TRAIL-β and TRAIL-γ: two novel splice variants of the human TNF-related apoptosis-inducing ligand (TRAIL) without apoptotic potential

Tumour necrosis factor (TNF) related apoptosis-inducing ligand (TRAIL/APO2L) is a recently identified member of the TNF family, which induces programmed cell death in a variety of neoplastic cell types, but not in most nonneoplastic cells. In this study, we report on the identification of two novel alternative splice variants of TRAIL in neoplastic and non-neoplastic human cells lacking either exon 3 (TRAIL-β) or exons 2 and 3 (TRAIL-γ). In both splice variants, loss of exon 3 resulted in a frame shift generating a stop codon with consecutive extensive truncation in the extracellular domain. Ectopic expression revealed a loss of proapoptotic potential for both alternative splice variants. In contrast to the predominantly cytoplasmatic localisation of GFP-tagged TRAIL-α and TRAIL-β, TRAIL-γ showed an additional association with the cell surface and nuclear membrane. In conclusion, alternative splicing might be involved in fine tuning of TRAIL-induced apoptosis and underlines the complexity of the TRAIL system.

TRAIL/APO2L (tumour necrosis factor related apoptosis-inducing ligand or TRAIL-a) is a recently identified type II transmembrane protein that belongs to the TNF family (Wiley et al, 1995;Pitti et al, 1996) and is processed proteolytically at the cell surface to form a soluble ligand (Mariani and Krammer, 1998). In contrast to other members of the TNF family, such as CD95L/FasL and TNF, TRAIL has been detected in a wide variety of both non-neoplastic and neoplastic cell types and tissues (Wiley et al, 1995;Pitti et al, 1996) and was found to induce programmed cell death in cancer cells, but not in most normal cells (Gura, 1997;Sheridan et al, 1997;Walczak et al, 1997;Ashkenazi et al, 1999;Dejosez et al, 2000). Consistently, TRAIL was shown to suppress growth of human tumours transplanted into mice (Walczak et al, 1999) and to improve survival of the tumour-bearing animals (Ashkenazi et al, 1999). Although repeated intravenous injections of TRAIL in primates did not cause detectable toxicity to non-neoplastic tissues (Ashkenazi et al, 1999), in vitro susceptibility to TRAIL-mediated apoptosis was reported recently for human hepatocytes (Jo et al, 2000).
Several studies found a correlation between TRAIL resistance and expression levels of TRAIL in human cancer cell lines and keratinocytes (Rieger et al, 1999;Zhang et al, 1999;Dejosez et al, 2000). Moreover, the presence of intracellular caspase-8 (Eggert et al, 2001) and antiapoptotic proteins such as c-FLIP Leverkus et al, 2000) and/or activation of the transcription factor NF-kB seem to determine TRAIL sensitivity (Jeremias et al, 1998;Goke et al, 2000;Keane et al, 2000;Bernard et al, 2001;Franco et al, 2001;Oya et al, 2001;Wuchter et al, 2001).
In this report, we describe the identification of two novel TRAIL splice variants, i.e., TRAIL-b and TRAIL-g, in non-neoplastic and neoplastic cells. The lack of exon 3 in TRAIL-b and of exons 2 and 3 in TRAIL-g results in massive truncation of the extracellular binding domain and loss of proapoptotic potential.

Cell lines and cultures
All RCC cell lines (n ¼ 30) used in this study were derived from typical representatives of the clear cell, chromophilic/papillary and chromophobe types of RCC, established in our laboratory as previously described . The cell lines were maintained with Dulbecco's modified Eagle's medium (DMEM, Gibco, Karlsruhe, Germany) supplemented with 10% fetal calf serum, penicillin and streptomycin. PBMCs were cultivated in RPMI medium (Gibco, Karlsruhe, Germany) supplemented with 10% fetal calf serum, penicillin and streptomycin. All cell lines and PBMCs were cultured at 371C in an atmosphere with 5% CO 2 .

Stimulation of PBMCs
Buffy coats of blood samples from four healthy independent donors were collected according to standard procedures. For stimulation, cells were seeded at a density of 5 Â 10 5 cells ml À1 and treated with 10 mg ml À1 concanavalin A (Sigma-Aldrich, Deisenhofen, Germany). Stimulated PBMCs were collected for RNA extraction after emergence of cell blasts (2 -4 days).

RNA extraction
Total RNA was isolated from the cultured RCC cell lines, PBMCs, mouse tissues and the mouse tumour cell line RAW 264.7 using the RNeasy kit (Qiagen, Hilden, Germany). The integrity of RNA (28S/ 18S ribosomal RNA) was controlled by electrophoresis followed by staining with ethidium bromide.

Reverse transcription (RT) and PCR amplification of TRAIL
For cDNA synthesis, 2 mg of total RNA was reversely transcribed in a final volume of 30 ml containing 25 mM of each dNTP (Stratagene, Heidelberg, Germany), 100 pmol random hexamer primer (Stratagene), 20 U of recombinant RNasin RNase inhibitor (Promega, Heidelberg, Germany) as well as 5 U of AMV reverse transcriptase (Promega) with the corresponding RT buffer. The RT reactions were incubated at 551C for 1 h.
Real-time PCR of murine TRAIL was performed on the LightCycler s (Roche Diagnostics, Mannheim, Germany). Amplification was performed in a total volume of 20 ml in the presence of 2 ml 10 Â SYBR Green Fast Start Reaction Mix (Roche Diagnostics), 3 mM MgCl 2 , 25 pmol of each 3 0 and 5 0 TRAIL-specific oligonucleotide (forward primer, 5 0 -GAA TCC CTG CAT TGG GAA GTC AGA-3 0 and reverse primer, 5 0 -GGA TCC TTA ATT AAA AAG GCT CCA AAG AAG-3 0 ) (GenBank accession number U37522) and 2 ml of cDNA (or water as negative control). Real-time PCR was carried out in glass capillaries with an initial denaturation step of 10 min at 951C, followed by 65 cycles of 0 s at 951C, 4 s annealing at 721C, and elongation for 36 s at 721C. Melting curve was directly performed after amplification with an initial denaturation of 0 s at 951C, a temperature delay at 821C for 30 s and a continous heating on 951C for 0 s with a slope of 0.11Cs À1 . To confirm the results, PCR products were electrophoresed on 3% agarose gels containing ethidium bromide and visualised under UV transillumination.

Sequence analysis
Bands of interest were excised from agarose gels and isolated using the QIAquick gel extraction kit (Qiagen), ligated into the pGEM-Tcloning vector (Promega) and cloned in accordance to standard protocols. Plasmid DNA was recovered employing the Plasmid Mini Kit (Qiagen), cycle sequenced using T7 or SP6 site-specific primers, and analysed with an ABI Prism 310 sequencing apparatus (Applied Biosystems, Weiterstadt, Germany).

Cloning of TRAIL coding sequences
To generate GFP-tagged constructs of TRAIL variants, the coding cDNA sequences of the three TRAIL variants were amplified by PCR as described above. PCR amplification products were cloned into the pGEM-T-vector as described above, digested with EcoRI and BamHI, ligated into the mammalian expression vector pEGFP-C1 (Clontech, Heidelberg, Germany) and cloned according to standard protocols. GFP-TRAIL-a construct was cloned as recently described (Kagawa et al, 2001).
Transfection of cultured cells, assessment of cell viability and apoptosis as well as analysis of subcellular localisation by confocal laser scanning microscopy Subconfluent TRAIL-sensitive HeLa cells were transiently transfected in 24-wells using polyfect transfection reagent, as described by the manufacturer (Qiagen). HeLa cells were incubated with the DNA and the transfection reagent for 24 h. To determine the number of cells that underwent apoptosis, HeLa cells were seeded on coverslips in 24-well plates, cultivated for 24 h and transfected as described above. After transfection, cells were washed twice with sterile PBS and fixed for 10 min in ice-cold methanol. Cells were treated with DAPI staining solution (0.1 mg ml À1 ) for 5 min and washed twice with PBS. Up to 300 cells were counted in five different fields per preparation using an Axioscop fluorescence microscope (Zeiss, Germany). The percentage of apoptotic cells was calculated as the ratio between the number of GFP-positive cells with fragmented nuclei and all GFP-positive cells. The data presented are the mean 7s.d. from three replicates.
Cell number was analysed using the colorimetric MTT assay as described before (Mosman, 1983) and measured on a spectrophotometric Titertek Multiskan plate reader (Labsystems, Finland) at 570 nm. The percentage of viable cells in each well was calculated as the ratio between the absorbance of TRAIL-a-, TRAIL-b-or TRAIL-g-transfected cells and the absorbance of vector control-transfected cells. The data presented are the mean 7s.d. from six replicates.
Statistical significance was considered for P-values of Student's t-test less than 0.05. Transfection efficiency was directly determined by the ratio of GFP-positive and -negative cells. The transfection efficiencies varied between 20 and 30% for HeLa cells.
To analyse the subcellular localisation of TRAIL-variants, HEK293 cells in the exponential growth phase were seeded onto sterile coverslips in six-well plates at a density of 1 Â 10 5 cells well À1 . After 24 h, cells were transfected as described above. Afterwards, cells were washed two times in PBS, fixed in ice-cold methanol for 10 min and incubated with endoplasmatic reticulum-specific fluorescence dye concanavalin A Alexa-564 (Molecular Probes, Eugene, USA) (25 mg ml À1 ) for 30 min. Coverslips were washed again in PBS, mounted and examined using a confocal laser scanning system TCS-NT (Leica, Bensheim, Germany).

Western blot analysis
Protein extracts from transiently transfected HeLa cells were isolated by disrupting cells in lysis buffer (100 mM NaCl, 10 mM Tris-HCl pH 7.6, 1 mM EDTA pH8 and protease inhibitor). Protein samples (up to 50 mg) were electrophoresed in 12% SDSpolyacrylamide gels at 70 mA for 4 h. Blotting to Opitran BA-S85 nitrocellulose membranes (Schleicher & Schuell, Dassel, Germany) was performed for up to 2 h at 650 mA in a tank of transfer buffer pH 8.3 (25 mM Tris-HCl, 1925 mM glycine, 20% methanol) using the Hoefer TE series Transphor electrophoresis unit (Hoefer Scientific Intsruments, San Fransisco, USA). To verify transfer efficiency and protein integrity, nitrocellulose membranes were stained with Ponceau S 0.2%. Afterwards, membranes were blocked in blocking buffer (100 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.2% Tween 20) plus 3% nonfat dry milk and 1% BSA for 2 h at room temperature. Immunodetection of GFP-TRAIL-fusion proteins was carried out by incubating membranes with a 1 : 2000 dilution of monoclonal mouse anti-GFP antibody JL-8 (Clontech, Palo Alto, USA) overnight at 41C. PARP cleavage was detected by incubating membranes with a 1 : 500 dilution of monoclonal mouse anti-human PARP antibody, clone AC10.5 (Pharmingen, San Diego, USA) overnight at 41C. Subsequently, membranes were incubated with a 1 : 1000 dilution with horseradish-peroxidase linked anti-mouse antibody (Amersham Pharmacia, Freiburg, Germany) for 1.5 h. Proteins were detected by incubation with Lumi-Light substrate (Roche, Mannheim, Germany). Equal loading was confirmed by bactin detection with a mouse anti-b-actin-specific antibody, clone AC-15 (Sigma-Aldrich, Taufeirchren, Germany). Data on X-ray films were quantified by densitometry with the program Tina 20 (raytest Isotopenmessgeräte GmbH, Germany).

Identification of two novel TRAIL splice variants
As revealed by RT -PCR analysis, all RCC cell lines (n ¼ 30) expressed TRAIL-a, irrespective of their histological types ( Figure 1A). Besides the expected amplification product of 371 bp, however, two additional bands of 328 and 190 bp were coamplified. These three amplification products were sequenced. The largest band was identified as regularly spliced TRAIL-a, whereas the 328 bp product was identified as a TRAIL variant with a deletion of 43 nucleotides corresponding to nucleotides 358 -400 of the TRAIL-a cDNA of the human TRAIL sequence (GenBank Accession no. U37518). The smallest PCR product (190 bp) lacked nucleotides 220 -357.
Although the genomic organisation and the exon -intron boundaries of TRAIL had recently been published (Gong and Almasan, 2000), we repeated the database analysis by running advanced BLAST, looking for a genomic clone that contains parts of the human TRAIL gene. This database search revealed that the Homo sapiens Chromosome 3-clone hRPK.44 A (Accession no. AC007051) contained the entire coding sequence of TRAIL. Subsequently, we determined the exon -intron organisation of the human TRAIL gene, which consists of five exons (Figure 2). Of note, we observed a difference in two nucleotides (position 143697 and 145990) marked by squares in Figure 2 when compared to the published sequence (Gong and Almasan, 2000). According to the genomic organisation, the 43 nucleotides deleted in the 328 bp long amplification product corresponded to exon 3 and, therefore, we designated this new alternative splice variant as TRAIL-b. The smaller band (190 bp) was found to lack both exon 2 (138 nucleotides) and exon 3 and, therefore, was designated as TRAIL-g. Additional amplification of genomic DNA by PCR showed no product and excluded the presence of pseudogenes.
To investigate whether the novel TRAIL splice variants are exclusively generated in neoplastic cells, we performed RT -PCR with unstimulated and stimulated PBMCs showing that the alternative splice variants TRAIL-b and TRAIL-g are expressed in non-neoplastic cells as well ( Figure 1B). No alternative TRAIL splice variants, however, could be detected in mouse liver and kidney or in the mouse tumour cell line RAW 264.7 ( Figure 1C). This observation strongly suggests that alternative splicing of TRAIL may be a species-restricted phenomenon.  Figure 2 Genomic organisation of the human TRAIL gene. The cDNA sequence of human TRAIL (accession no. U37518) is marked in bold letters within the Homo sapiens chromosome 3, clone hRPK.44 A 1 (accession no. AC007051). SD and SA sites at the exon -intron boundaries are underlined. Of note, differences between our analysis of the genomic organisation and a recent publication (Gong and Almasan, 2000) are marked by squares. The stop codon generated by alternative splicing of exons 2 and 3 is marked by a black-shaded square.

Splice donor and acceptor sites at the intron-exon and exon -intron boundaries of exons 2 and 3 are suitable for alternative splicing
Computational analysis with the Signal program (PC/GENE package) showed that exon 2 as well as exon 3 are flanked by splice donor (SD) and splice acceptor (SA) sites matching to the consensus sequence of common SD sites ({C/A}AG|GT{A/G}AGT) and SA sites ({T/C} 11 N{C/T}AG|G) as shown in Figure 3A. Both the SA sites at position 140298 -140457 of exon 2 and at position 143641 -143700 of exon 3 contain the AG bases of SA sites as well as the GT bases of the SD sites that are necessary for posttranscriptional splicing on intron -exon boundaries. The necessary GT bases of SD sites for splicing processes exist also at the exonintron boundaries of exons 2 and 3. Taken together, these data indicate that post-transcriptional processing of the TRAIL-pre-mRNA leads to the generation of three alternative splice products, i.e., TRAIL-a, TRAIL-b and TRAIL-g ( Figure 3B).

Alterations of protein domains in the novel TRAIL isoforms and functional implications
Computational analysis of TRAIL-a and its novel alternative splice variants showed that loss of exon 3 resulted in a frame shift generating a stop codon with consecutive truncation in the extracellular domains of both TRAIL-b and -g. This truncation resulted in a 98 amino-acids long TRAIL-b and a 52 amino-acids long TRAIL-g ( Figure 4A). Moreover, PSORT II analysis (Nakai and Horton, 1999) identified KIFL, a KKXX-like motif known as endoplasmatic reticulum membrane retention signal at the Cterminus of both TRAIL-b and -g. PROSITE scan (Bairoch, 1991) also identified potential protein kinase C phosphorylation sites between amino acids 92 -94 of TRAIL-b and amino acids 46 -48 of TRAIL-g, immediately adjacent to the potential endoplasmatic membrane retention signal. TRAIL-b also contains a potential cAMP-and cGMP-dependent protein kinase phosphorylation site between amino acids 89-and 92 ( Figure 4B). Although PROSITE scan analysis had identified a potential ER membrane retention signal at the C-terminus of both TRAIL-b and -g, PSORT II analysis predicted a preferential cytoplasmatic localisation for all TRAIL isoforms (data not shown). To further investigate the subcellular localisation of TRAIL-a and its alternative splice variants, GFP-fused TRAIL variants were ectopically overexpressed in HEK293 cells. Using concanavalin A Alexa-594 as a marker of endoplasmatic reticulum, confocal laser microscopy did not reveal a selective accumulation of TRAIL-b or TRAIL-g in the endoplasmatic reticulum ( Figure 5). However, TRAIL-g was found to produce a more intensive fluorescence of the cell surface and nuclear membrane when compared to the other splice variants.
Using HeLa cells known to be sensitive to TRAIL-mediated apoptosis (Harper et al, 2001), ectopic expression of TRAIL variants revealed marked differences in the apoptotic activity of TRAIL-a and its novel alternative splice variants: ectopic expression of TRAIL-a led to huge clusters of apoptotic cells, which were  (Thompson et al, 1994) and enhanced with Boxshade. Dark shading: residues are highly conserved; light shading: residues are less well conserved; not shaded: residues are not conserved. (B) Protein sequences were analysed by PROSITE scan. Potential domains and modification sites are marked. A frame shift because of alternative splicing results in a potential endoplasmatic reticulum membrane retention (ER) signal at the C-terminus adjacent to new potential PKC-phoshorylation sites in both TRAIL-b and -g. Light grey box ¼ newly generated ER-signal containing domain. ICD ¼ intracellular domain; ECD ¼ extracellular domain; TM ¼ transmembrane region; ER ¼ endoplasmatic reticulum membrane retention signal; cAMP ¼ cAMP-dependent protein kinase phoshorylation site; CK2 ¼ casein kinase II phosphorylation site; TYR ¼ tyrosine kinase phosphorylation site; PKC ¼ protein kinase C phosphorylation site.
not observed after transfection with the other TRAIL variants or the vector control ( Figure 5B). DAPI staining permitted the identification of fragmented or condensed nuclei indicative for apoptosis ( Figure 5C) and permitted a visual quantification of apoptosis (see Material and Methods). Using this assay, a significant decrease of apoptotic cells was observed after ectopic expression of GFP-TRAIL-b or GFP-TRAIL-g or the vector control when compared to GFP-TRAIL-a-transfected HeLa cells ( Figure 5D). The decrease of apoptotic potential observed for GFP-TRAIL-b and GFP-TRAIL-g was paralleled by a significant increase of cell viability as demonstrated by MTT assays ( Figure 5E). To further confirm the effects of the different TRAIL variants on apoptosis, PARP cleavage was analysed by a monoclonal mouse antibody, that recognises a 85 kDa cleavage product besides the 116 kDa uncleaved PARP. PARP cleavage was found to be enhanced three-fold by ectopic expression of TRAIL-a in comparison with the other TRAIL variants or the vector control ( Figure 5F). To ensure that the observed effects were based on protein expression of all TRAIL variants, the expression of the fusion proteins was confirmed by Western blot analysis using a mononoclonal mouse anti-GFP anbtibody ( Figure 5G).

DISCUSSION
In this study, we report on the identification of two novel splice variants of TRAIL, i.e., TRAIL-b and TRAIL-g, in RCC cell lines of all major histological subtypes. These previously unknown splice variants were not clearly detectable in RNase protection assays (Dejosez et al, 2000), because only 43 bp (TRAIL-b) or 181 bp (TRAIL-g) differences in length were found in comparison with the primarily identified TRAIL-a transcript (Wiley et al, 1995;Pitti et al, 1996). It was especially intriguing to detect these splice variants in non-neoplastic human cells as well, whereas no corresponding splice variants were found in different cell types of murine origin. Alternative splicing permits a high degree of protein diversity generating structurally and functionally distinct proteins that differ in their subcellular localisation, molecular targets or stability. Splice variants have previously been found to play a key role in the regulation of apoptosis as well, determining the actions of many apoptosis-related genes at all levels of apoptotic signalling pathways (Jiang and Wu, 1999). Thus, the CD95 (APO-1, Fas) receptor exists in membrane-bound and soluble isoforms which antagonistically affect apoptosis (Jiang and Wu, 1999). The Bcl-2 family encompasses genes, for example, Bcl-X and its different splice variants Bcl-X L /X S , which antagonistically determine the susceptibility to cell death signals (Boise et al, 1993). Finally, different isoforms of caspases (Srinivasula et al, 1999;Cote et al, 2001;Eckhart et al, 2001) and inhibitor of apoptosis (IAP) proteins Conway et al, 2000) are generated by alternative splicing. The novel TRAIL isoforms described in this study show an extensive loss of their extracellular binding domain, which plays a key role for trimeric stability, ligand -receptor binding capacity and apoptotic signalling (Pitti et al, 1996;Sheridan et al, 1997;Ashkenazi et al, 1999;Bodmer et al, 2000b;Kagawa et al, 2001). Consequently, the novel truncated TRAIL isoforms should be unable to form stabile dimeric or trimeric receptor -ligand complexes and fail to trigger TRAIL mediated-apoptosis as demonstrated by our transfection experiments.
At first sight, therefore, TRAIL-b and TRAIL-g may interfere with the fine tuning of TRAIL actions simply by reducing the amount of full-length ligand. In fact, the generation of truncated isoforms has been suggested as an additional mechanism to quantitatively regulate gene expression and action, because the mRNAs of all isoforms originate from the same pre-mRNA precursor pool (Jiang and Wu, 1999). Interestingly in this context, however, ectopically overexpressed TRAIL-g exhibited a more prominent association with the cell surface and the nuclear membrane when compared to both TRAIL-a and TRAIL-b. Further experimental work, therefore, will have to show whether these alternative splice variants actually play a passive regulatory role for TRAIL-mediated apoptosis only or acquired novel functional properties as well. In any case, the identification of two novel splice variants may have implications for our understanding of TRAILmediated apoptosis in neoplastic and non-neoplastic human cells.

ACKNOWLEDGEMENTS
We thank Sigrid Khalil, Michael Ringler and Martina Bellack for technical assistance. The results of this work are part of the PhD thesis of Thomas Krieg and the MD thesis of Andreas Krieg. This work was supported by a grant from the Deutsche Forschungsgemeinschaft and by the Stiftung für Altersforschung. Continued E-G This visual apoptosis assay (n ¼ 3) further confimed the loss of proapoptotic potential for GFP-TRAIL-b and GFP-TRAIL-g, which also became evident (E) from an increase of surviving cells in our MTT assays (n ¼ 6). (F) Immunodetection of the caspase-3-dependent 'death substrate' PARP showed enhancend cleavage only after ectopic expression of GFP-TRAIL-a. To ensure the detection of this cleavage product, protein lysates from TRAIL-treated control cells were separated on the same gel. The data presented are the mean 7s.d. from three independent transfections. (G) To show that the recombinant TRAIL variants are expressed on protein level, whole protein lysates of transfected HeLa cells were separated by SDS gels and immunodetection of GFP fusion proteins with anti-GFP-IgG was carried out after blotting on nitrocellulose membrane.