KSHV vFLIP Is Essential for the Survival of Infected Lymphoma Cells

Activation of nuclear factor (NF)–κB is a mechanism used by lymphomagenic viruses to prolong cellular survival and induce proliferation. EBV and human T cell leukemia virus 1 (HTLV-1) encode viral (v) oncoproteins, mainly LMP1 and Tax, that subvert cellular signaling pathways to constitutively activate NF-κB (1–4). Another lymphomagenic virus, Kaposi's sarcoma-associated herpesvirus (KSHV, also called human herpesvirus 8 or HHV-8), infects B cells and is present in B cell lymphoproliferative disorders, specifically primary effusion lymphomas (PELs; reference 5) and a subset of cases of multicentric Castleman's disease (6) and associated plasmablastic lymphomas (7). Like other virus-associated lymphomas, the transcription factor NF-κB is constitutively activated in PEL cells, and is essential for their survival (8). We postulated that expression of viral genes was a primary source of constitutive activation of NF-κB in PEL cells. Consistent with this has been our previous finding that types of NF-κB complexes, although varying in levels of intensity, are identical among different KSHV-infected PEL cell lines, which suggests a common source of activation.

Apoptotic cell death is a tightly regulated process that is crucial for the defense against certain infectious diseases and cancer. Interaction between immune effector cells as well as some infected cells can lead to apoptosis by the exogenous or death receptor pathway. Death receptors are a family of transmembrane proteins that belong to the TNF family of receptors (9). They all contain a cytoplasmic sequence called death domain (DD) that couples each receptor to caspases, inducing death signals that trigger apoptosis. Current understanding suggests that binding of extracellular death ligands to their receptors, such as the Fas ligand to Fas, results in conformational changes of preformed receptor clusters (10). This leads to recruitment of the bipartite molecule Fas-associated (FA)DD, composed of an amino terminal death effector domain (DED) and a carboxyl terminal DD. FADD binds to Fas via homophilic DD–DD interactions and recruits the upstream DED-containing procaspase 8 to the receptor by homophilic DED–DED interactions. This forms the death-inducing signaling complex (DISC), which autoactivates itself via proteolytic cleavage of caspase 8 and induces apoptosis by subsequent cleavage of downstream effector caspases.

Apoptosis is tightly regulated and can be inhibited at the receptor level (receptor endocytosis, soluble ligands, and decoy receptors) during signal transduction and at the effector stage (caspase inhibitors such as CrmA, p35, IAPs, BCL-2, etc.). The transcription factor NF-κB can inhibit apoptosis by controlling the expression of some of these antiapoptotic inhibitory proteins. A different and direct mechanism of apoptosis inhibition is mediated by a group of inhibitors of death receptor–induced apoptosis, called FLIP (FADD-like interleukin-1-β–converting enzyme [FLICE/caspase 8]-inhibitory protein; reference 11), but also having received the names of FLAME-1, I-FLICE, Casper, CASH, MRIT, CLARP, and Usurpin. FLIP proteins contain two DED domains and were shown to inhibit DED–DED interactions between FADD and procaspases 8 and 10. Two forms of cellular FLIP have been identified and characterized, both of which can inhibit caspase 8 activation at the DISC. These are cFLIP_L (long form) and cFLIP_S (short form), which correspond to splice variants. cFLIP_S consists of two DEDs with a short cytoplasmic tail, whereas cFLIP_L has an additional COOH-terminal caspase domain and resembles procaspase 8 in its overall structural organization, but lacks proteolytic activity (12). Only cFLIP_L has been found in human B cells by us (unpublished data) and others (13) when using antibodies that recognize both protein forms. Several viruses contain proteins with two DED with slight homology to cFLIP. The first described and most studied are MC159L from the Molluscum Contagiousum virus, E8 from the Equine Herpes Virus 2, and ORF71/K13 from KSHV (11, 14). Rhesus rhadinovirus and herpesvirus saimiri also contain FLIP homologues, but murine herpesvirus 68 does not. KSHV vFLIP has two DED domains, and structurally most closely resembles cFLIP_S. When transfected into a murine B cell lymphoma cell line, which was subsequently injected into immunocompetent mice, it facilitated tumor growth, suggesting that in confers cells with resistance to T cell–induced apoptosis (15). KSHV vFLIP has also been shown to be able to activate the IκB kinase (IKK) complex and NF-κB in a variety of ectopic overexpression systems, and to be present in IKK complexes in PEL cells (14, 16, 17). In addition, a recent paper by An et al. showed that vFLIP is responsible for IL-6 expression in a PEL cell line, where suppression of vFLIP resulted in a 100% reduction of AP-1 transcriptional activity and 70% reduction of NF-κB activity, providing the first direct evidence indicating that vFLIP is responsible for a significant proportion of the constitutive NF-κB activity observed in PEL cells (18).

In addition to KSHV vFLIP, three different viral transmembrane proteins have been shown to induce signaling that leads to NF-κB activation in ectopic expression systems. These are K1 (19), viral G protein–coupled receptor (vGPCR; reference 20), and K15 (21). We sought to confirm previous findings suggesting that vFLIP is responsible for initiating and maintaining the signaling cascade responsible for NF-κB in latently infected lymphoma cells, and to exclude a possible contributory role for K1, vGPCR, and K15 during latent infection. Our studies show that of these genes, vFLIP is largely responsible for the constitutive NF-κB activity in PEL cells. In addition, we show that KSHV vFLIP expression is essential for the survival of the infected lymphoma cells.

To clone KSHV vFLIP, viral Cyclin (vCYC), and cellular FLIPs (long and short forms), cDNA from the human PEL cell line BC3 was amplified using the following specific primers: vFLIP, 5′-GGGATGGCCACTTACGGGGTT-3′ and 5′-GGTGTATGGCGATAGTGTTGG-3′; vCYC, 5′-GGGATGGCAACTGCCAATAAC-3′ and 5′-ATAGCTGTCCAGAATGCGCGG-3′; cFLIP_L, 5′-GGGATGGCTGCTGAAGTCATCCAT-3′ and 5′-TATGTGTAGGA-GAGGATAAGTTTC-3′; and cFLIP_S, 5′-GGGATGGCTGC-TGAAGTCATCCAT-3′ and 5′-ACATGGAACAATTTCCA-AGGATTT-3′.

All the constructs were cloned into the eukaryotic expression vector pcDNA3.1 (Invitrogen). Correct orientation and sequence integrity was confirmed by nucleotide sequencing. vFLIP and cFLIP_L were also cloned in frame with a COOH-terminal 3X-FLAG in a CMV14 vector backbone (Sigma-Aldrich). Comparable expression of transfected plasmids was confirmed by RT-PCR or immunoblot with an anti-FLAG monoclonal antibody. vFLIP was also cloned into pTRUF2-Tet (22) by blunt end ligation into the single NsiI site, and the resulting plasmid called pTRUF2-tet/vFLIP.

Cell lines used were as follows: PEL cell lines were BC1, BC2, BC5 (KSHV⁺/EBV⁺), BC3, and BCBL-1 (KSHV⁺/EBV⁻). All of these were established in our laboratory from lymphomatous effusions as described previously (23, 24), except BCBL-1 (25) that was obtained from the AIDS and Cancer Specimen Bank. Burkitt lymphoma cell lines were Namalwa and BJAB, obtained from the American Type Culture Collection. An EBV-positive immunoblastic lymphoma (IBL) cell line was derived in our laboratory from an HIV-positive patient. A lymphoblastoid cell line (LCL9001) was obtained by infection of peripheral blood lymphocytes with EBV. Cells were grown in RPMI 1640 (GIBCO BRL) supplemented with 10% heat-inactivated FBS (Atlanta Biologicals) and 50 mg/ml gentamicin (Sigma-Aldrich). The human embryonal kidney 293T cell line was grown in DMEM supplemented with 10% FBS. Cells were cultured at 37°C in a humidified atmosphere with 5% CO₂. NF13 was established by electroporation of Namalwa cells with 10 μg of pTRUF2-tet/vFLIP. The cells were allowed to recover for 48 h in full culture medium, and 1 mg/ml of Geneticin (Invitrogen) was added. After 20 d under antibiotic selection, single cells approximated by limiting dilution were placed on irradiated fibroblast feeder layers (10⁴ to 2 × 10⁴ cells/well; 5,000 rad of cobalt γ-irradiation) in 96-well plates. After 2 wk, only wells with a single colony were expanded. 76 clones were screened for tetracycline inducibility-related protein expression by transfection with 10 μg of pTRE-Luc (CLONTECH Laboratories, Inc.), containing a tetracycline response element, and ± 2 μg/ml doxycycline (Doxy). Three clonal populations were selected having low basal transcription and up to 100-fold inducibility with the TRE-luciferase reporter and an NF-κB luc reporter, as well as by RT-PCR and immunoblot analysis for vFLIP. One of these, NF-13, was chosen to evaluate cellular gene expression.

Nuclear proteins were evaluated for binding to a ³²P-labeled oligonucleotide probe corresponding to the Igκ NF-binding sequence (a gift from H.-C. Liou, Weill Medical College of Cornell University) or Oct-1 (Promega) by EMSA. The methodology used was described previously in detail (8).

Total RNA was extracted with the RNAZOL solution (GIBCO BRL) according to the manufacturer's instructions. Poly(A)-selected RNA was isolated with the PolyATract mRNA Isolation System (Promega). RNA samples were treated with 2 U of RNase-free DNase I (Boehringer) before Northern blot analysis. 800 ng aliquots of poly(A)-selected RNA were separated electrophoretically in 1.5% denaturing formaldehyde agarose gel and blotted onto nylon filters (Dupont). After fixation, the blots were subjected to hybridization to PCR-generated double-stranded DNA probes that had been ³²P-labeled by the random primer extension method. To evaluate K15 expression, probes for the predominant (P) and minor (M) forms were used. Washing of the filters and autoradiography was performed according to standard methodology.

BJAB and Namalwa cells (5 × 10⁶) were transfected by electroporation (Gene Pulser II; Bio-Rad Laboratories) at settings of 270 mV and 975 μF in 0.8-cm cuvettes (Invitrogen). The amount of DNA transfected was equalized by addition of pcDNA3.1 control empty vector. All the transfections were performed in the presence of pRSV-RL vector (Promega), encoding renilla luciferase, to normalize the results for the transfection efficiencies. After 48 h, lysates were prepared using 1× cell culture lysis reagent as specified by the manufacturer (Promega). Luciferase assays were performed with a luminometer (Dynex Tech). The NF-κB firefly luciferase reporter derived from a Igκ promoter was provided by H.-C. Liou. Each experiment was independently performed a minimum of three times, and within each experiment each transfection was performed in triplicate.

Transient transfection of vCYC was performed using FuGene 6 transfection reagent (Roche). This method is highly efficient for transfection of BC-3 cells, achieving delivery into up to 90% of cells as determined by a control green fluorescent protein–expressing plasmid. BC3 cells were plated in RPMI 1640 without FBS at a density of 7.5 × 10⁵ cells/ml (2 ml in a six-well plate). 3 μl of FuGene 6 reagent were added in a small sterile tube containing serum-free medium to a total of 100 μl. Subsequently, 2 μg of pcDNA3.1/vCYC were added to the prediluted FuGene 6 and mixed gently. After 15 min of incubation, the mixture was added to the cells. This method was also used to transfect 3× FLAG-tagged vFLIP and cFLIP_L into BC-3 cells.

Transient transfection experiments were performed in 293T cells when the confluence of the cells was 30–50% using the Calcium Phosphate Transcription System (GIBCO BRL).

RNA duplexes were synthesized with 2-nt (2′deoxy) uridine overhangs by Dharmacon Research. The target region of vFLIP small interfering (si) RNA was 58 amino acids downstream of the start codon. The siRNA target sequence to vFLIP mRNA is 5′-AAGUGGUAUUGUUCCUCCUAA-3′. A control scramble siRNA duplex was obtained from Dharmacon (Scramble II Duplex). We performed the transfection of siRNA duplex using Oligofectamine Reagent (Invitrogen) and assayed for silencing 2 d after transfection. With this method, we achieved delivery into 60–80% of the cells in each of the lymphoma cell lines tested, as determined by transfection with a fluoresceinated control siRNA and flow cytometric analysis (unpublished data). For extended studies, the cells were transfected every 2 d, for a total of four transfections and an 8-d time course. Cells were plated in 12-well plates at a density of 7.5 × 10⁵ cells/ml. For each well, we used 1.68 μg siRNA duplex. We mixed 6 μl of 20 μM siRNA duplex with 100 μl of OPTI-MEM (Invitrogen). In another tube, 6 μl of Oligofectamine Reagent was mixed with 24 μl of Opti-MEM, incubated for 10 min at room temperature. These solutions were combined, mixed, and incubated for another 25 min at room temperature. 76 μl of fresh Opti-MEM were added to the tubes to obtain a final solution volume of 212 μl, which was added to cultured cells. Each RNA interference experiment was performed at least three times, and reporter assays were performed in triplicate within each experiment.

NF13 cells were plated at a density of 7.5 × 10⁵ cells/ml without Doxy (Sigma-Aldrich) or with 2 μg/ml Doxy for 48 h. Namalwa cells were also evaluated with and without 2 μg/ml Doxy as a control for drug effects. BC-3 cells that were untransfected, mock-transfected, or transfected with siRNA were collected 48 h after transfection. Lysates were prepared in standard radioimmunoprecipitation assay buffer supplemented with 1 μg/ml each of aproptinin, leupeptinin, and pepstatin; 0.5 mM phenylmethylsulfonyl fluoride; and 1 mM each NAV04 and NaF (Sigma-Aldrich). Lysates were prepared from three independent experiments and subsequently evaluated by Western blotting. The proteins were quantitated by the Bradford method, and 80 μg whole cell extract was loaded onto 7.5–12.5% SDS-polyacrylamide gels. After electrophoresis, semi-dry transfer to a polyvinylindene difluoride membrane (Millipore) was performed. Blots were probed with primary Ab overnight at 4°C. Horseradish peroxidase–conjugated secondary Ab was added after washing and was detected by an enhanced chemiluminescence system (Amersham Biosciences).

The following primary Abs were used: anti-vFLIP (provided by M. Collins, University College of London, London, UK; reference 26); anti-cFLIP, anti–cIAP-1, and cIAP-2 (Qbiogene); anti–Bcl-X, anti–Bcl-2, and anti–IL-6 (Santa Cruz Biotechnology, Inc.); anti-Caspase 3 (Upstate Biotechnology); anti-poly (ADP-ribose) polymerase (PARP) (BD Biosciences); anti–latency-associated nuclear antigen (LANA; ABI); anti-vCYC (Exalpha); and anti–β-actin and anti-FLAG (Sigma-Aldrich).

Cells were placed in culture at a density of 7.5 × 10⁵ cells/ml without anti-FAS Ab or with 800 ng/ml anti–human Fas ligand (B-R17; Upstate Biotechnology) for 24 h. Annexin V analysis was performed as described by the manufacturer (CLONTECH Laboratories, Inc.). In brief, 2 × 10⁵ cells were washed once with 1× binding buffer and resuspended in 200 μl of 1× binding buffer. 5 μl FITC–Annexin V was added, and the cells incubated for 15 min in the dark at room temperature and analyzed by flow cytometry using a FACSCalibur™ (Becton Dickinson).

We first sought to determine which of the KSHV-encoded genes that can activate NF-κB in ectopic expression systems is responsible for the constitutive NF-κB activity seen in PEL cells. We performed Northern blot analysis using probes for vGPCR, K1, K15 (not depicted), and vFLIP (Fig. 1), and compared the amount of transcripts in latently infected PEL cells (BC1, BC2, and BC3) to the level of NF-κB as measured by EMSAs in these three PEL cell lines. This analysis confirmed previous analyses (27, 28) indicating that K1, vGPCR, and K15 are lytic genes that are not significantly expressed by latently infected PEL cells, whereas vFLIP is expressed at significant levels during latency. In addition, we found that the transcript encoding vFLIP is present in PEL cells at levels that correlate with the degree of NF-κB DNA-binding activity (Fig. 1). Because vFLIP is encoded in a bicistronic mRNA together with vCYC, the levels of vCYC also parallel those of NF-κB. However, in our hands, vCYC cannot activate NF-κB (see Fig. 5 B). The protein levels of vFLIP also paralleled those of NF-κB, as indicated by an immunoblot analysis using a monoclonal antibody to vFLIP (Fig. 1). Although this simple correlation is insufficient to draw firm conclusions, it prompted us to further investigate the role of vFLIP in PEL.

Activation of NF-κB by vFLIP was evaluated in Namalwa cells by transient transfection assay, using a pcDNA3.1/vFLIP construct and an NF-κB firefly luciferase reporter derived from the Igκ promoter. We chose Namalwa because this is a KSHV-negative B cell lymphoma line with relatively low endogenous NF-κB, and therefore, corresponds to the same lineage as PEL, but does not contain endogenous vFLIP. The luciferase activity was measured 48 h after transfection and it was found to be markedly increased in the cells transfected with vFLIP as compared with the control cells transfected with NF-κB firefly luciferase reporter and empty vector (Fig. 2 A). This is the first demonstration that vFLIP can indeed activate NF-κB in human B cells. Notably, vFLIP was a more potent activator of NF-κB than saturating doses of TNFα. Cellular FLIP (the short and long isoforms, cFLIP_S and cFLIP_L) were also evaluated for comparison. These were cloned into the same vector (pcDNA3.1) and transiently cotransfected in Namalwa cells with the NF-κB luciferase reporter. vFLIP was a more potent activator of NF-κB than both forms of cFLIP (Fig. 2 B). When tested in 293T cells, vFLIP was also a more potent activator of NF-κB than TNFα, and the ability of vFLIP to induce NF-κB was at least 2-fold higher as compared with cFLIP_L and 10-fold higher than cFLIP_S (unpublished data). To assess the levels of the different FLIP proteins in the transfected cells using a single antibody, and to determine whether the differences in NF-κB activity may be explained by variability protein levels, we subcloned vFLIP and cFLIP_L into a vector in frame with a COOH-terminal 3X-FLAG. This also allowed us to examine the ability of these proteins to activate NF-κB when transfected into BC-3 cells and to distinguish them from the endogenous proteins. No significant differences in activity were found between FLAG-tagged and untagged proteins (unpublished data). The vFLIP-FLAG also showed higher activation of NF-κB than cFLIP_L-FLAG in BC-3 (Fig. 2 C) and 293T cells (not depicted). Proteins were extracted from these transfected cells, and immunoblot analysis using an anti-FLAG antibody was performed. We consistently found in BC-3 and 293T cells that the levels of cFLIP_L were approximately threefold higher than those of vFLIP. This is probably due to higher translation efficiency or stability of the cFLIP_L protein because these differences were not seen by RT-PCR. These results show that vFLIP has an intrinsically higher signaling capacity than cFLIP_L, and that the difference between viral and cellular proteins is larger than that apparent in the reporter assays when protein levels are taken into account.

To better characterize the effect of KSHV vFLIP in neoplastic B cells, this gene was cloned in pTruf2-Tet, an inducible plasmid expression vector developed in our laboratory based on a combination of the tetracycline-inducible system and adeno-associated virus (22) and was stably transfected in Namalwa cells. Selected and subcloned populations were tested for expression and inducibility of vFLIP after treatment with Doxy. Three clonal populations were selected, having low basal transcription and up to 100-fold inducibility. Results from a representative clone, designated NF13, are shown in Fig. 3. NF-κB activity was increased greater than two-fold in induced cells as compared with the uninduced control cells. Because NF-κB is a transcription factor that controls a number of cellular genes that are critical mediators of cell survival and proliferation, we evaluated the effect of vFLIP on cellular gene expression. We have identified a set of cellular genes that are antiapoptotic NF-κB targets in PEL cells. These include cFLIP_L (no cFLIP short was identified in PEL cells), cIAP-1, and cIAP-2. We have also found that BCL-X_L and BCL-2 are not NF-κB targets in PEL cells (unpublished data). Expression of vFLIP induced by Doxy in NF13 cells resulted in expression of cFLIP_L, cIAP-1, and cIAP-2, but no changes in BCL-X_L (Fig. 3). No changes in NF-κB activity or expression of cFLIP_L were seen when the parental Namalwa cells were treated with Doxy. These data suggest that virus-encoded vFLIP could inhibit apoptosis by inducing expression of a specific set of cellular NF-κB–dependent antiapoptotic genes.

We evaluated the role of vFLIP in PEL cells, which are naturally infected with KSHV and presumably a target of transformation by this virus, using RNA interference. We developed and tested a siRNA duplex that specifically targets the KSHV vFLIP sequence, but not cFLIP or any other known genomic sequence. This siRNA was transfected into a PEL cell line (BC-3) and a significant inhibition of vFLIP protein was obtained, as assessed by immunoblot analysis (Fig. 4 A). Although all experiments showed marked silencing by immunoblot analysis, it appears to be complete in some, partial in others, and seen by longer gel exposure. This is probably due to a threshold of detection by the antibody to vFLIP, where a small amount of residual vFLIP protein can no longer be detected. siRNA to vFLIP was able to inhibit basal NF-κB activity by ∼80% in BC-3 cells, indicating that a substantial proportion of the constitutive NF-κB activity is being induced by vFLIP (Fig. 4 B). Scramble siRNA used as a negative control did not have a significant effect on NF-κB activity and did not affect the levels of vFLIP protein. This result confirms a previously published work in which siRNA to vFLIP was found to significantly inhibit NF-κB reporter activity in the BCBL-1 PEL cell line (18). We also confirmed that siRNA to vFLIP inhibits transcription for an AP-1 reporter construct, as demonstrated previously (reference 18 and unpublished data).

To confirm the effect of vFLIP siRNA on NF-κB activity, we performed EMSAs. We evaluated BC-1 and BC-3 cells transfected with siRNA to vFLIP, using mock-transfected cells and cells transfected with scramble siRNA as controls. Nuclear extracts were made 48 h after transfection and evaluated for binding to a radiolabeled probe containing an NF-κB binding sequence. Silencing of vFLIP resulted in a marked decrease of NF-κB DNA binding activity (Fig. 4 C). In contrast, no significant changes were seen in the level of proteins binding to an oligonucleotide containing an octamer binding motif, indicating that not all transcription factors are affected by vFLIP down-regulation.

vFLIP is encoded in a bicistronic transcript also encoding for vCYC. In addition, vFLIP is also present in a tricistronic transcript encoding the LANA and also containing downstream vCYC and vFLIP sequences. Therefore, it was important to determine whether protein production of LANA and/or vCYC is affected by the siRNA to vFLIP. Our results indicate that LANA is unaffected by siRNA to vFLIP. The reason for this protection is unclear; LANA remains unchanged even after 8 d of siRNA treatment, so it is unlikely to be due to a longer half life of the protein. In contrast, expression of vCYC is dramatically reduced (Fig. 5 A). To overcome the down-regulation of vCYC, and to be able to dissociate the effects resulting from loss of vFLIP from those due to vCYC inhibition, we transfected vCYC as a monocistronic transcript and expressed it independently of vFLIP in BC-3 cells. We were able to achieve continuous expression of vCYC at levels comparable to those of untreated BC-3 cells (Fig. 5 A). The functional effects of treatment with siRNA to vFLIP described in the next paragraphs were also seen in vCYC-transfected BC-3 cells, where vCYC protein levels were maintained in the presence of siRNA to vFLIP, and thus can be attributed to vFLIP down-regulation. In addition, transfection of vCYC did not induce NF-κB, and did not affect the NF-κB activity induced by vFLIP (Fig. 5 B); thus, we did not find vCYC to be a contributing factor to NF-κB activation.

We assumed that a single transfection of siRNA into PEL cells would lead to incomplete and transient delivery, so we modified this approach by transfecting BC-3 cells every 2 d with siRNA, thus achieving sufficient and prolonged inhibition of vFLIP in PEL cells. Proteins were extracted at various time points during the course of this experiment, and the effect of vFLIP reduction on selected cellular proteins was evaluated. siRNA to vFLIP decreased the expression of the three NF-κB–dependent cellular antiapoptotic proteins examined, cFLIP, cIAP-1, and cIAP-2 (Fig. 6), although down-regulation of cIAP-1 was not immediate, perhaps due to a long half life of this protein. Reconstitution of vCYC did not affect this down-regulation of cellular proteins, as shown for cFLIP_L in Fig. 5 A. siRNA to vFLIP also decreased expression of cellular IL-6, another NF-κB–dependent factor, consistent with recently published findings (18). siRNA to vFLIP did not affect the levels of cellular BCL-2 and BCL-X_L proteins, which are not dependent on NF-κB in PEL cells (unpublished data), indicative of pathway specificity. These results show that vFLIP is largely responsible for sustaining the expression of cellular NF-κB–dependent antiapoptotic genes in PEL cells.

If NF-κB activity is essential for the survival of PEL cells, and vFLIP is responsible for this activity in latently infected cells, elimination of vFLIP should lead to apoptosis. However, it has never been shown that elimination of a single viral gene results in apoptosis of infected lymphoma cells. To test this, we have measured apoptosis by flow cytometry for annexin V after inhibition of vFLIP by RNA interference. Initially, we evaluated BC-3 cells transfected with siRNA for apoptosis up to 48 h after transfection. Because no significant cell death was seen, vFLIP BC-3 cells were transfected every 2 d (days 0, 2, 4, and 6), and annexin V was evaluated at 2-d intervals (days 2, 4, 6, and 8, before the next transfection), for a total of four transfections and an 8-d time course. The number of apoptotic cells increases over time when transfected with siRNA to vFLIP, and by day 8, a significant proportion of cells (∼80%) express surface annexin V. In contrast, cells similarly transfected with nonspecific (scramble) siRNA and mock-transfected cells remain largely viable (Fig. 7 A, 18 and 13% apoptotic cells). Reconstitution of BC-3 cells with vCYC did not protect cells from apoptosis, indicating that apoptosis results from vFLIP and not vCYC suppression. siRNA to vFLIP also resulted in partial cleavage of caspase 3 and PARP, confirming induction of apoptosis (Fig. 7 B). Because the possibility remained that multiple transfections were inducing stress and sensitizing the cell to apoptosis in the context of NF-κB inhibition, we determined whether apoptosis would also occur in cells that had only been transfected once (at day 0), or twice (at days 0 and 4). A similar proportion of cells underwent apoptosis as in cells transfected four times (Fig. 7 A), but in all experiments it took 6–8 d for this process to be obvious.

To exclude the possibility that the vFLIP siRNA induces apoptosis through toxicity or other nonspecific mechanisms mimicking those seen upon inhibition of NF-κB, we tested additional cell lines. siRNA was transfected every 2 d for 8 d into four additional PEL cell lines, containing KSHV and expressing vFLIP (BC-1, BCBL-1, BC-2, and BC-5), and into four cell lines lacking KSHV, which included two Burkitt lymphoma cell lines with low constitutive NF-κB activity (BJAB and Namalwa), as well as one EBV-positive immunoblastic lymphoma cell line established in our laboratory (IBL) and one lymphoblastoid cell line (LCL9001), both with high constitutive NF-κB activity. The results shown in Fig. 8 clearly demonstrate that transfection of siRNA to vFLIP only leads to apoptosis of the PEL cell lines, whereas the cell lines lacking endogenous vFLIP are not significantly affected by this treatment.

cFLIP has been shown to inhibit apoptosis by virtue of containing two DEDs, and preventing signaling from death receptors to the upstream caspases (29, 30). vFLIP has also been shown to inhibit Fas-mediated apoptosis and caspase activation when transfected into a cell line that is sensitive to death induction by this pathway (15). We evaluated whether vFLIP contributes to resistance to Fas-mediated apoptosis in PEL cells. We chose the sixth day time point that provides suboptimal apoptosis by vFLIP inhibition alone. We found that BC-3 cells are largely resistant to Fas-mediated apoptosis, but when treated with siRNA to vFLIP, there was a further increase in annexin V positivity when Fas was engaged by agonistic antibodies (Fig. 9). This increase was statistically significant (P = 0.001). Therefore, reduction of vFLIP not only leads to spontaneous apoptosis but it sensitizes PEL to extrinsic apoptotic stimuli.

Classic approaches to determine which proteins of oncogenic viruses are involved in transformation cannot be used to characterize KSHV. Although this virus is associated with specific lymphoma subtypes (suggesting a role in their pathogenesis), it is clearly not sufficient for their development and additional cofactors are necessary. One of these is probably EBV because a large proportion of PEL cells are coinfected by both viruses, but other cofactors are still unknown. A puzzling observation has been that in spite of the fact that KSHV encodes several genes that are transforming in vitro and/or are homologous to cellular oncogenes, in vitro infection of B cells does not transform them (31). This clearly distinguishes KSHV from EBV and may explain why KSHV-associated lymphomas are much less common than those associated with EBV infection. Nevertheless, once a lymphoma develops, when both KSHV and EBV are present, KSHV expresses a set of latent genes, whereas EBV latent gene expression is fairly restricted (32). In addition, alterations in common oncogenes and tumor suppressor genes are rare, in spite of multiple cytogenetic abnormalities (33, 34). These observations suggest that KSHV may be a sustaining force in PEL, and the results presented here support this notion.

We propose a model, illustrated in Fig. 10, whereby vFLIP may act by two mechanisms: (a) by preventing activation of the DISC complex, and (b) by activating the NF-κB signaling pathway, therefore inducing survival signals. The potential for vFLIP to directly inhibit DISC function is mainly suggested by homology to cFLIP long and short forms, which have been clearly shown to affect cleavage and/or release of caspase 8 (29, 30). However, there are minimal data to support a direct role of vFLIP in the DISC; a single paper indicates that vFLIP has the potential of playing such a role because it can bind and inhibit cleavage of procaspase 8 (35), but this interaction has only been shown in transfected HeLa cells and never documented in PEL cells. The second pathway, whereby vFLIP induces NF-κB and inhibits apoptosis by inducing expression of cellular prosurvival factors is clearly demonstrated by our data. We suggest that vFLIP forms an active signaling complex, which we propose to call a life-inducing signaling complex (LISC). At this time, it is unknown what elements are in this complex in naturally KSHV-infected B cells, and whether activity depends on the presence and/or activation of TNF receptors. PEL cells do not express many of the members of the TNF receptor family, and lack CD40, which is expressed in most other B cell types and tumors. However, they do express low levels or TNF-R2 and variable levels of Fas (CD95; unpublished data) and TRAIL receptors (36). Published work based on ectopic overexpression systems and PEL cell lines suggests that vFLIP can associate with the IKK complex, NIK, RIP, and TRAFs, suggesting that these are components of the life-inducing signaling complex (14, 16, 17). PEL cells express abundant TRAF-2 protein, and variable levels of other TRAFs, but no TRAF-1 (unpublished data). Binding of TRAF-2 to the complex containing vFLIP probably leads to activation of a kinase, perhaps RIP, MEKK1, or another as yet undefined kinase that in turn leads to phosphorylation and activation of JNK (18) and the IKK complex. In turn, this would lead to IκBα phosphorylation and degradation, followed by release of NF-κB, that in PEL cells is formed mainly by p65(Rel-A)/p50 heterodimers. Transcriptional activation of cellular prosurvival proteins, including cIAP-1, cIAP-2, and vFLIP, as well as cytokines such as IL-6, leads to protection of virally infected cells from spontaneous apoptosis. Of note, cFLIP feeds back into the same pathway by interfering with the function of the DISC. Our data indicating that vFLIP protects PEL cells from FAS-mediated apoptosis can be explained by direct interference of caspase activation by vFLIP, or by an indirect effect mediated by cFLIP and cIAP up-regulation.

In these studies, apoptosis was not immediate upon down-regulation of vFLIP, and significant apoptosis was only seen after 6 d of treatment with siRNA to vFLIP, whether these were transfected once, twice, or four times. A possible explanation is that some of the antiapoptotic proteins that are controlled by NF-κB have relatively long half lives, and their decrease is gradual upon vFLIP silencing, as seen in the case of cIAP-1. It is likely that a threshold exists in which a combination of antiapoptotic factors are no longer sufficient to protect the cells from apoptotic stimuli. For example, BCL-2 and BCL-X_L are present throughout the time course of our experiment, and they may play a protective role until down-regulation of other factors reaches a critically low level. In addition, we have examined expression of a set of antiapoptotic proteins that we identified in previous analysis as being NF-κB–dependent (unpublished data), but it is likely that additional factors play a role in this process. It is also possible that silencing of vFLIP does not truly induce spontaneous apoptosis, but rather sensitizes cells to extrinsic apoptotic stimuli, including stress and low levels of basal death receptor signaling. Therefore, it is very likely that down-regulation of vFLIP will also sensitize cells to cyototoxic compounds and greatly enhance their therapeutic effect.

Our data support the theory that vFLIP, out of all KSHV-encoded genes, is the gene most functionally analogous to that of EBV-encoded oncoproteins LMP-1 and HTLV-I–encoded Tax. Although obvious functional differences exist between these three proteins, they share expression during latency and the ability to activate NF-κB. Activation of NF-κB has emerged as a method by which EBV, HTLV-1, and KSHV protect the host cell from apoptosis to ensure establishment of viral latency and stable infection, with the unfortunate result of becoming a common pathway in viral lymphomagenesis.

Suppression of vFLIP induces apoptosis of PEL cells, demonstrating that this single gene encoded by KSHV is essential for the survival of infected lymphoma cells. This is the first demonstration that elimination of a single viral protein leads to the death of infected lymphoma cells. Our findings identify vFLIP as a disease-specific therapeutic target for the treatment of PEL and perhaps other KSHV-associated diseases.

This work was supported by National Institutes of Health grant CA68939 and a Leukemia and Lymphoma Society Translational Research grant (to E. Cesarman).