Structural biology of HIV integrase strand transfer inhibitors

Abstract

Integrase (IN) strand transfer inhibitors (INSTIs) are recent compounds in the antiretroviral arsenal used against HIV. INSTIs work by blocking retroviral integration, an essential step in the viral lifecycle that is catalyzed by the virally encoded IN protein within a nucleoprotein assembly called an intasome. Recent structures of lentiviral intasomes from simian and human immunodeficiency viruses clarified the INSTI binding modes within the intasome active sites and helped elucidate an important mechanism of viral resistance. The structures provide an accurate depiction of interactions of intasomes and INSTIs to be leveraged for structure-based drug design. Here we review these recent structural findings and contrast with earlier studies on prototype foamy virus intasomes. We also present and discuss examples of the latest chemical compounds that show promising inhibitory potential as INSTI candidates.

Inhibition of HIV integration as a pharmacological target

Untreated infection by the human immunodeficiency virus (HIV) causes a fatal disease called acquired immunodeficiency syndrome (AIDS) that destroys one’s immune system. Currently viral replication can be controlled by treatment with combinations of antiretroviral drugs (cART) that target essential steps of the viral lifecycle [1,2]. However, available drugs do not lead to a cure [3,4]. Drugs approved for the treatment of HIV infection [2] belong to five major classes, distinguished by their viral targets and mode of action: (i) nucleoside reverse transcriptase inhibitors [5], (ii) non-nucleoside reverse transcriptase inhibitors [5], (iii) protease inhibitors [6], (iv) entry inhibitors [7] or (v) integrase strand transfer inhibitors (INSTIs) [8].

INSTIs are the latest compounds developed within cART formulations. They work by blocking integration (see Box 1), thereby preventing the virus from establishing a permanent infection in the target cell. INSTIs are overall well tolerated by patients and are considered to be the most promising compounds for future development [8]. It is now over a decade since the first INSTI Raltegravir (RAL, licensed 2007) was approved by the U.S. Food and Drug Administration (FDA) [9]. Three other compounds have been approved in recent years: Elvitegravir (EVG, licensed 2012), Dolutegravir (DTG, licensed 2013) and Bictegravir (BIC, licensed 2018) [10-12]. Another INSTI Cabotegravir (CAB) was approved in March 2020 in Canada as a two-drug regimen [13]. Many reported cases of viral resistance variants (VRVs; see Glossary) lowering susceptibility to clinically used INSTIs generate the need to design new molecules that could potentially translate into therapies [14-17]. To facilitate design of improved compounds, it is important to understand INSTI binding at the atomic level and the structural basis of resistance causing mechanism(s). The first structures of the viral nucleoprotein assembly called an intasome from the prototype foamy virus (PFV) IN (Box 1), determined by protein crystallography provided important insights into the mechanism of retroviral integration and its inhibition by INSTIs [18-20]. Early this year, long-awaited high-resolution INSTI-bound structures of lentiviral intasomes [from HIV-1 (subsequently referred to as HIV) NL4.3 strain and SIV_rcm] were reported [21,22]. These advances were enabled by recent progress in the field of cryo-EM [23-25] and provide an accurate depiction of interactions of intasomes and INSTIs, including the location of solvent molecules. In this review, we will discuss and contrast the earlier and many novel findings corresponding to INSTI structural biology.

BOX 1. Intasomes drive retroviral integration.

Retroviral integration is the process by which retroviruses incorporate reverse transcribed viral DNA (vDNA) into the chromatin of the host cell (reviewed in [86]). Integration is catalyzed by a nucleoprotein assembly called an ‘intasome’, which consists of multimers of the viral protein integrase (IN) bound to vDNA ends. Two consecutive reactions catalyzed by IN proteins within the intasome must occur: 3′-processing and strand transfer (Box 1, Figure IA). During 3′-processing, di-or-trinucleotides are removed from the 3′ ends of vDNA leaving the reactive 3′-hydroxyls exposed. During strand transfer, IN joins the processed 3′-hydroxyl vDNA ends with phosphates on the opposing target DNA (tDNA) strands through S_N2-type nucleophilic substitution [87]. The process is then completed by the host cell’s DNA repair enzymes. The incorporated vDNA is referred to as a provirus within the host cell’s genome. Concerted integration is irreversible and establishes a permanent infection within the target cell. The requirement to catalyze two independent integration events – one for the U3 end and one for the U5 end (U3 and U5 are the 3’ and 5’ ends of viral DNA long terminal repeat (LTR) sequences, respectively) – implies that IN must minimally form a dimer on host DNA. The first intasome structures from prototype foamy virus (PFV) which has ~ 20% sequence identity to HIV IN, showed that PFV IN assembles into a tetrameric complex (Box 1, Figure IB) on the ends of two vDNA strands, with two of the four protomers catalyzing the integration reaction [18,19]. The mouse mammary tumour virus (MMTV) or Rous sarcoma virus (RSV) intasomes are octamers [88,89], and maedivisna virus (MVV) intasomes are hexadecamers [90]. Intasomes from HIV (Box 1, Figure IC) and simian immunodeficiency virus from red-capped mangabeys (SIV_rcm ; 75% sequence identity to HIV IN) appear to be more pleiotropic, forming different oligomeric states ranging from tetramers through hexadecamers, as well as polymers of incompletely assembled “proto-intasome” stacks that propagate linearly in a run-on manner [21,22,91]. Broadly speaking, the distribution of oligomeric states depends on a large variety of factors that influence intasome assembly in vitro. Despite the large range of distinct oligomeric assemblies, all intasomes share a conserved central region, termed the conserved intasome core (CIC) (Box 1, Figure I) [92]. The CIC contains the vDNA ends and two catalytic sites that perform concerted integration (Box 1, Figure I). Both catalytic sites are targeted by INSTIs (Box 1, Figure ID and IE). The vDNA ends must be recessed for INSTIs to bind (i.e. 3′-processing must have already occurred). Once bound, the INSTIs inhibit retroviral integration by directly competing for the tDNA binding site and blocking strand transfer [8]. In the structures of distinct intasome assemblies used to study INSTI binding, there are always two INSTIs bound to a CIC.

Box 1, Figure I.

(A) Schematic of intasome assembly. Under specific biochemical conditions, an oligomer of IN proteins (labeled IN_n) will assemble on the ends of vDNA to form the stable synaptic complex (SSC), which contains between 4 and 16 IN protomers. IN then cleaves two nucleotides from the 3′ ends of vDNA, forming the cleaved synaptic complex (CSC) that exposes the conserved free 3′-OH groups of the catalytically competent CA dinucleotides. CSC intasomes can capture target DNA (tDNA) to form the target capture complex (TCC), which will rapidly catalyze strand transfer to form the post-catalytic strand transfer complex (STC) in which the tDNA and the integrated vDNA are still bound to the intasome. INSTIs specifically bind to CSC intasomes and prevent the formation of the TCC, inhibiting catalysis. In the cell, the above steps occur within the larger context of the pre-integration complex (PIC). (B) Tetrameric PFV intasome (C) dodecameric HIV intasomes are shown. (D) Close-up of the intasome active site without and (E) with bound INSTI (red). The CIC is represented by surface and each protomer forming CIC is colored differently. Other protomers and DNA are in gray. Black arrows indicate the INSTI binding sites.

Development and structural biology of first-generation INSTIs

Attempts to identify HIV integration inhibitors date back ~ 25 years [26,27]. Two important discoveries toward identifying effective compounds were the demonstration that a higher-order nucleoprotein assembly is required for catalysis [28-30] and the delineation of two independent IN enzymatic activities, 3′-processing and strand transfer (Box 1) [31,32]. The initial efforts to target IN with small molecules [26,33,34] proved unsuccessful, because the assays were designed to mimic the 3′-processing (reviewed in [35]). The first effective IN inhibitors were identified using an assay based on the strand transfer [36,37]. Compounds arising from these successful screens were characterized by a β-diketo acid (DKA) moiety as their pharmacophore, e.g. L-731,988 (Figure 1A) [36], which ligate the two catalytic Mg²⁺ ions of IN (stabilized by conserved DDE motif).

Figure 1. Early INSTIs and molecular basis of their inhibitory activity.

(A) Extensive screening helped to identify first effective strand transfer inhibitors with β-diketo acid scaffold (marked in red) like compound L-731,988. (B) Chemical structures of the FDA-approved first-generation INSTIs, Raltegravir (RAL) and Elvitegravir (EVG). Magnesium chelating heteroatoms are in red and the halobenzyl group is in green. Central rings are numbered starting from the linker region that bridges the pharmacophore and halobenzyl groups. (C) Overall structure of PFV intasome showing differently colored IN subunits, viral DNA in grey, and active site regions with bound inhibitor Raltegravir (RAL) marked by arrows (PDB ID: 3OYA). (D) Close-up of the intasome active site in the apo form (PDB ID: 3OY9). The catalytic DDE residue triad, which is shown in red, consists of D128 (HIV D64), D185 (HIV D116), and E221 (HIV E152). For crystallography of this sample manganese (purple spheres) was used instead of magnesium to better resolve the ion positions within the intasome active site. (E) Zoom into the PFV IN active site, when bound to RAL (shown as magenta sticks). The RAL bound PFV intasome (PDB ID: 3OYA, vDNA, in grey). The catalytic DDE residue triad, which is shown in red, consists of D128 (HIV D64), D185 (HIV D116), and E221 (HIV E152). Three heteroatoms of the inhibitor’s pharmacophore bind the two Mg²⁺ ions (green spheres), and its fluorobenzyl group displaces the terminal adenine of vDNA (dA17) to occupy its original binding site. In consequence, the disrupted 3′-hydroxyl group is in its unreactive position (indicated by red arrow). The methyl-oxadiazole group of RAL makes a π–π stacking interaction with Y212 (corresponding to Y143 in HIV).

Modifications introduced by scientists from Merck and Japan Tabacco led to development of the first-generation INSTIs RAL (MK-0518) and EVG (JTK-303/GS-9137) that potently inhibit wild type (WT) HIV virus replication (Figure 1B) [9,38]. Their development and FDA approval came before there were any structural insights into their mechanism of action. The first crystal structures of PFV intasomes (Figure 1C) (which are more soluble, and thus more amenable to structural studies than their lentiviral counterparts), without (Figure 1D) and with (Figure 1E) INSTIs bound showed how they inhibit integration [18,19]. These structures revealed details of the PFV active site, including the role of the vDNA ends. PFV intasomes with RAL (Figure 1E) or EVG bound revealed the three important characteristics: (i) INSTIs contain a central pharmacophore with electronegative atoms positioned to engage the Mg²⁺ cofactors in the IN active site; (ii) INSTI binding displaces the 3′terminal adenine (dA17 in PFV) of the transferred vDNA strand (iii) INSTIs have a halogenated benzyl moiety that is attached to the core with a short linker. The benzyl moiety makes π–π stacking interactions with the nucleobase of the penultimate cytosine at the 3′ end of vDNA (Figure 1E) [18,20]. The interaction of INSTIs with the end of the vDNA explains why they bind and inhibit intasomes, but not free IN [37]. INSTI binding causes no major conformational changes in the intasome. Of note, the “intasome”, as the term is now used, refers to biochemically assembled complexes of IN and DNA. In the cell, such complexes would be part of a higher-order (and as yet poorly defined) nucleoprotein assembly called the pre-integration complex (PIC). Importantly, the PIC is a stable entity in cells that can persist for days, and INSTI binding accordingly stabilizes the molecules in the IN active site for prolonged periods of time [39]. The stability of the PIC in an infected cell, combined with the long dissociative INSTI half-life, renders the compounds highly active for HIV inhibition.

Resistance to first-generation INSTIs

Clinical use of RAL and EVG led to the expected development of IN mutants that conferred drug resistance. Some of the most common resistance pathways involved HIV IN residues Y143, Q148, and N155 (Y143H/R/C, Q148H/R/K, or N155H) (reviewed in [17]). Using the PFV intasome as a model system, structural efforts began shedding light on some resistance mechanisms [18,20]. RAL makes a stacking interaction with Y212 (corresponding to Y143 in HIV) [18,20] (Figure 1E). Mutant of Y143 in HIV results in the loss of π–π stacking of the methyl-oxadiazole group of RAL with the Y143 side chain [20]. This mechanism is further supported by the fact that Y143 mutants do not reduce susceptibility to INSTI compounds lacking the oxadiazole group, e.g. EVG [40]. Mutants that cause resistance to INSTIs reduce the ability of IN to carry out its normal functions. Although many of the mutants that arise in patients affect INSTI binding, some like G140S are compensatory mutants [41]. G140S improves the activity of HIV INs that carry VRVs at position 148: Q148H/K/R [42]. Crystal structures of PFV intasomes assembled using S209+S217Q/H and N224H (corresponding to G140, Q148 and N155 residues in HIV respectively) mutants have revealed slight but significant rearrangements in the IN active site region that were predicted to reduce INSTI binding affinity and/or increase the energy barrier for high-affinity binding [20]. However, the precise mechanism of resistance was clarified only recently through structures of the SIV_rcm intasome bound to the second-generation inhibitor BIC, which will be discussed later. There are likely other compensatory mutants, as evidenced from recent studies of highly favored (or entrenched) variants that develop after prolonged exposure to therapy [43].

Development and structural biology of second-generation INSTIs

To combat the rapidly emerging HIV variant strains resistant to first-generation INSTIs, new second-generation compounds were developed. An early candidate was MK2048 [40], later outperformed by compounds like DTG [12] (GSK1349572), BIC [10] (GS-9883) and CAB [44] (Figure 2A). Second-generation INSTIs inhibit broad panels of VRVs [13] and are now considered some of the best drugs in the clinic. Fundamental change in second-generation INSTIs is the introduction of a different central pharmacophore scaffold – from a bi-cyclic to tri-cyclic ring system responsible for metal chelation. Crystal structures of PFV intasomes bound with MK2048 or DTG showed that they occupy a similar space as RAL or EVG within the IN active site region, yet coordinate Mg²⁺ ions at slightly different and presumably more optimal angles [20]. As a result of this expanded scaffold, second-generation INSTIs make additional contacts with IN residues and accordingly retain potency against VRVs. In the recent SIV_rcm and HIV intasome structures [21,22], DTG and BIC (Figure 2B) make extensive contacts via their oxazinane/oxazepane rings, respectively, with N117 and G118 (in the IN β4-α2 loop), similarly to what was observed in a previous DTG-bound PFV intasome crystal structure [45]. DTG was 80-fold more potent against the HIV G140S/Q148H double mutant than a derivative drug counterpart that lacked the third ring, whereas both compounds inhibited WT HIV similarly [21]. Increased potency is attributed to longer dissociative half-life, as the analogue without the third ring more readily dissociated compared to BIC in molecular dynamics simulations [21]. Further, the flexibility of the oxazepane ring of BIC may help to retain protein-ligand interactions, which is hypothesized to explain the marginally improved performance of BIC versus DTG in some drug resistance mutant viral studies [13]. Another important factor contributing to second-generation INSTI success is the increased linker length connecting the central pharmacophore with the halobenzyl group. Consequently, the halobenzyl substituent can enter deeper into the pocket resulting from the displaced vDNA residue and more favorably interacts with the penultimate cytosine (Figure 2C). Moreover, the amide carbonyl of the halobenzyl linker is freed from metal chelation. This contributes to increased torsional flexibility of the linker, which helps to better adjust to altered mutant environments [45]. Finally, the nature and position of benzyl ring halogenation can influence inhibitor potency, as dihalo-substituted analogues are more potent than their mono-(or further)-substituted counterparts [46]. These combined changes into both DTG and BIC molecules substantially improved their potency, especially against VRVs.

Figure 2. Second-generation INSTIs and their binding characteristics.

(A) Chemical structures of selected second-generation INSTIs – MK2048, DTG, BIC and CAB. Magnesium chelating heteroatoms are in red and the halobenzyl groups are in green. Central rings are numbered starting from the linker region that bridges the pharmacophore and halobenzyl groups. (B) Superposition of BIC bound SIV_rcm intasome (PDB ID: 6RWM, green) and BIC bound HIV intasome (PDB ID: 6PUW, purple) shows extensive contacts between oxazepane ring of BIC (shown by red arrow) with N117 and G118 (in the IN β4-α2 loop) of the intasome. A10 Å zone of bound BIC is depicted. Residue labels follow the HIV/ SIV_rcm IN sequence order. Non-conserved residues in SIV_rcm vs HIV IN are shown in orange ball-and-stick representation. (C) Active site pocket of DTG (blue sticks) bound PFV intasome (PDB ID: 3S3M) superimposed with RAL (magenta sticks) bound PFV intasome (PDB ID: 3OYA) depicts the deeper penetrance of the halobenzyl substituent of DTG (depicted by the black arrow) when compared to RAL. (D) Structures of BIC bound SIV_rcm (PDB ID:6RWM, light green) and HIV (PDB ID:6PUW, purple) intasomes are superimposed. Selected water molecules are shown as small red spheres. (E) Structures of BIC bound G140S/Q148H SIV_rcm intasome (PDB ID: 6RWO, green) is shown overlaid with selected WT HIV intasome residues (PDB ID:6PUW, purple sticks), with small red spheres depicting water molecules. Upon introduction of G140S/Q148H changes, the W5 water (shown in panel D) providing secondary Mg²⁺ coordination shell is expelled from the active site (indicated by red arrow). An extended hydrogen bond network couples T138 to H148 forming a possible proton wire reinforcing the S140-H148 interaction. SIV_rcm residues I74 (L74 in HIV) and T97 are in close proximity to F121 which in turn is involved in van der Waals interactions with the carboxylate of D116. Readjustment of F121 side chain would result in perturbations to the metal-chelating cluster. Interatomic distances are given in Ångstroms and depicted with black dashed lines. In panels B-E the two Mg²⁺ ions are depicted as green spheres.

Recent cryo-EM structures of SIV_rcm and HIV intasomes determined in the presence of bound second-generation INSTIs revealed important differences to the active site that were previously unappreciated through structures of PFV intasomes [21,22]. While the immediate vicinity of the active site in proximity to the two Mg²⁺ ions is similar, there is substantial divergence further away from the metal ions. Analysis of active site residues found within 10 Å of bound BIC reveals 84% active site conservation between SIV_rcm and HIV (Figure 2B), 40% between SIV_rcm and PFV and 39% between PFV and HIV (37/44, 18/44 and 17/44 analyzed residues are conserved, respectively). Importantly, many of the VRVs that arise in patients treated with second-generation INSTI regimens [16] are scattered throughout the active site, including regions that are substantially distinct between PFV/HIV intasomes (some are also not conserved between SIV_rcm/HIV-1 INs and even between INs from HIV-1 isolates). Therefore, the recent works reached a complementary conclusion – that, although the previously used PFV model system was useful, it has limitations, and the newly acquired structural insights will benefit future drug design.

Resistance to second-generation INSTIs

Despite having markedly improved potency against VRVs arising in response to first-generation drugs, second-generation compounds are not immune to viral escape. The clearest evidence for this is the failure of DTG monotherapy (BIC monotherapy results have only been reported for short trials) [47-49]. Furthermore, complex patterns of VRVs containing three, four, and five IN amino acid substitutions are becoming increasingly more prevalent in the clinic. Changes at positions 50, 51, 66, 74, 92, 97, 118, 138, 140/148, 143, 147, 149, 151, 153, 155, 157, 230, and/or 263 are frequently encountered in response to treatment with DTG and are comprehensively documented in a recent review [16]. An up-to-date tally of all VRVs arising in response to both first- and second-generation compounds can be found in the Stanford University HIV Drug Resistance Databaseⁱ. Below, we discuss important variants from a structural biology perspective.

Q148H/K/R are some of the most widespread VRVs observed in patients and are often accompanied by G140S/A [50,51]. Early structures of mutant PFV intasomes could not fully explain the mechanism of resistance [20]. Recent structures of SIV_rcm intasome assembled using either WT (Figure 2D) or G140S/Q148H mutant IN (Figure 2E) and bound to BIC showed that there is a direct interaction between S140 and H148 [21]. The cumulative data revealed two important phenomena. First, the introduction of histidine (H) at position 148 expels a key water molecule (W5) located in the secondary coordination shell of the Mg²⁺ ions bridging two of the three catalytic carboxylates (D116, E152) and Q148. The same water molecule was also resolved in HIV intasome structures (Figure 2D) [22]. Second, the interaction with S140 increased electropositivity of H148 adjacent to E152, which redistributes the local charge around the Mg²⁺-ligand cluster and weakens the interaction between the drug heteroatoms and the metal ions. A similar phenomenon is thought to apply to arginine (R) or lysine (K) residues occurring at position 148 (Q148H/R/K pathway, [50,51]). Therefore, drug resistance is due to destabilization of Mg²⁺ coordination.

N155H was previously suggested to affect vDNA binding, and indirectly destabilize the bound INSTI [20]. Cryo-EM structures of SIV_rcm and HIV intasomes, reveal that N155 participates in the secondary coordination shell of the Mg²⁺ ions. Since histidine appears to be the exclusive variant arising in clinical settings [16], it is likely that the mechanism described for the G140S/Q148H mutant is extendable to N155H. Further structural and quantum mechanical studies should elucidate the precise mechanism. The binding to Mg²⁺ is a weakness of INSTIs, which was not fully apparent in earlier studies using PFV IN.

Long-range interactions within SIV_rcm structures led to hypotheses concerning resistance enhancing E138T or I74M/T97A [52]. An extended hydrogen bond network was found to couple T138 to H148 in the G140S/Q148H SIV_rcm intasome, forming a possible proton wire reinforcing the S140-H148 interaction (Figure 2E). Such explanation relies on a network hypothesis explored earlier for e.g. HIV protease [6], postulating that mutants located away from the active site can impact its center through the network of hydrogen-bonded interactions. Analysis of SIV_rcm residues I74 (isoleucine or leucine in HIV) and T97 revealed their close proximity to F121. This aromatic side chain is involved in van der Waals interactions with the carboxylate of D116, thus readjustment of F121 side chain would result in perturbations to the metal-chelating cluster (Figure 2E). This possible resistance mechanism would once more underline the fragility of INSTI binding.

Another factor to INSTI failure are VRVs of G118. The corresponding amino acid in PFV (G187) intasome structures was found in close van der Waals contact with bound inhibitors like MK2048. Likewise, the residue in SIV_rcm/HIV intasome structures was proximal to the third ring of DTG and BIC, respectively. Alteration to a bulky residue, such as G118R, would sterically hinder inhibitor binding [20,45]. The interaction is unique to the second-generation INSTIs, as G118R virus remains susceptible to RAL or EVG [53].

Binding of naphthyridine-based INSTI drug candidates

The naphthyridine scaffold is a pharmacophore frequently reported in compounds having various antiviral, antimicrobial or anti-inflammatory properties [54,55]. Early 1,6-naphthyridine compounds containing a carboxamide linker at position 7- (e.g. compound 22 (L-870,810); Figure 3A) were successful at inhibiting HIV IN-mediated strand transfer, but further development was halted due to toxicities [56]. A decade later, 1,8-naphthyridine derivatives emerged as attractive alternatives, with several potent and non-cytotoxic candidates for lead optimization (Figure 3A). The 1-hydroxy-2-oxo-1,8-naphthyridine-containing compounds had single digit nanomolar effective concentration 50% (EC₅₀) antiviral activities. The design was based on analogy to the biaryl-containing ribonuclease H (RNAse H) inhibitors possessing a bicyclic scaffold [57] and an observation that central hydroxyl amides allow for high affinity metal chelation. Analogues were further substantiated with the same 3-halobenzyl carboxamido group as in DTG [58], a modification that confers potency to second-generation INSTIs. Highly potent 1-hydroxy-2-oxo-1,8-naphthyridine-containing compounds were identified once hydroxyl at position 4- was substituted with an amino group (Figure 3A;>10-fold improvement). The primary amine forms an intramolecular hydrogen bond with the 3-carboxamide carbonyl, stabilizing the compound’s planar conformation and contributing to its potency [22,59]. The most promising pre-clinical drug candidates are compounds 4c and 4d (nomenclature based on [60]), which contain alcohol-derived 6- attachments to the 4-amino-1-hydroxy-2-oxo-1,8-naphthyridine core; other promising compounds include 4f [60], 6b and 6p (nomenclature: [61]), which contain a sulfonylphenyl or ester substituents at the 6- position, respectively (Figure 3A). The five compounds exhibit low cytotoxicity, inhibit WT HIV replication, and importantly, retain low nanomolar potency against a broad panel of VRVs [60]. Both 4c and 4d showed superior antiviral profiles in comparison to DTG, BIC, and CAB, with 4d emerging as the leading drug candidate [62].

Figure 3. Naphthyridine-based INSTI drug candidates.

(A) Chemical structures of 1,6-napthyridine-based (e.g. L-870,810) and 1,8-napthyridine-based compounds 4a, 4c, 4d, 4f, 6b and 6p. Heteroatoms chelating Mg²⁺ ions are depicted in red and blue. 4-amino substituent of the latter compounds is highlighted in purple. Fluorobenzyl groups are in green. Central rings are numbered starting from the linker region that bridges the pharmacophore and halobenzyl groups. (B) Superposition of 4d (PDB ID: 6PUY, light blue) and BIC (PDB ID: 6PUW, salmon) bound HIV intasome active site shows the compact pharmacophore of compound 4d allows it to bind closer to Mg²⁺ ions in comparison to BIC. (C) Superposition of 4d (PDB ID: 6PUY, light blue) and 4c (PDB ID: 6V3K, green) bound HIV intasomes with 4c bound PFV intasome (PDB 5FRN, grey) shows overall similar compound binding modes, with 6’ substituent of 4c exhibiting similar conformation to 4d while bound to HIV intasomes in comparison to distinct 4c orientation found previously in PFV intasome. Analogous IN and vDNA binding residues are labeled accordingly with HIV in green and PFV in black. 3′ terminal dA is hidden for clarity. (D) Comparison of compound 4f binding modes in PFV (PDB ID: 5FRO, light brown) and HIV intasomes (PDB ID: 6PUZ, pink). The 6’ substituent of compound 4f exhibits strikingly different conformation when bound to HIV intasome in comparison to PFV intasome. Analogous IN and vDNA binding residues are labeled accordingly with HIV in pink and PFV in brown. 3′ terminal dA is hidden for clarity. In panels B-D the two Mg²⁺ ions are depicted as green spheres.

To understand how the 4-series INSTI candidates bind in the active site, structures were determined for 4a, 4c, or 4f bound to PFV intasomes [60], as well as 4c, 4d, and 4f bound to HIV intasomes [22]. No mutant IN structures bound to 1,8-napthyridine-based compounds are currently available. The determined structures show that Mg²⁺ coordination is accomplished via the heteroatom triad of an N-hydroxyl group, a 2-oxo group and the 8-naphthyridine nitrogen. The halobenzyl group stacks with the penultimate cytosine of the vDNA similarly to other INSTIs. The compact scaffold of 4-series compounds allows them to bind closer to Mg²⁺ ions in comparison to second-generation INSTIs, which is observed in both PFV [60] and HIV [22] intasome-bound structures (Figure 3B). However, the binding modes of 1,8-naphthyridine compounds to PFV and HIV intasomes are not entirely conserved; important differences were observed in the position of the 6- substituent of 4c and 4f (and presumed for 4d), induced by local active site variations of the IN to which they bind [22] (Figure 3C and 3D). These small differences that are observed will have substantial implications to drug design. The precise geometry of hydrogen bonds between proteins and inhibitors, the exact solvation patterns and how they change upon ligand binding, and more generally the comprehensive set of interactions around the binding pocket will determine whether or not the ligand binds with acceptable affinity. Small changes to the local environment, including minor perturbations of the water network, can easily affect binding affinity by 1-2 orders of magnitude [63-66]. The exact implications of the altered orientation of the 6- substituents are not currently known. Nonetheless, the observed structural differences highlight the need to use the natural drug target (HIV IN here) to systematically study different drug candidates in the context of the WT and mutant IN protein for structure-based drug design.

High-resolution maps of HIV intasomes (−2.6-2.7Å within the CIC) also enabled determining water positions throughout the active site, with and without bound INSTIs. Comparison of the two indicated that displacement of loosely bound waters by the 6-hexanol of 4d offers an entropic gain and could, in part, account for its broad potency [22]. Water molecules were also resolved in the ~2.4-2.6Å BIC-bound SIV_rcm intasome active site [21]. Finally, it is important to note that the binding mode of 1,8-napthyridine compounds is deliberately designed to overlay with that of vDNA and/or opposing target DNA (tDNA) substrates. To maximize their potency against VRVs, these compounds occupy the molecular envelope of naturally occurring enzyme substrates, and they avoid any unnecessary protrusions outside of this natural substrate envelope (see Box 2) [22,60]. Ultimately, a detailed analysis of the protein-ligand interactions and water networks in lentiviral intasomes, together with insights from the natural substrate envelope, will aid in the development of future INSTI candidates that are broadly effective against VRVs.

BOX 2. Substrate envelope concept.

The “substrate envelope” concept was proposed by Schiffer and colleagues to rationalize why certain HIV protease inhibitors retain better efficacy against VRVs [6,93,94]. Enzyme activity depends on its ability to bind native substrates. If the inhibitor stays within the van der Waals volumes defined by the enzyme’s native substrates – the “substrate envelope” – then VRVs affecting the bound inhibitor will also affect the native substrates, and consequently viral fitness. This explains why newly designed INSTIs should bind entirely within the substrate envelope, a concept that is interchangeable with “substrate mimicry”. HIV protease inhibitors designed following this substrate mimicry concept retained potency against VRVs, whereas compounds violating the hypothesis and protruding beyond the substrate envelope were more susceptible to resistance, e.g. HIV protease inhibitor saquinavir [6].

In the HIV intasome, the substrate envelope is composed of both vDNA and tDNA (Box 2, Figure I). The HIV strand transfer complex intasome structure resolved the binding modes of two envelope components, the post-3′-processed vDNA substrate and host target DNA substrate (see PDB 5U1C) [91]. The native position of the third envelope component (dinucleotide fragment of pre-3′-processed vDNA substrate) has not yet been determined within an HIV intasome, thus its coordinates are taken from an overlaid PFV model. Developmental INSTI candidates that are broadly effective against mutant forms of IN, like compounds 4d or 4f, fit within the substrate envelope of HIV IN (Box 2, Figure IB).

Box 2, Figure I. Substrate envelope in HIV intasome.

(A) The substrate envelope of HIV intasome (shown as multicolor surface) comprises of van der Waals volumes of the post-3′-processed vDNA substrate from the HIV intasome (PDB 5U1C, dark red), a dinucleotide fragment of pre-3′-processed vDNA substrate from the PFV intasome (PDB 4E7I, light brown) and host target DNA substrate (PDB 5U1C, green). The intasome is colored by asymmetric unit (grey and dark grey). Panel A is reproduced from supplementary material of Passos et al. 2020 [22]. (B) Inset shows the zoom into the HIV intasome active site and fit of compounds 4d (PDB 6PUY, light blue) and 4f (PDB 6PUZ, pink) within the substrate envelope represented with the same surface/structure-color fashion as described in panel A. Within surface representation the corresponding DNA residues are shown as thin wire. The DDE motif residues are shown as sticks, and Mg²⁺ ions as green spheres.

We present two SIV/HIV intasome structures of the highest currently available resolution and provide a summary of the above-mentioned interactions in Figure 4, Key Figure. The two top panels show a detailed scheme of molecular interactions between the inhibitor and protein/vDNA residues within 5Å of the bound ligand. The two lower panels depict surface representations of the active sites and water molecules resolved within <5 Å distance from the respective INSTI. It is here well visible how the 6-substituent of 4d points towards the solvent exposed HIV intasome cleft. In both SIV/HIV, the good quality maps enabled resolving the metal ions and many stabilizing interactions, including relevant water molecules. Indeed, many of the waters are found in analogous positions between the two structures, further underlining the complementarity of the two works. Thus, the common insights from these new molecular signatures, including the exact ligand configurations in WT and mutant intasome forms, the position and orientation of coordinating residues, and the location of solvent molecules, will be crucial to future INSTI design efforts.

Figure 4, Key Figure. Summary of IN/DNA-INSTI interactions and representation of active site water networks based on two highest-resolution SIV_rcm and HIV intasome structures currently available in the PDB.

(A and B) Schemes showing ligand binding interactions (<5 Å away from the bound INSTI) for (A) SIV_rcm intasome bound to BIC (PDB ID: 6RWM) and (B) HIV intasome bound with compound 4d (PDB ID: 6PUY), presented using Maestro software (Schrödinger Release 2020-1: Maestro, Schrödinger, LLC, New York, NY, 2020). (C and D) Showing water molecules located <5 Å away from the bound inhibitor (shown as red spheres) in (C) SIV_rcm intasome bound to BIC (PDB ID: 6RWM, model in grey, BIC shown in orange) vs (D) HIV intasome bound to 4d (PDB ID: 6PUY, model in light blue, 4d shown in dark blue). In panels C and D the 3′-terminal dA is hidden for clarity, DDE motif residues are highlighted in bold and, Mg²⁺ ions and Cl⁻ are shown as green and blue spheres respectively.

Synthesis and evaluation of developmental INSTIs

Despite their potency against both WT virus and virus harboring single VRVs, results of clinical studies [16] and tissue culture experiments [10] have demonstrated that second-generation INSTIs (DTG and BIC) are comparatively ineffective against more complex VRV strains, such as the prevalent double mutant G140S/Q148H. This prompts the search for alternative INSTIs, especially those with complementary resistance profiles. Since the discovery of the first DKA inhibitors [36], many compounds have been screened, developed, and tested for inhibitory effects. Here, we consider studies and patents published within the past five years in which drug candidates have shown inhibitory effect at nanomolar range against both WT and single Q148H/K/R or prevalent double mutant G140S/A+Q148H/K/R variants. Such clinically relevant combination confers broad cross-resistance and appears to be a weakness common to all INSTIs [21]. Below, we describe selected strategies to novel INSTI design and briefly comment on the findings. Chemical molecules discussed are displayed in Figure 5.

Figure 5. Chemical structures of developmental INSTI candidates.

1 and 2 were developed based on RAL [67], 3 represents a 3-hydroxypyrimidine-2,4-dione (HPD) scaffold [68] and 4 shows a 2-hydroxyisoquinoline-1,3(2H,4H)-dione (HID) scaffold [69]. Compounds 5 and 6 present a hydroxyquinoline tetracyclic (HQT) scaffold [70], 7 and 8 are the tricyclic 2-pyridinone aminal lead molecules [72], and 9 and 10 are bridged tricyclic pyrimidinone-carboxamide derivatives [73]. Chemical groups expected to hydrophobically interact with penultimate nucleotide from the 3′ end of vDNA are colored in green. Heteroatoms expected to chelate Mg²⁺ ions are depicted in red and blue. Central rings are numbered starting from the linker region that bridges the pharmacophore and hydrophobic groups involved in nucleotide interactions.

Peese et al. used the chemical structure of RAL and replaced the canonical amide group with azole heterocycles to generate a more open geometry for the chelating heteroatoms to accommodate shifts in the positions of Mg²⁺ ions [67]. 1 and 2 displayed low EC₅₀ values against WT virus (6 and 10 nM, respectively) and improved potency against the double mutant G140S/Q148H (respectively 320 and 530 nM) in comparison to RAL. While the increase in EC₅₀ against the G140S/Q148H double mutant was substantial (~50-fold change, FC), the compounds remained effective against N155H (~0.3 FC) and Q148R (~5 FC) mutant viruses. A challenge will be to improve efficacy against double mutant G140S/A+Q148H/K/R strains.

Wu et al. explored pharmacophores with dual inhibitory effects against IN and RNaseH domain of reverse transcriptase (RT) to develop a 3-hydroxypyrimidine-2,4-dione (HPD) scaffold that was further optimized with distinct halobenzyl rings and other chemical groups [68]. 3 showed low EC50 values against WT virus (15 nM) and reasonable efficacy against the G140S/Q148H double mutant (195 nM). As the addition of bulky substituents such as a third ring [21] or aliphatic extensions [60], that point toward the solvent-exposed intasome cleft, contribute to retaining potency against a variety of mutants, similar modifications could provide future benefits to this group of compounds. Billamboz et al. also built upon the idea of dual-potency compounds targeting both IN and RNaseH and investigated the effects of halobenzyl to n-alkyl replacement within a 2-hydroxyisoquinoline-1,3(2H,4H)-dione (HID) scaffold [69]. The only alkyl substituent that conferred nanomolar inhibitory potency to the molecule was an n-hexane in 4. The hexane addition resulted in slightly worse EC₅₀ values when compared with the canonical halobenzyl analog (85 nM versus 56 nM) and higher IC₅₀ values for RNase H (5.35 μM and 0.36 μM). Surprisingly, HID compounds did not show any FC against Q148H, N155H or G140S/Q148H resistance mutants. This observation raises the question of whether the HID compounds are indeed bona fide INSTIs.

Velthuisen et al. combined the carbamoylpyridone third-ring pharmacophore found in successful second-generation INSTIs with a naphthyridinone scaffold to develop a hydroxyquinoline tetracyclic (HQT) scaffold [70]. Several promising compounds (5 and 6) showed low EC₅₀ values against WT (3 and 0.6 nM, respectively) HIV virus with only small changes against VRVs such as G140S/Q148H (respectively 2.6 and 3.7 FC). However, these compounds were comparatively cytotoxic (37 μM and 5 μM) when compared with CAB or BIC (>250μM) [13]. When bound to intasomes, the HQT scaffold is expected to extend towards the β4-α2 loop conferring superior potency against both WT and mutant viruses [71]. Therefore, this seems to be a promising avenue with potential to maintain long dissociative half-life even against VRVs.

Raheem et al. investigated a series of modifications based on the 2-pyridinone core of MK-0536, which produced novel tricyclic 2-pyridinone aminal lead molecules (7 and 8) [72]. These showed reasonable potency against WT (67 and 32 nM IC₅₀, respectively) and single mutant viruses (respective FCs for 7 and 8 against Q148R=1.5 and 1; Q148K=1 and 1 N155H=1.5 and 1). Both compounds contain bulky terminal substituents that superimpose well with the last ring from BIC. However, they possess a shorter linker, which may compromise stacking with the penultimate cytosine [71]. Although considerable effort has been made to improve the pharmacokinetic properties of both compounds, it would be valuable to test their efficacy against the prevalent G140S/Q148H double mutant.

Patel et al. examined modifications on the second ring of their lead compound and obtained a promising INSTI candidate 9 with low EC₅₀ for both WT (1 nM) and double mutant G140S/Q148H (10 nM) [73]. The bulky second ring is not large enough to mimic the third ring of second-generation INSTIs and establish interactions with the β4-α2 loop. A molecular docking experiment suggests that the oxamide group attached to the second ring is capable of interacting with the β5-α3 loop (Y143, N144, P145), nucleotide base of dA21, and solvent water molecules in that region. Further optimization with the addition of a meta-methyl group at the benzyl amide moiety provided additional potency to compound 10 with EC₅₀ of 0.7 nM (WT) and 3 nM (G140S/Q148H). The authors also suggest that the distal carbonyl on the oxamide group makes an H-bond interaction with the 3′OH of the terminal ribose of the vDNA, which may provide important stabilizing interactions with a conserved nucleic acid component of the substrate envelope. Future experimental structures of these compounds bound to HIV intasomes will be critical to define the proper conformations and relevant interactions associated with these inhibitors.

Other studies with INSTI candidates showing low EC₅₀ values against WT virus [74-76], including in the picomolar range [77], have also been reported. However, we chose not to elaborate on these efforts here, because the compounds were not evaluated against VRVs, which we argue is a critical gold standard moving forward. We encourage new studies to test the activity of their compounds against prevalent second generation VRVs, in particular the G140S/Q148H double mutant.

Concluding Remarks and Future Perspectives

INSTIs have become essential antiretrovirals in our arsenal against the HIV, and there continues to be much excitement for future development. Clinically approved drugs have improved the therapeutic landscape, and a single INSTI in combination with other drug(s) is now a standard within cART. However, these INSTI drugs do not act on all VRVs of HIV, laying out an unmet clinical need to develop better inhibitors. Structures of intasomes bound to INSTIs have elucidated their mechanisms of action, shed light on resistance, and are guiding current design efforts. Notably, all currently employed clinical drugs – including first- and second-generation INSTIs – as well as many developmental compounds are highly potent against the WT virus; whether they work against resistance variants is the underlying question (see Outstanding Questions). The history of HIV therapy can provide some insight here. Initial efforts to combat the virus were not effective, because the virus quickly escaped therapy. An important turning point was the advent of cART, which consists of a multi-drug regimen that remains effective against the majority of VRVs [78-80]. Modern treatment strategies are built on the idea of targeting dynamic viral populations and switching drug regimens as soon as VRVs are observed [81-83]. Effective cART blocks HIV replication, which in turn diminishes the development of new VRVs [84]. Therefore, in addition to continuing the search for novel drug targets (e.g. allosteric IN inhibitors [85]), a priority should be to develop compounds that are effective against VRVs, and future efforts need to focus on resistance. To that end, substrate mimicry can provide relevant guidance. The idea has contributed to protease inhibitors with broad efficacy against many VRVs and recently has shown applicability to INSTIs. Since many clinically important VRVs remain poorly understood, major efforts need to revolve around understanding molecular mechanisms of resistance to broad panels of INSTI-resistant viruses and using principles of substrate mimicry to guide design efforts. Structures, combined with computational, biochemical, and virology approaches, need to synergize to comprehensively address these issues. While resistance is particularly acute in HIV infection, it arises in response to therapeutic treatments to disease in general. Therefore, the ideas learned from HIV pertain more broadly to therapeutic design strategies and will help humanity stay ahead of pathogens and infectious diseases.

Outstanding Questions.

• What are the molecular mechanisms that underlie HIV resistance to integrase (IN) strand transfer inhibitors (INSTIs)?

• What are the most relevant interactions that allow an INSTI to maintain potency against the HIV G140S/Q148H double mutant?

• What are the best ways to design and/or optimize INSTIs that are broadly effective against drug resistant strains of HIV?

• What are the best scaffolds on which to develop broadly effective INSTIs?

• To what extent will staying within the substrate envelope help INSTIs broadly inhibit resistant strains of HIV?

• Is it possible to develop a single INSTI capable of maintaining potency against all known resistant variants? Or will developing drugs with complementary resistance profiles prove more pragmatic?

Highlights.

• The retroviral protein integrase forms large nucleoprotein assemblies called intasomes, which mediate the insertion of viral DNA into host chromatin, an essential step in viral replication.

• Advances in cryo-EM enabled high-resolution structures of intasomes, decipher modes of inhibitor binding, and locate solvent molecules within the active site surrounding the drug.

• In the decade since the first integrase inhibitor was approved by the FDA, viral resistant variants have emerged, demonstrating the need to develop more effective inhibitors.

• Future efforts to develop integrase inhibitors need to focus on drug resistant integrase mutants.

• Drug design will benefit from intasome-inhibitor structure-function analyses combined with complementary virological, biochemical, and computational approaches.

GLOSSARY

Compensatory mutants

secondary mutation that provides either a fitness or functional advantage to overcome limitations caused by a drug-resistance mutation.

DDE motif:

The HIV integrase active site is composed of invariant D64, D116 and E152 residues, collectively known as the DDE catalytic triad.

Viral resistance variants (VRVs)

organisms that lost sensitivity against a drug through genome mutation.

Effective concentration 50% (EC₅₀)

concentration of a drug that gives half-maximal response, used to measure inhibitory potency.

IC₅₀:

inhibitor concentration where the response (or binding) is reduced by half; measure of inhibitor potency.

Integrase (IN):

product of the pol gene, ~33 kDa polynucleotidyl transferase enzyme. The protein harbors three characteristic domains: (i) N-terminal domain (NTD) which coordinates zinc ions and forms 3-helical bundle, (ii) catalytic core domain (CCD) possessing an RNAse H fold and DDE catalytic triad, and (iii) the C-terminal domain (CTD).

Pharmacophore

part of chemical compound’s structure that is responsible for a particular biological or pharmacological interaction.

Pre-integration complex (PIC)

cellular nucleoprotein complex comprised of viral IN, viral DNA and host cell factors, which is responsible for the viral genome integration into the host cell DNA.

Retrovirus

RNA virus containing reverse transcriptase enzyme that converts its single-stranded RNA genome into double-stranded DNA for further integration into the host cell. The two characteristics that, together, distinguish retroviruses are reverse transcription and integration.

Ribonuclease H (RNase H)

family of non-sequence-specific endonuclease enzymes that hydrolyze solely the RNA component of RNA/DNA hybrids.

S_N2-type nucleophilic substitution

reaction mechanism where one bond is broken and another is formed synchronously within one step. S_N2 relates to its bi-molecular character.