HTLV-1 is prevalent across the globe. HTLV-1 is particularly endemic in some areas, including southwestern Japan, Central Australia, South America, the Caribbean islands, and sub-Saharan Africa (11). Approximately 5–10 million people are estimated to be infected with HTLV-1 globally (1). HTLV-1 is transmitted via infected lymphocytes from HTLV-1 carriers; breastfeeding and sexual contact are common routes of transmission where infectious lymphocytes are transferred to a new host (12–14).

HTLV-1 is mainly found in CD4⁺ T lymphocytes. An infected lymphocyte transmits HTLV-1 through cell-to-cell contact with other lymphocytes. HTLV-1 viral components, including its single-stranded RNA genome, are transferred to target cells through this junction (15). Recently, Hiyoshi et al. reported that the host factor M-Sec plays a critical role in efficient viral transmission (16). M-Sec induces membrane protrusions and establishes intercellular conduits (17). This is likely the molecular basis of what is known as the virological synapse in HTLV-1 infections (15).

HTLV-1 genomic RNA is reverse-transcribed in the target cell, and the resulting double-stranded DNA, 9 kb in size, is inserted into the host genome. The location at which the HTLV-1 provirus is inserted in each infection is not completely random. HTLV-1 favors genomic sites near genes, CpG islands, and chromatin regions with epigenetic marks associated with gene regulation (18). Unlike HIV-1 infection, where the reverse-transcribed HIV-1 genome is guided to actively transcribed genes by the host factor LEDGF (19), the mechanism by which HTLV-1 is preferentially integrated in these characteristic regions is currently unknown. Host factor PP2A has been identified as a binding partner of the HTLV-1 integration complex (20, 21). More studies are needed to elucidate the mechanisms underlying HTLV-1 integration preferences.

It has been postulated that HTLV-1 propagates rapidly in a new host during the early stages of infection. HTLV-1 is believed not to produce cell-free infectious viral particles in vivo. HTLV-1 increases the proviral copy number by a combination of de novo cell-to-cell infection and mitotic division in each infected cell. Each infected cell carries a single HTLV-1 provirus copy in its genome (22).

The expression of viral genes for HTLV-1 propagation in the new host elicits the host immune response. HTLV-1-infected cells will be lysed by cytotoxic T lymphocytes (CTLs) that are specific for viral antigens (23, 24, 25). Therefore, HTLV-1 propagation is counterbalanced by the host immune response, which in turn determines the set point of proviral load (PVL) in the host. PVL in ACs is approximately 1% [i.e., HTLV-1 is found in 1% of total peripheral blood mononuclear cells (PBMCs)]; PVL varies by 1,000-fold among ACs (26, 27). It is estimated that PVL in each individual is typically maintained by the mitotic division of cells in the chronic phase of infection (28). PVL positively correlates with the risk of developing ATL and HAM/TSP; that is, the risk of disease onset is greater with a higher PVL (26, 29).

HTLV-1 inserts its genome at a unique location on the host chromosome during de novo infection. Each infected cell carrying a single copy of the HTLV-1 provirus in the genome gives rise to a group of sister cells, or a clone, by mitotic division, which shares the same proviral integration site. Gillet et al. estimated the abundance of each clone, or clonality, in ACs and patients with ATL and HAM/TSP by quantifying the frequency of each provirus integration site using high-throughput sequencing (30). It is estimated that tens of thousands of unique HTLV-1-infected clones exist in a typical host. These clones persist for many years, from which a malignant clone emerges (31).

ATL is a malignancy characterized by clonal expansion of HTLV-1-infected lymphocytes, often with a PVL of >90% in acute ATL cases. It takes decades for a malignant clone to emerge from a typical HTLV-1 infection. Recently, two studies retrospectively performed exon sequencing of clinical samples to track gene mutations before ATL onset. These studies found, among other genes, recurrent mutations in CCR4, PLCG1, PRKCB, and NOTCH1 that precede the onset of ATL (32, 33). It is possible that HTLV-1 infection per se does not cause ATL. HTLV-1 infection prolongs the lifetime of infected lymphocytes, during which infected lymphocytes acquire a set of gene mutations and undergo malignant transformation.

HAM/TSP is another clinical entity associated with HTLV-1 infections. The PVL is significantly higher in HAM/TSP patients than in ACs (26). Monoclonal expansion is not observed in HAM/TSP; instead, it is envisaged that the number of clones increases, which accounts for the high PVL (30).

HTLV-1 tax and HBZ, as we describe in the next section, are the main viral factors that confer a growth advantage to infected cells. It appears that HTLV-1 performs two contradicting tasks: expressing viral genes to sustain the infected cells and avoiding CTL killing exerted by the host immune response. Therefore, understanding the regulation of HTLV-1 genes in vivo is crucial for understanding HTLV-1 infection and its associated diseases.

HTLV-1 viral genes are encoded in both the plus and minus strands of the provirus, which is 9 kb in size and is embedded in the host chromatin (Figure 1). HTLV-1 has two long terminal repeats (LTRs) at the 5′ and 3′ ends of its provirus. HTLV-1 gag, pol, and env, the essential retroviral genes, are encoded on the plus strand. HTLV-1 carries an additional genomic segment, referred to as pX, which is downstream of the env gene. The pX region encodes HTLV-1 tax, rex, and other accessory genes (34). The plus strand is transcribed from the promoter, which resides within the 5′ LTR. Alternative splicing yields mature mRNA for each gene. On the minus strand, HTLV-1 encodes the HTLV-1 bZIP factor, or HBZ, and its transcription is initiated within the 3′ LTR.

Distinct transcription factors operate in plus- and minus-strand transcription. Each LTR consists of three regions: U3, R, and U5. U3 contains binding sites for activating transcription factor (ATF), cAMP response element-binding protein (CREB), and activator protein 1 (AP-1) for plus-strand transcription. In contrast, minus-strand transcription is driven from U5 in the 3′ LTR by transcription factor Sp1 (35). HTLV-1 Tax, once it is produced, forms a complex with the transcription factors on the 5′ LTR and recruits CBP/p300, thereby enhancing viral gene transcription. Recently, another mechanism of plus-strand transcription was reported by Wang et al., where Yin Yang 1 (YY1) binds to the R region of HTLV-1 transcripts and enhances transcription initiation (36). Interestingly, transcriptional enhancement was not observed when the YY1-binding element was placed upstream of the transcription start site. It has been proposed that YY1 binds to HTLV-1 plus-strand transcripts, as opposed to DNA, and enhances transcription initiation.

HTLV-1 tax and HBZ have been extensively studied to understand the pathogenicity of HTLV-1 (37) (Figure 2). Tax binds to several proteins. For example, Tax binds to the nuclear factor κB (NF-κB) components and activates its pathway, resulting in the activation of inflammatory signaling. Conversely, HBZ has a counteracting effect on HTLV-1 tax. HBZ protein interferes with the NF-κB pathway (38). The HBZ protein binds to ATF and AP-1 transcription factors and inhibits their function (39). HBZ RNA also functions in the nucleus. Two recent studies performed RNA precipitation to identify chromatin regions targeted by HBZ RNA. Gazon et al. found that HBZ RNA binds to HTLV-1 LTR and displaces TATA-box binding protein, thereby suppressing the transcription of the plus strand (40). Ma et al. reported that HBZ RNA associates with the CCR4 promoter and enhances CCR4 expression (41). CCR4 is a chemokine receptor highly expressed in ATL (42) and HAM/TSP (43). CCR4 is an important molecule in HTLV-1 infection, not only because its mutation significantly contributes to the development of ATL as described above, but also because it serves as a marker for ATL and HAM/TSP (44, 45), and is targeted by the monoclonal antibody mogamulizumab, a clinically approved drug for the treatment of ATL and HAM/TSP (46, 47).

Despite the pleiotropic functions of HTLV-1 plus-strand gene tax, tax mRNA is often not detected in clinically isolated PBMCs (48, 49). This observation can be explained by three mechanisms: gene deletion, gene mutation, and DNA methylation near the 5′ LTR, where the plus-strand transcripts are encoded.

HTLV-1 proviruses are often defective in ATL. Tamiya et al. reported two types of defective proviruses: lacking the retroviral gag and pol segments or lacking the 5′ LTR (50). In some ATL cases, the deletion occurs before the HTLV-1 provirus is integrated into the genome (51). A proviral deletion was preferentially found at the 5′ end of the provirus, whereas the 3′ end of the provirus was unaffected. This raises the possibility that plus-strand genes are not essential, whereas the antisense HBZ gene is crucial for pathogenesis (52). More recently, Katsuya et al. reported a gene deletion in the 3′ LTR of the HTLV-1 provirus (53). The significance of losing the 3′ LTR, and hence HBZ expression, on HTLV-1 persistence and pathogenesis is yet to be investigated.

The second mechanism of gene silencing involves point mutation in the tax gene. A point mutation that introduces a premature termination codon in the tax mRNA results in the loss of functional Tax protein, a strong activator of its viral sense promoter (48, 54).

The third mechanism involves epigenetic modification. The cytosine residues in the 5′ LTR and the adjacent downstream region are highly methylated; although the 3′ LTR has a sequence identical to the 5′ LTR, the 3′ LTR is not methylated (55, 56). Currently, it is not known what regulates the contrasting DNA methylation patterns in the 5′ LTR and 3′ LTR. Recently, Satou et al. reported that HTLV-1 binds the host factor CTCF in the pX region upstream of the 3′ LTR (57). CTCF is responsible for transcriptional regulation, DNA insulation, and chromatin folding. Therefore, it was hypothesized that CTCF binding regulates DNA methylation in the pX region and keeps the 3′ LTR open for transcription. Cheng et al. showed that the boundary of DNA methylation moved beyond the CTCF-binding site toward the 3′ LTR without CTCF (58), whereas two CRISPR-mutated primary T cell clones eliminating the CTCF binding on the provirus reported no impact (59). It is possible that the effect of CTCF on regulating DNA methylation depends on the location at which the HTLV-1 provirus is integrated.

Although viral gene expression appears to be silenced during latency in HTLV-1 infection, a strong cellular immune response has been detected for plus-strand products such as Gag and Tax (60). This indicates that plus-strand transcription is not permanently silenced; however, HTLV-1 genes are expressed intermittently, which constantly evokes the host immune response against viral antigens. The apparent silencing of HTLV-1 plus-strand transcription in vivo is reversible. HTLV-1-positive PBMCs from HTLV-1-infected individuals initiate viral gene expression once they are isolated from the peripheral blood and cultured (24, 61, 62). Approximately half of HTLV-1-infected PBMCs reactivate plus-strand transcription, although this varies among HTLV-1-infected individuals (20 to 80%) (24, 63). Reactivation occurs rapidly within the first few hours of in vitro culture (63, 64). The plus-strand transcription reactivation is intense: about a hundred of transcripts are produced per hour in a single cell with the positive feedback of Tax protein (63). Kulkarni et al. showed that p38 MAP kinase and deubiquitylation of histone H2A in the HTLV-1 provirus are responsible for viral gene activation in ex vivo culture (65). The primary stimulation that ultimately leads to the activation and deubiquitylation of these factors is obscure. Any physical or chemical stress that PBMCs experience when drawn from the circulation may trigger spontaneous viral transcription reactivation. It is probable that HTLV-1 reactivates in breast milk in response to non-specific stimulation due to environmental changes.

Billman et al. recently applied single-molecule RNA fluorescence in situ hybridization (FISH) to detect viral transcripts in HTLV-1-infected cells in vitro (66). They used HTLV-1-infected cells freshly established and cultured from patient PBMCs (22). Single-molecule FISH detects diffraction-limited spots, each of which are from a single mRNA, thereby allowing for the absolute quantification of viral transcripts expressed in each cell (67). Using this technique, Billman et al. found that plus-strand genes are expressed in a transcription burst. Transcription is rare; however, once the genes are expressed, hundreds of transcripts are produced at a time. In contrast, the minus-strand transcripts contained per cell were much fewer (up to ∼10 molecules). A slight deviation from the Poisson distribution indicates that minus-strand transcription also occurs in a burst, yet it is much smaller. Stochastic HBZ transcription results in approximately 20% of a clonal population with no HBZ transcripts at a given time. The occurrence of the plus-strand transcription burst is associated with the progression to the G2/M cell cycle stage. Although the causation of these two events is not clear, the function of tax and HBZ genes suggests that HTLV-1 gene expression accelerates cell cycle progression.

The termination of a transcriptional burst is a common question in gene regulation. In the study by Billman et al. the occurrence of plus-strand transcription was significantly lower in HBZ-positive cells, in line with other observations that HBZ, in the form of protein (35, 68) or RNA (40), suppresses plus-strand expression.

It has been postulated that HTLV-1 genes are intermittently expressed in vivo. Does the rare expression of HTLV-1 plus-strand genes observed in vitro account for the viral expression in vivo? Therefore, if, and when PBMCs intermittently transcribe HTLV-1 plus-strand genes in vivo, are the frequency and intensity of expression similar to those observed in vitro? In a previous study where single-molecule RNA FISH was performed on hundreds to thousands of HTLV-1-infected PBMCs for each HTLV-1-positive subject, no intense plus-strand transcription burst was reported unless cultured in vitro (63). The frequency and intensity of plus-strand gene bursts in vivo should be much smaller than those observed in vitro.

There is an apparent discrepancy between PBMCs in vivo, fresh in vitro culture, and HTLV-1-infected cells maintained in vitro. Is it possible to translate the findings of HTLV-1 gene expression in vitro into the unseen nature of HTLV-1 gene expression in vivo?

First, there is a correlation between spontaneous plus-strand reactivation in HTLV-1-positive PBMCs in vitro and the expression of HTLV-1 genes in vivo. Patient-derived PBMCs contain many distinct clones of HTLV-1-infected lymphocytes. The provirus integration site is a strong determinant of spontaneous plus-strand transcription in vitro, and the degree of spontaneous expression in vitro is inversely correlated with the clonal abundance (18). This indicates that a clone that reactivates plus-strand transcription in vitro also transcribes HTLV-1 genes at high frequency in vivo, as the relatively small abundance of that clone is a result of CTL killing that recognizes viral expression (24).

It appears that cells in vitro, where a strong plus-strand transcription burst is observed, are in another equilibrium state that is different from what might otherwise be in vivo in which HTLV-1 is silenced. HTLV-1-positive PBMCs show transient, spontaneous reactivation of plus-strand transcription along the way throughout the circulation. It is possible that HTLV-1-positive lymphocytes express plus-strand genes when certain conditions are met in vivo, if not in peripheral blood, such as in lymph nodes or bone marrow (69), especially with the aid of local stimulatory signals from other cells in those compartments. This possibility is supported by the in vitro study by Kulkarni et al. that lower glucose availability and hypoxic conditions both enhance tax transcription (70). We are currently developing a microscopic technique to capture the transcription burst in each HTLV-1-positive clone in a given native tissue environment.

Finally, is the spontaneous expression of viral genes truly stochastic? If it is truly stochastic, then the expression is governed by the probabilistic binding of biochemical molecules under random thermodynamic fluctuations. Or if it is not otherwise, there should be unseen factors that determine the HTLV-1 gene expression. It is tempting to assay the transcription burst on an HTLV-1-infected cell line carrying multiple copies of the HTLV-1 provirus: if the multiple HTLV-1 copies burst at the same instance within a single cell, then this predicts that there are unseen factors that coordinate the HTLV-1 transcription initiation. The outcome of the in vitro study will be translated into an understanding of how HTLV-1 gene expression is regulated in vivo.

As HTLV-1 transmission requires cell contact, HTLV-1 propagates within the host by both clonal expansion of infected cells and de novo viral infection. In HTLV-1-infected individuals, cell-free virus particles are usually undetectable, and the plasma does not transmit the infection. Furthermore, PVL in PBMCs, which reflects the number of virus-infected cells, correlates with the risk of developing ATL and HAM/TSP (26, 29). It is therefore believed that HTLV-1 is almost entirely cell-associated in vivo, and clonal proliferation of infected cells predisposes individuals to ATL and HAM/TSP. Among HTLV-1 genes, tax and HBZ play a particularly important role in regulating the expression of viral and host genes as well as the activation and proliferation of host cells (71) (Figure 2). Tax induces the expression of serum response factor (SRF) and various cellular genes via transcriptional pathways, such as the NF-κB, CREB, and AP-1 pathways (72, 73). In contrast, the HBZ protein suppresses the transcription of the tax gene and the cellular pathways that Tax activates. HBZ RNA suppresses apoptosis by inducing survivin expression (74) and, therefore, promotes the proliferation of T cells (52). Thus, understanding how HTLV-1 regulates the expression of viral and cellular genes in vivo is key to elucidating the mechanisms of long-term survival and the persistence of HTLV-1 in infected individuals, which is closely related to the development of HTLV-1-associated diseases. The roles of Tax and HBZ in the pathogenesis of ATL and HAM/TSP are summarized in Figure 3.

Approximately 60% of ATL patients do not express tax mRNA in freshly isolated PBMCs (48). Tax is often repressed once ATL develops (75), whereas HBZ mRNA is expressed in all ATL cases (49), because HTLV-1 provirus is substantially silenced by proviral defects and/or epigenetic mechanisms (see section “Silent Infection of HTLV-1 in PBMCs in vivo”). These findings suggest that Tax is essential to initiate transformation, while HBZ has roles in promoting viral replication and cellular proliferation to maintain the transformed ATL cells when Tax expression is extinguished. If this is the case, it may be the most efficient method to escape HTLV-1-specific CTLs.

Previous reports have described the downregulation of microRNAs in ATL cells (76, 77). This may cause disordered gene expression at the transcriptional and post-transcriptional levels, thereby contributing to the development of ATL. The relationship between downregulation of microRNA and gene expression of tax and/or HBZ has not been reported; thus, it should be investigated in future studies.

In patients with HAM/TSP, the quantity of PVL in PBMCs is significantly higher than that in ACs and is well correlated with the concentration of neopterin in the cerebrospinal fluid (CSF) (26), a marker associated with cell-mediated immunity (78), and with disease progression (79). In individual HAM/TSP patients, PVL in CSF cells was higher than that in PBMCs, and the ratio of PVLs in CSF cells/PBMCs was significantly associated with clinically progressive disease and recent onset of HAM/TSP (80). Thus, HTLV-1 PVL is an important biomarker for HAM/TSP. Meanwhile, the total amount of HTLV-1 tax mRNA in PBMCs and mRNA expression level in HTLV-1-infected cells (mRNA/DNA ratio) were significantly higher in HAM/TSP patients than in ACs and correlated with PVL, Tax-specific CTL frequency, and disease severity of the patients (81). In addition, HBZ mRNA load was positively correlated with PVL, disease severity, and neopterin concentration in the CSF of HAM/TSP patients (82).

HTLV-1-specific CTLs are abundant in PBMCs of infected individuals, and their frequency is proportional to the PVL, indicating that HTLV-1 is not latent in vivo but is expressed persistently or at frequent intervals in infected individuals (3). Interestingly, in HTLV-1 infection, although the dominant antigen recognized by HTLV-1-specific CTLs is the Tax protein (83, 84), PVL and the risk of HAM/TSP are determined by the CTL response to poorly immunogenic HBZ proteins (85, 86). This is consistent with the idea that a persistent HTLV-1 infection establishes an equilibrium between viral replication and the host immune response, and that the response of HTLV-1 specific CTLs determines the equilibrium PVL and the risk of HAM/TSP. Meanwhile, it remains possible that the chronically activated anti-HTLV-1 CTLs found in patients with HAM/TSP contribute to systemic inflammation. Many host genes dysregulated by Tax and HBZ may activate and proliferate host cells and induce systemic inflammation in vivo.

In recent decades, it has become evident that chronic inflammation and tumor development are closely related (87). Infection, chemical substances, and injury can initiate tumorigenesis at the associated inflammation site. Alternatively, tumor induction triggers inflammation through the secretion of chemokines and inflammatory signaling molecules, creating a local microenvironment that supports further tumor development.

There is strong evidence suggesting that inflammation, induced by HBZ, promotes the development of ATL. HBZ-transgenic mice develop lymphoma, in which it was shown that HBZ induces Foxp3 expression and the proliferation of the regulatory T-cell phenotype (88). However, the Foxp3 expression in these cells is not stable; such cells secrete IFN-γ (Interferon-gamma) and promote systemic chronic inflammation (89). The incidence of lymphoma was significantly low in HBZ-transgenic mice lacking IFN-γ, suggesting that IFN-γ, alongside HBZ, promotes tumorigenesis in HTLV-1 infection (90). More recently, Higuchi et al. reported an unexpected observation; the deletion of IL-6, also a pro-inflammatory cytokine, increased the incidence of lymphoma development in their transgenic mice, and IL-10 was upregulated in HBZ-transgenic mice lacking IL-6 (91). IL-10 is an immunosuppressive cytokine and it is known to promote the proliferation of HTLV-1-infected T cells in vitro (92). Higuchi et al. have shown that IL-10 signaling is redirected to T-cell proliferation by HBZ modulating the STAT pathway.