Cellular and molecular mechanisms of HIV-1 integration targeting

Abstract

Integration is central to HIV-1 replication and helps mold the reservoir of cells that persists in AIDS patients. HIV-1 interacts with specific cellular factors to target integration to interior regions of transcriptionally active genes within gene-dense regions of chromatin. The viral capsid interacts with several proteins that are additionally implicated in virus nuclear import, including cleavage and polyadenylation specificity factor 6, to suppress integration into heterochromatin. The viral integrase protein interacts with transcriptional co-activator lens epithelium-derived growth factor p75 to principally position integration within gene bodies. The integrase additionally senses target DNA distortion and nucleotide sequence to help fine-tune the specific phosphodiester bonds that are cleaved at integration sites. Research into virus–host interactions that underlie HIV-1 integration targeting has aided the development of a novel class of integrase inhibitors and may help to improve the safety of viral-based gene therapy vectors.

Introduction

The lentivirus HIV-1, like all retroviruses, integrates the DNA copy of its RNA genome into a host cell chromosome. Integration provides a favorable environment for viral gene expression and ensures that both daughter cells receive a copy of the virus following cell division. Integration is mediated by the viral integrase protein, a specialized DNA recombinase that is incorporated into the virus during particle biogenesis from an infected cell. In the ensuing round of infection, reverse transcription and integration occur within the context of large subviral nucleoprotein structures that are termed the reverse transcription complex and preintegration complex (PIC), respectively [1, 2]. Within the PIC, the ends of the linear viral reverse transcript are held together by a multimer of integrase in a complex that is referred to as the intasome [3–6]. The active sites of two integrase protomers within the intasome interact with the viral DNA ends intimately, and it is these two active sites that promote the chemical steps of DNA recombination. The number of integrase protomers that comprise the intasome differs among the different retroviruses, from a minimum of four for the spumaviruses [6] to as many as 16 molecules for the lentiviruses [7, 8] (see [9] for a recent review).

Retroviral integrases possess two distinct catalytic activities, 3′ processing and strand transfer, which are required for integration. Integrase processes the viral DNA 3′ ends adjacent to conserved CA sequences, which most usually removes a dinucleotide from each end [10–13]. Integrase uses the resulting 3′-hydroxyl groups as nucleophiles to cut chromosomal DNA in staggered fashion, which at the same time joins the viral DNA CA_OH-3′ ends to target DNA 5′-phosphate groups [14]. The resulting DNA recombination intermediate is repaired by the host cell to yield a short 4–6 bp duplication flanking the integrated provirus, the sequence of which is derived from the staggered DNA cut (Fig. 1). Clinical integrase strand transfer inhibitors engage the intasome and displace the 3′-hydroxyl nucleophiles from the enzyme active site, disarming the nucleoprotein complex [6, 15]. See [16] for a recent review of retroviral integration and its inhibition.

Fig. 1.

Mechanism of retroviral DNA integration. A multimer of integrase (light orange) engages the ends of linear viral DNA (black lines), synapsing them together within the intasome complex. Integrase 3′ processing activity yields chemically reactive 3′-hydroxyl groups at the viral DNA ends. Following the binding of host DNA (blue lines) in the nucleus, integrase promotes the strand transfer or DNA joining reaction (staggered cut site in red). Repair of the DNA recombination intermediate yields a short duplication of host DNA (red lines) flanking the integrated provirus

Integration in the host genome is non-random, with aspects of chromatin structure/function that span from DNA sequence to nuclear architecture influencing the sites of DNA recombination. Different types of retroviruses display distinctive preferences for transcription units, promoter/enhancer regions, histone modifications, and transcriptional activity (see [17–19] for recent reviews). HIV-1 in particular favors the interior regions of genes that reside in relatively gene-dense, transcriptionally active regions of chromatin [20]. Gammaretroviruses, typified by Moloney murine leukemia virus (MoMLV), in contrast favor promoter/enhancer regions [21–23] with more modest preferences for transcriptional activity than HIV-1 [24]. Recent research has highlighted that HIV-1 integration targeting is in large part mediated by the interaction of two virus proteins with cognate cellular receptors [25]. One of these interactions is between integrase and lens epithelium-derived growth factor (LEDGF)/p75 [26], while the other is between the viral capsid protein and cleavage and polyadenylation specificity factor 6 (CPSF6) [27]. On the local level, nucleosomes [28–33] in addition to target DNA sequence [34–38] can influence the choice of integration site. Globally, HIV-1 preferentially integrates into chromatin located in the peripheral region of the nucleus [39–41] in association with nuclear pore complexes (NPCs) [42, 43]. Herein we review in detail the cellular and molecular mechanisms of HIV-1 integration targeting, focusing on what is known alongside questions for future research.

Integrase—a primary determinant of integration targeting

HIV-1 integrase purified following its expression in Escherichia coli is enzymatically active in vitro, revealing that human cellular proteins are dispensable for integrase catalytic function [13, 44]. However, because viruses from different genera of Retroviridae display distinct genome targeting preferences, it seemed reasonable to assume that integrase-binding cell factors played a role in integration targeting [45, 46]. Detailed review of the published literature [47–64] alongside the HIV-1 Human Interaction Database that is maintained at the National Center for Biotechnology Information [65] reveals that several hundred cellular proteins reportedly interact with HIV-1 integrase. Clearly, not all of these interactions can play an important role in HIV-1 infection, and a central tenet of viral interactome research is to determine which interactions are key to virus replication. Among the reported HIV-1 integrase interactors, LEDGF/p75 has been shown to play an important role in virus infection and integration targeting.

LEDGF/p75

From initial reports, it was unclear if LEDGF/p75 played an important role in HIV-1 biology. Efficient knockdown of LEDGF/p75 via RNA interference (RNAi) yielded approximate two- to fivefold infection defects in some studies [66, 67] where others failed to detect a specific replication defect [68, 69]. The estimated cellular content of LEDGF/p75 is several 100,000 molecules [69], which seemingly greatly outweighs the level required for the efficient integration of any given PIC. Consistent with this interpretation, deep knockdown of LEDGF/p75 by short-hairpin RNA [70] or disruption of the LEDGF/p75-encoding PSIP1 gene, which theoretically depleted all traces of the factor from the cell, yielded ~ tenfold or greater reductions in HIV-1 integration and virus infection [71–75]. CRISPR-Cas9 mediated knockout of PSIP1 in primary human CD4⁺ T cells yielded a relatively mild ~ twofold infection defect [76], a level consistent with early RNAi studies [66, 67]. This similarity could be due to residual LEDGF/p75 levels across studies, as Cas9 nucleoprotein complexes were introduced by electroporation, which may not transfect all cells in the population [76]. A significant reduction in the level at which HIV-1 targeted active genes during integration was observed even when knockdown cells supported normal levels of virus infection [25, 77], effects that were exacerbated via PSIP1 knockout [25, 71, 72, 78, 79].

LEDGF/p75 and its shorter p52 splice variant (collectively referred to as LEDGF) were described initially as transcriptional co-activators [80], and both interact with RNA splicing factors [79, 81, 82]. LEDGF/p75 plays a role in homologous DNA recombination [83] and mixed-lineage leukemia (MLL) [84–87]. LEDGF belongs to the hepatoma-derived growth factor (HDGF)-related protein (HRP) family that also includes HDGF and HDGF like (HDGFL) 1–3 [88]. The most common sequence element among these proteins is an N-terminal PWWP domain [88], which is a type of chromatin reader [89]; the LEDGF PWWP domain has a preference for histone H3 tri-methylated on Lys36 [90–92]. LEDGF/p75 at steady-state binds chromatin [26, 93, 94], which is mediated by the PWWP domain and additional charged elements within the N-terminal portion of the protein [95, 96]. LEDGF/p75 also displays affinity for supercoiled DNA [97] (Fig. 2a).

Fig. 2.

LEDGF/p75 and ALLINI structures and binding to integrase. a Schematic diagram of LEDGF/p75, highlighting different protein regions/domains. Chromatin binding is mediated by the PWWP domain, charged regions (CR) 1–3, two copies of an AT-hook DNA binding motif, and basic NLS [95, 96]. CR 2 and 3 confer binding to supercoiled DNA [97]. Numbers refer to domain boundary positions. b Schematic of HIV-1 integrase; numbers demarcate domain boundaries. Amino acids invariant across Retroviridae are indicated by single-letter code. c X-ray structure of the LEDGF/p75 IBD (magenta) bound at the HIV-1 integrase CCD dimer (cyan and green). Shown in sticks are LEDGF/p75 hotspot residues Ile365 and Asp366 as well as integrase residues that help mold the binding pocket. Dashed lines, hydrogen bonds. Blue, red, and yellow denote nitrogen, oxygen, and sulfur, respectively (protein database (PDB) accession code 2B4J [102]). d X-ray structure of ALLINI BI-D (magenta, with chemical structure shown above) bound to the integrase CCD dimer (PDB code 4ID1 [190]), oriented as in c. The carboxylic acid attached to position 3 of the quinoline ring via the tert-butoxy group makes the same hydrogen bond contacts with integrase (dashed lines) as LEDGF/p75 residue Asp366 (compare with c). Other labeling is as in c

The interaction of LEDGF/p75 with integrase is specific to the lentiviridae genus of Retroviridae [68, 98, 99]. An evolutionarily conserved domain in the C-terminal portion of LEDGF/p75, which is missing from LEDGF/p52, was termed the integrase-binding domain (IBD) because it was necessary and sufficient for binding to HIV-1 integrase in vitro [100] (Fig. 2a). The LEDGF/p75 IBD is a PHAT domain (for pseudo HEAT repeat analogous topology) composed of two helix–hairpin–helix motifs, and evolutionarily conserved hotspot residues Ile365 and Asp366 in the N-terminal hairpin were critical for integrase binding [101]. Retroviral integrase proteins harbor three common domains that are referred to as the N-terminal domain (NTD), catalytic core domain (CCD), and C-terminal domain (CTD) (reviewed in [9]) (Fig. 2b). The HIV-1 integrase CCD comprised the primary LEDGF/p75-binding determinant, while the NTD was required for high affinity binding [94]. An X-ray crystal structure of the IBD in complex with the HIV-1 integrase CCD revealed that the host factor engaged the CCD dimerization interface. The side chain carboxylic acid of Asp366 contacted backbone amides of residues Glu170 and His171 from one integrase monomer, while the adjacent LEDGF/p75 Ile365 residue mediated hydrophobic interactions principally with the other CCD molecule [102] (Fig. 2c). A subsequent structure with the 2-domain NTD-CCD construct from HIV-2 integrase revealed that electronegative side chains on one face of NTD helix 1 engaged conserved electropositive residues of the IBD [103]. Interestingly, reverse-charge LEDGF/p75 mutants restored partial infectivity to otherwise non-infectious HIV-1 integrase mutant viruses that harbored complementary reverse charge substitutions in NTD helix 1 [103, 104]. All-in-all, the structural basis of the LEDGF/p75-integrase interaction is fairly well understood. Additional structures that include the full-length host factor with the HIV-1 intasome [8] and nucleosomes would shed significant new light into the structural basis of integration targeting.

Both chromatin and integrase-binding activities are critical for LEDGF/p75 to function as an HIV-1 integration co-factor [70, 72, 105]. LEDGF/p75 normally functions to bring cellular IBD partners such as cell division cycle associated 7 like (a.k.a. JPO2) and MLL in proximity to chromatin [106], and thus HIV-1 hijacks this ancient chromatin-associated molecular beacon to fulfill its nefarious needs. Hybrid LEDGF/p75 constructs that swapped the N-terminal chromatin-binding elements for heterologous chromatin readers supported HIV-1 infection [107] and redirected integration to novel positions in the genome that were consistent with the known chromatin-binding properties of the substituted domains [108–111]. The plasticity of this approach was rather remarkable, as both promoter-proximal readers such as PHD (plant homeodomain) fingers as well as heterochromatin protein modules such as CBX1 and HP1α similarly supported HIV-1 infection. Genotoxic side effects of retroviral gene therapy that are linked to sites of MoMLV vector integration have led to fatal cases of childhood leukemia [112], so the ability to reprogram HIV-1 integration to near random via fusion proteins such as CBX1-LEDGF suggested possible novel approaches to safer integrating vectors. A major hurdle in such approaches is how to effectively introduce the hybrid LEDGF/p75 construct into the cell. LEDGF/p75 is inefficiently packaged into HIV-1 particles and the scant amount that is packaged is cleaved by the viral protease [113]; so the potential to hitchhike retargeting LEDGF/p75 into the cell as a virus vector component is an apparent non-starter. The need to introduce a hybrid construct into patient cells prior to a therapeutic lentiviral vector necessarily complicates the clinical utility of retargeting LEDGF/p75 constructs in human gene therapy [103, 104, 111, 114].

Other integrase-binding proteins

Although PSIP1 knockout significantly reduced the extent to which HIV-1 targeted genes during integration, the preference to integrate into genes remained much greater than would be expected based on random chance [71, 72, 78, 79]. Such observations suggested that other integrase-binding proteins might also play a role in HIV-1 integration targeting. SMARCB1 (SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily B, member 1) was the first cellular interactor of HIV-1 integrase to be published in the literature, which at the time was called INI1 for integrase interactor 1 [115]. As implicated by its name, SMARCB1 is a component of the ATP-dependent SWI/SNF chromatin remodeling complex [116], so it was from the get-go an obvious candidate for an integration targeting cofactor. Although SMARCB1 can influence the choice of HIV-1 integration site in in vitro reactions [117], it remains unclear if it plays a role in integration targeting during HIV-1 infection.

HRP protein family member HDGFL2 in addition to LEDGF/p75 harbors a functional IBD [100]; so it too was an obvious candidate for an integrase targeting cofactor. HDGFL2 was depleted from cells using RNAi [69, 70, 73] or gene knockout [74]. None of these manipulations significantly affected the level of HIV-1 infection or the selection of integration sites. Depleting HDGFL2 from PSIP1 knockout cells additionally reduced HIV-1 infection by ~ twofold, and further reduced integration into genes from the levels observed by sole PSIP1 knockout [74, 78]. However, since the sole depletion of HDGFL2 failed to significantly affect integration targeting, it seems unlikely that this integrase-binding factor plays a role in integration targeting under normal conditions of HIV-1 infection. Additional research is required to determine if any integrase-binding partner other than LEDGF/p75 plays a significant role in determining the profile of HIV-1 integration targeting.

HIV-1 capsid and integration targeting

A key biological parameter that distinguishes the lentiviruses from the gammaretroviruses is the requirement of cell cycling for virus infection. While HIV-1 can productively infect terminally differentiated, non-cycling cells, MoMLV infection is dependent upon cell division [118–121] (reviewed in [122]). This difference maps to PIC biology: HIV-1 PICs are efficiently transported through the cellular NPC in an energy-dependent manner [123], while MoMLV PICs require nuclear envelop breakdown such that their p12 Gag protein can latch onto chromatin for subsequent integration [124, 125]. The mechanistic details that underlie HIV-1 PIC nuclear import are still being worked out and exceed the scope of this review. In short, seminal work from Yamashita and Emerman took advantage of the biological distinction of HIV-1 versus MoMLV nuclear import mechanisms to construct chimeric viruses, which mapped the viral determinant of HIV-1 nuclear import to the capsid protein [126]. The use of such constructs first indicated a role for HIV-1 Gag, the polyprotein that includes capsid, in integration targeting [127]. HIV-1 capsid interacts directly with several host factors implicated in HIV-1 PIC nuclear import including nucleoporin (NUP) 153 [128], NUP358 [129], CPSF6 [130], and cyclophilin A (CYPA) [131–133], and each of these factors has additionally been shown to influence the choice of integration site [25, 129, 134–136]. Such observations highlight a potential mechanistic link between HIV-1 PIC nuclear import and integration targeting [137, 138].

HIV-1 capsid is a 2-domain protein composed of an NTD and CTD separated by a flexible linker. Capsid monomers assemble into hexameric or pentameric ring-like structures, which are the building blocks of the virion capsid shell [139, 140]. The capsid shell houses the ribonucleoprotein (RNP) guts of the virus, composed of viral RNA and viral proteins nucleocapsid, reverse transcriptase, and integrase (reviewed in [141]).

Capsid and integrase-binding host factors influence HIV-1 integration targeting in different ways. Integration in cells depleted for NUP358 was preferentially decreased in gene-dense regions of chromosomes as compared to gene bodies [134]. Similarly, capsid mutant viruses N74D and A77V, which are defective for binding to CPSF6 [27, 142], lost the targeting preference for gene-dense regions while retaining partial preference for genes [25, 129, 136, 142]. While LEDGF/p75 depletion via RNAi reduced integration into genes [25, 77], an initial report indicated this did not significantly affect the targeting of gene-dense regions [134]. Subsequent work indicated this too was likely due to residual LEDGF/p75 levels, as PSIP1 knockout reduced the targeting of HIV-1 to gene-dense regions of chromosomes as well as gene bodies [25, 136]. Thus, although there is some phenotypic overlap between capsid and integrase-binding proteins, cofactors that bind capsid principally influence the targeting of gene-dense regions, while LEDGF/p75 primarily influences intragenic targeting.

While loss of CPSF6 or NUP358 binding to capsid preferentially reduced the targeting of gene-dense regions, loss of CYPA binding yielded the opposite phenotype of enhanced gene-dense region targeting [129]. This phenomenon is not well understood, as NUP358 and CYPA bind the same loop region in the capsid NTD [132, 143, 144] (reviewed in [145]) (Fig. 3). While CYPA is composed of the sole cyclophilin domain, NUP358 is a relatively large protein composed of numerous domains including its C-terminal cyclophilin homology domain [146, 147]. One possibility is that other parts of NUP358 could interact with the capsid/PIC to account for the differential effect it versus CYPA has on HIV-1 integration targeting. Consistent with this interpretation, the N74D mutation in the CPSF6-binding pocket of capsid, which is distal from the CYPA-binding loop (Fig. 3), disrupted the interaction of the PIC with NUP358 in the cytoplasm of infected cells [148].

Fig. 3.

CPSF6 structure and interaction with capsid. a CPSF6 isoform 1, which lacks internal sequences encoded by exon 6. Regions that facilitate binding to cellular CPSF5, transportin 1 (TNPO1), transportin 3 (TNPO3), as well as HIV-1 capsid (CA), are indicated. RRM, RNA recognition motif; PRD, Pro-rich domain; RSLD, RS-like domain; numbers, domain boundaries. b X-ray structure of the HIV-1 capsid NTD highlighting approximate positions of NUP358 and CYPA (black dotted line) versus NUP153 and CPSF6 (gray dotted line) binding (PDB code 4XFY [223]). Shown in sticks are capsid residues that help mediate binding to the different host factors. Blue and red, nitrogen and oxygen, respectively. c Close-up view of CPSF6 residues 276–289 (magenta) bound to hexameric HIV-1 capsid (PDB code 4U0B [165]) highlighting hotspot residue Phe284 (italic type). Dashed lines represent hydrogen bonds. Other labeling is as in b

CPSF6

The roles of LEDGF/p75 and CPSF6 in integration targeting have been scrutinized through side-by-side analysis of isogenic sets of cells knocked down or knocked out for each or both factors [25]. In such studies, random targeting is calculated based on known annotation distribution within the human genome. While LEDGF/p75 knockdown in U2OS cells marginally reduced integration into genes, gene-tropic integration was impressively reduced to near random via CPSF6 knockdown. As residual levels of LEDGF/p75 can readily mask its role in integration [69, 70], the factors were subsequently knocked out alone or in tandem in HEK293T cells. While PSIP1 knockout reduced integration into genes from ~ 83 to ~ 63%, CPSF6 knockout yielded 57% gene targeting. Dual factor knockout reduced integration into genes to ~ 48%, just a few percentage points above the random value of 44.7%. In terms of gene density, each Mb of human chromatin harbors on average 8.7 genes, whereas HIV-1 targets regions that on average contain ~ 21 genes/Mb. PSIP1 knockout reduced this preference to ~ 14 genes/Mb. Impressively, CPSF6 knockout reduced it to 5.8 genes, well below the level expected by pure chance. Inspection of promoter proximal integration patterns further distinguished the roles of the integrase versus capsid-binding host factors. Consistent with prior reports [71, 72], PSIP1 knockout increased promoter proximal integration, from ~ 4 to 10% within a 5 kb window surrounding transcriptional start sites, while CPSF6 knockout reduced this metric to 1.8%, well below the random 4.0% value [25].

A number of additional genomic annotations were analyzed to further tease out unique contributions of LEDGF/p75 versus CPSF6 in HIV-1 integration targeting [25]. One of the more telltale was association with histone post-translational modifications. HIV-1 integration favors regions in the vicinity of activating epigenetic marks and disfavors regions nearby repressive marks [31, 149]. While PSIP1 knockout weakened each of these tendencies, both preferences were nevertheless maintained when compared to random chance values. In contrast, CPSF6 knockout flipped both preferences such that the virus now preferred repressive regions and disfavored regions nearby activating marks. Another telltale metric was genome-wide averaging of intragenic integration sites, which revealed a dominant role for LEDGF/p75 in positioning HIV-1 integration along gene bodies [25, 79]. We accordingly concluded that the primary role of the capsid–CPSF6 interaction was to steer the PIC away from heterochromatic regions of chromatin, while the LEDGF/p75–integrase interaction primarily positioned integration along gene bodies (Fig. 4) [25]. Such observations invoke a model whereby during its journey the PIC sequentially interacts with CPSF6 and then LEDGF/p75, but we currently do not know the order of binding events. The PIC may very well engage CPSF6 in the cytoplasm to enable its import through the NPC [27, 150–152] (Fig. 4), though this model is clouded by the fact that cytoplasmic accumulation of CPSF6 can potently restrict HIV-1 infection [27, 153–155]. The PIC is susceptible to LEDGF/p75 binding in the cytoplasm [156], but even here studies differ as to whether LEDGF/p75 is a component of the PIC [68, 157, 158]. Recombinant LEDGF/p75 protein can significantly stimulate the strand transfer activity of purified HIV-1 integrase [26, 96, 100, 103, 159, 160], yet PICs extracted from PSIP1 knockout cells notably support the wild-type level of strand transfer activity in vitro [72, 74]. Based on these observations, we favor a model whereby the HIV-1 PIC engages CPSF6 during or shortly after nuclear import, but need not engage LEDGF/p75 until it scans chromatin for integration sites (Fig. 4). Other scenarios that invoke cytoplasmic engagement or integrase–LEDGF/p75 binding in the nucleoplasm are also plausible. The integrase hexadecamer in the lentiviral intasome could be modeled to bind up to 16 LEDGF/p75 molecules, which may help the PIC target regions relatively enriched in LEDGF/p75 content [7]. Recent evidence suggests that HIV-1 preferentially integrates into genes that regulate T-cell activation, but the roles of LEDGF/p75 and/or CPSF6 in this process have yet to be addressed [161].

Fig. 4.

Cellular and molecular mechanisms of HIV-1 integration targeting. The HIV-1 PIC harbors viral DNA (black lines), integrase (IN, light orange ovals), and capsid (CA). CPSF6 may bind CA in the cytoplasm to facilitate PIC nuclear import in a NUP358 and NUP153-dependent manner; CPSF6 is shown in association with transportin 3 (TNPO3), one of its known β-karyopherin-binding partners. Following nuclear import, TNPO3 will recycle to engage additional transport substrates in the cytoplasm. The PIC utilizes both CPSF6 and LEDGF/p75 to target integration to active genes (represented as three colored exons separated by introns) preferentially located within the peripheral region of the nucleus (orange shade). A typical integration site (downward light green arrow) is enlarged below to indicate nucleotide sequence preference in International Union of Biochemistry base code (B: G, C, or T; D: G, A, or T; H: C, A, or T; N: G, C, A, or T; V: G, C, or A); opposing green arrows denote scissile phosphodiester bonds. PICs that cannot properly engage CPSF6, as represented by the N74D change in CA, enter the nucleus via an alternate route (marked ?) that may require NUP155 [27]. Such PICs prefer gene sparse regions and hyper target the peripheral region of the nucleus for integration [25, 129, 136, 142, 152]

CPSF6 is an SR-like protein composed of an N-terminal RNA recognition motif (RRM), a central Pro-rich domain (PRD), and a C-terminal RS-like domain (RSLD) [162] (Fig. 3a). Binding to capsid is mediated via the PRD [130, 163] and although a 15-mer peptide could bind the isolated HIV-1 capsid NTD, it displayed significantly higher affinity to capsid hexamers in vitro [164, 165]. The form of the capsid that engages CPSF6 or other nuclear transport factors during HIV-1 infection is unknown, but it seems likely that CPSF6 will minimally engage a hexamer. Additional work to determine the form of the capsid that is bound by CPSF6 during HIV-1 infection is warranted.

CPSF6 at steady state is nuclear, and a fusion protein composed of the RSLD and green fluorescent protein (GFP) accumulated in cell nuclei, revealing that the RSLD harbors a functional nuclear localization signal (NLS) [162]. An internal deletion mutant that lacked the PRD but retained the RSLD localized to both the cytoplasm and nucleus, indicating that sequences outside of the RSLD may also contribute to CPSF6 nuclear localization [162]. Consistent with this interpretation, two different β-karyopherin proteins, transportin 1 and transportin 3, were shown to bind sequences within the PRD and RSLD, respectively [150, 166]. It will be informative to determine if these CPSF6 sequences actually comprise functional NLSs.

CPSF6 is part of the cleavage factor I mammalian (CFIm) complex that regulates positions of polyadenylation in the 3′ untranslated regions of mRNAs [167–169]. CFIm is a hetero-tetramer composed of two copies of CPSF5 and two copies of either CPSF6 or CPSF7 [167, 168, 170]. Unlike LEDGF/p75, CPSF6-binding partners important for integration targeting that function downstream from its interaction with the PIC are unknown, but one obvious candidate was CPSF5. However, expression of a CPSF6 mutant that is defective for CPSF5 binding in CPSF6 knockout cells in large part restored HIV-1 integration targeting, indicating that the function of CPSF6 in integration targeting is independent from its role in polyadenylation regulation [171]. Consistent with this interpretation, CPSF6 did not preferentially direct integration into the genes for which it regulated polyadenylation [25]. CPSF6 resides predominantly in nuclear paraspeckles [162, 172] and is not known to directly bind chromatin. Nevertheless, a chromatin-immunoprecipitation sequencing (ChIP-Seq) dataset is available [173], and these sequences correlated with regions of CPSF6-dependent integration [25]. We accordingly expect that factor(s) that apparently bridge CPSF6 to chromatin, as evident by the ChIP-Seq dataset, could play a role in integration targeting. Additional work is required to clarify the mechanism of CPSF6 action in HIV-1 integration targeting.

Global integration targeting and nuclear architecture

Imaging-based studies have indicated that HIV-1 prefers to integrate into chromatin that resides within the peripheral region of the nucleus [39–41, 43, 152] (reviewed in [174]). LEDGF/p75 and NUP153 were reportedly both required for peripheral integration targeting, as each knockdown resulted in shifts of the viral DNA away from the periphery toward the central region of the nucleus [43, 175]. Results of two other studies, however, fail to support a significant role for LEDGF/p75 in the intranuclear localization of HIV-1 DNA [176, 177]. CPSF6 knockdown, or infection with binding defective capsid mutant viruses such as N74A or A105T, by contrast increased the accumulation of peripherally located proviruses [152]. The binding sites for NUP153 and CPSF6 on the capsid overlap [130, 164, 165], but in each case only relatively short peptides of the host factors have been crystallized with capsid (Fig. 3b, c). Thus, as previously mentioned for NUP358, regions of NUP153 or CPSF6 outside of the visualized peptides could potentially interact with capsid or PICs in unknown ways to effect the different fates of integration within the structure of the nucleus. Additional work is clearly required to assess the roles of the different integration targeting cofactors in HIV-1 PIC trafficking to architecturally distinct regions of the nucleus for integration.

Local chromatin features in integration targeting

Retroviral integration favors particular nucleobases at integration sites, although these preferences are rather subtle and thus only become evident through the alignment of multiple sequences [34, 35]. Spumavirus integrase cuts target DNA with a 4 bp stagger, and X-ray crystal structures of the spumaviral intasome with target DNA revealed a severe kink at the center of the 4 bp region [178]. The center of spumaviral integration sites is enriched in pyrimidine (Y)-purine (R) dinucleotides, which, due to base stacking, are naturally more flexible than YY and RR, or the most rigid RY dinucleotide [179]. Thus, the spumaviral intasome preferentially selects for target DNA sequences that can best bend to fit into the relatively confined space between two opposing integrase active sites [178]. Expanded preferential target DNA sequence analyses indicated that inherent bendability contributed to the integration site preferences of most retroviruses, and that viruses that cut target DNA with a 6 bp stagger more routinely utilized less bendable DNA than viruses that cut with 4 or 5 bp staggers [36, 37]. Substitutions of integrase residues that interact with target DNA in crystal structures [178] or molecular models [36, 180] not only altered nucleotide selectivity at integration sites [36, 178, 180], but also could retarget integration toward gene-sparse regions of chromatin [180].

Nucleosomes are favored over matched naked DNA templates for HIV-1 integration in vitro, with preferred sites mapping to outward regions of the distorted nucleosomal DNA major groove [29, 30]. Analysis of large numbers of integration sites indicated that HIV-1 is likely to retain this targeting preference during virus infection [31]. The cryo-electron microscopy structure of the spumaviral intasome bound to a mononucleosome revealed that the intasome induced the same target DNA bend that was observed previously with naked target DNA through the lifting of the DNA from the surface of the underlying histones [181]. Although these data suggest that retroviruses might universally integrate into preferred target DNA sequences on the surface of nucleosomes, other work has revealed differential effects of chromatin compaction on the in vitro activities of retroviral integrase proteins [32]. HIV-1 integrase in particular favored regions of lower nucleosome density under conditions where MoMLV and spumaviral integrase favored compact chromatin [32]. Thus, the HIV-1 intasome may require chromatin remodeling complexes such as SWI/SNF [115, 117] or FACT [64] to effectively access nucleosomal target DNAs during virus infection. Additional research that maps the integration sites of a variety of retroviruses such as HIV-1, MoMLV, and spumavirus in cells depleted for remodeling complexes should inform the extent that such complexes play a role in viral integration targeting.

Allosteric inhibitors of HIV-1 integrase activity

Over-expression of mutant LEDGF/p75 constructs that retained the IBD but substituted the N-terminal chromatin-binding elements for heterologous sequences such as GFP blocked HIV-1 infection at the integration step [70, 182]. Moreover, combining this approach with RNAi-mediated knockdown of LEDGF/p75 could inhibit HIV-1 infection > 500-fold [156]. These observations highlighted that small molecule inhibitors of the LEDGF/p75–integrase interaction could potentially possess antiviral activity.

Two different approaches led to the identification of what turned out to be highly similar inhibitors of the interaction of LEDGF/p75 with HIV-1 integrase. Whereas one utilized a high-throughput screen for integrase 3′ processing activity [183], the other used an in silico approach to identify molecules that mimicked the positions of key LEDGF/p75 interacting residues at the CCD dimer-binding site [184] (Fig. 2c, d). In both cases, micromolar lead compounds were developed into low nM inhibitors, with Boehringer Ingelheim compound BI 224436 evaluated in human clinical trials [185]. Such compounds go by various names in the literature. The Debyser laboratory coined LEDGINs for LEDGF/p75-integrase interaction site [184], whereas we prefer ALLINIs for allosteric-integrase inhibitors to represent the fact that such compounds bind far from the integrase active site and thus allosterically inhibit catalytic function [186]. Other terms include NCINIs for non-catalytic site integrase inhibitors [183, 185, 187] and INLAIs for integrase-LEDGF allosteric inhibitors [188].

ALLINIs possess a remarkable antiviral mechanism of action that is linked to their ability to hyper-multimerize integrase [189–192]. HIV-1 integrase in solution adopts numerous forms including monomer, dimer, tetramer, and higher-order multimer, the details of which in part depend on protein and salt concentration [160, 190, 193–199]. HIV-1 intasome assembly, which may occur via DNA-mediated tetramerization of the solution integrase tetramer [7], is exquisitely sensitive to perturbation of integrase multimerization such as that incurred via premature LEDGF/p75 binding [160, 200, 201]. ALLINI binding induces the formation of large integrase aggregates [191, 192, 202] through the bridging of neighboring protein molecules [203, 204]. Aggregation accounts for their anti-integrase activity, as the aggregates are unable to assemble with viral DNA to form functional integrase–DNA complexes in vitro [186]. And, when added to susceptible target cells, the compounds act as integrase inhibitors that specifically block the integration step of HIV-1 infection [184, 187–190, 205]. However, the compounds are across the board more potent inhibitors of HIV-1 particle maturation [187–190, 192, 206]. Particle maturation is particularly sensitive to changes in integrase, leading to the definition of a subclass of HIV-1 integrase mutant viruses, the class II mutants, to distinguish them from the class I mutants that mature normally and thus are specifically defective for the integration step of HIV-1 replication (see [207, 208] for reviews). ALLINI treatment phenocopies the maturation defect associated with certain class II integrase mutant viruses, which is highlighted by the mis-localization of the RNP complex to a region of the virion particle outside of the conical shell [187, 190, 206, 209, 210]. Such viral RNA is susceptible to degradation following infection [211], highlighting the common reverse transcription defect of ALLINI-treated and class II integrase mutant viruses [187–190, 192, 206–208]. HIV-1 integrase binds virion RNA in the virus, which is disrupted similarly by ALLINIs and class II mutations [212]. Thus, integrase binding to RNA may play a critical role in RNP localization during HIV-1 particle maturation [212], though the biochemical mechanism of how integrase–RNA binding orchestrates capsid shell morphogenesis around the RNP remains to be elucidated. It will also be telltale to ascertain if integrase plays a similar role in the morphogenesis of other retroviruses.

In the long run, it appears that inhibition of the LEDGF/p75–integrase interaction may have little to do with the mechanism of ALLINI action. Sites of HIV-1 integration are altered when susceptible target cells are treated with ALLINIs, highlighting that the compounds are likely effective inhibitors of the LEDGF/p75–integrase interaction under this infection condition [175, 192, 202]. However, ALLINI potency actually increases during the afferent arm of HIV-1 replication when LEDGF/p75 is depleted, indicating that LEDGF/p75 protects the intasome from the aggregation inducing effects of the molecules [74, 187–190, 192]. By contrast, LEDGF/p75 has no apparent influence on compound potency under conditions of virus assembly [75, 187–190, 192]. As LEDGF/p75 is not effectively incorporated into virions [113], it seems that the inability for LEDGF/p75 to bind integrase during viral late events unleashes the full potency of the compounds.

Conclusions and perspectives

Recent research has highlighted roles for both integrase- and capsid-binding proteins in HIV-1 integration targeting. The brunt of this data was derived from transformed cell lines such as U2OS and HEK293T due to the sensitivity of these cells to RNAi and CRISPR-Cas9 genetic manipulation. It therefore is important to extend this line of research to primary cells such as monocyte-derived macrophages (MDM) and CD4⁺ T cells to ascertain the extents that LEDGF/p75 and CPSF6 influence integration targeting under physiologically relevant conditions. As discussed, LEDGF/p75 has been depleted from primary T cells via electroporation of Cas9-guide RNA complexes [76], so mapping HIV-1 integration sites in these cells should be informative. Treating MDM with short-interfering RNA against CPSF6 or LEDGF/p75 yielded fairly modest knockdowns and reduced HIV-1 integration into genes, yet the targeting defects were, perhaps expectedly, much more modest than those seen by knocking down CPSF6 in U2OS cells or knocking out either factor [25]. Additional approaches to increase the potency of protein depletion in primary cells are warranted [76]. The A77V change in capsid, which disrupts binding to CPSF6, importantly ablated integration into gene-dense regions of chromatin in both transformed cells and MDM [142].

Patients on antiretroviral therapy harbor a reservoir of latently infected cells that rekindle systemic viremia upon cessation of therapy [213] (see [214] for review). Analysis of AIDS patients integration sites over time revealed the persistence of certain proviruses as evident by the clonal expansion of infected cells [215, 216]. As the brunt of viral DNA in patients is defective [217], it may be expected that many of the proviruses in clonally expanded cells are also defective [218], though in at least one patient a dominant cell clone was responsible for much if not all of the circulating virus [219]. What is currently unknown is whether proviruses, defective or otherwise, drive clonal expansion due to subtle growth advantages instilled by the particular site of integration, though this is surely a popular working model (see [220] for review). Significant research is currently dedicated to HIV cure, where cells that harbor latent proviruses would either be permanently locked down for viral expression or transcriptionally activated and then purged from the body (reviewed in [221]). The site of HIV-1 integration can influence both entry into latency and transcriptional reactivation by latency-reversing agents (LRAs) [222]. Interestingly, suppression of LEDGF/p75-mediated integration targeting by RNAi knockdown or ALLINI treatment increased the number of latent viruses and decreased the extent of LRA reactivation [175]. As CPSF6 shields the HIV-1 PIC from targeting heterochromatin [25], it will be of interest to determine the latent state of proviruses formed in the absence of this targeting pathway and how such viruses respond to LRAs. The cellular and molecular mechanisms that drive HIV-1 integration targeting are inextricably linked to the persistence of particular proviruses in AIDS patient reservoirs and research into HIV cure.

Abbreviations

PIC

Preintegration complex

MoMLV

Moloney murine leukemia virus

LEDGF

Lens epithelium-derived growth factor

CPSF6

Cleavage and polyadenylation specificity factor 6

RNAi

RNA interference

MLL

Mixed-lineage leukemia

HDGF

Hepatoma-derived growth factor

HRP

HDGF-related protein

HDGFL

HDGF like

CR

Charged regions

IBD

Integrase-binding domain

PHAT

Pseudo HEAT repeat analogous topology

NTD

N-terminal domain

CCD

Catalytic core domain

CTD

C-terminal domain

PHD

Plant homeodomain

SMARCB1

SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily B, member 1

INI1

Integrase interactor 1

NPC

Nuclear pore complex

NUP

Nucleoporin

CYPA

Cyclophilin A

RNP

Ribonucleoprotein

RRM

RNA recognition motif

PRD

Pro-rich domain

RSLD

RS-like domain

IN

Integrase

GFP

Green fluorescent protein

NLS

Nuclear localization signal

CFIm

Cleavage factor I mammalian

ChIP-Seq

Chromatin-immunoprecipitation sequencing

Y

Pyrimidine

R

Purine

LEDGIN

LEDGF/p75-integrase interaction site

ALLINI

Allosteric integrase inhibitor

NCINI

Non-catalytic site integrase inhibitor

INLAI

Integrase-LEDGF allosteric inhibitor

MDM

Monocyte-derived macrophages

LRA

Latency-reversing agent

PDB

Protein database

TNPO1

Transportin 1

TNPO3

Transportin 3

CA

Capsid