Introduction

Adult T-cell leukemia/lymphoma (ATLL) is defined as a mature T-cell neoplasm caused by human T-cell leukemia virus type 1 (HTLV-1) [1]. ATLL has a long incubation period, and only less than 5% of HTLV-1 carriers develop ATLL in their lifetime [1, 2]. According to Shimoyama [3], ATLL is classified into four clinical subtypes: acute, lymphoma, smoldering, and chronic type. Of the four clinical subtypes, acute, lymphoma, and chronic types, are especially known to have poor prognosis [4, 5].

Previous studies with deep sequencing have revealed that most HTLV-1-infected T cells contain a single copy of integrated HTLV-1 provirus [6] and that each host contains a large number (often 104–105) of distinct HTLV-1-infected T-cell clones [6, 7]. Moreover, the integration sites of the provirus could be involved in clone selection and clinical subtype [7].

Like other complex retroviruses, HTLV-1 provirus encodes long terminal repeat (LTR) on both 5′ and 3′ sides, and structural genes (gag, pol, env) downstream of 5′LTR [8, 9]. Furthermore, regulatory genes (tax and rex) and accessory genes (p12, p13, p30, and HTLV-1 bZIP factor (HBZ)) are encoded in pX region [8, 9]. Of these viral genes, HBZ is only encoded on the minus strand of the provirus and transcribed from 3′LTR. On the other hand, other viral genes including tax are encoded on the plus strand and transcribed from 5′LTR [8, 9].

Tax plays an important role in the early stages of tumorigenesis through various mechanisms [10,11,12,13,14,15,16]. It is considered to be involved in tumorigenesis in vivo [17]; however, the expression of tax in ATLL could be suppressed due to genetic and epigenetic alterations [18,19,20,21]. While tax mRNA is frequently undetectable in ATLL cells, it has been reported that HBZ mRNA is expressed in almost all ATLL cells [22]. HBZ suppresses the Tax-mediated transcription from 5′LTR by interacting with CREB-2 [23]. On the other hand, HBZ promotes cell proliferation and migration, [22, 24,25,26] and induces T-cell lymphoma in vivo [27]. Interestingly, HBZ has different functions as RNA and protein [28].

From these results, both HBZ and tax are considered to be deeply associated in the pathogenesis of ATLL. However, there are scant data concerning HTLV-1-related mRNA including HBZ and tax in human formalin-fixed, paraffin-embedded (FFPE) tissue samples. In this study, we detected HBZ and tax mRNA on FFPE tissue samples using in situ hybridization (ISH), to investigate their association with clinicopathological characteristics in ATLL patients.

Materials and methods

Patients and samples

In this study, 88 biopsy samples from newly diagnosed ATLL patients were examined, which are included in our previous studies [29,30,31]. Tissue microarrays (TMAs) including all 88 samples were created with a 2-mm core diameter. Each sample was reviewed according to the World Health Organization classification [1] by two experienced hematopathologists (HM and KO). The use of patient materials and clinical information was approved by the Research Ethics Committee of Kurume University and was in accordance with the Declaration of Helsinki.

In situ mRNA analysis

ISH was performed on FFPE samples using RNAscope 2.5 HD Reagent Kit-BROWN [Advanced Cell Diagnostics (ACD), Hayward, CA] according to the manufacturer’s protocols. In brief, 2.5-μm-thick sections from the TMA samples were created. All sections were baked at 60 °C for 1 h and deparaffinized; then other pretreatments were performed appropriately. Hybridization was performed at 40 °C for 2 h using HybEZ hybridization oven (ACD). HBZ-specific probe (ACD) and tax-specific custom-designed probe (ACD, targeting 7394-7812 of U19949.1) were used. Hs-PPIB (ACD) and DapB (ACD) were used for positive and negative control of ISH assay, respectively (Supplementary Fig. 1). ISH of HBZ and tax was validated using MT-4 (HTLV-1-immortalized cell line) and Jurkat (T-cell acute lymphoblastic leukemia cell line). Amplification and detection of signals were performed properly, and then hematoxylin was used for counterstain. All samples were scanned by using Aperio ScanScope AT2 (Leica Biosystems, Vista, CA, USA). Dot-like signals were counted at high magnification (40 diameters) on ten randomly selected fields. The number of signals per 1000 ATLL cells were calculated. High expression was indicated when not less than the median value of HBZ or tax signals was stained.

Immunohistochemical analysis

Immunohistochemistry (IHC) was performed as previously reported [29,30,31]. The antibodies used for IHC targeted CD4, CD30, Ki-67 (MIB-1), CCR4, FoxP3, GATA3, IRF4, HLA class I, β2 microglobulin (β2M), HLA class II, PD-1, and PD-L1 (Supplementary Table 1). As previously reported [29,30,31], neoplastic PD-L1 (nPD-L1) was considered positive when 50% or more neoplastic cells were stained. PD-L1-positive nonmalignant stromal cells were counted as microenvironmental PD-L1 (miPD-L1), and miPD-L1 was considered as positive when 10 or more nonmalignant stromal cells were stained per high power field (HPF) [29,30,31]. Other than PD-L1, it was considered as positive when 30% or more neoplastic cells were stained [29,30,31]. PD-1-positive tumor infiltration lymphocytes (TILs) were counted in up to five representative HPFs, and the average and median values were calculated as previously reported [29,30,31].

Statistical analysis

Clinicopathological characteristics of ATLL patients were compared by Fisher’s exact test (2-sided), Mann–Whitney’s U test, and Spearman’s rank correlation analysis. Overall survival (OS) was defined as the time from the day of diagnosis to the day of death or last follow-up. Progression-free survival (PFS) was defined as the time from the day of diagnosis to the day of first progression or death. OS and PFS were estimated using the Kaplan–Meier method and compared using the log-rank test. Univariate and multivariate analyses for survival time were performed using the Cox proportional regression model. P < 0.05 was considered as statistically significant. All statistical analyses were performed by EZR ver. 1.32 [32].

Results

Clinicopathological characteristics of ATLL patients

Table 1 summarizes the clinicopathological characteristics of 88 newly diagnosed ATLL patients. The median age was 66 years (range: 35–85), including 54 men and 34 women. The median follow-up period was 12.1 months (range: 0–105.2). CR or CR unconfirmed was achieved in 29% (22/76) of patients. According to Shimoyama classification [3], 1% (1/71); 44% (31/71); and 55% (39/71) of patients were of smoldering, acute, and lymphoma types, respectively. No patient was of chronic type in this study. High or high-intermediate scores of international prognostic index (IPI) were observed in 56% (46/82) of patients. High scores of Japan Clinical Oncology Group prognostic index (JCOG-PI) were observed in 39% (33/84) patients.

Table 1 Clinicopathological characteristics of ATLL patients.
Full size table

Regarding IHC of tumor-immunity-related proteins, patients were positive for both HLA class I and β2M for membrane (39% [30/78]); HLA class II (34% [28/82]); neoplastic PD-1 (nPD-1) (16% [14/87]); and nPD-L1 (5% [4/87]). PD-1-positive TIL ranged from 0 to 187.0 counts/HPF with an average value of 7.4 counts/HPF and with a median value of 0 count/HPF. miPD-L1 ranged from 0 to 171.0 counts/HPF with an average value of 27.2 counts/HPF and with a median value of 22.0/HPF.

ISH of HBZ and tax mRNA

HBZ and tax mRNA were evaluated by ISH for 88 ATLL samples shown in Table 1. Figure 1 shows the histogram and scatter plot of HBZ and tax mRNA. HBZ signals ranged from 0.4 to 4013.1/1000 ATLL cells, with an average value of 916.7/1000 ATLL cells and with a median value of 795.2/1000 ATLL cells. Tax signals ranged from 0.1 to 891.2/1000 ATLL cells, with an average of 80.4/1000 ATLL cells, and a median of 5.1/1000 ATLL cells. Both HBZ and tax signals showed extremely high expression in a few cases. There was no significant correlation between HBZ and tax signals (ρ = −0.00367, P = 0.973; Fig. 1c). Representative samples of HBZ and tax signals were presented in Fig. 2a, b, respectively. Target-specific signals showed dot-like pattern and were observed in ATLL cells.

Fig. 1: Distribution of HBZ and tax mRNA signals.
figure 1

a Histogram of HBZ mRNA signals. The median value of HBZ signals (/1000 ATLL cells) was 795.2 (range: 0.4–4013.1). b The median value of tax signals (/1000 ATLL cells) was 5.1 (range: 0.1–891.2). c Scatter plot of HBZ and tax mRNA signals. There was no significant correlation between HBZ and tax signals (ρ = −0.00367, P = 0.973).

Full size image
Fig. 2: Representative results of ISH.
figure 2

a ISH of HBZ mRNA. Target-specific signals showed dot-like pattern. The signals were expressed in a large number of ATLL cells, with some cluster formation (arrow). b ISH of HBZ mRNA. 1 or 2 signals were expressed per ATLL cell (arrow). c ISH of HBZ mRNA. A few ATLL cells expressed the signals, and the intensity of some signals was weak (arrow). d ISH of tax mRNA. At most 1 or 2 signals were expressed per ATLL cell, and the intensity of many signals was weak (arrow). e ISH of tax mRNA. The signals were expressed only in a small number of ATLL cells (arrow). Original magnification is ×1000 for all panels.

Full size image

Clinicopathological comparison according to the expression of HBZ mRNA

HBZ has been known to be expressed in all ATLL cells. However, the clinicopathological characteristics due to the difference in the expression level have remained unknown. Therefore, we showed a clinicopathological comparison between the high-expression group and the low-expression group of HBZ (Table 2). Notably, the low-expression group of HBZ displayed significant increases in skin lesions (P = 0.0283), Ann Arbor stage III or IV (P = 0.00696) and IPI high-intermediate or high (P = 0.0461). Small or medium cell variants, compared with other variants, were significantly more frequent in the low-expression group of HBZ (P = 0.000771).

Table 2 Comparison of clinicopathological characteristics and HBZ mRNA expression.
Full size table

HBZ is known to be closely associated with immune-suppressive phenotypes of HTLV-1 infected cells [33]. Therefore, we evaluated the expression level of HBZ and the phenotypes of ATLL. HBZ-transgenic (Tg) mice and in vitro analysis showed that HBZ induces Foxp3 expression [33], however, no significant association was observed between the expression level of HBZ and FOXP3 expression in this study (Table 2). In addition, there was no difference between the expression level of HBZ and the other protein expression (Table 2). However, by comparing each protein expression frequency with the expression level of HBZ, weak but significant correlation was observed between HBZ and IRF4 (ρ = 0.326; P = .0288), between HBZ and nPD-L1 (ρ = −0.264; P = 0.0135), and between HBZ and miPD-L1 (ρ = −0.223; P = 0.038) (Fig. 3).

Fig. 3: Representative correlation between HBZ and clinicopathological findings.
figure 3

a HBZ and IRF4 (ρ = 0.326; P = 0.0288). b HBZ and miPD-L1 (ρ = −0.223; P = 0.0380). c HBZ and nPD-L1 (ρ = −0.264; P = 0.0135).

Full size image

Clinicopathological comparison according to the expression of tax mRNA

Regarding the expression level of tax, we first performed clinicopathological comparison between the high-expression group and the low-expression group of tax (Table 3). Notably, the high-expression group of tax displayed significant increases in lactate dehydrogenase (LDH) activity (P = 0.00209), splenomegaly (P = 0.00721), bone marrow (BM) involvement (P = 0.0295), CD30 positivity (P = 0.00434), nPD-1 positivity (P = 0.0385) and PD-1-positive TIL (P < 0.0001). In addition, the high-expression group of tax displayed significant reductions in male (P = 0.0481) and radiation (P = 0.0298).

Table 3 Comparison of clinicopathological characteristics and Tax mRNA expression.
Full size table

Subsequently, we compared each protein expression frequency with the expression level of tax; then, weak but significant correlation was observed between tax and age (ρ = 0.225; P = 0.00352), between tax and CD30 (ρ = 0.303; P = 0.00435), between tax and GATA3 (ρ = 0.237; P = 0.0263), between tax and β2M for membrane (ρ = −0.281; P = 0.0298), between tax and nPD-1 (ρ = 0.259; P = 0.0156), between tax and nPD-L1 (ρ = 0.297; P = 0.00527), and between tax and PD-1-positive TIL (ρ = 0.397; P = 0.000138) (Fig. 4).

Fig. 4: Representative correlation between tax and clinicopathological findings.
figure 4

a tax and Age (ρ = 0.225; P = 0.0352). b tax and CD30 (ρ = 0.303; P = 0.00435). c tax and GATA3 (ρ = 0.237; P = 0.0263). d tax and β2M (ρ = −0.281; P = 0.0298). e tax and nPD-1 (ρ = 0.259; P = 0.0156). f tax and nPD-L1 (ρ = 0.297; P = 0.00527). g tax and PD-1-positive TIL (ρ = 0.397; P = 0.000138).

Full size image

OS in ATLL patients according to the expression of HBZ and tax mRNA

We evaluated the association between the expression level of HBZ and prognosis, however, there was no significant difference in OS between the high-expression group and the low-expression group of HBZ (Log-rank P = 0.834; Supplementary Fig. 2A). Likewise, there was no significant difference in OS between the high-expression group and the low-expression group of tax (Log-rank P = 0.365; Supplementary Fig. 2B).

Clinicopathological characteristics and OS in ATLL patients with extremely high expression of tax mRNA

Tax is not expressed in many ATLL patients [18,19,20,21]. However, Tax has been reported as a viral oncoprotein [10,11,12,13,14,15,16]; it is highly immunogenic and can be the target for cytotoxic T lymphocytes (CTLs) [34,35,36,37,38,39,40]. In this study, we identified seven patients with extremely high expression of tax (not less than 400 signals/1000 ATLL cells). Among these pathological characteristics, only 1/7 patients (14%) were positive for HLA class I for membrane and no patient was positive for β2M for membrane. The group of 400 or more tax signals displayed significant reductions in the membranous positivity of HLA class I and β2M compared with the group of less than 400 tax signals (P = 0.0385, 0.0124, respectively) (Table 4). The group of 400 or more tax signals also displayed significant increases in LDH activity (P = 0.0170) and splenomegaly (P = 0.0326). These group (n = 6, median survival time [MST] 7.7 months, 95% confidence interval [CI] [4.7–NA]) had significantly inferior OS compared with the group with less than 400 tax signals (n = 78, MST 22.6 months, 95% CI [13.7–41.7]) (P = 0.0499; Fig. 5).

Table 4 Clinicopathological comparison between tax signals ≧400 group and tax signals <400 group.
Full size table
Fig. 5: OS of ATLL between the two tax expression groups.
figure 5

The group of 400 or more tax signals displayed significant inferior OS compared with the group of less than 400 tax signals (P = 0.0499).

Full size image

Prognostic factors in ATLL patients

We analyzed the prognostic factors affecting OS and PFS in ATLL patients (Supplementary Tables 3, 4). In OS (n = 74), univariate analyses identified the following variables as prognostic factors: age over 70 (hazard ratio (HR), 2.196; 95% CI, 1.209–3.986; P = 0.00974), miPD-L1 expression (HR, 0.545; 95% CI, 0.307–0.970; and P = 0.0390), HLA class I and β2M expression (HR, 0.494; 95% CI, 0.253–0.964; and P = 0.0386) and HLA class II expression (HR, 0.398; 95% CI, 0.205–0.772; and P = 0.00640). Multivariate analyses revealed that age over 70 remained a significant prognostic factor (HR, 2.262; 95% CI, 1.171–4.369; and P = 0.0150). In PFS (n = 74), univariate analyses identified tax signals not less than 400 as a prognostic factor (HR, 12.570; 95% CI, 2.186–72.340; and P = 0.00457), but was not remain as a significant prognostic factor in multivariate analyses (P = 0.158).

Discussion

In this study, we demonstrated three new findings on FFPE samples in ATLL patients as follows: (i) some patients showed low expression of HBZ; (ii) some clinicopathological characteristics including antitumor immunity were significantly associated with expression of HBZ and tax; and (iii) the extremely high-expression group of tax was significantly associated with the loss of HLA class I/β2M and poor prognosis.

In our in situ mRNA analysis, the median value of HBZ signals was about 800/1000 ATLL cells with low-expression group of HBZ observed. Satou et al. reported that HBZ mRNA was universally expressed in almost all ATLL cells [22]. HBZ mRNA suppresses cell apoptosis via transcription of survivin [25] and promotes cell proliferation and migration [22, 24,25,26]. On the other hand, HBZ protein has different functions from HBZ mRNA; HBZ protein promotes transcription of immunity-related genes including FoxP3, interleukin 2 (IL-2), and IL-10; and promotes cell proliferation and apoptosis [28]. In this study, it cannot be completely ruled out that the low-expression group of HBZ was observed due to RNA degradation in FFPE samples [41]. However, a recent study with integrated molecular analysis reported that some ATLL patients displayed low expression of HBZ [42]. Furthermore, another study reported that HBZ mRNA expression was suppressed in some HTLV-1-infected cell lines and fresh ATLL cells [43].

This study revealed that the high-expression group of HBZ displayed significant reduction in skin lesion, progressive Ann Arbor stage and HI or more risk of IPI; these results were different from the previous studies with model mice [24, 27]. Sato et al. reported that skin lesion of CD4-positive T cells increased in the Tg mice expressing HBZ on CD4-positive T cells, compared to non-Tg mice [27]. These results are suggested to be derived from HBZ-mediated cell proliferation in vitro and in vivo [22, 24,25,26,27], and from HBZ-mediated migration by inducing CCR4 via GATA3 [24]. However, no significant differences were observed between GATA3, CCR4, HBZ and organ infiltration in this study (data not shown). More detailed studies of HBZ expression will validate the association between HBZ expression and biology in ATLL tissue samples.

Unlike HBZ, the median value of tax signals was around 5/1000 ATLL cells in the present study: tax expression was suppressed in most patients. However, a small number of patients displayed extremely high expression of tax. A recent study showed that most patients displayed low expression of tax and that 1/57 (1.8%) ATLL patients displayed high expression of tax [42]. In this study, the high-expression group of tax displayed significant increases in splenomegaly and BM involvement. Tax has various functions [11], including activation of the transcription of HTLV-1-related genes from 5′LTR [10], collapse of cell cycle checkpoint [12], activation of nuclear factor kappa B (NF-κB) [13, 14], and inducing genomic instability and chromosomal aneuploidy via TAX1BP2 [15] and RanBP1 [16]. Especially, NF-κB was activated in tax-Tg mice with ATLL-like phenotype [17]. Splenomegaly and infiltration into many organs including BM and spleen are observed in tax-Tg mice and severe combined immunodeficiency mice transferred lymphoma cells from tax-Tg mice [17]. For the first time, we found the association between tax and splenomegaly, and between tax and BM involvement in ATLL tissue samples. More detailed studies of tax expression are needed to elucidate the mechanism of these associations. In addition, 5′-LTR deletion [18] and epigenetic alterations including DNA methylation of 5′-LTR [19,20,21] may occur in the patients with low expression of tax. Further studies are needed to validate biological significance of tax expression level in ATLL tissue samples.

In this study, weak but significant positive correlation was observed between tax and nPD-1, between tax and nPD-L1, and between tax and PD-1-positive TIL. Moreover, the group with 400 or more signals of tax displayed poor prognosis and reductions in HLA class I, β2M or both for membrane (HLAm+β2Mm+). However, 400 or more signals of tax did not have the prognostic significance in multivariate analysis. Tax is the most immunogenic of HTLV-1-associated molecules [34,35,36,37,38,39] and Tax-specific CTL response was activated in some ATLL patients who achieved complete response (CR) after allogeneic hematopoietic stem cell transplantation [40]. Tax-specific CTLs are being applied clinically [44], however, “T-cell exhaustion” [45, 46] is induced due to PD-1 expression on Tax-specific CTLs [47]. CTLs recognize HLA class I/β2M [48].The mechanism of HLA class I/β2M loss in this study is suggested to be hypermethylation, loss-of-function mutations and copy number deletion of HLA class I/β2M genes [42]. We have previously reported that ATLL patients with HLAm+β2Mm+ is significantly associated with high expression of miPD-L1 and a good prognosis compared to patients in other groups [30]. There was no significant association with miPD-L1 expression in this study. Various numbers of PD-1-TIL were observed in this study (Supplementary Fig. 3), and we also analyzed the prognostic significance of PD-1-TIL positivity; there were no significant differences (data not shown). Mahgoub et al. have reported that sporadic and transient Tax expression by various cytotoxic stresses in cooperation with HBZ is critical for survival of the whole population [49]. We calculated tax/HBZ ratio (Supplementary Fig. 4) and investigated the clinicopathological features; there were no new significant differences (data not shown). In this study, we showed for the first time that the high expression of tax is important in evading antitumor immunity including loss of HLA class I/β2M in ATLL tissue samples. Furthermore, the high expression of tax may be associated with poor prognosis, so Tax-targeted treatment such as Tax-targeted vaccine therapy [44] should be performed carefully. Comprehensive analysis with tissue microenvironment including various immune cells is necessary for elucidating the biological significance of tax expression in antitumor immunity.

Weak but significant positive correlation was observed in HBZ and IRF4 in the present study. HBZ suppresses gene expression downstream of the classical pathway of NF-κB including IRF4 due to inhibiting the binding of p65 to DNA and inducing the degradation of p65 [50]. However, in ATLL patients, that gain-of-function mutations are highly enriched for T-cell receptor/NF-κB signaling, including IRF4 [42]. Moreover, a recent study with functional screening in ATLL cell lines revealed that HBZ promoted the expression of basic leucine zipper ATF-like transcription factor 3 (BATF3) and its downstream targets including IRF4, and that both BATF3 and IRF4 were necessary for the gene expression and proliferation of ATLL [51]. Further studies are expected to clarify the detailed mechanism of HBZ and IRF4 expression in ATLL.

There were some limitations in the present study. First, this study was conducted only with patients diagnosed by biopsy samples. Because previous HTLV-1-clonality analysis suggests that HTLV-1 clone size is different in peripheral blood and in lymph nodes [52], further studies are needed to validate our results, using all clinical subtypes of ATLL patients diagnosed by various organ samples. Second, this study is conducted only with mRNA regarding HBZ and tax, and is not conducted with gene mutation analysis. More detailed studies conducted with integrated analysis including DNA, RNA and protein on ATLL tissue samples are required. Last, sample size is relatively small in this study although several statistical differences were found in the present study. Further large-scale studies are needed to confirm our results.

In conclusion, we demonstrated for the first time that the expression of HBZ and tax mRNA is associated with clinicopathological characteristics including antitumor immunity on ATLL tissue samples. These results suggested that detailed in situ mRNA analysis on FFPE samples may identify the association between the expression of HTLV-1-related mRNA and biology. Multidisciplinary analysis using DNA, RNA, and protein of ATLL and HTLV-1 is necessary for more-detailed analysis of the pathophysiology and for applying the treatment and prevention of ATLL.