Functional Analysis of Aberrantly Spliced Caspase8 Variants in Adult T-Cell Leukemia Cells Free

Adult T-cell leukemia/lymphoma (ATL) is one of the most aggressive T-cell malignancies caused by infection with human T-cell leukemia virus type 1 (HTLV-1). HTLV-1 is transmitted vertically via breastfeeding from infected mothers to infants and horizontally by sexual transmission. HTLV-1–infected cells are immortalized and subsequently transformed into ATL cells 50 to 60 years after infection. It is considered that some HTLV-1 carriers develop into relatively mild indolent-type ATL, that are the smoldering and chronic types, then further progress to aggressive-type ATL such as acute and lymphoma types. Although ATL cells of patients with indolent-type ATL show monoclonal expansion, the number of the monoclonal cells usually remains stable, whereas those of patients with aggressive-type ATL show uncontrolled cell growth with constitutive activation of NFκB. The molecular abnormalities that trigger malignant transformation of ATL cells remain to be clarified. We demonstrated various molecular disorders in ATL cells, such as abnormal epigenetic modifications (1), disorders of transcription factors (2), and drastic dysregulations in gene expression profiles (3–5). We previously reported overexpression of structurally abnormal transcripts in ATL cells (2, 6), indicating malfunctions of splicing machinery in ATL cells.

Caspase8 (Casp8) has been well documented to have dual roles in cell death and survival, that is, as an extrinsic apoptotic mediator and as a cell-proliferative NFκB activator (7). cFLIP is the negative regulator of Casp8, although FLIP and Casp8 both belong to the same group of proteins called the death effector domain (DED) containing protein family (8–10). Both Casp8 and cFLIP have two DEDs in the N-terminus, whereas only Casp8 has a functional caspase domain in the C-terminus, and cFLIP has a pseudo-caspase domain without activity (8, 9). For activation of apoptotic cascade, Casp8 is required to form a homodimer for self-cleavage. It is known that v/cFLIP forms a heterodimer with Casp8 through their DEDs, thus inhibit Casp8. The DED-containing proteins, including FLIP and Casp8, are known to activate NFκB. vFLIP from Kaposi sarcoma–associated herpes virus/human herpes virus 8 (KSHV/HHV8) was shown to activate NFκB by phosphorylating IκBα, provably via interactions with NIK and IKKα/β (11). cFLIP activates NFκB by interacting with TRAF2 via its pseudo-caspase domain (12, 13). Later, it was demonstrated that the two DEDs of cFLIP were enough to activate NFκB (14). As for Casp8, NFκB activation is mediated through the interaction with RIP, NIK, IKKα, and IKKβ (15). In addition, Shikama and colleagues (16) demonstrated that Casp8 interacts with RIP, NIK, and TRAF2 via its DED-containing prodomains. Interestingly, both authors showed significantly higher NFκB activations by recombinant Casp8 prodomain (2DEDs) than the full-length Casp8.

In ATL cells, an unusually spliced form of CASP8 mRNA with 136-nt insertion of intron 8 between exon 8 and exon 9 was reported and named CASP8L mRNA (17, 18). Because of a nonsense codon within the insertion, CASP8L mRNA is expected to encode a truncated Casp8 isoform of 270 aa, lacking the C-terminus catalytic domain but retaining two DEDs at the N-terminus (reported as isoform7 or β3. Casp8L hereafter in this study). Later, it was demonstrated that two DEDs competed for binding to FADD with WT-Casp8 and inhibit Fas-mediated apoptosis (19), although the influences of Casp8L on NFκB activation have never been examined.

In this study, we thoroughly surveyed quantitative and qualitative abnormal expression patterns of the CASP8 gene at transcriptional and protein levels in PBMCs of HTLV-1 carriers and patients with ATL. Then we clarified the functional changes of these abnormal Casp8 isoforms due to mutations, compared with the original functions of WT-Casp8, which plays a critical role in optimization of apoptosis and cell growth in a normal cell.

All cell lines used in this study were obtained and maintained as reported previously (2). Briefly, Jurkat, CEM (T-ALL patient-derived T-cell lines), TL-Om1, ATN1, MT-1 (ATL patient-derived T-cell lines), MT-2, and C91/PL (HTLV-1-immortalized T-cell lines) were maintained in RPMI (Gibco, Thermo Fisher Scientific) containing 10% FBS (Gibco, Thermo Fisher Scientific) at 37°C with 5 % CO₂. HeLa (cervical cancer-derived epithelial cell line), 293T (human embryonic kidney-derived cell line containing the SV40 T-antigen), and HEK293FT (derivative of 293T) cells were cultured in DMEM (Nissui Pharmaceutical Co., LTD.) containing 10% FBS (Gibco, Thermo Fisher Scientific) at 37°C with 5% CO₂.

The cell lines used in this study were obtained and authenticated as follows. TL-Om1, derived from malignant T cells from a patient with ATL, was kindly provided by Dr. K. Sugamura (Tohoku University, Miyagi, Japan). We authenticated within 6 months of this study by FISH and HTLV-1 provirus-specific real-time PCR and confirmed that this cell line contained 1 copy/cell of HTLV-1 provirus at the site of 1p13 of chromosome 1 in the genomic DNA, maintaining the original characteristics of TL-Om1 (20). C91/PL was a kind gift from Dr. Maria-Isabel Thoulouze (The Pasteur Institute, Paris, France), who authenticated and confirmed that this cell line contains one HTLV-1 full-length provirus/cell and produces infectious HTLV-1 virions within 6 months of this study. MT-2, HTLV-1 transformed T-cell line, was kindly provided by Dr. H. Hoshino, (Gunma University, Gunma, Japan). We authenticated within 6 months of this study by the proviral integration-site sequencing technique (21), and confirmed that this cell line contains 10 copies of HTLV-1 provirus/cell. We also confirmed that this cell line produced infectious HTLV-1 viral particles and expressed viral Gag–Tax fusion protein, which are major characteristics of MT-2 cells. Jurkat was obtained within 6 months of this study from the ATCC, who authenticated this cell line by short tandem repeat (STR) analysis. HEK293FT was obtained more than 6 months before this study from Riken Cell Bank, who conducted a series of authentication analysis on this cell line, including growth–curve analysis, cell adhesion analysis, isozyme analysis, and STR analysis. Just after arrival to our laboratory, HEK293FT cells were cultured for several days in the conditioned medium and divided to 20 cryo-tubes as live cell stock, then stored in liquid N₂ until use. ATN1, MT-1, CEM, HeLa, and 293T cells were gifted by the Japanese Foundation for Cancer Research (JFCR) more than 6 months before this study. We authenticated these cell lines by karyotype analysis within 6 months of this study.

Peripheral blood mononuclear cells (PBMC) from patients with ATL, HTLV-1 carriers, and healthy volunteers were a part of those collected with written informed consent as a collaborative project of the Joint Study on Prognostic Factors of ATL Development (JSPFAD). The project was approved by the Human Genome Research Ethics Committee in the Institute of Medical Sciences, The University of Tokyo (IMSUT). Clinical information of patients with ATL and healthy volunteers enrolled in this study is shown in Supplementary Table S1.

For detection of proteins in Western blotting, the following primary antibodies were purchased from the indicated companies; Casp8 (sc-73526, Santa Cruz Biotechnology, Inc.), GST (GE27-4577-01, Merck KGaA), and β-actin (sc-69879, Santa Cruz Biotechnology, Inc.). For the secondary antibodies, anti-mouse IgG antibody or anti-goat IgG antibody conjugated with horseradish peroxidase (KPL) was used.

Total RNA was extracted from PBMCs isolated from a patient with smoldering-type ATL using Isogen (Nippon Gene Co., Ltd.) following the manufacturer's protocol. Extracted total RNA was subjected to reverse transcription to cDNA using SuperScript II (Invitrogen, Thermo Fisher Scientific). Semiquantitative-PCR was then performed by Platinum Taq DNA Polymerase High Fidelity (Invitrogen, Thermo Fisher Scientific) to amplify the complete coding region (1,440 bp) or intron 8–containing cDNAs with a reverse primer designed in the intron 8 (834 bp). Amplicons were separated by electrophoresis in 2% agarose gels. Then bands were exerted from the gels to extract DNAs, which were TA-ligated to pGEM-T-Easy vector (Promega Corp.) with the Ligation kit (Takara Bio Inc.) for 1 hour at 16°C, followed by transformation to Escherichia coli (DH-5α). Eight clones each were bacterially amplified and sequenced (Fasmac). Sequences were aligned with WT-CASP8 mRNA sequence (Variant B, NM_033355) to clarify the structures of CASP8 transcript variants in ATL cells.

Total RNA was extracted from cell lines or primary PBMCs using Isogen (Nippon Gene Co., Ltd.) following the manufacturer's protocol. Extracted total RNA samples were subjected to reverse transcription using SuperScript II (Invitrogen, Thermo Fisher Scientific), followed by quantitative real-time PCR using SYBR Premix Ex Taq and Thermal Cycler Dice (both from Takara Bio Inc.) or by semiquantitative PCR. Supplementary Table S2 shows primer sequences used for the real-time PCR.

To detect WT-CASP8 and CASP8L mRNAs, we designed the forward primer in the exon 8 and the reverse primer in the exon 9 of CASP8 mRNA, respectively (Fig. 2A). With these primers, WT-CASP8 and CASP8L cDNAs are detected as 234 bp and 370 bp bands, respectively. Total RNA was extracted from HTLV-1–uninfected T-cell lines (Jurkat and CEM) and ATL patient–derived HTLV-1–infected T-cell lines (TL-Om1, ATN-1, MT-1), followed by reverse transcription and PCR by Platinum Taq DNA Polymerase (Invitrogen, Thermo Fisher Scientific) with the primers mentioned above.

To evaluate the preference of intron8 inclusion or exclusion among cell lines with/without HTLV-1 infection, we constructed the CASP8 mRNA splicing reporter plasmid. As shown in Fig. 2B, the partial fragment of CASP8 mRNA from the exon 8 to exon 9, including the first 136 nt of the intron 8 with intact splicing sites was PCR amplified and inserted to pCDNA 3.1C-His vector (Invitrogen, Thermo Fisher Scientific), followed by cDNA fragments encoding EGFP and mCherry (Clontech, Takara Bio, Inc.) in the different frames. The Kozak sequence and ATG were inserted at the start of the whole fragment; thus, mRNAs transcribed from this plasmid would be translated from the first ATG. If the intron 8 is spliced-out (WT splicing pattern), the start codon will be in-frame with EGFP. If the intron 8 is included (CASP8L splicing pattern), the start codon will be in-frame with mCherry. The naturally occurred stop codon within the intron 8 region of CASP8L mRNA was deleted in this reporter plasmid, allowing the read-through to mCherry. Therefore, with this reporter plasmid, preference of WT- (EGFP expression) or CASP8L-(mCherry expression) splicing pattern in tested cells can be detected in tested cell lines by flow cytometry.

The cDNAs of the full-length CASP8, CASP8L (including the first 136 bp of the intron 8 between exon 8 and exon 9), CASP8-ΔE4 (without the exon 4), and CASP8-ΔE7 (without the exon 7) were PCR amplified from the cDNA library of PBMCs from a patient with smoldering-type ATL for constructions of expression plasmids of WT-Casp8, Casp8L, Casp8-ΔE4, and Casp8-ΔE7 isoforms. Each cDNA fragment was inserted to pME-FLAG, pEGFP-C1 (Clontech, Takara Bio, Inc.), pMEG-2, and pmCherry-C1 (Clontech, Takara Bio, Inc.) vectors for the expression of FLAG-, EGFP-, GST-, and mCherry-tagged Casp8 isoforms, respectively.

For the construction of mutant Casp8 isoforms (F122G/L123G) to destruct the filamentation via DEDs (22, 23), point mutagenesis was conducted with PrimeStar (Takara Bio, Inc.) in WT-CASP8, CASP8-ΔE4, CASP8-ΔE7, and CASP8L in pEGFP-C1 and in pmCherry-C1. The detailed primary structures of Casp8 isoforms examined in this study are described in Supplementary Fig. S1.

HEK293FT cells were seeded at the concentration of 8 × 10⁴ cells/mL and 200 mL/well in 48-well culture plate one day before transfection. At 24 hours after seeding, 10 ng reporter plasmid (6×NFκB-binding sites-pGL4.10), 5 ng RSV-Renilla-luciferase plasmid, and 200 ng of Casp8-expressing plasmid for a well were cotransfected by polyethylenimine (PEI). At 24 hours after transfection, Renilla and firefly luciferase activities were measured using the dual luciferase assay system (Promega Corp.) with the Centro LB 960 luminometer (Berthold Technologies GmbH & Co. KG). Detected firefly-luciferase activity was divided by corresponding Renilla-luciferase activity to normalize the transfection efficiency.

Induction of apoptosis by Casp8 isoforms was analyzed by flow cytometry. EGFP-tagged Casp8 isoforms (WT-Casp8, Casp8L, Casp8-ΔE4, and Casp8-ΔE7) were overexpressed in HEK293FT cell and in Jurkat cells. For apoptosis assays, cells were stained with PE-AnnexinV in the binding buffer (all from PE-AnnexinV Apoptosis Detection Kit, BD Biosciences, Thermo Fisher Scientific) for 20 minutes at room temperature before analyses. All flow cytometry analyses were conducted with FACSCalibur (BD Biosciences, Thermo Fisher Scientific), with appropriate instrument settings and compensation, if necessary. Obtained data were analyzed by FlowJo (BD Biosciences, Thermo Fisher Scientific).

The status of cellular NFκB activity was observed by subcellular localization of EGFP-tagged RelA. mCherry-tagged Casp8 isoforms (WT-Casp8, Casp8L, Casp8-ΔE4, and Casp8-ΔE7), together with EGFP-tagged RelA were cooverexpressed in the HEK293FT cell and in Jurkat cells. Then, subcellular localizations of mCherry-Casp8 and EGFP-RelA were observed by a fluorescent microscope (Olympus IX73, Olympus Corp.) and a laser-scanning microscope (Nikon A1, Nikon Corp.). The EGFP-tagged RelA expression plasmid (RelA-pEGFP-N1) was a kind gift from Dr. Makoto Nakashima, (The University of Tokyo, Tokyo, Japan).

The fluorescent resonance energy transfer (FRET) analysis is a method to evaluate the distance between two donor- and acceptor-dye molecules. FRET is a transfer of excitation energy from donor to acceptor molecule (calculated as FRET efficiency), which occurs effectively only when the distance between the two-dye molecules is less than 10 nanometers, that is, they are most likely physically interacting.

FRET analyses were conducted to observe the interaction between mCherry-Casp8 and HA-IKKγ. The mCherry-Casp8 expression plasmid and HA-IKKγ expression plasmid were cotransfected to HEK293FT cells, and further incubated at 37°C for 96 hours. The pCDNA3HA-NEMO (24) was a gift from Dr. Kunliang Guan via Addgene (plasmid #13512; http://n2t.net/addgene:13512; RRID: Addgene_13512). The cells were fixed in 4% paraformardehyde in PBS for 15 minutes, then blocked in the blocking buffer (3% BSA in PBS) for 1 hour. Then, the cells were immunostained with anti-HA antibody (#561-5, MBL Co., LTD.; 1:100 in the blocking buffer) for 1 hour followed by Alexa Fluor 488-goat anti-rabbit IgG (#A11034, Molecular Probes, Thermo Fisher Scientific; 1:500 in the blocking buffer) for 1 hour before the FRET analysis.

The FRET analysis was also conducted to investigate the interaction between EGFP-WT-Casp8 and mCherry-Casp8 isoforms. The expression plasmid of EGFP-WT-Casp8 and a mCherry-Casp8 (WT-Casp8, Casp8L, Casp8-ΔE4, or Casp8-ΔE7) were cotransfected into HeLa cells by Lipofectamine 2000 (Invitrogen, Thermo Fisher Scientific). To eliminate nonspecific interaction between EGFP/mCherry and Casp8, we prepared HeLa cells expressing EGFP-WT-Casp8/mCherry, and EGFP/mCherry-WT-Casp8 as negative controls. After transfection, HeLa cells were further incubated for 24 hours, then fixed in 4% paraformardehyde in PBS for 15 minutes prior to FRET analyses.

To calculate the FRET efficiency, the acceptor-photobleaching method was used. LSM710 confocal microscope (Carl Zeiss AG) was used for imaging of Alexa488, EGFP, and mCherry, by excitation at 488 and 543 nm, respectively, as well as acceptor-photobreaching of mCherry by the excitation laser of 543 nm in the prepared cells. The FRET efficiency was calculated as follows; FRET efficiency (%) = [1−(prephotobreaching intensity of Alexa488 or EGFP)/(postphotobreaching intensity of Alexa488 or EGFP)] × 100.

We identified several transcript variants of CASP8 mRNAs in primary ATL cells, including previously reported WT-CASP8 mRNA (NM_033355 or Var. B encoding Isoform B), CASP8L (encoding Isoform 7 or β3), and CASP8-ΔE7 without exon7 (NM_033358 or Var. E encoding isoform5 or β1), together with unreported Casp8-ΔE4 without exon4 (Fig. 1A). The Casp8L mRNA has a premature termination codon (PTC) in the inserted intron 8 region. Deletion of exon 4 yields a PTC in exon 5, whereas deletion of exon 7 results in a PTC in exon 8. Therefore, all CASP8 variants, beside WT-CASP8 mRNA, identified in this study encode truncated proteins without the caspase catalytic domain encoded in exon 9 and exon 10 (Fig. 1A). The WT-CASP8, CASP8L, CASP8-ΔE7, and CASP8-ΔE4 mRNAs encode the full-length WT-Casp8 (479 aa), Casp8L hereafter (270 aa with two DEDs,), Casp8-ΔE7 hereafter (235 aa with two DEDs), and Casp8-ΔE4 hereafter (108 aa, a newly identified isoform with one DED), respectively (Figs. 1A and 3A; Supplementary Fig. S1).

Next, the expression levels of aberrant CASP8 mRNAs were measured by real-time PCR with primers specifically designed for detection of each splicing variant (Fig. 1B). Total and each CASP8 mRNA levels increase significantly from healthy donors, asymptomatic carriers (AC) to patients with ATL. Especially, patients with acute-type ATL tended to have higher levels of Casp8L, CASP8-ΔE4, and CASP8-ΔE7 mRNA levels than patients with indolent-type ATL, ACs, and healthy donors. We also examined protein expression levels of Casp8 isoforms in primary samples by immunoblotting (Fig. 1C). Pro-Casp8 (WT-Casp8) levels were higher in normal PBMCs compared with PBMCs from patients with chronic-type and acute-type ATL. The Casp8L isoform was detected at higher densities in patients with chronic-type ATL, although the reason is not clear at present. On the other hand, Casp8-ΔE4 and Casp8-ΔE7 were detected only in ATL samples but not in normal samples (Fig. 1C), which agreed with the mRNA levels of these CASP8 variants (Fig. 1B).

We examined the WT-CASP8 and CASP8L mRNA expression patterns by semiquantitative RT-PCR in HTLV-1 infected/uninfected T-cell lines (Fig. 2A). WT-CASP8 mRNA was detected at the similar levels in all tested cell lines, whereas CASP8L mRNA was detected at higher levels in HTLV-1 infected ATL patient-derived T-cell lines compared with HTLV-1–uninfected ALL patient-derived T-cell lines. Splicing patterns of CASP8 mRNA in HTLV-1–related cell lines were further analyzed by the reporter system described below, which has been constructed on the basis of CASP8 mini-gene. As shown on the top of Fig. 2B, the WT-CASP8 mRNA type splicing results in the in-framed EGFP expression, whereas the CASP8L mRNA type splicing results in a frameshift and expression of mCherry. Analyses by flow cytometry showed that the population size of mCherry-expressing cells was more massive in HTLV-1–infected cell lines compared with HTLV-1–uninfected cell lines (Fig. 2B; Supplementary Fig. S2). These results demonstrate that the CASP8L-type splicing pattern is dominant compared with the WT-CASP8-type splicing pattern in HTLV-1–infected cell lines, especially in ATL-derived cell lines, such as TL-Om1 and ATN1.

All identified CASP8 transcript variants bear PTC, thus encode mutant Casp8 isoforms without the C-terminal catalytic domain (Fig. 3A). Casp8L (270 aa) and -ΔE7 (235 aa) consist of only two DEDs, and Casp8-ΔE4 (108 aa) retains of only the first DED (Fig. 3A; Supplementary Fig. S1). It has been reported that the two DEDs of Casp8 are enough to activate the NFκB pathway (13). Therefore, we examined the influences of these Casp8 isoforms on activation of NFκB. Figure 3B shows that Casp8L and Casp8-ΔE7 significantly activated NFκB, whereas WT-Casp8 and Casp8-ΔE4 did not compare with the mock control. Moreover, Fas stimulation by Fas-Ab (CH-11) increased the magnitude of NFκB activation by WT-Casp8, Casp8L, and Casp8-ΔE7, indicating that the stimulation through Fas was converted to NFκB activation (Fig. 3C).

The Casp8L, Casp8-ΔE4, and Casp8-ΔE7 are expected to function dominant-negatively because they have one or two intact DED(s) and interrupt the formation of WT-Casp8 homodimer through DEDs, which is essential to trigger apoptosis. Thus, we examined the influences of mutant Casp8 isoforms on apoptosis inhibition. Figure 4A shows that the population size of AnnexinV(+) cells is significantly increased in WT-Casp8–expressing cells, whereas it was not significantly increased in Casp8L, Casp8-ΔE4, or Casp8-ΔE7–expressing cells compared with mock (EGFP expressing) cells in HEK293FT cells. In Jurkat cells, Casp8L, Casp8-ΔE4, or Casp8-ΔE7 still induced apoptosis, but to less extent compared with WT-Casp8 (Fig. 4B). In addition, Supplementary Fig. S3A shows that the population size of late-apoptotic cells [AnnexinV(+)/7AAD(+)] is also significantly increased only in the cells overexpressing WT-Casp8 compared with the mock control cells in HEK293FT cells (Supplementary Fig. S3A). In Jurkat cells, the same tendency as HEK293FT was observed except for Casp8-ΔE4, which somehow induced the late-apoptotic cells to the same extent as WT-Casp8 (Supplementary Fig. S3B). These data indicate that Casp8L, Casp8-ΔE4, and Casp8-ΔE7 show lower proapoptotic activities compared with WT-Casp8. The EGFP-tag does not obstruct the NFκB activation by Casp8L and Casp8-ΔE7 (Fig. 4C).

Subcellular localization of mCherry-tagged Casp8 isoforms revealed that apoptosis-defective/NFκB-activating isoforms, Casp8L and Casp8-ΔE7, form condensed structures or so-called “death-effecter filaments (DEF)” in HEK293FT cells and in Jurkat cells (Fig. 5A; Supplementary Fig. S4). We examined the implication of DEF formation and NFκB activation by cooverexpression of EGFP-RelA and mCherry-Casp8 isoforms in HEK293FT cells and in Jurkat cells. Activation of NFκB was detected as the nuclear localization of RelA. HEK293FT cells demonstrate a significant increase of the nuclear localization rate of RelA in the cells overexpressing mCherry-Casp8L or mCherry-Casp8-ΔE7, which show condensed or filament-like structures, respectively (Fig. 5B and C; Supplementary Figs. S4 and S5). On the other hand, inactive cytoplasmic localizations of RelA were observed in the cells with mCherry-WT-casp8 or mCherry-Casp8-ΔE4, which showed uniform subcellular distributions without filament-like structures (Fig. 5B; Supplementary Fig. S5). Similar results were observed in Jurkat cells (Supplementary Fig. S6).

Upon confirming the positive correlation between DEF formation of Casp8 isoforms and NFκB activation, we further tested whether NFκB activation is DEF-dependent. Dickens and colleagues (22) and Fu and colleagues (23), demonstrated elegantly in protein structural studies that F122 and L123 of Casp8 are critical for the filament formation. According to their reports, we constructed expression plasmids of double-mutated (F122G and L123G) WT-Casp8, Casp8L, and Casp8-ΔE7. Figure 6A and Supplementary Fig. S7 show that the double mutations of F122G and L123G in Casp8L and Casp8-ΔE7 successfully disrupt DEF formations of these isoforms in agreement with the previous reports (22, 23). Figure 6B demonstrates that the DEF disruption significantly suppresses NFκB reporter activities in Casp8L and Casp8-ΔE7. Unexpectedly, F122G/L123G mutagenesis in WT-Casp8 enhanced formation of condensed subcellular structures, together with a slight, but significant increase of NFκB reporter activity (Fig. 6A and B). Moreover, the EGFP-RelA was localized in nuclei of the cells with Casp8L and Casp8-ΔE7, whereas it was localized in cytoplasm of the cells with the DEF-disrupted Casp8L (F122G/L123G) or Casp8-ΔE7 (F122G/L123G; Fig. 6C). Consequently, these results support a notion that formations of condensed structures or DEFs by Casp8L and Casp8-ΔE7, respectively, enhance the nuclear localization of RelA, thus NFκB activation.

Finally, the FRET analysis between Casp8 and IKKγ (NEMO) demonstrates that Casp8L and Casp8-ΔE7 showed significantly higher affinities to IKKγ than WT-Casp8 and Casp8-ΔE4 (Fig. 7A). Moreover, such high affinity between IKKγ and Casp8L or Casp8-ΔE7 was reduced to the mock control level by DEF disruption (Fig. 7A). These data indicate that Casp8L and Casp8-ΔE7 physically interact with IKKγ, which may enhance the formation of IKK complex, thus NFκB activity.

Figure 7B demonstrates subcellular localization of EGFP-WT-Casp8, when coexpressed with mCherry-Casp8 isoforms. The results showed that the pattern of subcellular localization of WT-Casp8 is remarkably influenced by that of coexpressed isoforms. WT-Casp8 alone showed the uniform distribution pattern, whereas it colocalizes with other Casp8 isoforms. The FRET analysis between EGFP-WT-Casp8 and mCherry-Casp8 isoforms demonstrated that Casp8L, Casp8-ΔE4, and Casp8-ΔE7 indeed bind to WT-Casp8 (Fig. 7B, right). Importantly, the binding affinities were higher between WT-Casp8 and other isoforms compared with that of two WT-Casp8s. Therefore, it is highly likely that these mutant Casp8 isoforms function dominant-negatively by binding to WT-Casp8 via DED(s), thus inhibiting the formation of homodimer of WT-Casp8s, which is necessary for the induction of apoptosis.

In this study, we demonstrate for the first time that a wide variety of CASP8 transcript variants are expressed in PBMCs from patients with ATL. Overexpression of CASP8L mRNA was reported in patients with ATL (17, 25). In addition to CASP8L, we identified CASP8 mRNAs without the exon4 (CASP8-ΔE4) or the exon7 (CASP8-ΔE7; Fig. 1A). We focused on the CASP8L, CASP8-ΔE4, and CASP8-ΔE7 mRNAs, because all of them encoded Casp8 isoforms without the C-terminal caspase domains; thus, their functions were expected to be similar to c-FLIP, the Casp8 inhibitor, with a pseudo-caspase domain (Fig. 3A).

Evaluation of CASP8-splicing variants by quantitative-PCR with variant-specific primers in PBMCs from HTLV-1 carriers and patients with ATL demonstrated that abnormal CASP8-splicing variant levels are significantly increased in HTLV-1-carriers and patients with ATL (Fig. 1B). Importantly, the levels of these aberrant CASP8 transcript variants were increased in HTLV-1 infection and in an ATL progression–associated manner, thus these splicing variants can be sensitive biomarkers for HTLV-1 infection to T cells and malignant transformation/progression of HTLV-1–infected T cells.

Alternative-splicing sometimes yields aberrant transcripts containing nonsense codons (26), which are selectively eliminated via the nonsense-mediated mRNA decay (NMD) pathway. All CASP8L, CASP8-ΔE4, and CASP8-ΔE7 mRNAs contain nonsense codons; thus, they are expected to be degraded by NMD. Previously, we reported that the NMD activity was significantly suppressed in HTLV-1 infected cells by the function of the viral mRNA-binding protein, Rex, and also in ATL patient-derived T-cell line, TL-Om1 (27). Such malfunctions in NMD in HTLV-1–infected cells and in ATL cells may enhance stabilization of aberrant PTC-containing CASP8 transcript variants in ATL cells.

We have also reported aberrant splicing patterns in MYB mRNA (2) and HELIOS mRNA (6) in ATL cells. These data let us speculate that the global splicing machinery might be deregulated in ATL cells. Indeed, Fig. 2A demonstrates that the level of CASP8L mRNA is higher in ATL patient–derived cell lines compared with HTLV-1–uninfected cells. Also, the CASP8 splicing reporter analysis showed that CASP8L-type splicing events were more dominant in HTLV-1–related T-cell lines, compared with HTLV-1 uninfected T-cell lines (Fig. 2B; Supplementary Fig. S2). These results propose a possibility that HTLV-1 infection to T cells, as well as their malignant transformation to ATL cells, may cause the dysregulation of the global splicing machinery.

CASP8 transcription variants in ATL cells encode C-terminus truncated mutant Casp8 isoforms. The CASP8L and CASP8-ΔE7 mRNAs encode Casp8 isoforms only with two DEDs and CASP8-ΔE4 mRNA encodes a Casp8 isoform only with one DED (Fig. 3A). The cFLIP, a well-known Casp8 inhibitor, also consists of two DEDs and inactive caspase-like domain (8, 9). cFLIP inhibits Casp8 function by forming a heterodimer with Casp8 via DEDs and activates NFκB also through DEDs (14). Because Casp8L and Casp8-ΔE7 only contain two DEDs, we speculated that these Casp8 isoforms might be functionally similar to cFLIP (Fig. 3A). As expected, Casp8L and Casp8-ΔE7 with two DEDs significantly activated NFκB, whereas Casp8-ΔE4 with one DED did not (Fig. 3B), which were in agreement with previous studies using recombinant Casp8 isoforms with one or two DED(s; refs. 15, 16). Moreover, Fas-antibody (CH-11) treatments, mimicking FAS-L stimulations, induced even higher NFκB activations in cells overexpressing WT-Casp8, Casp8L, or Casp8-ΔE7, indicating that Fas signaling was transduced to NFκB activation via these Casp8 isoforms (Fig. 3C).

Figure 4A demonstrates that the AnnexinV(+) early apoptotic cell population is significantly increased in WT-Casp8–overexpressing HEK293FT cells compared with mock-transfected cells, as expected. On the contrary, the apoptotic cell population was not significantly changed in the cells overexpressing Casp8L, Casp8-ΔE4, and Casp8-ΔE7 in HEK293FT cells. In Jurkat cells, the AnnexinV(+) cell population was smaller in Casp8L, Casp8-ΔE4, or Casp8-ΔE7 expressing cells compared with WT-Casp8 expressing cells, although these Casp8 isoforms still induced apoptosis (Fig. 4B). The size of late-apoptotic cell population [AnnexinV(+)/7AAD(+)] was reduced in Casp8L, Casp8-ΔE4, or Casp8-ΔE7–expressing cells compared with WT-Casp8–expressing cells in HEK293FT (Supplementary Fig. S3A). A similar result as HEK293FT was observed in Jurkat cells, except for Casp8-ΔE4, which induced the late-apoptosis to the same extent as WT-Casp8 (Supplementary Fig. S3B). These data suggest that the Casp8 mutants have reduced activities to induce apoptosis compared with WT-Casp8. Homodimerization of functional WT-Casp8 is required to induce apoptosis. Those short aberrant isoforms may bind to WT-Casp8 through their DED(s), thus acting as an obstacle in the formation of the homodimer of WT-Casp8s.

EGFP- or mCherry-tagged Casp8 isoforms revealed unique subcellular localization and structure of each isoform in HEK293FT cells and Jurkat cells (Figs. 4A and 5A; Supplementary Fig. S4). WT-Casp8 and Casp8-ΔE4 showed uniform subcellular distributions, whereas Casp8L and Casp8-ΔE7 formed condensed- or filament-like structures in cytoplasm. Especially, Casp8-ΔE7 showed extensive elongations of filament-like structures in HEK293FT cells and in Jurkat cells (Fig. 5A; Supplementary Fig. S4). The cytoplasmic structure formed by interactions among DEDs, namely DEFs, was already discovered in 1990s (28). The critical role of filament formation in Casp8 activity and the detailed structural mechanism for the formation of the DEF has not been fully uncovered until recently (22, 23, 29). The death-inducing signaling complex (DISC) formation is critical to initiate the extrinsic apoptotic pathway. DISC is composed of Fas, FADD, and pro-Casp8 or pro-Casp10. Until recently, the primary component of DISC comprises of a ternary complex of Fas, FADD, and pro-caspase. Casp8/10 is expected to be dimerized with proximal Casp8/10 within the DISC assembly for induction of autocleavage of each catalytic domain, which triggers the downstream apoptotic pathway. According to this model, the DISC contains Fas, FADD, and Casp8 in 1:1:1 ratio. However, Dickens and colleagues (22) reported that DISC forms a higher-order assembly of more than 700 kDa, and FADD is substoichiometric compared with Casp8, indicating that DISC contains a more Casp8 molecules than FADD molecules. On the basis of their results, the authors proposed a new model, that is, “Casp8 DED chain formation in DISC” model, in which, Fas/FADD complex (in 1:1) binds to Casp8 molecules forming a chain-like structure. The authors elegantly proved that hypothesis by structural analysis. Importantly, they also demonstrated that the DED chain–forming activity of Casp8 is critical for its function. Most recently, Fu and colleagues (23) revealed more detailed and realistic procedure of Casp8 tandem-DED-filament formation by using the cryo-EM system. These structural studies were conducted with artificially designed recombinant Casp8-DEDs. Our current study showed that naturally occurring Casp8L and Casp8-ΔE7 also formed DEFs. It is also noteworthy that the full-length Casp8 does not form such filament-like structures, because the catalytic subunit of the full-length Casp8 masks the DED-interacting site; thus, extensive filamentation may be inhibited (22). Indeed, the WT-Casp8 showed uniform subcellular distribution in this study as well (Figs. 4A and 5A).

Because Casp8L and Casp8-ΔE7 activated NFκB by nuclear translocation of RelA in HEK293FT cells and in Jurkat cells (Fig. 5B and C; Supplementary Figs. S5 and S6), we further investigated the implication of the DEF formation to NFκB activation (Fig. 6). It was reported that a filament-like structure of Bcl10 functioned as a scaffold to enhance formation of the IKK complex, to activate NFκB (30) and probably for signal amplification (31). The direct relationship between the DEF formation by a mutant Casp8 isoform and NFκB activation has not been tested. Staal and colleagues (32) mentioned in their review that Casp8-DEDs might function as scaffolding platforms for component proteins of the NFκB pathway. In agreement with their hypothesis, we demonstrated that F122G/L123G mutagenesis in Casp8L and Casp8-ΔE7 completely disrupted the filament-like structures (Fig. 6A; Supplementary Fig. S7) and significantly reduced NFκB activities (Fig. 6B). EGFP-RelA showed the cytoplasmic (inactive) distribution in the cells with Casp8L (F122G/L123G) and Casp8-ΔE7 (F122G/L123G; Fig. 6C). Thereby, the impairments of DEF formations in Casp8L and Casp8-ΔE7 resulted in significant reductions in their abilities to activate NFκB. Unexpectedly, WT-Casp8 (F122G/L123G) showed speckle-like condensed structures, together with a significantly elevated NFκB activity (Fig. 6A and B). These results support the notion that the formation of condensed structures or DEFs by Casp8L or Casp8-ΔE7, respectively, has a strong correlation with NFκB activation. Indeed, the FRET analysis between Casp8 and IKKγ (NEMO) demonstrated that Casp8L and Casp8-ΔE7 showed strong physical interactions with IKKγ in the cells, which were reduced to the negative-control level in DEF-disrupted Casp8L (F122G/L123G) and DEF-disrupted Casp8-ΔE7 (F122G/L123G; Fig. 7A). These data indicate that Casp8L and Casp8-ΔE7 enhance the NFκB activity probably by physical assistance in effective assembly of the IKK complex.

Finally, we examined the interaction between WT-Casp8 and mutant Casp8 isoforms in the cell (Fig. 7B). The EGFP-tagged WT-Casp8 alone showed a typical uniformed subcellular distribution, whereas it colocalized with mCherry-Casp8 isoforms (The left of Fig. 7B). These data indicate that Casp8 isoforms bind to WT-Casp8, probably through DED(s) in the cell. The FRET analyses between WT-Casp8 and mutant-Casp8 isoforms confirmed that these Casp8 isoforms indeed physically interacted with WT-Casp8 (Fig. 7B, right). It is noteworthy that the binding affinity between WT-Casp8 and a mutant Casp8 isoform of all types was higher than that between two WT-Casp8s. Thus, it is highly likely that these mutant Casp8 isoforms function dominant-negatively by preventing the homodimerization of WT-Casp8, which is essential for induction of apoptosis.

In summary, we demonstrated here that aberrant CASP8 transcript variants were overexpressed in primary PBMCs of HTLV-1 carriers and patients with ATL. These aberrant CASP8 transcripts encode mutant Casp8 isoforms with intact DED(s), but without the caspase domain. Through characterization and functional analysis of these mutant Casp8 isoforms in HEK293FT and Jurkat cells, we showed that Casp8L and Casp8-ΔE7, which have only two DEDs, formed condensed structures or extensively elongated filament-like structures (DEFs). DEF-destructed Casp8L or Casp8-ΔE7 showed decreased NFκB activity together with decreased interaction with IKKγ, compared with native Casp8L or Casp8-ΔE7, respectively. Those results let us speculate that these structures may function as scaffolds for IKK complex formation, which enhances NFκB activation. Also, these abnormal Casp8 isoforms, including Casp8-ΔE4, are highly likely to act dominant-negatively by interrupting homodimerization of WT-Casp8s, which is essential for induction of apoptosis. In view of increasing levels of these abnormal CASP8 transcripts in PBMCs of HTLV-1 carriers and patients with ATL, we propose a possibility that the resultant abnormal Casp8 isoforms may cause imbalanced apoptosis in these primary cells. Further examinations are essential to evaluate the effects of these Casp8 isoforms in the HTLV-1–infected T cells and ATL cells in HTLV-1 carriers and patients with ATL, respectively.

No potential conflicts of interest were disclosed.

Conception and design: K. Nakano, T. Watanabe

Development of methodology: K. Nakano

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Nakano, M. Iwanaga

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Nakano, A. Utsunomiya, T. Watanabe

Writing, review, and/or revision of the manuscript: K. Nakano, A. Utsunomiya, K. Uchimaru, T. Watanabe

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Utsunomiya

Study supervision: K. Nakano, K. Uchimaru, T. Watanabe

This work was supported by grants-in-aid for scientific research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan [to K. Nakano (15K06827); to T. Watanabe (16H06277); and to M. Iwanaga and T. Watanabe (16H05248)], and from Japan Agency for Medical Research and Development (to T. Watanabe; 18fk0108039h0002).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.