Novel Cellular Genes Essential for Transformation of Endothelial Cells by Kaposi's Sarcoma–Associated Herpesvirus Free

Kaposi's sarcoma–associated herpesvirus (KSHV) is involved in the development of lymphoproliferative diseases and Kaposi's sarcoma. The oncogenicity of this virus is reflected in vitro by its ability to transform B cells and endothelial cells. Infection of dermal microvascular endothelial cells (DMVEC) transforms the cells from a cobblestone-like monolayer to foci-forming spindle cells. This transformation is accompanied by dramatic changes in the cellular transcriptome. Known oncogenes, such as c-Kit, are among the KSHV-induced host genes. We previously showed that c-Kit is an essential cellular component of the KSHV-mediated transformation of DMVEC. Here, we test the hypothesis that the transformation process can be used to discover novel oncogenes. When expression of a panel of KSHV-induced cellular transcripts was inhibited with antisense oligomers, we observed inhibition of DMVEC proliferation and foci formation using antisense molecules to RDC1 and Neuritin. We further showed that transformation of KSHV-infected DMVEC was inhibited by small interfering RNA directed at RDC1 or Neuritin. Ectopic expression of Neuritin in NIH 3T3 cells resulted in changes in cell morphology and anchorage-independent growth, whereas RDC1 ectopic expression significantly increased cell proliferation. In addition, both RDC1- and Neuritin-expressing cells formed tumors in nude mice. RDC1 is an orphan G protein–coupled receptor, whereas Neuritin is a growth-promoting protein known to mediate neurite outgrowth. Neither gene has been previously implicated in tumorigenesis. Our data suggest that KSHV-mediated transformation involves exploitation of the hitherto unrealized oncogenic properties of RDC1 and Neuritin.

Kaposi's sarcoma–associated herpesvirus (KSHV) or human herpesvirus 8 (HHV8) is an oncogenic γ-2 herpesvirus (1). KSHV is the etiologic agent of Kaposi's sarcoma, a mesenchymal tumor consisting of abnormal blood and lymphatic vessels (2). Kaposi's sarcoma tumors are complex, multifocal lesions characterized by spindle cells of endothelial origin and infiltrating inflammatory cells (T cells, B cells, and monocytes; ref. 3). In addition, KSHV is associated with primary effusion lymphoma or body cavity–based lymphoma (PEL/BCBL), a B-cell non-Hodgkin's lymphoma characterized by pleural, pericardial or peritoneal lymphomatous effusions without a contiguous tumor mass. KSHV is also present in multicentric Castleman's disease, which can take the form of angiofollicular lymph node hyperplasia (solid tumor), or a multisystem generalized lymphoadenopathy with immunologic abnormalities (3). Kaposi's sarcoma is the most frequent malignancy in patients with acquired immune-deficiency syndrome (AIDS; ref. 4).

In the Kaposi's sarcoma lesion, KSHV infects the majority of spindle cells as well as lesional endothelial cells and infiltrating leukocytes (5–8). The majority of spindle cells harbor the KSHV genome in a latent form, with a small percentage of infected cells entering the lytic cycle and producing infectious virus (7–11). KSHV genes with the potential to deregulate cellular growth have been described and several excellent reviews on this subject exist (12, 13). Despite the fact that KSHV proteins with oncogenic potential have been identified, the mechanisms of virus-induced tumorigenesis remain unclear.

To identify oncogenic pathways associated with Kaposi's sarcoma, we developed an in vitro culture system that allows efficient and sustained KSHV infection of dermal microvascular endothelial cells (DMVEC; ref. 14). This tissue culture system recapitulates many of the transforming and cytologic aspects of KSHV infection in vitro such as spindle cell morphology and expression of cell surface markers characteristic of Kaposi's sarcoma lesions. As observed in vivo, lytic replication occurs spontaneously in only a small percentage of infected DMVEC in vitro. Lytic replication can however be induced in a larger percentage of infected cells by treatment with phorbol esters. This system is thus ideally suited to the study of virus-host interactions as well as the cellular events that take place during KSHV-mediated transformation.

Using the DMVEC system, we employed DNA microarray technology to identify changes in host cell gene expression that occur in response to a typical KSHV infection (15). This analysis showed that DMVEC harboring predominantly latent KSHV exhibited substantial changes in host cell gene expression. We further showed that the KSHV-induced cellular oncogene, c-Kit, was an essential component of the KSHV transformation process. Inhibition of c-Kit signal transduction by the small molecule inhibitor STI571 (Gleevec), or by dominant-negative c-Kit constructs, prevented KSHV-induced spindle cell formation. Furthermore, ectopic expression of c-Kit induced morphologic changes in uninfected DMVEC that resembled those seen with KSHV infection. Therefore, we concluded that c-Kit might be a valid therapeutic target for the treatment of Kaposi's sarcoma. This hypothesis is supported by the recent finding that treatment of cutaneous Kaposi's sarcoma in AIDS patients with Gleevec resulted in clinical and histologic regression of lesions (16).

Our findings thus suggested that a well-known oncogene was important for KSHV-associated tumorigenesis. The complex nature of cellular transformation, together with the fact that multiple host transcripts are dysregulated by KSHV, suggests that other KSHV-induced cellular genes might also play a role in DMVEC transformation. Unlike c-Kit however, specific small molecule inhibitors do not exist for the vast majority of cellular gene products. Recent advances in antisense and RNA interference technology create the opportunity to knockdown the transcript levels of selected genes of interest (17, 18). Therefore, we used antisense technology to examine whether or not a selected panel of KSHV-up-regulated genes was involved in viral transformation of DMVEC. Combining transcriptional profiling with antisense gene knockdown led to the identification of two KSHV-induced genes, RDC1 and Neuritin, whose expression proved essential for DMVEC transformation by KSHV. Neither RDC1, an orphan G protein–coupled chemokine receptor nor Neuritin, a growth-promoting protein regulating neurite outgrowth, have previously been implicated in cancer. We now show that RDC1 and Neuritin transform NIH 3T3 cells and induce tumors in nude mice, suggesting that they might play an important and unanticipated role in Kaposi's sarcoma tumorigenesis.

Viruses and cell culture. KSHV-infected DMVECs were established as previously described (14). DMVEC were maintained in endothelial-SFM growth medium (Life Technologies, Gaithersburg, MD) supplemented with 10% human AB serum (Sigma, St. Louis, MO) and 25 μg/mL endothelial cell growth supplement (Fisher Scientific, Pittsburgh, PA). The KSHV viral stocks were derived from the BCBL-1 cell line (contributed by D. McGrath and D. Ganem, AIDS Research and Reference Program, Division of AIDS, National Institute of Allergy and Infectious Diseases, NIH) as previously described (14). KSHV-infected DMVEC were monitored by immunofluorescence and used in microarray experiments when >90% of the cells expressed the latent nuclear antigen-1 LANA-1 (ORF73).

Human tissue. Four-millimeter punch biopsies of cutaneous Kaposi's sarcoma lesions were obtained with informed consent from HIV-positive Kaposi's sarcoma patients following human experimental guidelines of the U.S. Department of Health and Human Services and the Beth Israel Deaconess Medical Center. Tissue was immediately submerged in RNAlater (Ambion, Inc., Austin, TX) for preservation before RNA isolation. RNA was extracted after mechanical separation of tissue and lysing dispersed cells in RNA isolation buffer.

RNA isolation and labeling. RNA was purified from DMVEC or dispersed tissue using RNeasy spin columns (Qiagen, Inc., Valencia, CA). RNA quality was assessed with an Agilent bioanalyzer. RNA labeling was done as outlined in the Affymetrix labeling protocol for Gene Chips (Affymetrix GeneChip Expression Analysis Technical Manual rev. 3. 2001). Five micrograms of RNA were converted into double-stranded cDNA by priming with an oligo(dT)-T7 primer which allows for second strand synthesis by T7 polymerase. The cDNA was labeled with biotin using an RNA labeling kit (ENZO) to produce biotin-labeled cRNA transcripts. The final transcripts were analyzed with an Agilent bioanalyzer for consistency in transcript length and yield. Fifteen micrograms of labeled RNA were fragmented and hybridized to Human Affymetrix chip HG-U95A by the Gene Microrarray Shared Resource at the Oregon Health and Science University.³

After the hybridization and washing steps, gene chips were scanned with an HP GeneArray Scanner.

Data analysis. Gene Chip data was analyzed with Affymetrix GeneChip analysis software (M.A.S.5.0). Comparisons were made between passage-matched KSHV-infected DMVEC and mock-infected DMVEC in two separate experiments. For each data set, difference calls and fold changes were compiled and filtered into text-delimited format for import into Microsoft Excel. Initial filters removed all absent (A) calls and genes with no change (NC) calls across both data sets. The final set contained genes with at least a 2-fold change in one of the two data sets with increase (I), marginally increase (MI), decrease (D), and marginally decrease (MD) calls. The exported data was further filtered to include only genes with a 2-fold change in both experiments. Gene descriptions, annotations and functional groups were updated with data from the Netaffx analysis center.⁴

Functional groups were assigned from data obtained from Gene Ontology and data compiled at Source. ⁵ Genes with unknown function were searched on Public Medline for any recent experimental evidence on their possible functions. The complete data sets, including decreased genes, are posted on a web site.⁶

Quantitative PCR. Real-time PCR was done on an ABI 7700 sequence detection system (Applied Biosystems, Foster City, CA). To normalize gene expression between mock and infected samples, 18S RNA, β-actin, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were assessed. GAPDH was shown to have the least variation between different samples and was used as the normalizing gene in quantitative analysis. Total RNA was treated with Dnase I (Dnase Free, Ambion) before synthesis of cDNA by random hexamers and Superscript II (Invitrogen, Carlsbad, CA). The following primers were selected by using Primer Express software (Applied Biosystems): RDC1 383F, CTGCGTCCAACAATGAGACCT and 452R, CCGATCAGCCACTCCTTGA; KIA1036 4804F, AGCCAAGAAGTTGACCACGTG and 4926R, AGGTGCACACATTCACACAGG; Osteopontin (SPP1) 462F, CCTGCCAGCAACCGAAGTT and 537R, AACCACACTATCACCTCGGCC. The remaining primers have been previously described (15). Reactions were done using SYBR Green PCR core reagents. Relative expression values between mock and infected samples were calculated by the comparative cycle threshold (c_t) method as previously described (15). Dissociation curves were done after each amplification run to control for primer-dimers. Absolute standard curves were generated from plasmids encoding Neuritin and RDC1 (Open Biosystems, Huntsville, AL).

Morpholino treatment and focus inhibition assay. Phosphorodiamidate morpholino antisense oligonucleotides (PMO-AS) were designed and synthesized by Gene Tools LLC (Philomath, OR) as a Special Delivery formulation, which pairs the morpholino with a partially complementary DNA oligo. A weakly basic delivery agent, ethoxylated polyethlyenimine (EPEI) is then used to deliver the anionic morpholino/DNA duplex into the cytosol. PMO-AS were received as sterile, lyophilized solids (300 nmol/L) that were solubilized in sterile water to obtain 0.5 mmol/L stock solutions. To load a 35-mm dish of DMVEC at 90% confluence, 5 μL of stock solution was mixed with 5 μL of EPEI delivery solution in 590 μL H₂0 for 20 minutes at room temperature and mixed with 1.5 mL of serum-free RPMI. Endothelial growth media was removed from DMVEC, and replaced with the PMO-AS/EPEI/RPMI delivery solution for 3 hours in a 37°C incubator. After incubation, cells were rinsed, complete endothelial growth medium was added, and cells were cultured for up to 10 days without subculture, with medium replacement every 48 to 72 hours. This protocol allows for the characteristic post-confluent growth and development of multilayered spindle cell aggregates (foci) in KSHV-infected cultures. Cells were monitored daily under the phase microscope to evaluate PMO-AS inhibition of the development of these foci. At the conclusion of an assay, cells were fixed in 2% paraformaldehyde and stained with a MAB against CD31 (1:100; DakoCytomation California Inc.,Carpenteria, CA) followed by a goat-anti-mouse FITC second conjugate. Junctional staining of cells with CD31 accentuated the disorganized multilayered nature of the spindle cell aggregates compared with the flat, organized profile of contact-inhibited cells.

Small interfering RNA treatment and focus inhibition assay. Small interfering RNA (siRNA) oligonucleotides were designed using the oligoengine design tool (Oligoengine, Seattle, WA).⁷

RNA olionucleotides were purchased from Oligoengine or Dharmacon, Inc. (Lafayette, CO). The following sequences were used: RDC1-234, AACATGCCCAACAAAAGCGTC; RDC1-597, AAGAAGATGGTACGACGTGTC; NEURITIN-258, AAAGATATCTG ATTAATTCCA. The 21-nucleotide RNA oligos were resuspended according to manufacturer's instructions to obtain either 20 or 50 μmol/L solutions and delivered to DMVEC by transfection. For transfection, cells were seeded into 35-mm plates for overnight incubation, and transfections were done at ∼80% confluence according to published methods (19). Briefly, from a 20 μmol/L siRNA stock solution, 12 μL of siRNA were added (240 pmol) to 200 μL Opti-MEM (Life Technologies). After a 10-minute incubation, 12 μL Oligofectamine (Invitrogen) diluted in 48 μL Opti-MEM was mixed with the diluted siRNA and incubated for an addition 20 minutes. The complex was added dropwise to a 35-mL plate containing 1 mL of complete endothelial growth medium. Cells were monitored for up to 14 days for evidence of foci formation, as described for the PMO-AS assays. Control Cy3-Luciferase GL2 Duplex siRNA (Dharmacon) was used to monitor transient transfection and the duration of siRNA expression in endothelial cells. Cy3 siRNA was visible for over 3 weeks post-transfection (data not shown).

2,3-Bis[2-methoxy-4-nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilide inner salt assay. Metabolism of the tetrazolium salt 2,3-bis[2-methoxy-4-nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilide inner salt (XTT) to a water-soluble formazan dye by viable cells gives a quantitative determination of relative cell number. The XTT assay was thus used to quantitatively assess the effect of PMO-AS treatment on KSHV focus formation and post-confluent growth. For these assays, KSHV-infected and uninfected DMVEC were seeded into 96-well flat-bottomed Primaria trays at 10⁴ cells per well, loaded with PMO-AS after overnight equilibration, and cultured for 72 hours. For the final 4 hours of culture, 50 μL XTT (Roche Diagnostics, Indianapolis, IN) were added to each well and the absorbance of each well was recorded by an ELISA reader between 450 and 500 nm. To verify efficient (>80%) PMO-AS uptake in the 96-well format, cells were treated with a FITC-tagged morpholino (FITC-PMO) and uptake visually assessed under a fluorescence microscope. This FITC-PMO also served as a control for any nonspecific effect of the PMO-AS/EPEI complex on DMVEC growth. As a positive control, cells were treated with a c-Kit PMO-AS previously shown to inhibit KSHV-induced transformation (20). For NIH 3T3 cells, the XTT assay was done in 96-well plates at a cell density of 10⁴ cells per well, and absorbance measured at 48 hours.

Production of stable NIH 3T3 cell lines. A hemagglutinin (HA) tag was introduced onto the NH₂ terminus of RDC1 by PCR with primers specific for the full-length RDC1 sequence (accession BC036661). The following primers were used: NH₂ terminus with HA and Xho site 5-CCGCTCGAGATGTACCCATACGATGTTCCAGATTACGCTGATCTGCATCTCTTCG-3 and COOH terminus 5-ATCTCATTTGGTGCTCTGCTCCA-3. The tagged sequence was then cloned into the expression vector pcDNA3.1 and used to transfect NIH 3T3 cells. To clone Neuritin (accession NM 016588), the GPI anchor was removed and an HA-tag was inserted at the COOH terminus before cloning into pcDNA3.1. The following primers were used: NH₂ terminus with Xho site 5-CCGCCTCGAGCGGATGGGACTTAAGTTGAACGGCA-3 and COOH terminus 5-ATCTTATTAAGCGTAATCTGGAACATCGTATGGGTAGCTGGTGAAGGAAAGCCAGGTCGCTAAAGCT-3. The cloned genes were sequenced to confirm absence of mutations and the presence of the HA epitopes. The HA epitope was detected by immunofluorescent staining with a monoclonal antibody against HA (Sigma). To stabilize Neuritin expression, cells were treated with Brefeldin A (10 μg/mL) 16 hours before staining. To generate stable mass cell line clones, plasmids were transfected into NIH 3T3 mouse fibroblasts and selected with 750 μg/mL G418 (Cellgro, Herndon, VA). Mass clone cultures were subcloned into 96-well plates to isolate individual colonies. NIH 3T3 cells were maintained in DMEM (Life Technologies) containing 5% bovine calf serum (HyClone, Logan, Utah). Full-length Neuritin cDNA plasmid was subcloned from a full-length cDNA plasmid (clone MGC:3391 IMAGE:3605775) into pcDNA 3.1.

Soft agarose assays. Five thousand stably transfected NIH 3T3 cells were plated in 1.5 mL of DMEM with 5% bovine calf serum and 0.3% melted agarose onto a 3-mL bottom layer of 0.6% agarose medium per well of a 6-well dish. The cells were fed every 3 days with several drops of medium, and colonies were photographed after 2 to 3 weeks. For soft agarose assays with mass culture NIH 3T3 cells, 10,000 cells were plated.

Tumorigenicity in nude mice. Seven-week-old B6.Cg-Foxn1^nu/J nude mice were obtained from The Jackson Laboratory (Bar Harbor, MA) and maintained according to the Oregon Health and Science University (OHSU) Division of Animal Resources Standard Operating Procedures. OHSU, Mice were injected s.c. in the right flank with early passage NIH 3T3 cell lines stably expressing RDC1, Neuritin, Rasv12 (as a positive control for tumor formation), or the pcDNA 3.1 vector only (as a negative control). Two independent clones of RDC1- and Neuritin-expressing cells were used. Cells were prepared for injection by trypsinizing and suspending in 10% fetal bovine serum/MEM, followed by two washes in sterile PBS. Cells were resuspended in PBS to a concentration of 1.5 × 10⁷ cell/mL, and 200 μL of the suspension was injected s.c. according to an OHSU Institutional Animal Care and Use Committee–approved protocol. Two groups of four mice each received the two Neuritin clones. A group of four mice and a group of five mice received RDC1 clones 1 and 2, respectively. Control pcDNA 3.1 vector cells were injected into the left flank of two groups of four mice that had received either RDC1 or Neuritin in the right flank. Two additional mice were injected in the right flank only with control Ras-expressing cells. Tumor production and size were monitored twice weekly. Mice were humanely sacrificed when tumors reached ∼1 cm in diameter. Tumor volumes were calculated using the following formula: width² × length × 0.52 (21). Tumors were excised and RNA was isolated by a modified tissue isolation protocol with RNeasy columns (Qiagen).

Two previous studies that described transcriptional changes in long-term KSHV-infected primary or life-extended DMVEC (15, 22) relied on cDNA libraries that were either displayed on nylon filters or glass arrays. To confirm and extend these studies with a different methodology, we compared RNA samples from KSHV-infected and passage-matched mock-infected DMVEC using the Affymetrix HG-U95A Gene Chip, which contains 12,626 unique probe sets. Before harvesting cells, we verified the establishment of latent KSHV infection by monitoring spindle formation and the expression of KSHV LANA-1 (ORF73) but not KSHV lytic antigens (ORF59 and ORFK8.1A/B) by immunofluorescence as described (14). After ∼4 weeks, when >90% of the cells developed prominent spindle morphology and were confirmed as being latently infected, RNA was isolated from parallel cultures in T75 flasks. We repeated this procedure with an independent infection to acquire a second set of microarray data. Sample comparisons were done using Affymetrix MAS 5.0 software, which calculates fold changes and difference calls of increase (I), decrease (D), marginal increase/decrease (MI or MD), or no change (NC) based on a statistical algorithm. For our analysis, we focused only on up-regulated genes that were scored as (I) and showed at least 2-fold up-regulation in both microarray experiments. The complete data set including down-regulated genes is available.⁶ There were 97 commonly up-regulated genes in the two independent KSHV experiments (online Supplementary Table S1). Of these, 36 genes have been previously shown up-regulated on cDNA arrays (15, 22). The up-regulated genes were divided into functional groups using the GeneOntology Database as well as recent experimental evidence from published literature. The majority of genes could be grouped into genes involved in signaling, oncogenesis, angiogenesis, or transcription. Whereas the table likely contains genes that are important for various aspects of KSHV biology and virus-mediated alteration in cellular function, the challenge is to extract those genes that are of biological importance for a specific pathway. Here, we focus on identifying cellular transcripts that play a role in KSHV-mediated transformation of DMVEC as characterized by spindle cell morphology and foci formation.

Consistent with our previous observations, one of the up-regulated genes on the U95A array was c-Kit, which has been shown essential for the transformation of DMVEC (15). Our current goal was to identify additional KSHV-induced genes involved in the transformation of DMVEC. One strategy to find new genes involved in tumorigenesis would be to inhibit the expression of the suspected gene and examine whether inhibition influenced the ability of KSHV-infected cells to form foci, without affecting the growth of noninfected DMVEC. To examine if this strategy would work in principle, we selected eight up-regulated genes from different functional classes, inhibited their expression using PMO-AS antisense oligonucleotides, and monitored morphologic changes in the PMO-AS-treated versus untreated cultures. Genes were primarily selected based on criteria such as their known or suspected involvement in oncogenesis or their strong induction by KSHV, but we also selected some genes with no known function that fell into the latter category. Induction of these genes by KSHV was confirmed by quantitative PCR (Table 1). Genes selected for inhibition by antisense molecules are listed in Table 2 and are briefly described in the following paragraphs.

RDC1 or chemokine orphan receptor 1. This G protein–coupled receptor was the most strongly induced cellular transcript in this study as well as other GeneChip experiments done in our laboratory. Its up-regulation was also observed previously using cDNA array analysis (15, 22). We have also confirmed RDC1 induction by real-time quantitative reverse transcription-PCR (qRT-PCR). For qRT-PCR experiments, we tested both the original RNA samples, as well as two independent infections and passage matched controls. Remarkably, a >500-fold change, representing a difference of more than nine cycle thresholds (c_t), was observed between KSHV-infected cells and mock controls (Table 1). One c_t difference represented a 2-fold change. The difference between the GeneChip and real-time results probably reflects a saturation of hybridized transcripts on the chip versus a larger dynamic range of detection for real-time PCR. The extraordinarily strong induction of RDC1 by KSHV was the basis for selecting this molecule for further analysis.

Lim-domain only 2. Lim-domain only 2 (LMO2) is a transcriptional regulator that is part of a pentameric activator complex known to regulate c-Kit transcription (23, 24). LMO2 was consistently up-regulated in cDNA and Affymetrix microarrays and this up-regulation was previously confirmed by qRT-PCR (15).

Osteopontin (SPP1). The cytokine osteopontin is consistently induced in colon cancer and is thought to be involved in metastasis (25, 26). We confirmed its up-regulation by qRT-PCR (Table 1).

c-Mer (MERTK). Similar to c-Kit, c-Mer is a receptor tyrosine kinase (27). Up-regulation of c-Mer was previously observed on cDNA arrays and was confirmed by qRT-PCR (15).

Neuritin 1 (NRN1; Neuritin). Neuritin is GPI-anchored protein that controls morphologic changes in neurons, particularly neurite outgrowth and branching of neuritic processes (28). Up-regulation of this gene was previously observed by cDNA array and confirmed by qRT-PCR (15).

In addition to these genes with known function described above, we also selected a completely unknown gene, KIAA1036, for our validation studies. KIAA1036 transcripts were up-regulated on the Affymetrix GeneChip and confirmed by qRT-PCR (Table 1).

To inhibit the KSHV-mediated up-regulation of selected transcripts, we chose to use PMO-AS. These molecules have been shown to effectively knock down genes in numerous developmental biology models including sea urchin, Xenopus laevis, zebrafish, and mouse (18). Earlier generations of DNA-based antisense molecules acted via RNase H degradation of mRNA, were relatively nonspecific, could require several oligomers for effective target knockdown, and often displayed considerable toxicity. In contrast, PMO-AS act to sterically block translation of mRNA with high specificity, allowing the design and use of a single molecule targeted to the start codon. In addition PMO-AS are highly resistant to enzymatic degradation and exhibit low toxicity.

To determine whether PMO-AS molecules are able to enter DMVEC and effectively block target gene expression, we previously designed a PMO-AS against c-Kit and showed that this PMO could inhibit c-Kit protein expression in KSHV-infected DMVEC (20). In these experiments, treatment of KSHV-infected DMVEC with the c-Kit-specific PMO-AS also inhibited foci formation and allowed maintenance of a confluent but contact-inhibited monolayer for up to 10 days of post-treatment culture. Inhibition of foci formation was similar to that seen when KSHV-infected DMVEC were treated with Gleevec, a small-molecule inhibitor of c-Kit signal transduction (20). These results suggest that PMO-AS are taken up by endocytosis and diffuse to the cytoplasm where they are able to stay for prolonged periods of time. The ability of the PMO-AS to remain effective in the culture for extended periods was an important feature because approximately a week is required for the post-confluent growth of KSHV-infected cells that gives rise to multilayered foci. Moreover, the PMO-AS block target protein expression very efficiently as evidenced by the effective reduction of c-Kit surface expression specifically in PMO-AS-loaded cells (20).

Using the focus inhibition assay described above, we examined whether treatment of KSHV-infected DMVEC with PMO-AS molecules against the eight KSHV-up-regulated genes selected (Table 2) had any effect on the ability of the cells to form multilayered foci. For these assays, KSHV-infected DMVEC monolayers were cultured to 90% confluence and then loaded with each PMO-AS, or maintained as unloaded controls. Cultures were observed daily until control cultures had developed prominent multilayered aggregates of spindle cells. Typically, this took 5 to 10 days with some variation between replicate experiments. Each experiment was done at least thrice and was done with mock-infected and KSHV-infected DMVEC. As expected, mock-infected DMVEC maintained a quiescent monolayer at confluence, and PMO-AS treatment of uninfected cells occurred without any adverse effect. This observation allowed us to verify that none of the gene knockdown treatments induced nonspecific toxicity in DMVEC. The majority of tested PMO-AS had no effect on the ability of KSHV-infected DMVEC to grow post-confluence and form disorganized, multilayered foci. These results are summarized in Table 2. In contrast, treatment of KSHV-infected DMVEC with PMO-AS against RDC1 and Neuritin had a dramatic effect on phenotype equivalent to that previously seen with the c-Kit PMO-AS (Figure 1; ref. 20). Cells grew to confluence but remained strictly contact-inhibited, and cell shape resembled that of normal DMVEC as compared with the typical spindle profile of KSHV-infected cells. Cells treated with a PMO-AS against microfibrillar-associated protein (MFAP) had no effect.

Next, we wanted to confirm that PMO-AS inhibition of focus formation corresponded with reduced cellular proliferation. This was our assumption because the development of multilayered cell foci requires additional cell growth that would not occur if contact inhibition was maintained. To test this hypothesis, we employed a colorimetric cellular proliferation assay based on the conversion by metabolically active cells of the tetrazolium salt XTT to a colored formazan product. Unlike uninfected endothelial cells, when KSHV-infected cells are plated at confluence in 96-well trays, they are not contact inhibited but continue to grow. This growth can be accurately reflected by metabolism of XTT added to each well and measurement of absorbance at 570 nm 4 hours later (data not shown). For the PMO inhibition assays, KSHV-infected cells were seeded just below confluence into 96-well trays overnight, PMO-AS-loaded and allowed to grow post-confluence for an additional 72 hours with addition of XTT for the final 4 hours of culture. To verify efficient (>80%) PMO-AS uptake in the 96-well format, cells were treated with a FITC-tagged PMO-AS (FITC-PMO) and uptake visually assessed under a fluorescence microscope. This PMO-AS also served as a control for any nonspecific effect of the PMO-AS/+EPEI complex on DMVEC growth. As a positive control, cells were treated with a PMO-AS against c-Kit previously shown to inhibit foci formation by KSHV-infected DMVEC. All treatments were done in quadruplicate. As summarized in Table 2, the majority of the PMO-AS had no effect on cell growth, likely reflecting the inability of these PMO-AS to inhibit KSHV-induced focus formation. In contrast, the PMO-AS molecules that targeted c-Kit (included as a positive control), RDC1, and Neuritin all dramatically inhibited cell growth. As expected, XTT metabolism by contact-inhibited DMVEC was less than with untreated KSHV-infected DMVEC, but no deleterious effect on their metabolic activity was observed with any of the PMO-AS treatments (data not shown). Thus, the XTT assay provided a quantitative confirmation of the ability of PMO-AS against RDC1 and Neuritin to inhibit the transformed growth patterns of KSHV-infected DMVEC thus further establishing a role for RDC1 and Neuritin in KSHV tumorigenesis.

To ascertain whether PMO-AS treatment could be influencing cell phenotype through an effect on KSHV lytic replication, cells treated with PMO-AS against c-Kit, RDC1, and Neuritin were evaluated by immunofluorescent staining for expression of latent (LANA-1/ORF73) and early lytic (ORF59) viral proteins as previously described (14). All cultures maintained expression of ORF73 and, expression of ORF59 was consistently <2% of cells in culture (data not shown). Thus, PMO-AS-associated changes in culture phenotype seemed a direct result of inhibition of KSHV-induced cellular genes and were not influenced by loss of the viral genome or induction of lytic replication.

The only way to prove that PMO-AS treatment actively inhibits the translation of a given protein is to monitor a decrease in expression of the target protein, as was shown for c-Kit in our preliminary PMO-AS validation studies (14). However, for most of the genes we selected for PMO-AS knockdown studies, antibodies were not currently available. Thus, as an independent method to verify the results obtained with PMO-AS, we used RNA interference with siRNA (29). Because siRNA treatment results in the specific destruction of a target mRNA, proof for its effective action can be obtained on the RNA level. To verify that RNA interference was able to inhibit gene expression as well as transformation of DMVEC, we first examined the effect of siRNA against c-Kit using the same techniques (immunostaining to verify inhibition of protein expression; post-confluent growth and phenotype in the focus inhibition assay) successfully established for Kit-PMO-AS treatment (30). siRNA against c-Kit inhibited both c-Kit protein expression and focus formation (30). Thus, we obtained siRNAs specific for RDC1 and Neuritin and did qRT-PCR on siRNA-transfected KSHV-DMVEC to examine if siRNA treatment reduced the levels of RDC1 and Neuritin mRNA. As shown in Fig. 2A, two RDC1 siRNA molecules were tested and either a 56% or a 25% reduction of RDC1 mRNA was observed relative to control DMVEC transfected with a FITC-tagged control siRNA, or treated with transfection reagent alone. Similarly, Neuritin siRNA-treated cells yielded a 63% reduction in mRNA levels relative to controls. These results are consistent with previous observations using siRNA and measuring mRNA levels (31). Next, we determined if treatment of KSHV-infected DMVEC with siRNA against RDC1 or Neuritin inhibited focus formation. KSHV-infected DMVEC were transfected with the test siRNAs or with control siRNA. Cells were observed daily for up to 14 days until control cells had developed prominent multilayered foci as described for the PMO-AS assays. As illustrated in Fig. 2, cells transfected with either Neuritin or RDC1 siRNA maintained contact-inhibited growth through 2 weeks of post-confluent culture, whereas control cells development multicell aggregates of spindle cells. Interestingly, treatment with the RDC1 siRNA #1 allowed more stringent maintenance of contact inhibition, which correlated with the larger decrease (56%) in RDC1 mRNA levels seen with this siRNA. In conclusion, this independent antisense method confirmed that RDC1 and Neuritin are essential for KSHV-induced transformation of DMVEC.

The results above are consistent with RDC1 and Neuritin being essential for the KSHV-mediated transformation of endothelial cells. To investigate if RDC1 and/or Neuritin were also sufficient for cellular transformation, we generated stable clones of NIH 3T3 cells expressing Neuritin or RDC1 and examined their phenotype. For a control, we also generated clones obtained by transfection with the vector plasmid pcDNA3.1. In addition, we introduced the known oncogenes Ras and the KSHV chemokine receptor ORF74 into NIH 3T3 cells. To monitor expression, Neuritin and RDC1 were tagged with the influenza HA epitope. Immunofluorescence with HA-specific antibodies of representative stable transfectants indicated high levels of expression of RDC1 or Neuritin respectively (Fig. 3A). To assess the cellular morphology of transfectants, clones obtained from the stable transfectants were grown under soft agarose. NIH 3T3 cell clones expressing Neuritin exhibited neurite-like outgrowths and arborization (Fig. 3A and B). This phenotype is reminiscent of the previously reported ability of a recombinant Neuritin construct lacking a GPI anchor to enhance neurite formation in neurons upon expression in hippocampal cultures (28). Removal of the GPI anchor would lead to Neuritin being secreted rather than retained at the cell membrane. To ensure that the morphologic changes seen in Neuritin-expressing NIH 3T3 cells (Fig. 3A) were also observed with GPI-linked Neuritin, we produced a full-length Neuritin protein. Upon transfection into NIH-3T3 cells, the full-length GPI-linked Neuritin induced a branching morphology comparable to that seen with the secreted construct lacking the anchor (Fig. 3C). This type of morphology was not evident in the vector control transfectants or in cells transfected with RDC1, Ras, or KSHV ORF74 (data not shown). However, we noted that NIH 3T3 clones stably transfected with RDC1 grew faster in soft agarose than vector controls or Neuritin transfectants. To quantify this increased growth, we measured the metabolic activity of two RDC1 clones by performing an XTT assay after 48 hours. As shown in Fig. 3D, both RDC1 clones examined showed a statistically significant increase in growth compared with vector transfectants. As expected, KSHV ORF74 and Ras oncogenes also showed a significant growth advantage. In contrast, Neuritin-transfected NIH 3T3 clones showed either no change or a decrease in their metabolic activity. These observations suggested that RDC1 increased the growth rate of NIH3T3 cells, whereas Neuritin altered their morphology.

Anchorage-independent growth is a common phenotype of transformed cells. To examine anchorage-independent growth, RDC1- and Neuritin-transfected NIH 3T3 clones were grown in soft agar. As shown in Fig. 4, four independently derived RDC1 clones averaged ∼100 colonies each, whereas two different Neuritin cell lines clones gave rise to 126 and 65 colonies in the soft agar assay. This represented a 30-fold increase over NIH 3T3 clones with the vector pcDNA 3.1 alone. For comparison, a Ras oncogene–expressing cell line produced 300 colonies. Colonies obtained from Ras and RDC1 transfectants grew to a higher density and size compared with Neuritin clones. This data is graphically represented in Fig. 4A. The adjacent pictures are representative of individual colonies counted. Thus, it seems that RDC1 conferred a more robust transformed phenotype to the NIH 3T3 cells than did Neuritin. Mass cultures of NIH 3T3 cells transfected with either Neuritin or RDC1 were also assessed with the soft agar assay. As shown in Fig. 4B, mass cultures of Neuritin- or RDC1-expressing cells formed significantly more cell colonies then those expressing the vector-only pcDNA 3.1 control.

To assess the ability of RDC1 and Neuritin to induce tumor formation in vivo, NIH 3T3 clones expressing RDC1 or Neuritin were injected s.c. into the right flanks of nude mice. Two independent clones of RDC1 cells (clone 1, n = 4; clone 2, n = 5) or Neuritin cells (clone 1, n = 4; clone 2, n = 4) were used. Control NIH 3T3 cells transfected with pcDNA 3.1 vector alone were injected into the left flanks of two of the groups (n = 8). NIH 3T3 cells expressing the Ras oncogene were injected into two additional mice for a positive control. Tumor volumes were measured at weeks 3, 4, and 5 post-injection, at which time the experiment was terminated. These results are shown in Fig. 5A, whereas Fig. 5B shows an example of a tumor induced by injection of either clone of RDC1 or Neuritin at week 4. As expected, mice injected with Ras-expressing cells developed visible tumors by week 2, whereas those injected with RDC1- and Neuritin-expressing cell lines had their first visible tumors at week 3. One of eight mice injected with cells expressing the pcDNA 3.1 vector control developed a small tumor at week 4 on the left flank. All nine mice injected with RDC1 clones developed large tumors by week 4. This result confirms the tumorgenicity potential of RDC1 and supports its transforming role in the KSHV infected cells. All four mice injected with one of the Neuritin clones (clone 2) developed tumors, whereas only one of four inoculated with the other clone developed a tumor starting at week 4 with obvious growth by week 5 when the experiment was terminated. The presence of RDC1 and Neuritin RNA was confirmed by qRT-PCR from excised tumors (data not shown). The Neuritin-induced tumors were generally smaller than the RDC1-induced tumors (Fig 5A and B). This and the fact that not all injections of Neuritin-expressing cells resulted in tumors by 5 weeks suggest that the tumorgenic potential of Neuritin in vivo is lower than that of RDC1.

Our data suggest that RDC1 and Neuritin might be important for the development of Kaposi's sarcoma tumors. To examine if RDC1 and Neuritin are present in Kaposi's sarcoma tumors, we measured their transcript levels in two tumor samples by quantitative PCR. Samples were obtained with informed consent by punch biopsy of lesions from Kaposi's sarcoma patients. RNA was extracted by mechanically separating the tissues and lysing the cells in RNA isolation buffer. Both samples showed levels of RDC1 and Neuritin that were comparable to KSHV-infected DMVEC but higher than uninfected DMVEC (Fig. 5C). Therefore, we conclude that both Neuritin and RDC1 are well expressed in Kaposi's sarcoma tumors and could thus be therapeutic targets.

We have examined the role of cellular genes in transformation induced by KSHV in an in vitro endothelial cell culture system, which recapitulates many of the features of Kaposi's sarcoma. Other cell systems are also available for the in vitro study of KSHV biology, each with its own characteristics and particular strengths (32–34). Compared with systems that rely on primary endothelial cells, the advantage of using life-extended DMVEC is that parallel cultures can be easily maintained in the absence of KSHV. Moreover, genes that are involved in general mitosis are likely to be induced already in this culture system so that the transforming process might be more visible in DNA arrays. Another major advantage of this system is that the morphologic changes that are imposed by KSHV are very obvious and easy to monitor. Cells change to a spindle cell phenotype and pile up to form multilayered cell foci when confluent monolayers are not passaged but allowed to grow post-confluence. None of these changes occurs in uninfected cells, which maintain a cobblestone appearance and grow as a contact-inhibited single cell monolayer that becomes quiescent at confluence. Importantly, the morphologic changes characteristic of KSHV-infected DMVEC require the presence of certain cellular pathways as previously exemplified with the proto-oncogene c-Kit (15). The relative ease with which the inhibition of the development of this transformed phenotype can be monitored makes this system amenable to large-scale functional genomics approaches.

DNA microarrays were used to identify transcripts that are up-regulated in KSHV-infected cells. We selected our gene products from three different microarray experiments based on two different kinds of cDNA arrays (15, 22), as well as Affymetrix arrays (this study). In these three studies, a total of >25, 000 probes were used to interrogate KSHV-infected DMVEC. However, considerable overlap exists between the different array platforms, meaning that the number of individual genes tested will be lower. These DNA microarray studies were used as a prescreening step to select potentially important candidates for KSHV-mediated oncogenesis. We further focused on up-regulated transcripts because it is easier to inhibit gene function on a large scale than to force gene expression. From the total of several hundred up-regulated genes in all three experiments, we further selected those that were up-regulated in several independent experiments. Seven genes with known function were selected based on inference that they could play a potential role in transformation; one gene with unknown function was also selected.

A two-step antisense approach was applied to examine the role of the selected genes in KSHV-mediated transformation of DMVEC. PMO-AS were used in the first screen and of the eight tested, two had a dramatic effect on inhibition of KSHV-associated focus formation. Whereas the PMO-AS molecules have several advantages, as discussed above, they have the disadvantage that it is not possible to verify their functionality at the level of RNA. Thus, we cannot exclude the possibility that the lack of effect observed with some of the other PMOs could have been due to our inability to inhibit the translation of the targeted RNA. However, previous experiments in these cells showed a very dramatic reduction of protein levels by PMO-AS suggesting that PMOs work very efficiently in this cell system (20). Thus, it seems more likely that the reason for the lack of activity for most of the PMO-AS molecules is that the corresponding gene products do not play an essential role in KSHV-induced transformation. In contrast, knockdown of two genes, RDC1 and Neuritin, showed a clear inhibition of the transformed phenotype, whereas having no deleterious effect on the phenotype or viability of uninfected DMVEC. The fact that this phenotype inhibition was observed with two completely different methods of gene silencing also rules out the possibility that it was caused by an unrelated effect of either the siRNA or the PMO-AS. Moreover, the degree of inhibition was dependent on the efficiency of each siRNA in reducing the level of the corresponding mRNA (see Fig. 2). Thus, we conclude that RDC1 and Neuritin are essential components of the KSHV-induced transformation of DMVEC.

Given our previous observation that c-Kit is not only essential but also sufficient for spindle cell formation of DMVEC, it is possible that RDC1 or Neuritin are involved in the up-regulation of c-Kit, are induced by c-Kit, or otherwise intersect with the Kit-associated pathway. However, preliminary microarray profiling experiments of DMVEC transduced with an adenovirus expressing c-Kit suggest that neither RDC1 nor Neuritin are induced, at least by enforced expression of c-Kit. In contrast, two of the other genes tested, LMO-2 and Osteopontin, were induced by adenovirus-expressed c-Kit.⁸

Experiments to examine if Neuritin or RDC1 induce c-Kit are currently under way.

Both RDC1 and Neuritin are relatively unknown proteins. Neuritin, also known as candidate plasticity gene 15 (CPG15), was discovered in differential screens for genes regulated in the adult hippocampus during neural stimulation (28, 35). Neuritin is a small 145–amino acid, GPI-anchored protein with a signal sequence. A recombinant secreted version of Neuritin, similar to the one used in our studies, has been shown to promote neurite outgrowth and arborization in neural cultures (28). The X. laevis homologue of Neuritin represents a growth-promoting protein that seems to regulate spatial and temporal control of neuronal structure (35). Thus, the induction of Neuritin has to date only been associated with nervous system development and intercellular signaling, which promotes synaptic maturation and axon arborization (36). In our system, Neuritin changed the morphology of NIH 3T3 cells and supported their anchorage-independent growth but did not seem to increase proliferation. We observed morphologic changes upon expressing either a truncated, secreted, or full-length version of Neuritin in NIH 3T3 cells, suggesting that GPI-linked Neuritin might give rise to a secreted version via phospholipase cleavage as observed for many GPI-linked proteins (37). Further studies will help to clarify this question. These data suggest that the previously observed function of Neuritin in regulating cellular structure is also applicable in nonneuronal cells. In addition, the expression of Neuritin in NIH 3T3 cells in nude mice promoted tumor formation. Overexpression of Neuritin in cancer cells might thus contribute to changes in cell morphology during oncogenesis, which may contribute to tumor formation via pathways that have yet to be elucidated.

The orphan G protein–coupled receptor RDC1 was by far the strongest KSHV-induced cellular gene in our studies. The ligand for RDC1 is unknown. Previously, work identifying the hormones calcitonin-related peptide or adrenomedullin as putative ligands (38) are no longer considered valid; thus, RDC1 is still considered an orphan receptor (39, 40). Sequence homology and a genomic localization in close proximity to the CXC chemokine receptors CXCR2 and CXCR4 strongly suggests that RDC1 might be a CXC chemokine receptor (41–44). Similar to other chemokine receptors, RDC1 was also shown to act as a coreceptor for HIV during in vitro transfection in the NP/CD4 cell line (45). Interestingly, the ORF74 gene of KSHV encodes a chemokine receptor (46). This receptor displays the highest homology to the cellular interleukin 8 receptor but also to other chemokine receptors including RDC1 (47). Moreover, ORF74 transforms NIH 3T3 cells (46). It is thought that this transformation occurs independent of ligand because ORF74 can signal constitutively without added ligand (47). However, constitutive signaling can be increased or decreased by adding exogenous chemokines (48). Transformation by persistently activated GPCRs has also been observed in other systems (49, 50). Thus, it is tempting to speculate that RDC1 also has the capacity to transform cells independently of ligand. Experiments to address this question are currently under way. ORF74 can induce endothelial cell transformation via autocrine vascular endothelial growth factor receptor-2/KDR activation (51) and induce Kaposi's sarcoma lesions in avian leukosis virus (ALV) receptor transgenic mice infected with a recombinant ALV expressing ORF74 (52). Whether or not RDC1 does similar functions remains to be shown. It is also possible that the two GPCRs reciprocally augment their oncogenic function. For instance, it is conceivable that RDC1 does some of the functions of the viral GPCR during the latent phase of infection because the viral GPCR is expressed during the lytic cycle of the virus. The remarkable tumor forming ability exhibited by RDC1-expressing NIH 3T3 cells in nude mice strongly supports a role for this gene in KSHV-associated tumorigenesis. The mechanism and molecular pathways by which KSHV induces RDC1 as well as Neuritin are currently not known. RDC1 is induced by hypoxia (53) and by tumor necrosis factor-α (54).⁹

Thus, it could be that RDC1 is up-regulated in the context of a proinflammatory response that is triggered by the latent transcript vFLIP (ORF71) of KSHV (55). In addition to playing a role in transformation, RDC1 might thus also contribute to the chronic inflammation of Kaposi's sarcoma tissue.

The discovery of genes that are essential for KSHV-mediated transformation could be useful in developing novel treatments for Kaposi's sarcoma. Proof-of-principle for this is the recent observation that Gleevec might be useful to treat Kaposi's sarcoma (16). Both Neuritin and RDC1 are likely to act at the cell surface, which might facilitate their targeting. Moreover, RDC1 falls into one of the most commonly targeted group of receptors because many currently available drug targets are GPCRs (56). Furthermore, because the KSHV-induced c-Kit is a well-known oncogene that is involved in many different cancers, it is also conceivable that both Neuritin and RDC1 will likewise be involved in the development or progression of other human cancers.

Grant support: PS1-RR00163 and RO1-CA099906 (A.V. Moses and K. Früh) and Virogenomics, Inc. (a company that may have a commercial interest in the results of this research).

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.

We thank the Affymetrix core of the Gene Microarray Shared Resource for excellent services, the center for Biostatistics Computing for providing data storage and retrieval, and Jeff King (Virogenomics) for help with the data analysis.