Retroviral intasomes arising

Abstract

Retroviral DNA integration takes place within the context of the intasome nucleoprotein complex. X-ray crystal structures of functional spumaviral intasomes were previously revealed to harbor a homotetramer of integrase, and it was generally believed that integrase tetramers catalyzed the integration of other retroviruses. The elucidation of new structures from four different retroviruses over the past year has however revealed this is not the case. The number of integrase molecules required to construct the conserved intasome core structure differs between viral species. While four subunits suffice for spumaviruses, α- and β-retroviruses require eight and the lentiviruses use up to sixteen. Herein we described these alternative architectures, highlighting both evolutionary and structural constraints that result in the different integrase-DNA stoichiometries across Retroviridae.

Introduction

Integration of the linear viral DNA that is made by reverse transcription into host cellular chromatin is a critical step in the retroviral life cycle. To accomplish integration, the viral enzyme integrase (IN) catalyzes two consecutive reactions, 3′ processing and strand transfer. During 3′ processing, IN hydrolyzes the nascent viral DNA adjacent to invariant CA dinucleotides to yield recessed CA_OH-3′ ends. In the nucleus IN uses the 3′-hydroxyl groups to cleave chromosomal target DNA, which at the same joins the viral DNA ends to host DNA 5′-phosphates. Both reactions occur via S_N2 transesterification, assisted by a pair of divalent metal cations in the IN active site [1,2].

Retroviral integration occurs within the context of intasome nucleoprotein complexes that consist of an IN multimer engaging the two viral DNA ends. The initial assembly is most usually referred to as the stable synaptic complex (SSC), which becomes the cleaved synaptic complex (CSC) after 3′ processing. The target capture complex (TCC) forms when the CSC engages chromatin, and 3′ end joining yields the post-catalytic strand transfer complex (STC, Figure 1). Disassembly of the STC and DNA repair of the hemi-integrant, both of which are carried out by host enzymes, results in the formation of a stable provirus. The process of integration is highly similar to that of DNA transposition, and the retroviral INs belong to an extended superfamily that includes prokaryotic and eukaryotic DD(E/D) transposases (reviewed in [3]).

Figure 1.

Retroviral integration and intasome nucleoprotein complexes. Shown are representative PFV intasome structures (SSC, pdb accession code 4e7h; CSC, pdb code 3oy9; TCC, pdb code 3os2; STC, pdb code 3os0). The intasomes consist of purified recombinant IN protein and synthetic oligonucleotides that model the U5 ends of viral DNA. Two DNA binding IN protomers are painted blue and green, whereas supportive IN molecules are cyan. Adjacent DNA end sequences are color-coded to match the transferred magenta and non-transferred orange viral DNA strands in the structures. Short vertical arrows, scissile phosphodiester bonds. Target DNA is shown in grey; during strand transfer, the DNA is cleaved with a 4 bp staggered cut.

Prototype foamy virus intasome structures

The first retroviral intasome to be characterized structurally was the prototype foamy virus (PFV, from the spumavirus genus) CSC in 2010. Determined using X-ray crystallography, this structure was transformative in that it provided the initial glimpse of a functional IN multimer and also elucidated the mechanism of action of the clinical HIV IN strand transfer inhibitors [4]. Additional structures, which visualized the remaining PFV complexes along the integration pathway, were similarly reported over the ensuing 2 years (Figure 1) [2,5,6]. Today, the PFV intasome can be considered a minimalist assembly with the lowest order IN-to-viral DNA stoichiometry among retroviral intasomes characterized thus far [7]. Accordingly, we will overview its salient features before discussing the more complex assemblies that were elucidated only very recently.

Retroviral IN proteins are composed of three or four conserved domains joined via highly divergent flexible linkers. The N-terminal domain (NTD, a three-helical bundle stabilized by coordination of a Zn²⁺ ion), catalytic core domain (CCD, harboring the DDE catalytic triad within an RNase H fold), and C-terminal domain (CTD, a beta-barrel highly similar to SH3 and Tudor domains) are common to all of the proteins. The canonical domains do not have independent functions; each of them participates in essential protein-protein and protein-DNA interactions within the intasome structure. The INs from the spuma-, ε-, and γ-retroviruses additionally carry the NTD extension domain (NED), which is involved in interactions with viral DNA.

The PFV intasome is a 2-fold symmetric complex harboring four IN subunits, with viral DNA ends synapsed between a pair of IN dimers [4]. The inner IN molecules from each dimer (green and blue in Figures 1 and 2) are responsible for all interactions with viral DNA and accordingly provide their active sites to catalyze the 3′ processing and strand transfer reactions. The inner subunits interact with each other across the synaptic interface, and an underlying theme of the intasome is the swapping of the inner NTDs to function in trans with the opposing CCDs (Figure 2). The associated CTDs, which are referred to as synaptic CTDs, engage target DNA during integration [5] and help to rigidly bridge the halves of the intasome together (Figure 2). The outer IN subunits (cyan) fasten to the catalytic protomers via the CCD-CCD dimerization interface [4,8].

Figure 2.

Retroviral intasome structures and the CIC. The PFV structure is an underside view of the CSC relative to Figure 1, labeled for individual IN domains. The other intasome structures (MMTV CSC, pdb code 3jca; RSV STC, pdb code 5ejk; tetrameric HIV-1 STC, pdb code 5u1c; MVV CSC; pdb code 5m0q) are similarly oriented, size matched for common CIC components (colored). Synaptic CTDs that are donated from flanking IN dimers are shown in yellow. The portions of these structures that do not form part of the CIC are deemphasized by light gray. Viral DNA strands are black, and the target DNA strands from the RSV and HIV-1 STC structures, as well as the LEDGF/p75 IN-binding domain from the HIV-1 dodecamer, were omitted for clarity. Numbers indicate the number of IN molecules within each structure.

Characterization of a series of PFV intasomes revealed that the viral nucleoprotein architecture persists throughout the integration pathway without global perturbations. Hydrolysis of the 3′ dinucleotide(s) from viral DNA ends opens the intasomal active sites for the engagement of host DNA. The latter binds across the intasome structure, making contacts with both synaptic CTDs in a sharply bent conformation to provide the IN active sites with access to target phosphodiester bonds for strand transfer (Figure 1). Furthermore, the deformation may allow accumulation of conformational strain in target DNA, thought to be responsible for the ejection of the nascent virus-target DNA bond from the active site [2,5]. This subtle reconfiguration prevents the reversal of transesterification, which, by itself, is an energy-neutral process. More detailed reviews of the PFV intasome structure and the mechanism of retroviral DNA integration can be found elsewhere [9,10].

Surprising diversity of retroviral intasome architectures

Solution-based measures indicated that tetrameric IN forms could be the functionally relevant multimers for α-retroviral [11,12] and lentiviral [13–16] integration. Although arguments and some experimental evidence for larger species were also presented [15,17,18], the homo-tetrameric architecture observed in the PFV intasome was expected to be a universal feature of the retroviral integration machinery. However, recent intasome structures from four orthoretroviruses revealed both striking complexity and diversity. Reported from February 2016 to January 2017, the new structures elucidated intasomes from the β-retrovirus mouse mammary tumor virus (MMTV) [19], α-retrovirus Rous sarcoma virus (RSV) [20], lentivirus maedi-visna virus (MVV) [7], and lentivirus human immunodeficiency virus 1 (HIV-1) [21]. As three of the studies utilized single particle cryo-electron microscopy [7,19,21], much of the work benefited from the ongoing revolution in this structural biology platform (reviewed in [22]).

Whereas the PFV intasome is constructed from four IN molecules, eight were found in the MMTV and RSV structures (Figure 2). The IN assemblies within the latter two structures can be viewed as tetramers of dimers (Figure 2). Four core IN molecules situate in positions analogous to those in the PFV IN tetramer, forming the conserved catalytic IN core dimer (blue and green chains in Figure 2) with a pair of outer subunits (cyan). As in the PFV intasome, the catalytic subunits exchange their NTDs across the synaptic interface. The intasome is completed by insertion of CTDs, in these cases from the flanking IN dimers. An immediately obvious feature of the MMTV and RSV assemblies is the presence of a PFV intasome-like core, which we term the conserved intasome core (CIC, displayed in color in Figure 2). The factor necessitating the flanking subunits is the length of the linker that connects the CCD and CTD in α- and β-retroviral INs. While the CCD-CTD linker in PFV IN comprises 49 residues, which is sufficiently long to position the inner IN CTDs at the critical synaptic positions (Figure 2), the analogous regions in α-retroviral and β-retroviral INs harbor only 8 amino acids [19]. Accordingly, the CTDs of the catalytic IN protomers cannot assume the synaptic positions to complete the CIC structure. Instead, the α- and β-retroviruses evolved to utilize CTDs from additional IN molecules to fulfill the critical synaptic roles (Figure 2, yellow). As RSV and MMTV INs form stable dimers in solution [19,23,24], the resulting intasomes require two extra dimers, accounting for the observed octameric assembly. Results of biochemical complementation experiments indicate that the flanking IN subunits are required for MMTV IN strand transfer activity in vitro [19].

Analysis of representative IN sequences across Retroviridae indicated that the lengths of ε- and γ-retroviral IN CCD-CTD linkers, which are 55 and 61 residues, respectively, may support CIC formation from just four IN molecules, akin to the spumaviruses. The lengths of lentiviral and δ-retroviral IN CCD-CTD linkers, at ~19–22 residues, are intermediary to the comparatively short and long lengths noted for all other retroviruses [19]. The CCD-CTD linker of HIV-1 IN can be modeled sufficiently stretched to allow formation of the minimalist PFV-like tetrameric intasome [25,26]. However, the required fully extended conformation is incompatible with IN crystallography studies that observed this region in a compact α-helical configuration [27,28]. Moreover, the CCD-CTD linker is predicted to be α-helical across divergent members of the lentiviral genus [7].

HIV-1 intasome preparations revealed extensive heterogeneity, which we suspect resulted from the use of a mutant IN protein [21]. Wild-type HIV-1 IN displays limited solubility in isotonic salt-containing buffers (100–150 mM NaCl) and is minimally active with oligonucleotide viral DNA substrates in the absence of added cofactors such as the common lentiviral host factor LEDGF/p75 [18,29,30]. Given that PFV IN is much more amenable to in vitro studies [4,31,32], Craigie and colleagues assessed over 30 small DNA binding domains fused to the N-terminus of HIV-1 IN in an attempt to mimic the NED of the spumaviral protein. The screen identified Sso7d, an archaeal histone-like protein, which greatly improved the solubility and activity of the fusion construct [30]. While the mechanism of this effect is not fully understood, the mutation reduced the propensity of the protein to form multimers in solution [21]. Importantly, however, Sso7d-IN could support integration in the context of HIV-1 infection [30]. Perhaps owing to the approach of mimicking PFV IN, the Sso7d-IN construct formed a tetrameric IN intasome, although higher-order species were also observed in the assembly reactions [21]. The CIC in the tetrameric structure is made from the same components that formulate the PFV intasome (Figure 2), with the important caveat that the critical CCD-CTD linker could not be visualized in the electron density. The addition of the α helical IN-binding domain of LEDGF/p75 stabilized the higher-order species, including a dodecameric assembly that could be refined to high resolution [21]. In this complex, the CIC is formed between two tetramers, with the synaptic interface completed by insertion of CTDs from flanking IN subunits (Figure 2).

MVV intasomes assembled using wild-type protein harbor sixteen IN subunits that assume a tetramer-of-tetramers architecture. Although LEDGF/p75 was essential for the formation of the MVV intasome, the host factor dissociated during purification of the nucleoprotein complex. Just as in the case of the other characterized intasomes, the CIC is found at the center of the structure (Figure 2), and four IN tetramers participate in its formation. As in the dodecameric Sso7d-IN intasome, two tetramers provide a catalytic IN subunit each, and the synaptic CTDs are provided by flanking subunits. However, in contrast to the Sso7d-IN structure, the flanking subunits are each part of a complete tetramer. Overall, the dodecameric HIV-1 Sso7d-IN and the hexadecameric MVV intasomes are remarkably similar, although the former is missing an IN dimer from each flank. All of the CCD-CTD linkers in both structures assume nearly straight α-helical configurations, in agreement with crystallographic studies [7,27,28]. The CTD-CTD dimerization interface originally discovered via nuclear magnetic resonance spectroscopy is another previously underappreciated HIV-1 IN structure that is observed in the higher-order lentiviral intasome structures, but absent in the tetrameric Sso7d-IN intasome [7,21,33,34]. The remarkable conservation of the quaternary features in face of limited primary structure conservation between HIV-1 and MVV INs, which share less than 30% amino acid sequence identity [16], strongly argues for the relevance of the higher-order lentiviral assemblies. Although more work is clearly required, the results of mutagenesis experiments are consistent with the importance of the higher-order HIV-1 intasome during virus infection [21]. Nevertheless, the barebones tetrameric Sso7d-IN assembly, refined to a near-atomic resolution, provides the essential details of the CIC structure and the enzyme active sites much needed for IN strand transfer inhibitor development. Modifications that include an intact active site (an E152Q IN mutant was used in the recent work) and processed viral DNA end substrate (without target DNA) will optimally serve this purpose.

The compact α helical configuration of the CCD-CTD linkers in the higher-order lentiviral intasomes precludes the positioning of inner core protomer CTDs at the critical synaptic positions. Thus, the physical distance that can be spanned by inner core CCD-CTD linkers, prohibitively short for α- and β-retroviruses due to limited residue content and likewise short for the lentiviruses due to α helix formation, underlies the requirement for additional flanking IN protomers to complete the respective CIC structures. The propensity of lentiviral INs to form large multimers, such as tetramers, in the absence of DNA explains the resulting higher-order architectures, in particular for the MVV hexadecamer.

Collectively, the recent structural studies highlight a remarkable evolutionary plasticity in the construction of the intasome. While, in principle, one could envisage a functional intasome made of only two catalytic IN subunits, the smallest complex observed to-date is that of PFV, a bare-bones tetrameric CIC. Interestingly, the NED, NTD, and CTD of the outer IN subunits, which are disordered in all PFV intasome crystal structures [2,4–6], are dispensable for catalytic function in vitro [35]. However, the outer IN subunits were implicated in interactions with nucleosomal DNA during integration into chromatinized targets [36]. Therefore, the tetrameric CIC may represent the minimally functionally competent integration machine, although representative structures from three more retroviral genera are yet to be resolved.

As has been noted some time ago [37], DNA transposases tend to form functional multimers only on their DNA substrates, which is explained by the key roles that protein-DNA contacts play in the formation of transpososomes [38–40]. Applied to integration, the functional intasome would be constructed from available species in solution. Indeed, the available structures contain a consistent pattern, whereby intasomes assemble via the tetramerization of the predominant solution form of IN [7]. Thus, PFV IN is monomeric [31,32], whereas RSV and MMTV are predominantly dimers [19,23,24]. Although HIV-1 IN monomers can be detected, especially with Zn²⁺-free apo protein [41–43], tetramers were the prominent species reported by most studies [14,16,21,44–46], while MVV IN exists in a tetramer-octamer equilibrium in solution [7]. This model will benefit from additional IN biophysical characterizations and structural elucidation of representative intasomes from the remaining three retroviral genera.

Conclusions and perspectives

Recent work has significantly expanded our appreciation of retroviral intasome structure and function, as we now know that the CIC can form from a total of four, eight, or sixteen IN protomers. Based on intensive, decades-long approaches from several laboratories, elucidation of the lentiviral intasome structures represent a substantial leap forward. The elucidation of further intasome structures from wild-type primate lentiviral IN proteins should help to further clarify the structural basis of HIV DNA integration and the generation of clinical resistance to IN inhibitors. We also hope that current advancements will inspire characterization of the intasomes from the remaining retroviral genera, in particular from human T-lymphotropic virus 1, a δ-retrovirus that causes severe pathology.

One particular puzzling aspect of retroviral integration that has not been addressed is the mechanism of intasome assembly. Although we interpret the Sso7d-IN tetramer and dodecamer multimers as likely side effects of the heterologous fusion protein, the work raises other intriguing possibilities. It seems possible that the Sso7d IN tetramer could represent an intermediate in intasome assembly or disassembly. If so, the IN CCD-CTD linker would need to adopt two radically different forms, extended versus α-helical, during transition to (or from) the higher-order hexadecamer.

While the structural explanation for the seemingly excessive architectures of orthoretroviral intasomes is relatively straightforward, their functional consequences remain a mystery. One hypothesis, which could merit rigorous testing, is that higher-order stoichiometry facilitates high valency interactions with host chromatin, which could help the viral integration machinery to sense local enrichment of cues associated with transcriptional competency [7].

Highlights.

• Retroviruses use 4, 8, or 16 integrase molecules for a common intasome core structure

• A divergent linker region that connects two domains influences higher-order assembly

• Intasomes seemingly represent the tetramerization of predominant solution IN species

• Lentiviral structures will facilitate inhibitor design and drug resistance studies