The sense behind retroviral anti-sense transcription

The ability of HIV-1 to encode an antisense protein was suspected as early as 1988, when a conserved ORF, later named asp-1, was identified in the region complementary to the env gene in many HIV-1 strains [5]. Since then, many studies have confirmed the expression of ASP-1 RNA and protein in vitro in various HIV-expressing cell types, including monocyte-derived macrophages, dendritic cells, and T cells [6‐10]. Antibodies recognizing an antisense protein derived from the env gene have also been identified in HIV⁺ patients [11]. Recently, ASP-1 has been shown to induce autophagy, which may explain its low abundance in HIV-1 infected cells and the inherent difficulty in detecting this protein [12]. ASP-1 has been postulated to utilize this pathway to enhance HIV-1 replication, as wild-type, but not mutated forms of this antisense protein, resulted in optimal viral replication through stimulation of autophagy [12]. In contrast, antisense transcript variants corresponding to proviral asp have been shown to inhibit HIV-1 replication in vitro in acutely and chronically infected cell lines, as well as in HIV-1 infected human PBMCs [8]. This suggests that HIV-1 antisense transcription may have a critical role in establishing viral latency. Moreover, ASP-1 can form stable cytosolic aggregates, which may sequester essential cellular proteins, and thus modulate the function of cellular pathways in HIV-1 infected cells [12]. Nonetheless, the precise functions of viral antisense RNA and protein during HIV-1 infection in vivo remain to be clearly elucidated.

The 3′ LTR promoter is crucial for driving antisense transcription in HIV-1. It has been experimentally demonstrated that HIV-1 antisense transcription initiates at multiple positions in the U3 region of the 3′ LTR, as well as at other downstream regions within the antisense strand [8]. Other studies have reported that HIV-1 antisense transcription is initiated in the U5 region of the 3′ LTR [7, 13]. The multiplicity of transcription initiation sites may be a consequence of the lack of a TATA box in the HIV-1 3′ LTR [6], in which case transcription is initiated through alternative core promoter elements called initiator (INR) motifs (YYANWYY) [7, 14]. The presence of these multiple INR motifs serves to explain the variability observed in transcription initiation sites reported by different studies.

In comparison to that of its sense transcription, the regulation of HIV-1 antisense transcription is not well understood. Nonetheless, several host transcription factors have been shown to play key roles in inducing transcription from the antisense strand of HIV-1, including Specificity Protein-1 (Sp1) [15, 16] Upstream Stimulating Factor (USF) [17], and Nuclear Factor-kappa B (NF-κB) [8, 16]. Mutagenesis of a conserved USF site has been shown to diminish the activity of the HIV-1 antisense promoter in reporter constructs [17]. The HIV-1 3′ LTR also contains several conserved NF-κB binding sites, which are known to drive HIV-1 antisense transcription. Point mutations in these sites have been demonstrated to down-regulate the activity of the antisense promoter in HIV-1 LTR reporter constructs [8, 15, 17]. NF-κB-activating agents, such as phorbol 12-myristate 13-acetate (PMA) and tumor necrosis factor α (TNFα), are also known to induce HIV-1 antisense transcription, likely through increased binding of NF-κB on the viral 3′ LTR [6, 8, 15]. In line with this finding, HIV-1 antisense LTR reporter plasmids containing mutated κB sites demonstrate decreased responsiveness towards PMA and TNFα stimulation [6, 8]. PMA stimulation of antisense transcription in a luciferase-expressing HIV-1 proviral DNA clone has also been demonstrated in both transfection and infection experiments [6]. In comparison, TNFα-mediated induction of HIV-1 antisense transcription remains debatable, as other studies have failed to replicate this phenomenon [6]. Thus, accumulating evidence illustrates an important role of NF-κB in the induction of HIV-1 antisense transcription. However, our bioinformatics analysis of the HIV-1 antisense promoter suggests many additional transcription factors likely contribute to the overall regulation of negative-sense transcripts in this retrovirus (Fig. 1, Table 1).

In addition, there is evidence of single nucleotide polymorphisms in transcription factor binding sites within the LTRs of different HIV-1 subtypes [18‐21]. This sequence variation leads to subtype-specific differences in proviral gene expression, thereby imparting unique biological characteristics to a given strain. This is well illustrated for the 5′ LTR of HIV-1 subtype E, which harbors a shift from an NF-κB to a GABP binding site [20]. Abolished NF-κB binding to this LTR does not lead to a loss of promoter function in vitro; instead, gain of GABP binding to this mutated NF-κB site enhances Tat-mediated HIV-1 gene expression in several cell types, as well as improves virus replication in SuPT1 cell line [20]. Thus, variability in transcription factor binding sites in retroviral LTRs can have a positive impact on proviral gene expression under certain conditions – this may serve to enhance retroviral fitness and spread. Likewise, subtype-specific variations in the HIV-1 3′ LTR may also exert another layer of control over the proviral antisense transcription.

Further, retroviral proteins can also modulate proviral antisense transcription. Conflicting results have been reported on the potential role of the HIV-1 accessory protein Tat in regulating antisense transcription. Tat has been reported to enhance HIV-1 antisense transcription in cell lines co-transfected with a Tat expression vector and luciferase reporter plasmids containing HIV-1 3′ LTR but not 5′ LTR [6]. However, in studies utilizing luciferase reporter constructs containing both HIV-1 3′ LTR and 5′ LTR, Tat was not shown to alter antisense transcription [22, 23]. Likewise, in a separate study utilizing HIV-1 3′ LTR luciferase reporter plasmids, overexpression of Tat did not influence antisense luciferase activity [8]. Thus, the role of Tat in regulating HIV-1 antisense transcription remains controversial and should be confirmed in the context of full length proviruses integrated during HIV-1 infection, rather than in artificial reporter assays. As there is no evidence of TAR RNA synthesis during antisense transcription, the mechanism by which Tat influences the activity of the HIV-1 antisense promoter remains unknown. It is possible that the interaction of Tat with cellular transcription factors, such as Sp1, modulates their binding to the HIV-1 3′ LTR [6], which may affect the extent of HIV-1 antisense transcription.

A large portion of our knowledge on retroviral antisense transcription stems from studies of Human T-Lymphotropic Viruses (HTLV), particularly HTLV-1. The HTLV-1 antisense genomic strand encodes a basic leucine zipper (bZIP)-containing protein, designated HBZ. Although HTLV-1 is capable of infecting different cell types in vitro, HBZ protein is mainly detected in CD4⁺ T cells in vivo [3, 24, 25]. This cell-type specific expression of HBZ has been shown to play a variety of roles in the pathogenesis of HTLV-mediated T-cell leukemia (reviewed in [3, 26]). For instance, HBZ transforms T-cells into a cancerous phenotype, in part by enhancing the expression of chemokine receptor CCR4 in this cell type, which promotes T-cell proliferation and migration [27]. HBZ also inhibits HTLV-1 sense transcription by recruiting essential transcription factors, such as CREB, away from the proviral sense promoter – this process facilitates HTLV-1 latency in infected T cells [3]. HBZ also affects many other cellular processes, including host gene expression, innate immune signaling, apoptosis, autophagy, and DNA repair – all of which further influence the pathology of the HTLV-1 infection (reviewed in [3]). Similar to HTLV-1, HTLV-2, HTLV-3, and HTLV-4 are equally capable of producing antisense proteins – APH-2, APH-3, and APH-4, respectively – though their functions have not been clearly elucidated [4, 26, 28‐30].

Despite the extensive research focused on deciphering the role of HTLV-encoded antisense proteins in disease pathogenesis, there are a limited number of studies aimed at understanding the regulation of HTLV antisense transcription at the level of the proviral 3′ LTR. It has been demonstrated that HTLV-1 hbz is transcribed starting from the 3′ LTR of the HTLV-1 provirus [31‐33]. Initiation of transcription is possible at several different positions within the R and U5 regions of the 3′ LTR [31]. Like HIV-1 antisense promoter, the HTLV-1 3′ LTR is a TATA-less promoter harboring many INR motifs, thus leading to a multitude of antisense transcription initiation sites [31, 34, 35]. The transcription of hbz relies heavily on three Sp1 sites in the U5 region of the proviral 3′ LTR [23, 31, 34, 36]. In luciferase assays, HTLV-1 antisense promoter activity is markedly reduced upon mutation of single or multiple Sp1 sites [34]. The same study identified binding sites for GATA binding protein-2 (GATA-2), cAMP responsive element binding protein (CREB), activating protein 1 (AP-1), and nuclear factor-1 (NF-1) in the HTLV-1 3′ LTR. However, mutations of each of these sites only reduced promoter activity slightly in luciferase assays [34]. Other cellular transcription factors, including activating transcription factor (ATF), CCAAT-enhancer binding protein (C/EBP), and histone acetyltransferase p300 have also been shown to bind the HTLV-1 antisense promoter in HTLV-1 transformed cell lines, as well as in cells derived from patients with Adult T-cell Leukemia/Lymphoma (ATL) [3, 37]. Thus, these transcription factors are postulated to play a role in regulating antisense HTLV-1 transcription. Whether they promote or inhibit antisense gene expression remains to be elucidated. In addition, T-cell factor 1 (TCF-1) and Lymphoid enhancing factor 1 (LEF-1) have been shown to slightly enhance hbz transcription and HTLV-1 3′ LTR activation in luciferase assays [38]. In line with these studies, bioinformatics analysis of the consensus HTLV-1 3′ LTR has not only confirmed the presence of intact binding sites for the aforementioned transcription factors, but has also revealed putative sites for numerous other antisense transcriptional regulators (Fig. 2, Table 1). Interestingly, some of the identified binding sites for transcription factors, notably ATF, CREB, and NF-I, are unique to the 3′ LTR of HTLV-1, and are not predicted within the HIV-1 3′ LTR. Thus, HTLV-1 antisense gene expression is likely regulated by a multitude of cellular and retroviral transcription factors. There is an evident need for future research characterizing the transcriptional regulators that broadly and selectively modulate antisense gene expression in the various tissue types targeted by retroviruses.

The transcription of HTLV-1 proviruses is further modulated by the antisense-encoded HBZ protein. HBZ binding to the 5´ LTR of HTLV-1 promotes viral latency by suppressing sense transcription [39]. Conversely, HTLV-1 antisense transcription is positively regulated by HBZ. HBZ has the capacity to form heterodimers with a cellular transcription factor JunD [36]. Co-expression of JunD and HBZ has been shown to significantly increase HTLV-1 3′ LTR activity in luciferase assays as compared to the expression of JunD or HBZ alone [36]. Also, luciferase activity was not enhanced with HBZ overexpression in knockout cells lacking JunD [36]. It was further shown that HBZ/JunD dimers are recruited to Sp1-bound regions of the HTLV-1 3′ LTR, due to the interaction of JunD with Sp1 [36]. Accordingly, mutation of one of these Sp1 sites in the HTLV-1 reporter construct, or the overexpression of Sp1 mutants lacking DNA-binding ability, resulted in a significant decrease in luciferase expression [36]. Therefore, HTLV-1 antisense transcription is regulated through interactions between HBZ, JunD, and Sp1 at the 3′ LTR.

As suggested for HIV-1 Tat interaction with its 3´ LTR, the HTLV-1 accessory protein Tax can also up-regulate the proviral antisense transcription. Overexpression of Tax has been shown to markedly enhance luciferase activity from transiently expressed, as well as stably integrated, HTLV-1 3′ LTR reporter constructs in human cell lines [40]. Tax responsive elements (TxREs) containing near-consensus CREB binding sites have been reported in the HTLV-1 antisense promoter [3, 34, 40]. Mutations of these TxREs, which render them incapable of interacting with CREB, exhibited a dramatically reduced luciferase activity from the 3′ LTR in the presence of Tax [34, 40]. Thus, viral Tax protein has been shown to drive HTLV-1 antisense transcription by cooperating with CREB at TxREs at the 3′ LTR. In stark contrast, several reports using similar methodology, but different host cells, have not detected Tax-mediated regulation of viral antisense transcription [23, 32, 41]. Thus, the discrepancies between these studies suggests that Tax-mediated regulation of antisense gene expression likely depends on the cell type being investigated, and consequently, the availability of cell-specific transcription factor complexes required for this process. This would be consistent with similar cell-type specific observations surrounding Tax-dependent transactivation of HTLV-1 sense transcription [41].

Among human endogenous retroviruses, antisense transcriptional regulation of ERV9 loci is the best understood. Several cellular transcription factors are known to induce the expression of antisense RNA from the U3 region (referred to as the U3 AS RNA) of the ERV9 LTR. This includes CREB, glucocorticoid receptors (GR), IRF, signal transducers and activators of transcription (STAT), and activating protein 2 (AP-2) [47]. Interestingly, the AUUGG motifs within the ERV9 antisense transcripts have been experimentally demonstrated to interact with and sequester select cellular transcription factors – NF-Y, p53 and Sp1 [47]. We have predicted the presence of similar motifs in the antisense RNA originating from the ERVK 3’ LTR (data not shown). By sequestering the aforementioned cellular transcription factors, ERV9 U3 AS RNA serves to repress the expression of genes involved in cell cycle activation, thereby inhibiting uncontrolled cellular proliferation. Accordingly, deregulation of this ERV-derived antisense RNA has the potential to promote tumor formation and propagation [47]. Thus, the production of endogenous AS RNA decoys may be an important phenomenon among endogenous retroviruses, and may serve essential regulatory and protective functions for their human hosts.

The human genome is ubiquitously populated with ERVK sequences including solitary LTRs and partial proviral sequences, as well as full-length proviruses. Solitary LTRs are the most abundant ERVK elements within the human genome, and are estimated to number over 25,000 [48]. They are frequently present in close proximity to our genes, and therefore may be involved in the regulation of neighbouring genes by acting as promoters or enhancers. It is estimated that at least 50% of human-specific ERVK (HML-2) LTRs serve as promoters for the transcription of human genes [49]. ERVK LTRs have been experimentally shown to activate the expression of promoter-less reporter genes in luciferase assays when inserted in both forward and reverse orientations, indicating their bidirectional promoter activity [48]. Such bidirectional activity lends plausibility to antisense viral RNA transcription mediated by the 3′ LTR of ERVK.

Recently, several ERVK loci present outside human intronic regions have been demonstrated to exhibit transcription of the proviral antisense strand in prostate cancer cell lines. These include ERVK(I), ERVK-106, an un-named ERVK within locus 7q34, and multiple loci of solo LTRs [50]. When inserted in an opposite transcriptional orientation to that of their host intron, antisense transcription of ERVK proviruses can be explained as a consequence of host gene transcription. In contrast, the basis of transcription of the antisense strands of ERVK loci, such as ERVK-106, situated outside of human genes remains unclear. Though the regulation of antisense transcription driven by the ERVK 3′ LTR is poorly understood, it is likely mediated by a complex of TFs binding to the 3′ LTR (Fig. 3).

As a first step to better understand putative antisense ERVK transcription from its 3′ LTR, we performed an extensive bioinformatics analysis of 92 full-length ERVK HML-2 sequences, and predicted intact and conserved binding sites for numerous human transcription factors within 3′ LTRs of human-specific ERVK HML-2 proviruses (Fig. 3 shows transcription factor binding sites on five prototypic ERVK 3′ LTRs, and Table 1). Similar to antisense promoters of other retroviruses, conserved signatures of ERVK 3′ LTR include absence of a TATA-box and the presence of multiple conserved INR motifs scattered throughout the LTR (Fig. 3). This suggests that these putative alternative core promoter elements may initiate transcription from the ERVK proviral antisense strand at multiple sites. Additionally, in the absence of a TATA box (TATAAA), select subtypes of exogenous retroviruses, such as HIV-1 subtype E, have been shown to utilize a CATA box (CATAAA) to initiate proviral gene transcription [51]. Since the ERVK 3′ LTR contains a conserved putative CATA box (Fig. 3), ERVK may similarly use this promoter element to initiate antisense transcription.

It is noteworthy that we identified multiple Sp1 binding sites in the ERVK 3′ LTR, as they are critical for inducing transcription from TATA-less promoters. This is due to Sp1 recruitment of transcription factor II-D (TFII-D) which promotes the formation of transcription initiation complexes [16]. Since Sp1 binding to both HIV-1 and HTLV-1 antisense LTRs activates expression of their respective antisense proteins, this ubiquitous transcription factor may also have a key role in driving antisense transcription from the TATA-less ERVK 3′ LTR. Indeed, the ERVK 3′ LTR is laden with multiple conserved potential Sp1 binding sites; a total of 15 were identified, a similar density to that found of the HTLV-1 antisense promoter (Fig. 3). The ERVK antisense promoter harbors putative binding sites for other transcription factors known to induce the activity of 3′ LTRs of exogenous retroviruses and ERV9. These include docking sequences for STAT, AP-2, AP-1, USF, GATA, NF-Y, ATF, CBP, TCF, LEF, and E-box (Fig. 3).

In addition, the ERVK 3′ LTR contains multiple putative NF-κB binding sites (Fig. 3). Thus, this pro-inflammatory transcription factor may drive ERVK antisense transcription under select conditions, as documented for exogenous retroviruses [17, 26]. Another interesting feature of the ERVK 3′ LTR is the presence of two conserved consensus interferon stimulated response elements (ISREs). We have recently demonstrated the ability of IRF1 and NF-κB p65/p50 to synergistically enhance transcription from the ERVK sense promoter in the presence of select pro-inflammatory cytokines [52, 53]. Of note, ERVK 3′ LTR ISRE sequences are more similar to canonical ISREs as compared to their 5′ LTR counterparts, suggesting stronger IRF/NF-κB binding potential [54, 55]. Thus, inflammatory stimuli that enhance the activity of IRFs and NF-κB have the potential to provide an additional level of regulation on the ERVK antisense transcriptome.

Bioinformatics analysis further revealed select transcription factor binding profiles unique to ERVK antisense promoters, notably the presence of hormone responsive elements that were absent from both HIV-1 and HTLV-1 3′ LTRs. This includes the presence of putative binding sites for androgen (AR), estrogen (ER), glucocorticoid (GR), and progesterone (PR) receptors (Fig. 3, Table 1). Since these hormonal receptors are known to drive the activity of the ERVK 5′ LTR [53, 56, 57], they may also modulate the activity of its 3′ LTR.

The TF profile of the ERVK 3′ LTR may also point to cell-specific activation of antisense transcription. As with the HTLV-1 3′ LTR, the ERVK 3′ LTR contains putative binding sites for FOXP3, a transcription factor specific to regulatory T cells. Expression of antisense HTLV-1 transcripts in T regulatory cells is associated with the development of Adult T-cell Leukemia (ATL) [3, 24], potentially suggesting a shared mechanism for ERVK-associated leukemia [58]. In addition, several GATA family transcription factor binding sites are found within the ERVK antisense LTR. Owing to the importance of GATA family transcription factors in regulating immune cells [59], our data indicate that antisense ERVK expression may be modulated in hematopoietic cells, Moreover, impairment of GATA transcription factors are a hallmark of many cancers [60]. If ERVK were to employ an antisense product whose expression was i) driven by GATA proteins, and ii) held a similar latency-inducing function as HTLV-1 HBZ, the lack of GATA protein expression in cancers could explain the enhanced expression of sense-encoded ERVK protein products in transformed cells [58, 61].

Understanding the activity of the ERVK 3′ LTR promoter may be the key to elucidating the basis of antisense transcription of endogenous proviruses; however, it should be noted that not all ERVK LTRs are equally intact. Evaluation of human-specific and older 3′ ERVK LTRs in the HML-2 family reveals conserved, alternative and unique TF binding site profiles, when comparing recent and older provirus LTRs (data not shown). Therefore, developing an understanding of ERVK antisense transcription, especially in the context of specific genomic loci, is an area of research that clearly requires more investigation.