Sites of Retroviral DNA Integration: From Basic Research to Clinical Applications

Abstract

One of the most crucial steps in the life cycle of a retrovirus is the integration of the viral DNA (vDNA) copy of the RNA genome into the genome of an infected host cell. Integration provides for efficient viral gene expression as well as for the segregation of the viral genomes to daughter cells upon cell division. Some integrated viruses are not well expressed, and cells latently infected with HIV-1 can resist the action of potent antiretroviral drugs and remain dormant for decades. Intensive research has been dedicated to understanding the catalytic mechanism of integration, as well as the viral and cellular determinants that influence integration site distribution throughout the host genome. In this review we summarize the evolution of techniques that have been used to recover and map retroviral integration sites, from the early days that first indicated that integration could occur in multiple cellular DNA locations, to current technologies that map upwards of millions of unique integration sites from single in vitro integration reactions or cell culture infections. We further review important insights gained from the use of such mapping techniques, including the monitoring of cell clonal expansion in patients treated with retrovirus-based gene therapy vectors, or AIDS patients on suppressive antiretroviral therapy (ART). These insights span from integrase (IN) enzyme sequence preferences within target DNA (tDNA) at the sites of integration, to the roles of host cellular proteins in mediating global integration distribution, to the potential relationship between genomic location of vDNA integration site and retroviral latency.

Retroviruses are parasites that exploit the replication machinery of their hosts to form stable genetic domains, allowing their consistent replication throughout the lifespan of the infected host cell. Retroviral infection is currently incurable for multiple reasons. First, compounds developed to inhibit specific steps in the viral replication cycle lose potency over time due to the emergence of resistance mutations within the drug target. Second, even under pressure from suppressive ART, a latent reservoir of infected cells persists – the origin of which is not fully recognized. Latently infected cells are capable of persisting primarily because the integrated provirus is not expressed, making it invulnerable to current treatment methodologies. Therefore, the process of integration underlies the current incurability of retroviral infection. Over the last four decades, a wealth of knowledge has been uncovered about the catalytic process of integration, the involvement of host cells in facilitating this process, and the virus genus-specific distributions of integration sites in animal cell genomes. The key to developing this knowledge has been the ability to specifically recover and characterize retroviral integration sites from the bulk of background host genomic DNA (gDNA). Methodologies to accomplish this feat have been exponentially optimized through time. Initial success was gained by detecting proviruses in different locations within infected tissue culture cells, whereas the current state of technology can resolve millions of integration events from a single infection experiment. In this review we summarize the historical improvement of integration site detection techniques, as well as important insights about the mechanics of the integration process. These insights have significantly impacted not only antiretroviral efforts, but also the understanding of cancer gene networks and the development and improvement of gene therapy vectors.

The catalytic process of retroviral integration

Soon after entering a susceptible target cell, the viral enzyme reverse transcriptase replicates the two plus-sense copies of the viral RNA (vRNA) genome into a linear, double stranded DNA molecule containing a copy of the viral long terminal repeat (LTR) at each end (Hu and Hughes, 2012). The INs catalyze two separate activities, which are known as 3′-processing and DNA strand transfer, to accomplish vDNA integration (Craigie and Bushman, 2012). The catalytic process (depicted for HIV-1 in Figure 1) begins as units of IN within the viral preintegration complex (PIC) multimerize upon the nascent vDNA termini to form the intasome, the IN-vDNA complex from wherein IN catalyzes integration (Chen et al., 1999, Li et al., 2006, Hare et al., 2010, Hare et al., 2012). The X-ray crystal structure of the prototype foamy virus (PFV) intasome has revealed the active configuration of this IN to be a tetramer (Hare et al., 2010, Maertens et al., 2010, Hare et al., 2012), and HIV-1 (Wang et al., 2001, Faure et al., 2005, Li et al., 2006, Bera et al., 2009, Krishnan et al., 2010, Kessl et al., 2011) and α-retroviral avian sarcoma-leukosis virus (ASLV) (Yang et al., 2000, Bao et al., 2003) IN have also been proposed to function as tetramers. IN cleaves the LTR ends of the vDNA adjacent to conserved 5′-CA-3′ dinucleotides during 3′-processing to yield recessed, chemically reactive CA_OH-3′ hydroxyl groups (Fujiwara and Mizuuchi, 1988, Roth et al., 1989, Brown et al., 1989, Pauza, 1990, Lee and Coffin, 1991, Hare et al., 2012). Once inside the cell nucleus, the intasome docks with host cellular tDNA to form the pre-catalytic target capture complex (TCC), with the recessed CA_OH ends of vDNA poised for chemical attack (Hare et al., 2012, Maertens et al., 2010). During DNA strand transfer, the CA_OH-3′ termini are used by IN to cleave the complementary strands of tDNA in a staggered fashion and simultaneously join the vDNA ends to the tDNA 5′-phosphates (Engelman et al., 1991). Due to the nature of the staggered tDNA cleavage, the resulting DNA recombination intermediate contains single stranded tDNA gaps adjacent to the joined vDNA ends (Fujiwara and Mizuuchi, 1988, Brown et al., 1989). These single stranded gaps are repaired by host cell enzymes subsequent to disassembly of the strand transfer complex (STC), which for HIV-1 yields a 5-bp target site duplication (TSD) flanking the stably integrated provirus (Vincent et al., 1990, Vink et al., 1990). The spacing between staggered top and bottom tDNA strand cuts during strand transfer differs depending on the type of retrovirus, and TSDs accordingly range in length from 4-bp for spumaviruses and γ-retroviruses to 6-bp for α-, β-, and δ-retroviruses (Shimotohno et al., 1980, Dhar et al., 1980, Hughes et al., 1981, Majors and Varmus, 1981, Neves et al., 1998, Kim et al., 2010). The size of the TSD that is generated by ε-retroviral integration is currently unknown.

Figure 1. The mechanism of retroviral integration.

The integration process begins in the host cell with formation of the intasome, which consists of multimerized IN bound to the viral LTR ends. The LTR is composed of U3, R, and U5 sections, each of which contains unique elements responsible for mediating viral gene transcription (Pereira et al., 2000). While within the cytoplasm, IN cleaves nucleotides (specifically two nucleotides for the depicted HIV-1 integration pathway) from each 3′ end of the vDNA adjacent to invariant 5′-CA-3′ dinucleotides during 3′-processing. Completion of 3′-processing leaves a recessed and chemically reactive hydroxyl group at each vDNA 3′ end. Lentiviral PICs are actively imported into the cell nucleus (denoted by an asterisk in figure), while γ-retroviruses can only enter the nucleus after dissolution of the nuclear membrane during cell division [reviewed in: (Matreyek and Engelman, 2013)]. Within the nucleus, IN binds tDNA and utilizes the reactive vDNA 3′-hydroxyl groups to simultaneously cleave the tDNA phosphodiester backbone and insert the vDNA molecule, through the process of strand transfer. IN subunits cleave the top and bottom tDNA strands in a staggered fashion. The length of stagger differs among retroviruses; HIV-1 IN cleaves tDNA with the depicted 5-bp stagger. After disassembly of the strand transfer complex, host cell enzymes repair the DNA recombination intermediate to yield the integrated provirus flanked by a host DNA TSD. See main text for descriptions of the roles that other cellular factors play in the process of retroviral DNA integration. Please note that a color version of Figure 1 is available online.

The lentiviruses, which include HIV-1, can efficiently infect non-dividing target cells whereas other types of retroviruses, typified by the γ-retrovirus Moloney murine leukemia virus (Mo-MLV), cannot (Lewis et al., 1992, Roe et al., 1993, Lewis and Emerman, 1994). This biology is reflective of the mechanisms used by different types of retroviruses to access the nuclear environment: whereas HIV-1 PICs are actively transported through the nuclear pore complex (NPC) (Bukrinsky et al., 1992), Mo-MLV PICs require the dissolution of the nuclear membrane in order to access the chromosomal targets for vDNA integration [reviewed in: (Matreyek and Engelman, 2013)]. The ability to usurp the nuclear transport system of the cell is apparently shared by additional types of retroviruses. For examples, the β-retrovirus mouse mammary tumor virus (MMTV) has been reported to transduce growth-arrested cells as efficiently as HIV-1 (Konstantoulas and Indik, 2014), whereas ASLV possesses a phenotype that is intermediary to those of Mo-MLV and HIV-1 (Hatziioannou and Goff, 2001, Katz et al., 2002). HIV-1 integration tends to occur into chromatin that is affiliated with the nuclear periphery, indicating that PIC transport through the NPC and integration may be mechanistically linked (Di Primio et al., 2013, Marini et al., 2015, Lelek et al., 2015).

Early techniques for detecting retroviral integration

Techniques for retrieving proviral DNA from infected cells and mapping the integration sites have matured greatly over the last four decades, leading to exponential increases in the efficiencies through which sites were recovered (Figure 2). The capacity of cellular gDNA fragments that were made by restriction endonuclease digestion to form infectious virus was established in some of the earliest work (Battula and Temin, 1977, Battula and Temin, 1978). As activity mapped to several gDNA fragments, these studies importantly proved that integration could occur at more than a single site in the host genome. Experimental resolution was incrementally increased in subsequent studies when the digested and electrophoretically separated gDNA was subjected to Southern blotting using vDNA probes (Keshet and Temin, 1978, Steffen and Weinberg, 1978, Bacheler and Fan, 1979, Cohen et al., 1979, Ringold et al., 1979, Peters et al., 1986). The increased sensitivity of these experiments further solidified the understanding that retroviruses can integrate into multiple gDNA locations, but distinguishing between random integration and integration into a large number of preferred sites was not possible at this point. Preliminary insight into integration distribution was gleaned as host gDNA was fractionated based on various parameters such as length or GC content prior to Southern blotting (Franklin et al., 1983, Salinas et al., 1987). This for the first time allowed assignment of integration sites to certain chromosomes or GC-rich gDNA regions.

Figure 2. The evolution of techniques used to determine retroviral integration sites.

Integration site recovery techniques are listed in chronological order from top to bottom, corresponding to a general timeline shown at the left of the figure. The average numbers of integration sites recovered in studies employing these techniques, which exponentially increased over time, are illustrated on a logarithmic axis at the right side of the figure. The initial step of host cell gDNA fragmentation by restriction endonuclease digestion is shared by all of the listed techniques, though sonication has recently become a desirable method of shearing. Please note that a color version of this figure is available online.

The ultimate marriage of next-generation sequencing and ligation-mediated PCR

DNA sequencing technology was obviously critical to obtain integration site sequences. Initial sequences of retroviral integration sites were determined by Maxam Gilbert chemical cleavage (Maxam and Gilbert, 1977) or dideoxy sequencing (Sanger et al., 1977, Maat and Smith, 1978) using fractionated gDNA fragments following restriction endonuclease digestion (Shimotohno et al., 1980, Shimotohno and Temin, 1980, Dhar et al., 1980, Hughes et al., 1981, Fitts and Temin, 1983). These studies laid important groundwork for the classification of different retroviral TSDs, and for the first time highlighted that integration was related to DNA transposition as transposition was known to generate short TSDs flanking these mobile genetic elements (Calos et al., 1978, Johnsrud et al., 1978, Kleckner, 1979). The subsequent development of the polymerase chain reaction (PCR) provided a boon for the burgeoning field of retroviral integration site selection, as the technique greatly simplified the recovery and characterization of specific vDNA-gDNA junctions (Figure 2). Inverse PCR (Ochman et al., 1988, Triglia et al., 1988) was an early application of the technique to integration site sequencing. Specifically, it was used to amplify self-ligated gDNA fragments using outward-facing primers complementary to a known sequence, for example the viral LTR, prior to molecular cloning and dideoxy sequencing (Silver and Keerikatte, 1989, Jin et al., 2002). Using forward and reverse PCR primers that annealed to two different known sequences afforded linear amplification of vDNA-gDNA junctions. An early example of this approach included providing Mo-MLV with a characterized tDNA substrate by co-infecting cells with the SV40 polyomavirus that reaches high copy number as an unintegrated episome in cell nuclei (Pryciak et al., 1992). SV40-specific primers could be used in tandem with Mo-MLV LTR primers to amplify integration sites that occurred into this excess tDNA source. An alternative to viral co-infection involved designing primers to pseudo-randomly prime off of gDNA sequences when integration happened to occur within range (Sorensen et al., 1993, Butler et al., 2001, Gentner et al., 2003). Such work formed the basis of the so-called Alu real-time PCR assays for determining the bulk level of HIV-1 integration in cell culture and in patient samples (Brussel et al., 2005, De Spiegelaere et al., 2014). However, the major breakthrough in recovery efficiency of individual integration sites occurred with the development of ligation-mediated PCR, or LM-PCR (Rosenthal and Jones, 1990, Cavrois et al., 1995). In this method gDNA is digested with restriction enzymes, and then an oligo-cassette (also called linker) containing a compatible “sticky” end is ligated to all digestion fragments. PCR amplification of integration sites is then accomplished using primers complementary to the viral LTR and the linker sequence (Vandegraaff et al., 2001, Schroder et al., 2002, Wu et al., 2003). Along with the completion of the human, mouse, and other host genome reference sequences, this marriage of library construction and DNA sequencing led to a wealth of insights into the integration site distribution patterns of different types of retroviruses (Schroder et al., 2002, Wu et al., 2003, Mitchell et al., 2004).

The combination of LM-PCR amplification with powerful next-generation sequencing (NGS) platforms such as pyrosequencing from 454 Life Sciences (Margulies et al., 2005, Wang et al., 2007) and DNA cluster-based sequencing from Illumina (Gillet et al., 2011) sent the number of recoverable integration sites skyrocketing, from the low hundreds prior to the use of NGS to ultimately hundreds of thousands and millions (LaFave et al., 2014, De Ravin et al., 2014, Maskell et al., 2015) (Figure 2). Substituting sonication for restriction endonuclease digestion as the way to fragment gDNA (Gillet et al., 2011) has additionally afforded the monitoring of identical retroviral integration sites from unique gDNA sources. With NGS, it can be difficult to ascertain if the repeated determination of a unique integration site was due to preferential PCR amplification during library construction or something more interesting, for example the presence of the exact same integration site among different cells in the queried population. Sonication, which cleaves gDNA in a relatively sequence-independent manner, has afforded the identification of unique linker attachment points on multiple sequence reads with the identical integration site (Gillet et al., 2011, Maldarelli et al., 2014, Wagner et al., 2014).

Retroviruses integrate at multiple gDNA locations

As the initial details of retroviral replication were being uncovered (Temin, 1971, Bishop, 1975, Vogt, 1977) it became a point of interest to determine where in the host genome integration occurred and whether the distribution of proviral DNA was random, biased toward certain genomic regions, or perhaps targeted to a limited set of loci. Southern blotting as mentioned provided some insight, as digested gDNA exhibited multiple different restriction patterns when queried with virus-specific probes (Kettmann et al., 1979, Gilmer and Parsons, 1979). The detection of differently sized restriction fragments implied integration into unique regions of the genome, and therefore integration was clearly not always directed to the same place. But did this mean that integration occurred randomly, or rather that specific regions or annotations were favorable? A primer extension-based assay that queried the frequency of integration into a handful of regions of the chicken cell genome indicated that most of that genome was accessible for integration (Withers-Ward et al., 1994), but numerous studies using finer mapping techniques have since determined that distribution is nonrandom by uncovering virus-specific preferences for particular tDNA sequences and chromosomal annotations.

Mechanism of retroviral tDNA base preferences

Retroviruses exhibit reproducible tDNA nucleotide preferences at integration sites (Shih et al., 1988, Chou et al., 1996, Bor et al., 1996, Stevens and Griffith, 1996, Leclercq et al., 2000, Holman and Coffin, 2005, Wu et al., 2005, Valkov et al., 2009, Ballandras-Colas et al., 2013, Serrao et al., 2015). Although these base preferences are relatively weakly conserved, the mechanics behind their selectivity has been resolved for different retroviruses. The groundbreaking crystallization of PFV TCC and STC structures revealed that IN accommodates tDNA with a severe central kink, which facilitates strand transfer by unstacking the base pair at the very center of the targeted stretch of DNA to place scissile phosphodiester bonds at the IN active sites for strand transfer (Maertens et al., 2010). Alignment of PFV integration sites accordingly revealed preference for the most flexible dinucleotides (pyrimidine/purine, or YR) – and selection against the most rigid dinucleotides (purine/pyrimidine, or RY) – at the central position. Target DNA distortion did not involve IN interactions with the flexible, unstacked central base pair, but rather with distal nucleotides at either end of the 4-bp TSD region. PFV IN residues Ala188 and Arg329 mediated side chain-specific contacts with the tDNA, and the sequencing of in vitro integration reaction products that were generated using IN mutant proteins resulted in tDNA base preferences that were altered in ways predicted by the nucleoprotein interactions observed in the crystal structures (Maertens et al., 2010).

HIV-1 IN-tDNA interactions analogous to those illuminated in the PFV intasome crystal structures have been shown to similarly motivate the selection of specific, flexible dinucleotides at the centers of these integration sites (Serrao et al., 2014). As mentioned previously, HIV-1 integration generates 5-bp TSDs, which a dinucleotide step analysis revealed to on average be composed of RYXRY (where X is any base). As was determined for PFV, this specific signature enforces for flexible YR dinucleotides at the two center positions of the 5-bp TSD while selecting against rigid RY dinucleotides at these positions. The structural mechanics of HIV-1 base preferences also resembled those of PFV, as HIV-1 IN residue Ser119 (analogous to PFV IN residue Ala188) was responsible for determining analogous base preferences relative to the points of vDNA insertion (Serrao et al., 2014). This finding was in line with prior (Appa et al., 2001, Harper et al., 2001, Nowak et al., 2009) and subsequent (Demeulemeester et al., 2014) studies that implicated Ser119 in the process of HIV-1 tDNA site selection. The analogous residue in Mo-MLV IN, Pro187, plays the same role as Ala188 in PFV IN and Ser119 in HIV-1 in determining tDNA base selectivity (Aiyer et al., 2015).

A meta-analysis of thousands of integration sites generated by 12 different retroviruses has revealed significant enrichment for flexibility signatures at the central positions of integration sites across the studied viruses (Serrao et al., 2015). The extent of central tDNA flexibility was moreover inversely proportional to TSD length. The examined viruses harbored a neutral, compact amino acid at the position analogous to Ala188 in PFV, and the polarity of the amino acid side chain correlated with the positioning of base preference significance relative to the points of vDNA insertion – a finding that was since confirmed by analyzing the behavior of mutants of the non-polar Pro187 side-chain in Mo-MLV IN (Aiyer et al., 2015). Taken together these studies imply that, though retroviral INs have structurally evolved to target unique nucleotide signatures, the common functional purpose of integration site base preferences may be to generate strand transfer-facilitating central tDNA distortion within the TCC. In this vein, the degenerate nature of tDNA base preference conservation at retroviral integration sites in large part reflects the multitude of nucleotide combinations that on average spawn central (YR) flexibility signatures. Moreover, retroviruses that generate 6-bp TSDs may need to bend tDNA less rigorously to facilitate strand transfer than the viruses that generate 4-bp and 5-bp TSDs (Serrao et al., 2015).

Integration is favored within flexible, nucleosome-bound tDNA

Some of the earliest reports of linkage between integration sites and genomic features involved the association of Mo-MLV and ASLV integration with DNase I hypersensitive sites and actively transcribed regions of the genome (Robinson and Gagnon, 1986, Vijaya et al., 1986, Rohdewohld et al., 1987, Scherdin et al., 1990, Mooslehner et al., 1990, Lewinski et al., 2006, Roth et al., 2011). The level of cellular gene transcriptional activity has since been demonstrated to positively correlate with the integration targeting of additional retroviruses (Zoubak et al., 1994, Jordan et al., 2001, Schroder et al., 2002, Mack et al., 2003, Mitchell et al., 2004, Lewinski et al., 2005, Trobridge et al., 2006, Monse et al., 2006, Kang et al., 2006, Moalic et al., 2006, Hacker et al., 2006). Bent DNA – and specifically DNA bent around nucleosomes – is a highly favored substrate for integration (Milot et al., 1994, Pruss et al., 1994a, Muller and Varmus, 1994, Pruss et al., 1994b, Wang et al., 2007, Roth et al., 2011, Serrao et al., 2015, Naughtin et al., 2015). This may seem counterintuitive, as on the one hand DNase I hypersensitivity and gene transcription are both correlated with open, accessible chromatin (which would be logical for IN to be attracted to), while nucleosomes on the other hand are found with the highest density in heterochromatin [see (Bell et al., 2011) for review]. Also intriguing is that chromatin density in the regions surrounding integration sites can negatively modulate the DNA strand transfer activity of some retroviral INs, while having the opposite effect on other proteins (Taganov et al., 2004, Lesbats et al., 2011, Benleulmi et al., 2015). It nonetheless appears that, regardless of viral preferences for chromatin density in the region surrounding the average integration site, distorted phosphodiester bonds are generally favored by IN at the points of integration (Maertens et al., 2010, Serrao et al., 2014, Serrao et al., 2015, Aiyer et al., 2015), and nucleosomes provide for natural sources of DNA distortion. The mechanism of how nucleosome structure could accommodate the natural propensity to integrate into flexible tDNA sequences was recently revealed through the cryo-electron microscopy structure of the PFV intasome in complex with a bound nucleosome (Maskell et al., 2015). The structure revealed novel interactions between IN amino acids and one H2A-H2B heterodimer, as well as a secondary gyre of wrapped tDNA that was separate from where integration occurred. These interactions afforded a ~7 Å separation of the tDNA sequence from the histone surface, allowing the formation of the severe central bend in tDNA that was observed in the X-ray crystal structure of the TCC with naked tDNA (Maskell et al., 2015).

It is noteworthy that the propensity to integrate into nucleosomal tDNA may not be shared among all retroviruses. In particular, tDNA sequences surrounding the sites of porcine endogenous retrovirus (PERV) and MMTV integration did not exhibit characteristics of nucleosomal DNA, indicating that these two viruses on average select for sites that are not associated with chromatinization (Serrao et al., 2015). While MMTV integration yields 6-bp TSDs, PERV generates 4-bp TSDs. PERV accordingly seems to seek out optimally distorted tDNA without the assistance of nucleosome docking, possibly due to unique IN-tDNA contacts mediated by this intasome.

Retroviruses exhibit unique preferences for genomic annotations

The answer to the question of “why” some retroviruses (Mitchell et al., 2004, Narezkina et al., 2004, Faschinger et al., 2008, de Jong et al., 2014, Konstantoulas and Indik, 2014) favor integration in or nearby genes that exhibit relatively high expression levels is not completely understood, though enhancement of viral transcription is a logical assumption (Jahner and Jaenisch, 1980, Fincham and Wyke, 1991, Zoubak et al., 1994, Jordan et al., 2001, Jordan et al., 2003, Bushman, 2003, Lewinski et al., 2005, Meekings et al., 2008). Despite this, a lot has been learned about the question of “how” integration targeting is accomplished. Cellular cofactors that engage viral PICs can guide integration to distinct areas of the host genome [recently reviewed in (Kvaratskhelia et al., 2014, Craigie and Bushman, 2014, Debyser et al., 2015)]. Dating back to the initial discoveries that HIV-1 and Mo-MLV integration sites cluster within gene bodies (Schroder et al., 2002) and promoters (Wu et al., 2003), respectively, measurements of integrations within RefSeq genes or in proximity to promoter-associated features such as transcription start sites (TSSs) and CpG islands have been used as traditional metrics to quantitatively compare integration site distribution phenotypes among different viruses. Figure 3 provides a summary of the genomic integration site preferences of multiple retroviral genera with respect to the structure of an average transcription unit. Gammaretroviruses such as Mo-MLV exhibit the greatest extent of TSS and CpG targeting of all viral genera (Scherdin et al., 1990, Wu et al., 2003, Mitchell et al., 2004, Moalic et al., 2006, Tsukahara et al., 2006, Dong et al., 2007, Kim et al., 2008, Moalic et al., 2009, De Ravin et al., 2014, LaFave et al., 2014, Serrao et al., 2015), while the lentiviruses target RefSeq transcription units and gene-dense regions of the genome more so than other retroviruses (Schroder et al., 2002, Mack et al., 2003, Mitchell et al., 2004, Crise et al., 2005, Hacker et al., 2006, Kang et al., 2006, Monse et al., 2006, Marshall et al., 2007). The majority of gene-tropic lentiviral integrations occur within introns (Han et al., 2004, Ikeda et al., 2007, Shan et al., 2011, Maldarelli et al., 2014, Wagner et al., 2014, Cohn et al., 2015), presumably due to the fact that introns are generally much larger than exons. The α-, β-, and δ-retroviruses display more random distributions relative to genes and promoter-associated annotations than do either the lenti- or γ-retroviruses (Mitchell et al., 2004, Narezkina et al., 2004, Derse et al., 2007, Faschinger et al., 2008, Konstantoulas and Indik, 2014), while the spumavirus PFV displays the unique phenotype to preferentially avoid gene bodies yet target promoters at a level that is intermediate to the γ- and lentiviruses (Trobridge et al., 2006, Nowrouzi et al., 2006, Maskell et al., 2015). Thus, it is apparent that different cellular forces help to govern the integration site targeting of different classes of retroviruses.

Figure 3. General integration site preferences of different retroviral genera.

Representational integration site patterns with respect to the structure of the average host cell gene are depicted by vertical arrows, with each arrow type corresponding to a different viral genus (or group of genera in the case of α-, β-, and δ-retroviruses). Gammaretroviruses preferentially target active promoters and enhancers (Wu et al., 2003, LaFave et al., 2014, De Ravin et al., 2014) while lentiviruses integrate primarily within the bodies of actively transcribed genes (Schroder et al., 2002). Spumaviruses avoid genes and rather target intergenic regions, while α-, β-, and δ-retroviruses exhibit nearly random distributions. As mentioned in the text, additional features not depicted in the figure can also influence the integration sites of various viruses, such as gene transcription level, proximity to the NPC, and local nucleotide sequence.

The involvement of LEDGF/p75 in tethering lentiviral integration to genes

Integration site targeting by host factors is most thoroughly understood at present for the lentiviruses. A variety of cellular proteins have been characterized for their ability to bind HIV-1 IN, stimulate its catalytic activity in vitro, and to potentially play a role in HIV-1 infection [reviewed in (Van Maele et al., 2006, Engelman, 2007)], but none more so than lens epithelium-derived host factor/p75 (LEDGF/p75). LEDGF/p75 was originally discovered to associate with ectopically expressed HIV-1 IN in HEK293T cells and to stimulate HIV-1 IN catalytic activity in vitro (Cherepanov et al., 2003). Subsequent work revealed that the binding of LEDGF/p75 to IN is lentivirus-specific (Llano et al., 2004, Busschots et al., 2005, Cherepanov, 2007) and that LEDGF/p75 is critical for efficient lentiviral replication (Llano et al., 2006a, Vandekerckhove et al., 2006, Shun et al., 2007, Schrijvers et al., 2012a, Fadel et al., 2014).

LEDGF/p75 functions as a bimodal tether to target the lentiviral intasome to specific regions of chromatin. On one side of the tether is the N-terminal region of LEDGF/p75, which contains a PWWP (Pro-Trp-Trp-Pro) domain – a type of Tudor domain – as well as charged regions and a pair of AT-hook DNA binding motifs that work together to confer constitutive chromatin association (Turlure et al., 2006, Llano et al., 2006b). The PWWP domain in particular recognizes histone 3 tails that are tri-methylated on Lys-36 (H3K36me3) (Pradeepa et al., 2012, Eidahl et al., 2013), an epigenetic mark that positively correlates with transcriptional elongation (Bannister et al., 2005) and sites of HIV-1 integration (Roth et al., 2011). The C-terminal side of the tether contains the IN-binding domain (IBD), which is necessary and sufficient to bind HIV-1 IN in vitro (Cherepanov et al., 2004). Single amino acid substitutions at either end of the tether, for example the W21A change in the PWWP domain or the D366N change in the IBD, are sufficient to effectively break tethering function and render the mutant LEDGF/p75 protein unable to support HIV-1 integration (Cherepanov et al., 2005, Llano et al., 2006a, Shun et al., 2007, Shun et al., 2008, Botbol et al., 2008).

The involvement of LEDGF/p75 in genic integration distribution was validated by depleting the protein from cells prior to lentiviral infection using either RNA interference (Ciuffi et al., 2005) or gene knockout (Shun et al., 2007, Marshall et al., 2007, Schrijvers et al., 2012a). Under these conditions integration into transcription units was reduced, concomitant with increases in integration in the vicinity of normally disfavored promoter-associated features such as TSSs and CpG islands. Although lentiviral infectivity could be considerably reduced by LEDGF/p75 depletion, significant levels of virus infection remained, and the distributions of the residual proviruses were far from random. This implied that although LEDGF/p75 plays an important integration co-factor role, it is not essential for lentiviral infection or integration, and that additional host cell factors likely play supporting roles in directing integration to genic regions. LEDGF/p75 is a member of the hepatoma-derived growth factor (HDGF) related protein (HRP) family, which is defined by similarities amongst N-terminal PWWP domains (Izumoto et al., 1997). One other HRP family member, HRP2, additionally harbors an IBD (Cherepanov et al., 2004), which made it a prime suspect for a role in gene-tropic lentiviral DNA integration. Removal of HRP2 from LEDGF/p75 knockout cells, either by RNA interference or through its own knockout, added to the overall HIV-1 infectivity defect and altered the pattern of integration sites more towards random (Wang et al., 2012, Schrijvers et al., 2012b). However, residual targeting of integration to genes was still apparent, and knockout of HRP2 on its own failed to negatively affect either HIV-1 titer or integration site usage (Wang et al., 2012). The fact that LEDGF/p75 needed to be removed from the system to bring out a subsidiary role for HRP2 indicates that HRP2 normally plays little if any role in determining the sites of HIV-1 integration. At the same time, it became clear that IN is not the only viral determinant of HIV-1 integration site selection, as the interplay among the viral capsid (CA) and CA-binding host proteins has also been shown to play a significant role.

The role of HIV-1 CA-binding host factors in integration targeting

CA forms the protective core that encases retroviral vRNA genomes in association with nucleocapsid protein within the virus particles. Though the core is likely to partially dissolve during the early phase of infection to enable reverse transcription (Hulme et al., 2011, Yang et al., 2013) and trafficking to the nucleus, it is apparent that some fraction of HIV-1 CA remains associated with the PIC as it enters into the cell nucleus (Hulme et al., 2015, Peng et al., 2015). A number of host cell factors that associate with CA, for example cyclophilin A (CypA), the β-karyopherin transportin 3 (TRN-S2 or TNPO3), and nucleoporins (NUPs) 358 and 153, can influence the distribution of HIV-1 integration. Whereas depletion of TNPO3, NUP358, or NUP153 preferentially reduced integration targeting to gene dense regions of chromosomes, disrupting the CA-CypA interaction – either through the addition of a small molecule cyclosporine inhibitor or by genetically altering the CA protein by G89V or P90A mutation – yielded the opposite effect; that is, these manipulations increased the targeting of the virus to gene dense regions (Ocwieja et al., 2011, Schaller et al., 2011, Koh et al., 2013, Di Nunzio et al., 2013). A separate CA mutation, N74D, which rendered HIV-1 resistant to the negative effects of TNPO3, NUP153, or NUP358 depletion (Lee et al., 2010), on its own steered integration to regions of significantly less gene density than the wild-type (WT) control virus (Schaller et al., 2011, Koh et al., 2013). Thus, the N74D mutation in large part phenocopied the effects of depleting TNPO3, NUP358, or NUP153. The N74D virus also mildly resisted the depletion of LEDGF/p75, as it retained relatively efficient infectivity while exhibiting less genic integration targeting than the WT virus (Koh et al., 2013). It has recently been observed that NUP153 and translocated promoter region (TPR), which together comprise the NPC basket (Umlauf et al., 2013), seem to cooperate to maintain chromatin architecture that is favorable for HIV-1 replication by helping to arrange actively transcribed chromatin proximal to the NPC (Marini et al., 2015, Lelek et al., 2015). However, it is currently unclear whether TPR exerts this effect by binding to the PIC through a viral component or perhaps a known cellular binding partner, such as LEDGF/p75. All together, this work underscores the complexity of the interactions that lentiviral PICs utilize to guide integration to favored sites, which are predominantly if not exclusively mediated through interactions of the viral CA and IN proteins with host factors. Additional work that co-depletes CA- and IN-binding cellular factors from cells may reveal whether random or near-random HIV-1 integration distribution is achievable under these circumstances.

BET proteins influence γ-retroviral integration target site selection

As discussed above, lentiviruses are not alone in utilizing interactions with host cell proteins to target integration to specific gDNA features. Additional examples include the retrotransposons Ty5 and Tf1 from yeast, which are directed by host proteins to target heterochromatin (Xie et al., 2001, Zhu et al., 2003) and promoters (Leem et al., 2008), respectively. Also, the catalytic activities of γ-retroviral INs are stimulated through the binding of bromodomain and extraterminal (BET) proteins, with BET-mediated bimodal tethering underlying the characteristic predilection of these viruses to integrate adjacent to promoter-associated annotations (Sharma et al., 2013, Gupta et al., 2013, De Rijck et al., 2013, Larue et al., 2014, Aiyer et al., 2014, El Ashkar et al., 2014, Serrao et al., 2015). Unlike the LEDGF/p75 bimodal tether and lentiviral DNA integration, small molecules exist that compete for BET protein binding to chromatin, and these compounds have been shown to diminish Mo-MLV infectivity and integration targeting without preventing the BET-IN interaction (Sharma et al., 2013, Gupta et al., 2013, De Rijck et al., 2013). This observation has potential implications for the future of ART, as small molecules that reduce infectivity by binding to and modulating the activity of cellular factors at positions that are distinct from the locations of cell-virus factor interaction sites would be expected to remain potent in the face of viral mutations. Readers are directed to several comprehensive reviews that have been recently published on the role of host proteins in directing lentiviral and γ-retroviral integration for additional details (Kvaratskhelia et al., 2014, Craigie and Bushman, 2014, Debyser et al., 2015).

Towards safer vector integration profiles

Understanding the mechanisms that govern the targeting of retroviral integration to specific regions of gDNA has far-reaching implications for the development and improvement of gene therapy technologies. Gene therapy has faced several hurdles over the years, including the integration of Mo-MLV-based vectors into unintended genomic sites that led to oncogene activation in an unacceptable number of cases (Marshall, 2002, Marshall, 2003, Hacein-Bey-Abina et al., 2003, Howe et al., 2008, Hacein-Bey-Abina et al., 2008). Gammaretrovirus- and lentivirus-based vectors display the integration site profiles described above for the parental viruses (Hematti et al., 2004, Laufs et al., 2006, Deichmann et al., 2007, Schwarzwaelder et al., 2007, Cattoglio et al., 2007, Mantovani et al., 2009, Wang et al., 2010), so customizing vector integration profiles to target “safe” regions of the genome is a current topic of interest. The design of completely safe vectors is not within reach at present, but knowledge of retrotransposon, lentivirus, and γ-retrovirus tethering mechanisms has been exploited to create prototype vectors with redirected, potentially less dangerous insertion profiles. For example, Mo-MLV vectors with lowered propensity to target promoter-proximal regions have been generated by mutating the C-terminal tail of IN to prevent the BET protein interaction (Aiyer et al., 2014, El Ashkar et al., 2014), and the mutant IN viruses generated similar distribution profiles as those observed using small molecules that block the BET-chromatin association. However, the distributions of the IN mutant viruses were still significantly different from random, with residual promoter proximal targeting evident.

A different approach to potentially safer retroviral-based gene therapy includes the development of self-inactivating (SIN) vectors, whereby the LTR U3 promoter and enhancer region is deleted (Schambach et al., 2006, Thornhill et al., 2008). A recent clinical trial employing Mo-MLV-based SIN vectors to correct X-linked severe combined immunodeficiency revealed that, though the genomic distribution of the vector integration sites was indistinguishable from that of enhancer-containing Mo-MLV-based vectors, integrations adjacent to proto-oncogenes and genes associated with adverse events in earlier trials was significantly reduced (Hacein-Bey-Abina et al., 2014). It is not fully understood why a SIN vector would exhibit selective reductions in integration near proto-oncogenes without a noticeable global change in integration site distribution, but a potential explanation is that the Mo-MLV LTR enhancer sequence contributed to the in vivo selection of engrafted cells containing such integrations, possibly through stimulating gene expression and ultimately cell growth rate. As the above-mentioned trial was comprised of only eight subjects, it will be informative to study the utility of SIN vectors to correct human genetic ailments using larger sample sizes moving forward.

Retargeting integration with customized chromatin tethers

The development of custom-targeted vectors is another option to avoid integration in genomic areas that promote unwanted side effects. Such customization research studies were originally conducted in vitro, employing HIV-1 or ASLV IN fused to DNA binding proteins such as λ-repressor (Bushman, 1994) or bacterial LexA (Goulaouic and Chow, 1996, Katz et al., 1996). The catalytic activity of these fusions resembled that of the WT INs, and integration into λ-repressor- and LexA-binding sequences was observed, thus providing optimism that IN protein hybrids could eventually be useful in vivo. Such experiments did in large part prove successful in the context of tissue culture infection, as vectors containing IN fused to zinc finger proteins (Bushman and Miller, 1997, Tan et al., 2006) displayed evidence for the expected altered integration distribution profile. However, while fusions of IN to large proteins or protein domains can exhibit WT-like levels of IN catalysis in vitro, the infectivity of vectors carrying such fusions is significantly diminished in cell culture (Bushman and Miller, 1997, Tan et al., 2006), thus limiting their potential clinical utility. An alternative approach to redirect retroviral integration targeting is to alter the chromatin binding activity of the IN-binding chromosomal DNA tether rather than the IN protein itself. Such approaches have been successfully developed for both LEDGF/p75 (Ciuffi et al., 2006, Ferris et al., 2010, Gijsbers et al., 2010, Silvers et al., 2010) and BET (De Rijck et al., 2013) fusion proteins. The fusion protein construct in these cases was expressed in target cells rather than packaging within the cargo-size-limited virion, thus permitting efficient redirection of integration into altered gDNA regions. Despite these promising results, the clinical utility of these constructs is complicated by the fact that these hybrids would have to compete with ubiquitously expressed endogenous LEDGF/p75 or BET proteins in target cells. An approach to circumvent this limitation has been developed by altering IN and LEDGF/p75 amino acid side chains that mediate the protein-protein interaction. The interaction between the IBD and the N-terminal domain of lentivirus IN is in particular mediated through salt-bridge contacts (Hare et al., 2009), and reversing the charge of the side chains (from acidic to basic for IN, and vice-versa for the complementary interacting LEDGF/p75 mutant) afforded selective IN mutant viral infection and integration in cells expressing the reverse-charge LEDGF/p75 variant (Hare et al., 2009, Wang et al., 2014). Although intriguing, the methodology has the disadvantage of requiring expression of the customized LEDGF/p75 partner in target cells.

Given the slow pace at which potential translational applications have evolved, interest in developing directed integration targeting vectors through modifying either IN or host cofactors may very well wane due to the development of other more promising gene editing technologies. The basis of the prokaryotic adaptive immune system that recognizes and degrades exogenous genetic elements (the CRISPR/Cas system) has been manipulated and applied toward customized editing in a variety of species including human embryos and non-human primates (Jinek et al., 2012, Niu et al., 2014, Liang et al., 2015, Chen et al., 2015, Wan et al., 2015). The details of this system involve a customized guide RNA directing a CRISPR-associated (Cas) nuclease to genomic DNA in a sequence-specific fashion, leading to a double-stranded DNA cut at virtually any desired location in the genome (Cong et al., 2013). Although the CRISPR/Cas system is particularly promising for eventual clinical applications, similar to the zinc finger protein and transcription activator-like effector nuclease/TALEN (Miller et al., 2011) technologies that preceded it, the threat of low-frequency off-target events remains a potential hurdle (Pattanayak et al., 2011, Mussolino et al., 2011, Cho et al., 2014, Fu et al., 2014). Lentiviral vectors have notably been designed to introduce CRISPR/Cas system components into target cells (Kabadi et al., 2014, Albers et al., 2015).

Retroviruses as tools to identify cancer genes through insertional mutagenesis

Although retroviruses preferentially target particular tDNA nucleotides and genomic annotations for integration, the bulk of the nucleotides and genes within the host genome are available as insertion targets (Carteau et al., 1998). The tendency of retroviruses to integrate at innumerable gDNA locations has been harnessed to identify cancer-related genes through insertional mutagenesis [reviewed in (Mikkers and Berns, 2003, Carlson and Largaespada, 2005, Ranzani et al., 2013)]. Initial observations that tumors from multiple different birds contained evidence of ASLV integration adjacent to a specific cellular gene (c-myc) suggested that insertion of the viral promoter adjacent to this gene resulted in its enhanced expression, thereby leading to neoplastic transformation and positive selection for the provirus-harboring cells (Neel et al., 1981, Hayward et al., 1981, Fung et al., 1981, Noori-Daloii et al., 1981, Robinson and Gagnon, 1986). It was subsequently shown that retrovirus-mediated activation of additional genes could also stimulate tumor formation (Nusse and Varmus, 1982, Peters et al., 1983, Seiki et al., 1984, Sorensen et al., 1996), which led to the realization that retroviruses could be used as tools to probe entire genomes for theoretically all regions relevant to cancer. With completion of the Mouse Genome Project and development of powerful integration site sequencing methodologies, it became possible to conduct in vivo genetic screens using insertional mutagenesis, which is also referred to as retroviral tagging (Li et al., 1999, Hansen et al., 2000, Lund et al., 2002, Mikkers et al., 2002, Suzuki et al., 2002). These original high-throughput screens recovered hundreds of recurrent integration sites in different tumor samples (Akagi et al., 2004), thus leading to the classification of oncogenesis-relevant signaling pathways. The ability to focus on large networks of proteins as potentially being associated with cancer allowed for comprehensive follow-up biochemical studies, thus representing a major strength of the technique. One procedural limitation, however, has been the underrepresentation of identified tumor suppressor genes (TSGs), which are arguably as important for study as proto-oncogenes. This drawback results from the fact that powerful enhancer features in retroviral promoters cause most integrations to be activating rather than inactivating. Furthermore, whereas insertional mutagenesis proceeds via integration nearby one copy of a proto-oncogene, both copies of the TSG in general need to be inactivated to promote tumorigenesis (Knudson, 1971).

Surveillance of repopulating stem cell expansion during gene therapy

Aside from putative oncogene identification, retroviral tagging has been used to follow the repopulation dynamics of hematopoietic cells (Williams et al., 1984, Dick et al., 1985, Lemischka et al., 1986). It was ultimately discovered that individual retrovirally-transduced stem cell clones exhibited extensive variability in terms of lifespan and proliferative capacity (Guenechea et al., 2001). The clonal dominance of certain cellular subsets (depicted in Figure 4A) may be due to increased fitness caused specifically by integration-mediated transcriptional dysregulation of gene(s) involved in cell survival or self-renewal. Initial evidence described that integrations clustering within the transcriptional regulator gene Evi1 in mice and nonhuman primates were associated with long-term clonal persistence without signs of dominance or abnormal proliferation (Barjesteh van Waalwijk van Doorn-Khosrovani et al., 2003, Kustikova et al., 2005, Calmels et al., 2005), despite the fact that Evi1 expression had been correlated with leukemic disease progression (Barjesteh van Waalwijk van Doorn-Khosrovani et al., 2003, Buonamici et al., 2004). Surveillance of transduced cell engraftment in subsequent human gene therapy trials clarified that not only were integrations within EVI1 and other DNA replication- and cancer-related genes like PRDM16, SETBP1, CCND2, and HMGA2 recurrently present in recipient patients, but cell clones harboring these integrations had extensively expanded to dominate over time (Li et al., 2006, Deichmann et al., 2007, Cavazzana-Calvo et al., 2010). Interestingly, a β-thalassemia gene correction study in mice (Ocwieja et al., 2011), which employed the same vector utilized in the above-cited human trial (Cavazzana-Calvo et al., 2010), failed to find evidence for clonal expansion of cells harboring vectors integrated near genes involved in growth control. Therefore, taken together, it seems possible that species-specific factors play tangible roles in determining whether retroviral vector integrations influence host cell survival, engraftment, or proliferation. It is noteworthy that recent gene therapy trials utilizing optimized lentiviral vectors to correct metachromatic leukodystrophy and Wiskott-Aldrich syndrome have achieved efficient gene correction in multiple human patients without any evidence for clonal expansion among the tens of thousands of integration sites analyzed (Aiuti et al., 2013).

Figure 4. Clonal expansion of provirus-harboring cells.

A) General schematic of cellular clonal expansion. From an initially diverse population of cells, certain clones can exhibit relatively increased viability or growth rate, thus leading to persistence and ultimate enrichment. In this example the orange cell clone is dividing slowly or not at all, while the green cell clone exhibits an intermediate rate of expansion, and the blue cell clone expands most quickly. Thus in the final cell population, the blue clone predominates. B) Host cells harboring a particular integration site(s) can also become enriched over time within infected patients. The fact that clonally expanded, provirus-harboring cells can persist in the face of suppressive ART suggests that additional factors beyond integration site-mediated stimulation of cell growth rate contribute to clonal persistence. Two suggestions have been that the resident proviruses within the expanded cell clones are 1) nonfunctional due to non-tolerable insertions or deletions or 2) are molecularly intact but are expressed at a relatively low level (or not at all). In either case, the clones would escape viral replication-associated cytopathic effects and cell lysis (orange clone), as well as host cell recognition and purging by HIV-specific cytotoxic T-cells (green clone), leading to the enrichment of particular integration sites over time (blue clone). Neither of these possibilities, though, has thus far been definitively proven. A color version of this figure is available online.

Following integrated proviruses to classify host cell clonal expansion

Interestingly, cellular clonal expansion has recently been observed with provirus-harboring cells in HIV-1-infected patients undergoing ART, and the most highly expanded clones correspondingly exhibited strong integration clustering in cancer-related genes such as BACH2 (Ikeda et al., 2007, Maldarelli et al., 2014, Wagner et al., 2014). The presumption is that enhancement of cell cycle-regulating gene transcription by HIV-1 proviruses allows certain cells to persist in the face of ART for long periods of time and expand at a higher rate than cells that harbor less growth-promoting integration events. Accordingly, it has been suggested that the clonally expanded cells form a significant component of the latent viral reservoir (Maldarelli et al., 2014, Wagner et al., 2014), though a subsequent report failed to find supporting evidence for this model (Cohn et al., 2015). Consistent with the suggestion that clonally expanded cells can contribute to residual viremia is the fact that during suppressive ART the residual HIV-1 virion population shifts from diverse sets of variants to one or a small number of persisting clusters with genetic uniformity (Bailey et al., 2006, Josefsson et al., 2013, Kearney et al., 2014).

The means by which clonally expanded cells persist in the face of ART while other cells die is currently unknown, though there are two likely possibilities (Figure 4B). First, in addition to proliferating at a relatively increased rate, these clones could express the virus at an extremely low level. In this case the provirus would be duplicated continuously while remaining relatively dormant and thereby escaping not only drug treatment but also the added dangers of host cell recognition by HIV-specific cytotoxic T lymphocytes and viral protein-induced cytopathic effects. Though not pertinent to clonally expanded cells specifically, attempts have been made to quantify the relationship between integration site and latent provirus expression, and the results between systems have been inconsistent. In some cell culture models “latent” clones selectively harbor proviruses in heterochromatin (Jordan et al., 2001, Jordan et al., 2003), suggesting that low proviral transcription is a function of the surrounding genomic environment. Alternatively, replication-competent non-transcribed proviruses from treated patients have been found with high frequency in active transcription units (Han et al., 2004, Ikeda et al., 2007, Ho et al., 2013), while still other cell culture models failed to detect any chromatin feature that served as better latency predictors than pure chance (Sherrill-Mix et al., 2013). Therefore, the extent to which the chromatin environment of any given latent provirus contributes to viral transcriptional activity, host cell clonal expansion, and persistence is currently unresolved.

A second, equally plausible explanation for clonal cell resilience during ART could be that the proviruses that reside within these cells are defective due to inactivating insertions, deletions, or other mutations. In this case cells could continue to proliferate during ART without virion production while intact, integrated LTRs would still be sensed by integration site sequencing methodology. These integration sites would appear to be longitudinally enriched, since the bulk of other integrated proviruses would be eliminated over the course of time on treatment. Both sides of the coin have landed face-up in this research realm as well. Some studies have observed that up to 12% of clonally expanded proviruses are fully functional (Ho et al., 2013), while others have reported that every provirus that was analyzed contained an inactivating mutation of some sort (Josefsson et al., 2013, Imamichi et al., 2014, Cohn et al., 2015). Only a relatively low number of full-length proviruses have been sequenced when all of these studies are merged, mainly due to the laborious procedures required to obtain full HIV-1 genome (~10 kb) sequencing reads. A consensus in this realm may be thus reached in the near future when sample sizes are ultimately increased. This is becoming a possibility due to promising new technologies such as Single Molecule, Real-Time (SMRT®) Sequencing from Pacific Biosciences, which allows long and continuous DNA sequencing reads in a multiplexed format (Dilernia et al., 2015). The application of this particular technology to high-throughput HIV-1 full-genome sequencing is currently being optimized, and it holds the promise of becoming an invaluable new tool for following HIV-1 population dynamics over time in infected patients.

Perspectives

All retroviruses replicate in a similar fashion, reverse transcribing their RNA genome, integrating the vDNA product into host cell chromatin, and ultimately budding out of the host cell to infect a new target. However, different patterns and preferences evident at many steps of the infection process highlight close similarities and major differences among viral genera. These patterns are particularly evident at stages surrounding the integration step of the life cycle. As one example, γ-retroviruses preferentially target gene promoter and active enhancer regions, while viruses like MMTV integrate in a largely random fashion throughout the genome. Such differences not only provide an important springboard for understanding the dynamics of viral infection, but also for the gradual development of clinically applicable gene therapy vectors. Alternatively, the use of retroviruses as tagging agents has led to huge leaps forward in both the recognition of complex gene networks involved in cancer, as well as in the understanding of infused stem cell repopulation dynamics. And most recently the ability to distinguish clonally expanded proviruses from PCR-amplified sequence duplicates has begun to shed light on the elusive components of the latent HIV-1 reservoir. This important basic science and translational research has been made possible by extremely powerful, specific, and high-throughput integration site sequencing methodologies combined with thoroughly annotated reference genomes. These techniques have progressed from electrophoretic resolution of a handful of differently sized provirus-containing restriction fragments to the precise mapping of millions of unique tDNA sequences. To say that the methods have come a long way would be an understatement, but there is still ample room for further improvement. One weakness of the current state of integration site sequencing technology is that only about 30% of all LTRs within a DNA library can be expected to be amplified and recovered in a single sequencing run, even from the best of library preparations (unpublished observations). This potential limitation can be addressed by repeatedly sequencing the same library preparation and/or by sequencing multiple parallel libraries prepared from the same gDNA source. Further improvements in assay sensitivity should allow for more translational questions to be asked and answered as to the relationship between retroviral integration site and host disease state.