFT
To identify the mutations responsible for SFT-induced resistance, we generated a panel of HIV-1 NL4-3 Env mutants carrying the amino acid substitutions either singly or in combination ( Table 1). The corresponding pseudoviruses were then produced and used in a single cycle infection assay to determine the inhibitory activity of SFT. Among the single mutations, V38A, A47I, and Q52R conferred 7.93-, 4.58-, and 39.49-fold resistance to SFT, respectively, whereas N126K had minor effect on the inducer (1.7-fold). Apparently, the mutant viruses carrying the combined mutations resulted in dramatically increased -fold changes in resistance to SFT, as shown by the mutants with V38A/A47I (41.61-fold), V38A/A47I/N126K (81.2-fold), V38A/A47I/Q52R (1107.19-fold), and V38A/A47I/Q52R/ N126K (1316.01-fold). Therefore, the mutant virus with V38A/ A47I/Q52R/N126K mutations might finally dominate the resistance phenotype to SFT.

Cross-resistance of HIV-1 mutants to newly designed fusion inhibitors that contain the M-T hook structure
We previously demonstrated that adding an M-T hook structure to the CHR-derived peptide fusion inhibitors

HIV-1 resistance to sifuvirtide
could greatly enhance their binding and antiviral activities as well as the genetic barriers to resistance (24,25,(27)(28)(29)(30)(31). Herein, it was interesting to know the cross-resistance profile of SFT-induced HIV-1 mutants to the MTSFT. As shown in Table 1, the single mutations (V38A, A47I, Q52R, and N126K) conferred little resistance to MTSFT, whereas the combined mutations (V38A/A47I, V38A/A47I/Q52R, V38A/ A47I/N126K, and V38A/A47I/Q52R/N126K) resulted in the resistance with dramatically decreased -fold changes relative to that of SFT. In particular, the single Q52R mutation could result in ϳ40-fold resistance to SFT, but it had no obvious effect on the inhibitory potency of MTSFT, verifying the importance of the M-T hook structure in overcoming the resistance.
Afterward, we also characterized the cross-resistance of SFT-induced mutants to the short-peptide fusion inhibitor 2P23 and its lipid derivative LP-19. As illustrated in Fig. 1, 2P23 is a 23-mer peptide with the M-T hook residues at its N terminus, and LP-19 has been further developed by conjugating a C16 fatty acid group to the C terminus of 2P23. Both inhibitors mainly target the gp41 pocket rather than the T20-resistant site and possess potent inhibitory activities against HIV-1, HIV-2, and simian immunodeficiency virus (SIV) (23,32). Impressively, we found that all of the mutant viruses with single or multiple mutations had no or minor resistance to 2P23 and LP-19.

Effects of the resistance mutations on the binding stability of fusion inhibitors
To understand the mechanisms underlying the SFT-selected resistance, we sought to determine the effects of the resistance mutations on the binding stability of inhibitors. To this end, the NHR-derived peptide N36 with a WT sequence (N36 wt ) or its mutants with single or multiple amino acid substitutions (N36 V38A , N36 A47I , N36 Q52R , N36 V38A/A47I , and N36 V38A/A47I/Q52R ) were synthesized as target surrogates, and the interactions between SFT and target mimic peptides were initially measured by CD spectroscopy. As shown in Fig. 3 and Table 2, CD spectra of all of the peptide pairs displayed typical double minima at 208 and 222 nm, which indicated that SFT interacted with each of the N36 peptides to form 6-HB structures. The V38A and Q52R mutations resulted in decreased ␣-helical contents of the 6-HBs, but the A47I mutation might enhance the ␣-helicity. Defined as the midpoint of the thermal unfolding transition (T m ), the thermostability of 6-HBs demon-strated the binding stability of SFT. It was found that V38A and Q52R severely reduced the T m value, whereas A47I conversely increased it.
We also examined the helical interaction and binding stability of all of the other inhibitors by CD spectroscopy (Table  2). Consistently, the V38A and Q52R mutations reduced the ␣-helical content and T m value of the 6-HBs formed by C34, T20, and MTSFT, whereas the A47I mutation did enhance their ␣-helicity and thermostability. In sharp contrast, V38A and Q52R had no apparent effect on the interactions of 2P23 and LP-19, as shown by the ␣-helical contents and T m values of their 6-HBs, whereas A47I still enhanced the binding stability of inhibitors. As shown by N36 with combined mutations, the A47I mutation could counteract the negative effects of the V38A and Q52R mutations for the binding of inhibitors.

Table 1 Resistance profiles of HIV-1 mutants to SFT and the first-generation fusion inhibitors
The experiment was performed in triplicate and repeated three times. Data are expressed as means Ϯ S.D. -Fold change in the IC 50 was determined relative to the WT level. The t test was performed to judge the significance of the difference between the WT and mutants, and p Ͻ 0.05 values are indicated in boldface type.

HIV-1 resistance to sifuvirtide Effects of SFT-induced mutations on Env-mediated cell fusion and virus entry
We next focused on investigating the effects of SFT-induced resistance mutations on the functionality of viral Env glycoprotein by two approaches. First, the infectivity of HIV-1 NL4-3 pseudoviruses carrying single or combined mutations was determined by a single-cycle infection assay, in which the cell entry efficiency of WT virus was normalized to 100% and the relative virus entry of other mutants was calculated accordingly. As shown in Fig. 4A, all of the pseudoviruses with the Env carrying single resistance mutations (V38A, A47I, or Q52R) or combined mutations (V38A/A47I or V38A/A47I/Q52R) showed markedly decreased cell entry efficiency. Apparently, the Q52R mutation dramatically impaired the functionality of viral Env, and the secondary N126K mutation exhibited a compensatory effect. Second, we measured Env-mediated cell-cell fusion by a dual split protein (DSP)-based fusion assay. It was found that single or combined mutations of V38A and A47I had minor effect on the Env's fusion activity, but Q52R resulted in a dramatic reduction. Consistently, N126K mutation efficiently displayed a compensatory role to the V38A/A47I and V38A/ A47I/Q52R mutants (Fig. 4B). The fusion capacity of diverse Envs was also measured at different time points, and the results supported the observation above (Fig. 4C).

The resistance phenotypes were not associated with the expression and processing of viral Env glycoprotein
It was critical to know whether the resistance mutations affected the expression and processing of the Env glycoproteins, thereby resulting in the phenotypes of viral infectivity and Table 2 Effects of SFT-induced mutations on the ␣-helicity and thermostability of 6-HB structure The ␣-helicity and thermostability of 6-HBs were measured by CD spectroscopy, and the experiments were repeated at least two times to verify the results. NA, not applicable for calculation due to too low (T20) or too high (LP-19) T m values. Ϫ24,881 75 NA

Figure 4. Effects of SFT-induced mutations on the functionality of HIV-1 Env.
A, the WT and mutant HIV-1 NL4-3 pseudoviruses were normalized to a fixed amount by p24 antigen, and their relative infectivity was determined in TZM-bl cells using a single-cycle infection assay. Forty-eight hours after infection, the luciferase activity was measured and corrected for background. The luciferase activity of WT HIV-1 NL4-3 was treated as 100%, and the relative activities of other mutant viruses were calculated accordingly. B, relative fusion activity of WT and mutant HIV-1 NL4-3 Envs was determined by a DSP assay. The Env-transfected HEK293T cells were used as effector cells, and U87-CD4-CXCR4 cells were used as target cells. Similarly, the luciferase activity of WT Env was treated as 100%, and the relative activities of other mutant Envs were calculated accordingly. C, fusion activity of the WT and mutant HIV-1 Envs was determined at different time points by a DSP assay. For both viral entry and fusion, data were derived from the results of three independent experiments and are expressed as mean and S.D. (error bars). t test was performed to compare the WT and mutants; * and **, p Ͻ 0.05 and p Ͻ 0.01, respectively.

HIV-1 resistance to sifuvirtide
resistance. To this aim, we first applied the human anti-gp120 antibody VRC01 and anti-gp41 antibody 10E8 in a capture ELISA-based method, as both antibodies recognize the highly conserved epitopes and possess broadly reactive virus-neutralizing activity, and their reactivity can refer to the expression profile of gp160/gp120/gp41 in the culture media and lysates of transfected cells. As shown in Fig. 5, VRC01 and 10E8 reacted equivalently with the lysates of HEK293T expressing WT and mutant Env glycoproteins that theoretically comprise gp160, gp120, and gp41. Concordantly, VRC01 detected the same level of secreted gp120 proteins in the cell culture supernatants. Further, the expression and processing of the Env glycoproteins were characterized by a Western blotting assay, in which the human anti-HIV polyclonal antibody HIV-IG or mAb HY54 was used to probe the cleaved and uncleaved Env glycoproteins (Fig. 6). The results verified that the resistant mutants had no obvious changes in the protein expression and processing (Fig.  S2). As expected, N126K could abolish the glycosylation site in gp41 (33). We also detected the WT and mutant Env glycoproteins expressed on the surface of transfected cells by flow cytometry and immunofluorescence assays. As shown in Fig. 7 and Fig. S3, no significant differences were observed either. Combined, the data suggested that the observed resistance phenotypes were not determined by the expression profile of viral Env glycoprotein.

The resistance mutations might change the conformation of gp41 core structure
The 6-HB conformation was initially determined by the NHR-derived peptide N36 and the CHR-derived peptide C34 (4), thus representing the core structure of gp41 that plays an essential role in viral fusion and entry. Here, we were intrigued to know whether SFT-selected mutations affected the functionality of viral Env through changing the conformation of gp41 core structure. We also applied two approaches that are routinely used in our laboratory. First, we used a native PAGE (N-PAGE)-based method to visualize the 6-HB complexes formed between diverse N36 and C34 peptides. As shown in Fig. 8, N36 and its mutants showed no band in the native gel because they could migrate up and off the gel due to their net positive charges, whereas the negatively charged C34 and its N126K mutant (C34 N126K ) showed specific bands. Whereas N36 or its mutants were mixed with C34 or C34 N126K , there existed specific bands corresponding to the 6-HB complexes. Densitometric analysis suggested that all the resistance mutations had little effect on the formation of 6-HBs (Fig. S4).
Subsequently, three 6-HB conformation-specific monoclonal antibodies (NC-1, 17C8, and 2G8) were used as probes in an ELISA-based assay (34 -36). As shown in Fig. 9, whereas the V38A mutation might not significantly affect the reactivity of 6-HB with three antibodies, the A47I mutation resulted in a 6-HB with largely increased reactivity, and the Q52R mutationbearing 6-HB only displayed marginal reaction. The binding  HIV-1 resistance to sifuvirtide epitopes of three antibodies have not been finely characterized, but previous studies suggested that NC-1 might mainly target the gp41 NHR (37,38), whereas the critical binding residues of 17C8 and 2G8 were localized at the N-terminal portion of the NHR and the C-terminal portion of the CHR (39). Thus, the reaction patterns of diverse 6-HBs with three antibodies implied the increased or decreased exposure of antigenic epitopes, which might correlate with the structural changes of 6-HB core.

The resistance mutations changed the ␣-helicity and stability of NHR helix
Based on the results above, we sought to delineate the effects of the resistance mutations on the ␣-helicity and thermostability of N36 peptide in the absence of inhibitors, with the intention to infer the structural properties of the viral NHR helices. As measured by CD spectroscopy (Fig. 10, A and B), the WT N36 showed an ␣-helical content of 79% and a T m value of 56°C. In sharp contrast, N36 with the V38A mutation only displayed 18% ␣-helicity, and its T m could not be defined; N36 with A47I displayed 66% ␣-helicity, but its T m increased to 62°C; N36 with Q52R displayed 40% ␣-helicity with a T m of 46°C. Interestingly, N36 with V38A/A47I double mutations showed 66% ␣-helicity with a T m of 60°C, and N36 with V38A/ A47I/Q52R triple mutations had 47% ␣-helicity with a T m of 49°C, which suggested that the A47I mutation could counteract the disruptive effects by the V38A and Q52R mutations.
We recently reported the selection and characterization of HIV-1 variants resistant to the CHR-derived fusion inhibitors MTSC22EK and SC29EK, which also identified key resistance mutations in the inhibitor-binding NHR helices of gp41 (34 -36). To conceptualize that the resistance mutations can directly change the conformation and stability of the gp41 NHR helices, thereby critically determining the binding affinity of inhibitors

HIV-1 resistance to sifuvirtide
and HIV-1 infectivity, we back-checked the effects of MTSC22EK-and SC29EK-induced resistance mutations on the ␣-helical structure of N36 itself by CD spectroscopy. In general, the single or combined resistance mutations could markedly reduce the ␣-helicity and thermal stability of N36 (Fig. 10, C-F). For example, the combination of two MTSC22EK-induced mutations, E49K and L57R, resulted in the ␣-helical content and T m value not applicable for calculation. As an exception, a single N43K mutation rather increased the ␣-helical content of N36 but it did not affect its T m value.

The secondary N126K mutation stabilized the 6-HB structure of SFT-resistant mutants
The N126K substitution in the CHR of gp41 easily emerges in the drug-resistance selection, and it has been considered a secondary mutation that can compensate for the fusion kinetics of resistant viruses. Again, our above data indicated that the N126K mutation exhibited obvious compensatory roles to the fusion and entry capacity of the V38A/A47I and V38A/A47I/ Q52R mutants. Therefore, we sought to further characterize its effect on the binding of C34 with diverse target surrogates, from which we could infer the interactions between the NHR and CHR helices of mutant viruses. As determined by CD spectroscopy, all of the 6-HBs formed between the peptides C34 N126K and N36 or its mutants exhibited relatively increased T m values, which indicated their enhanced ␣-helical thermostability (Fig.  11). For instance, whereas the N36 V38A -C34 -based 6-HB had a T m of 54°C, the N36 V38A -C34 N126K -based 6-HB had a T m of 59°C, and whereas the N36 Q52R -C34 -based 6-HB had a T m of 56°C, the N36 Q52R -C34 N126K -based 6-HB had a T m of 60°C. It should be noted that the A47I mutation-mediated 6-HB  . Effects of SFT-induced mutations on the conformation of gp41 core structure. The reactivity of 6-HBs formed by C34 and N36 or its mutants with conformation-dependent mAbs NC-1 (A), 17C8 (B), and 2G8 (C) was tested by ELISA. The peptide mixture was used to coat the plate wells at 10 g/ml, and the final concentration of a tested mAb was 5 g/ml. Data were derived from three independent experiments and are expressed as mean and S.D. (error bars). A t test was performed to compare the WT and mutants; * and **, p Ͻ 0.05 and p Ͻ 0.01, respectively.

HIV-1 resistance to sifuvirtide
thermostability could be further enhanced by the N126K mutation (Fig. 11C).

Structural basis of SFT-induced resistance
We previously determined the crystal structure of SFT in complex with the target peptide N36 (Protein Data Bank accession number 3VIE), which provided a structural basis for the mechanism of action of SFT (26). Herein, we analyzed the resistance mutations based on the atomic interactions of the 6-HB structure formed by N36 and SFT. Clearly, the residue Val 38 is located at the e position of a helical wheel model, and its hydrophobic side chain interacts with a hydrophobic patch at the terminal portion of SFT, which is formed by the side chains of Asn 145 , Glu 146 , and Leu 149 (Fig. 12A). Because the alanine side chain is much smaller than a valine side chain, the V38A mutation could significantly reduce the area of the interaction interface between V38A and the hydrophobic patch; hence, it accounted for the loss of binding affinity of SFT. In the helical wheel, the residue Ala 47 stays at a g position, where its side chain is located near a hydrophobic patch in the middle region of SFT constituted by the side chains of Ile 131 (distance 4.6 Å), Ile 134 (distance 3.8 Å), and Leu 135 (distance 4.4 Å) (Fig. 12B). Due to the small side chain of Ala 47 , there is a gap between Ala 47 and the hydrophobic patch of SFT. Replacing Ala 47 by isoleu-cine increased the size of the side chain and filled the space between the N36 trimer and SFT; thus, the A47I mutation serves to increase the interfacial area between the NHR trimer and SFT. Because isoleucine has a stable ␤ branched side chain and it is deeply buried inside the hydrophobic core of 6-HB, the A47I mutation may also contribute to the stabilization of the NHR trimer itself, which is consistent with the elevated T m values of 6-HB or the isolated N36 bearing A47I determined by CD spectroscopy. The side chain of residue Gln 52 , which is located at the e position in the CHR helix, mediates several specific hydrogen bonds with both SFT and the adjacent NHR simultaneously (Fig. 12C). Whereas the N⑀2 atom of Gln 52 donates a hydrogen bond (distance, 3.0 Å; angle, 149º) to the OH group of residue Tyr 132 of SFT, the O⑀1 atom of Gln 52 accepts another hydrogen bond (distance, 2.5 Å; angle, 160º) from the N⑀2 atom of Gln 51 on the adjacent NHR helix. Additionally, the O⑀1 atom of Gln 52 accepts the third hydrogen bond from the C␦1 atom (distance, 2.7 Å; angle, 152º) of Ile 131 . Therefore, Gln 52 not only directly interacts with SFT but also stabilizes NHR trimeric coiled-coil; thus, replacing Gln 52 with an arginine could result in the loss of binding affinity with SFT as well as the destabilization of the NHR trimer itself.

HIV-1 resistance to sifuvirtide
From the virus perspective, we analyzed the resistance mutations on the crystal 6-HB structure of C34/N36 (Protein Data Bank accession number 1AIK), which contains the native sequences of gp41 core. As anticipated, the V38A mutation has similar effects on the interaction of N36 and C34, as C34 and SFT share a hydrophobic patch formed by the side chains of Asn 145 , Glu 146 , and Leu 149 . Clearly, the side chain of Val 38 interacts with the side chain of Ile 37 to form a hydrophobic layer (Fig.  12D), whereas three Ile 37 residues on N36 helices interact with each other, which critically determines the conformation and stability of the inner NHR helices; thus, the V38A mutation would dramatically destroy the helical structure of N36 itself. As compared with that of the N36 -SFT complex, the space and distance between Ala 47 on N36 helices and the hydrophobic patch formed by Ile 131 (distance 5.3 Å), Leu 134 (distance 4.3 Å), and Ile 135 (distance 5.2 Å) on C34 helices are much larger (Fig.  12E); thus, it is conceivable that the A47I mutation would enhance the helical interactions of N36 and C34 more efficiently, consistent with the difference in the increased T m values of N36 A47I -SFT (2°C) and N36 A47I -C34 (9°C). Similar to the structure of N36 -SFT, the side chain Gln 52 can form hydrogen bonds with Gln 51 on another N36 and Ile 131 on C34, but it does not contact with His 132 (corresponding to Tyr 132 on SFT) due to a far distance (Fig. 12F). However, the Q52R mutation might not only disrupt the critical hydrogen bond network in the middle region of 6-HB but also introduce an electrostatic repulsion with His-132.
To gain insights into the mechanism of the secondary N126K mutation-mediated compensation for the 6-HB stability and viral infectivity, we also analyzed its potential effects on the atomic interactions of viral 6-HB as mimicked by N36 and C34 peptides. As shown in Fig. 12G, Asn 126 is located at the c position in the CHR helix, and its site chain faces the outside of the binding surface of C34; thus, it is difficult to define the function of Asn 126 in the interhelical interaction of 6-HB structure; however, we can see that a water molecule links Asn 126 and Glu 119 together via hydrogen bonds, which would stabilize the N-terminal PBD of C34 and thus facilitate its interactions with the deep pocket on the NHR helices. Conceivably, the relatively longer and positively charged side chain of lysine might disrupt the water-mediated connection but introduce salt bridges or hydrogen bonds with the negatively charged residues Glu 119 and/or Glu 123 , which can strengthen the intra-and interhelical interactions. We expect a crystal structure for the N36 -C34 N126K complex that can delineate the effect of the N126K mutation in detail.

Discussion
In this study, we dedicated our efforts to exploring the genetic pathway and underlying mechanisms of HIV-1 resistance to the peptide fusion inhibitor SFT. Three primary mutations (V38A, A47I, and Q52R) on the NHR helix and one secondary mutation (N126K) on the CHR helix of gp41 were identified as conferring high resistance to SFT alone or in combination, with a mutant virus with V38A/A47I/Q52R/N126K possibly dominating the resistance. The cross-resistance profiles of SFT-induced mutants to the drug T20, the template C34, and three newly designed fusion inhibitors (MTSFT, 2P23, and LP-19) were also characterized. Impressively, the single and combined SFT-induced mutations did confer cross-resistance to T20 and C34, but they were largely tolerated by the M-T hook structure-modified SFT and short-peptide inhibitors mainly targeting the gp41 pocket site. In the underlying mechanisms, the V38A and Q52R mutations could reduce the binding stability of SFT, C34, and MTSFT, but they had no obvious effects on the binding of 2P23 and LP-19, and the A47I mutation, in sharp contrast, enhanced the binding of inhibitors. Furthermore, the results show that the primary resistance mutations could impair the functionality of viral Env to mediate cell entry/fusion and change the conformation of the gp41 core as well as its internal NHR helices, whereas the secondary N126K mutation did display compensatory roles in both the virus entry and the 6-HB stability. Therefore, the present studies have provided important information for understanding the mechanism of HIV-1 resistance to gp41-dependent fusion inhibitors and would help in the development of new anti-HIV drugs.
The peptide drug T20 remains the only membrane fusion inhibitor available for treatment of viral infection, which has shown effectiveness as a salvage therapy for HIV/AIDS patients who failed to respond to antiretroviral therapeutics that include reverse transcriptase inhibitors and protease inhibitors; however, T20 behaves with relatively weak antiviral activity and a low genetic barrier to inducing drug resistance. The resistance profile of T20 was initially characterized by in vitro selection of escaping HIV-1 variants, which revealed that the 36 GIV 38 motif in the inhibitor-binding site was a hot spot (18). Later, a number of clinical studies demonstrated that T20-resistant mutations appeared predominately within the amino acid Gly 36 -Leu 45 stretch on the NHR of gp41, such as G36D/V/S, I37T, V38A/ E/M, Q40H, N43D/K, and L54M mutations (13)(14)(15)(16)(17)(18). There have been tremendous efforts to develop next-generation fusion inhibitors that can overcome the drawbacks of T20, resulting in a group of new peptides with significantly improved pharmaceutical profiles (19 -24, 28). Meanwhile, considerable effort was also devoted to the selection and characterization of HIV-1 variants resistant to newly designed fusion inhibitors, which delineated their resistance evolution pathways and underling mechanisms (34, 35, 40 -47). Previously, Liu et al. (42) identified a panel of SFT-induced resistance mutations largely overlapping with the T20-resistant sites, including I37T, V38A/M, Q41H/K/R, and N43K. As noticed, the concentration of SFT was escalated to ϳ1,000 nM after 9 -15 generations of in vitro virus passage, and the most resistant HIV-1 variants with combined mutations (e.g. I37T/V38A and I37T/N43K/N126K) displayed ϳ100-fold increased resistance to SFT (42). In fact, our research project was originally focused on comparing the genetic resistance barriers of the M-T hook structuremodified peptide inhibitors rather than identifying the resistance mutations in detail (25); however, we had been intensely curious about the HIV-1 variants that could survive in the presence of ϳ10,000 nM SFT inhibitor. Therefore, the present study was performed and surprisingly found a different panel of SFTresistant mutations, including two sites (A47I and Q52R) that had not been defined previously. In brief, three primary mutations (V38A, A47I, and Q52R) located at the inhibitor-binding site and one secondary mutation (N126K) located at the CHR of gp41, corresponding to four HIV-1 variants, contributed to the resistance phenotype. Two mutant viruses carried a single V38A or A47I mutation, whereas two mutant viruses evolved with combined mutations (V38A/A47I/N126K and V38A/ A47I/Q52R/N126K). From the sequence characterization, one might speculate a genetic resistance pathway with the following scenario. A single V38A or A47I mutation emerged first, which resulted in a modest resistance (5-8-fold); subsequently, the V38A/A47I double mutations emerged, thus conferring a significantly increased resistance (ϳ80-fold). It might require a prolonged selection, but the appearance of Q52R mutation on the background of V38A/A47I could dramatically boost the

HIV-1 resistance to sifuvirtide
resistance phenotype (Ͼ1,000-fold). Apparently, the mutant virus with two or three combined mutations (V38A/A47I or V38A/A47I/Q52R) was accompanied by the second N126K mutation, which also contributed to the resistance levels. Therefore, it is conceivable that the mutant virus with V38A/ A47I/Q52R/N126K would finally dominate the resistance.
In this study, we found that SFT-induced primary mutations, either singly or in combination, mediated considerable crossresistance to the drug T20 and the template peptide C34, except for the A47I mutation, which could result in an increased susceptibility to T20 (ϳ3-fold). Interestingly, all of the HIV-1 mutants with a single mutation (V38A, A47I, or Q52R) had no obvious resistance to the M-T hook structure-modified SFT, and all of the HIV-1 mutants with combined mutations (V38A/ A47I, V38A/A47I/Q52R, V38A/A47I/N126K, and V38A/A47I/ Q52R/N126K) displayed dramatically decreased resistance -fold changes (Table 1). Actually, we are still wondering why the single Q52R mutation could render a ϳ40-fold resistance to SFT but did not affect the inhibitory activity of MTSFT. More impressively, both the single and combined mutations did not confer significant resistance levels to the short-peptide fusion inhibitors 2P23 and LP-19, which also contain the M-T hook structure and mainly target the gp41 pocket site. It is worth noting that Q52R did confer a mild resistance to 2P23, but this phenotype could be mitigated by its combination with other mutations and by a 2P23-based lipopeptide . Taken together, the presented data have further verified that both the M-T hook structure and lipid conjugation are important strategies to design HIV-1 fusion inhibitors possessing markedly improved antiviral activity against WT and mutant viruses, especially for short peptides that target the gp41 pocket rather than the T20-and SFT-resistant sites. It will be interesting to select and characterize HIV-1 mutants resistant to MTSFT, 2P23, and LP-19, which would definitely help in our understanding of the resistance mechanisms of diverse HIV-1 fusion inhibitors.
Previous studies have described multifaceted mechanisms of HIV-1 resistance to peptide fusion inhibitors, including small amino acid-mediated reduced contact, large amino acidmediated steric obstruction, acidic amino acid-mediated electrostatic repulsion, basic amino acid-mediated electrostatic attraction, and disruption of hydrogen bonds and hydrophobic contacts (48). It was also proposed that kinetically restricted entry inhibitors have reduced sensitivity to affinity-disrupting resistance mutations (49,50). Here, we investigated the underlying mechanisms of SFT-induced resistance mutations from two aspects: their effects on the binding stability of inhibitors and on the functionality of viral Env glycoprotein. Clearly, the V38A and Q52R mutations reduced the binding stability of SFT, C34, and MTSFT but not that of 2P23 and LP-19; however, the A47I mutation could further stabilize the binding of each inhibitor. Based on the crystal structures of SFT and C34 bound to a target surrogate, we can see that the residue Val 38 makes critical contacts with the residues Asn 145 , Glu 146 , and Leu 149 of SFT. Thus, the V38A mutation would destroy the interhelical interactions through a mechanism of small amino acid-mediated reduced contact; the residue Ala 47 has hydrophobic interactions with Ile 131 , Ile 134 , and Leu 135 on SFT; the A47I mutation would introduce an extended side chain that enhances the interhelical interactions; the residue Gln 52 critically stabilizes the NHR helices via a hydrogen bond network and interacts simultaneously with Tyr 132 and Ile 131 on SFT; and the Q52R mutation would disrupt both the intra-and interhelical interactions, with a mechanism of large amino acidmediated steric obstruction or basic amino acid-mediated electrostatic repulsion/attraction. From a viral perspective, the resistance mutations could severely impair the ␣-helicity and stability of viral NHR helices and/or 6-HB conformation, which might correlate with the functionality of the Env glycoprotein to mediate viral cell fusion and entry. It is conceivable from an evolutionary angle that the mutant virus would like to find a balance among the mutations, resistance, and viral fitness.
In conclusion, our studies have demonstrated the underlying mechanism of HIV-1 resistance to SFT and verified the importance of the M-T hook structure in overcoming the resistance. Finally, there are several points that should be addressed or considered in future studies. First, can the primary resistance mutations induced by CHR-derived fusion inhibitors universally result in a structural change in the NHR region of gp41? In other words, do the changes in the conformation and stability of NHR essentially determine the resistance by interfering with the binding of inhibitors and the interaction of viral NHR and CHR helices? We are in the process of examining the effects of T20-, C34-, and SC34EK-induced resistance mutations, which may help to generalize a proof of concept; Second, can some of the resistance mutations in the NHR region also play compensatory roles in viral infectivity? To this point, the A47I and E49K mutations did behave like the secondary N126K mutation that can enhance the NHR-CHR interactions. Third, how do we explain the fact that whereas the A47I mutation enhances the binding of SFT, it still mediates the resistance? Now, one may speculate that this mutation can enhance the interaction of viral NHR and CHR helices more efficiently, thus outcompeting the binding of inhibitor. When enhancing the binding of T20 more efficiently, the A47I mutation does increase the sensitivity of drug (Table 1). Fourth, many mutations in the NHR and CHR of gp41 were previously reported to affect the expression and processing profiles of viral Env glycoproteins, but our studies showed that the fusion inhibitor peptide-induced resistance mutations had no such effects. Thus, it will be interesting to clarify this point as a common phenomenon or a misleading coincidence. Fifth, it is critical to characterize whether fusion inhibitor-selected NHR mutations impact the structure and function of HIV-1 Rev response element, as described by previous studies (51,52). Additionally, as the fitness of the resistant viruses to SFT seems to be very poor compared with the parental virus, we are wondering if these viruses can survive in vivo.

Cell lines and culture conditions
HEK293T cells were purchased from the American type culture collection (ATCC) (Manassas, VA); TZM-bl indicator cells stably expressing CD4 and CCR5 along with endogenously expressed CXCR4 were obtained from John C. Kappes and Xiaoyun Wu through the AIDS Reagent Program, Division of HIV-1 resistance to sifuvirtide AIDS, NIAID, National Institutes of Health (17). Both cells were cultured in complete growth medium that consisted of Dulbecco's minimal essential medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin-streptomycin, 2 mM L-glutamine, 1 mM sodium pyruvate, and 1ϫ minimum Eagle's medium nonessential amino acids (Gibco/Invitrogen) and were maintained at 37°C in 5% CO 2 .

Synthesis of fusion inhibitor and target mimic peptides
A panel of fusion inhibitor peptides (SFT, T20, C34, MTSFT, 2P23, and LP-19) and target mimic peptides (N36, N36 V38A , N36 A47I , N36 Q52R , N36 V38A/A47I , and N36 V38A/A47I/Q52R ) were synthesized with a standard solid-phase Fmoc (N-(9-fluorenyl) methoxycarbonyl) method as described previously (30). All peptides were protected by N-terminal acetylation and C-terminal amidation. To synthesize LP-19, the template peptide 2P23 contains a lysine residue at its C terminus with a Dde side-chain protecting group, enabling the conjugation of a C16 fatty acid that requires a special deprotection step in a solution of 2% hydrazinehydrate/N,N-dimethylformamide. The peptides were purified to a purity of Ͼ95% by reversed-phase HPLC and characterized for correct amino acid composition by MS. Concentrations of the peptides were measured by UV absorbance and a theoretically calculated molar extinction coefficient based on tryptophan and tyrosine residues.

Generation of HIV-1 Env mutants
A panel of HIV-1 NL4-3 Env-based mutants was generated by site-directed mutagenesis as described previously (53). In brief, two primers contained the desired mutation and occupied the same starting and ending positions on opposite strands of plasmid. DNA synthesis was performed by PCR in a 50-l reaction volume using 1 ng of denatured plasmid template, a 50 pM concentration of upper and lower primers, and 5 units of the highfidelity polymerase PrimeStar (TaKaRa, Dalian, China). PCR amplification was performed for one cycle of denaturation at 98°C for 5 min, followed by 18 cycles of 98°C for 15 s and 68°C for 15 min, with a final extension at 72°C for 10 min. The amplicons were digested with DpnI at 37°C for 1 h, and DpnI-resistant molecules were recovered by transforming Escherichia coli strain DH5␣ to antibiotic resistance. The introduced mutations were confirmed by DNA sequencing.

Single-cycle infection assay
The backbone plasmid pSG3 ⌬env that encodes an Env-defective, luciferase-expressing HIV-1 genome was obtained from John C. Kappes and Xiaoyun Wu through the AIDS Reagent Program, Division of AIDS, NIAID, National Institutes of Health (17). HIV-1 pseudoviruses were generated as described previously (53). In brief, 293T cells were cotransfected with pSG3 ⌬env and a plasmid expressing Env glycoprotein. Supernatants were harvested 48 h after transfection, and tissue culture ID 50 values were determined in TZM-bl cells. Peptides were prepared in graded concentrations, mixed with 100 tissue culture ID 50 of viruses, and then incubated for 1 h at room temperature. The mixture was added to TZM-bl cells (10 4 cells/ well) containing DEAE (final concentration 15 l/ml) and incubated 48 h at 37°C. The luciferase activity was measured using luciferase assay reagents and a luminescence counter (Promega, Madison, WI).

DSP-based cell fusion assay
A DSP-based cell-cell fusion assay was performed to determine the fusogenic activity of viral Env as described previously (54,55). The plasmids DSP 1-7 and DSP 8 -11 were kindly provided by Zene Matsuda at the Institute of Medical Science of the University of Tokyo (Tokyo, Japan) Briefly, a total of 1.5 ϫ 10 4 293T cells (effector cells) were seeded on a 96-well plate, and a total of 8 ϫ 10 4 U87-CXCR4 cells (target cells) were seeded on a 24-well plate. On the following day, 293T cells were transfected with a mixture of an Env-expressing plasmid and a DSP 1-7 plasmid, and U87-CXCR4 cells were transfected with a DSP 8 -11 plasmid. Twenty-four hours post-transfection, the target cells were resuspended in 300 l of prewarmed culture medium containing EnduRen live-cell substrate (Promega) at a final concentration of 17 ng/l and then transferred to each well of the effector cells with equal volume. The cell mixture was spun down, and luciferase activity was measured by a luminescence counter (Promega).

CD spectroscopy
CD spectroscopy was conducted to determine the ␣-helicity and thermostability of the isolated or combined peptides as described previously (22). CD spectra were acquired on a Jasco spectropolarimeter (model J-815) using a 1-nm bandwidth with a 1-nm step resolution from 195 to 260 nm at room temperature and corrected by subtraction of a solvent blank. The ␣-helical content was calculated from the CD signal by dividing the mean residue ellipticity ([]) at 222 nm by the value expected for 100% helix formation (Ϫ33,000 degrees cm Ϫ2 dmol Ϫ1 ). Thermal denaturation was performed by monitoring the ellipticity change at 222 nm from 20 to 98°C at a rate of 1.2°C/min.

Detection of viral Env glycoprotein by capture ELISA
A capture ELISA was conducted to determine the effects of the introduced mutations on the expression and processing of viral Env glycoprotein, as described previously (35). Briefly, the wells of an ELISA plate were coated with a sheep anti-gp120 antibody (D7324) at 10 g/ml and blocked by 3% BSA. Cell lysates or culture supernatants (50 l) of Env-transfected cells were added to the wells and incubated at 37°C for 1 h. After extensive washes, 50 l of human anti-gp120 mAb VRC01 or anti-gp41 mAb 10E8 was added at 10 g/ml and incubated at 37°C for 1 h. The bound antibodies were detected by horseradish peroxidase-conjugated goat anti-human IgG. The reaction was visualized by adding 3,3,5,5-tetramethylbenzidine, and the A 450 was measured.

Detection of 6-HBs by conformation-specific antibodies
To detect the 6-HBs formed by N36 and C34 peptides, three 6-HB conformation-specific mAbs (NC-1, 17C8, and 2G8), which react with the N36 and C34 complex but not the isolated peptides, were applied in an ELISA as described previously (34 -36). NC-1 was kindly provided by Shibo Jiang (Lindsley F. Kimball Research Institute of the New York Blood Center, New York) (56); 17C8 and 2G8 were kindly provided by Yinghua

HIV-1 resistance to sifuvirtide
Chen (School of Life Sciences of Tsinghua University, Beijing, China) (39). Briefly, the isolated or mixed peptides were coated to the ELISA wells at 10 g/ml and blocked with 3% BSA. After washing, the anti-6-HB mAb diluted at 5 g/ml was added to the wells and incubated at 37°C for 1 h. After washes, the bound antibodies were detected by horseradish peroxidase-conjugated anti-mouse IgG, the reaction was visualized by the addition of 3,3,5,5-tetramethylbenzidine, and the A 450 was measured.

Western blotting
To verify the expression and processing profile of viral Env glycoprotein by Western blotting, the lysates or culture supernatants of transfected cells were centrifuged at 20,000 ϫ g at 4°C for 15 min to remove insoluble materials. Equal amounts of total proteins were separated by 10% SDS-PAGE and then transferred to a nitrocellulose membrane. After blocking with 5% nonfat dry milk solution in Tris-buffered saline (TBS, pH 7.4) at room temperature for 1 h, the membrane was incubated with the human anti-HIV polyclonal antibody HIV-IG (obtained through the AIDS Reagent Program, Division of AIDS, NIAID, National Institutes of Health) or human anti-gp120 mAb HY54 (which was generated in our laboratory) overnight at 4°C. The following day, the membrane was washed three times in TBS-T and then incubated with IRDye 800CW goat anti-human IgG at room temperature for 2 h. The membrane was then scanned using the Odyssey IR imaging system (LI-COR Biosciences, Lincoln, NE). Band intensities were analyzed using ImageJ software (National Institutes of Health).

Flow cytometry assay
The WT or mutant Env glycoproteins expressed on the surface of transfected cells were detected by flow cytometry. Briefly, HEK293T cells were transiently transfected with a specific plasmid and harvested at 36 h after transfection. After two washes with PBS, the human anti-gp120 mAb VRC01 was added to the cells and incubated at 4°C for 1 h. Then cells were washed twice and incubated with Alexa Fluor 488 anti-human IgG (Life Technologies, Inc.) at 4°C for 1 h. After three washes, cells were resuspended in PBS and analyzed by a FACSCalibur (BD Biosciences).

Immunofluorescence assay
An Env-encoding plasmid was transiently expressed in HEK293T cells by transfection. After 36 h, the cells were immobilized by 4% paraformaldehyde, and then 1% BSA was added to block nonspecific binding of the antibodies. After washing twice with PBS, VRC01 was added to the cells at a final concentration of 20 g/ml and incubated at 4°C overnight. After three washes, bound antibodies were detected by Alexa Fluor 488 anti-human IgG and observed under an immunofluorescence microscope.

N-PAGE
N-PAGE was conducted to determine the interactions between N36 and C34 with native or mutant sequences as described previously (22,57). Briefly, N36 was mixed with C34 with the final concentration of each peptide at 40 M and then incubated at 37°C for 30 min. The sample was added with Trisglycine native sample buffer at a ratio of 1:1 and then loaded onto 10 ϫ 1.0-mm Tris-glycine gels (20%) at 25 l/well. Gel electrophoresis was carried out with 100-V constant voltage at 4°C for 3 h. The gel was then stained with Coomassie Blue and imaged with a Bio-Rad imaging system.