Efficacy of bortezomib in a direct xenograft model of primary effusion lymphoma

Case Report.

An 86-y-old HIV-negative white male with extensive cardiovascular disease and renal failure presented with weight loss, shortness of breath, and a right pleural effusion. A pleural tap led to a diagnosis of B-cell PEL based on the presence of atypical lymphoid cells that stained positive for CD45, CD30, CD38, CD138, HLA-DR, HHV-8 (LANA1), and Epstein–Barr encoded RNA (EBER) (Figs. S1 and S2 and Table S1). CD3, CD19, CD20, and CD79a staining was negative. A monoclonal rearrangement of the B-cell receptor and a complex karyotype (Fig. S1B) confirmed the diagnosis. The patient received doxorubicin (Adriamycin) and bortezomib therapy and demonstrated a marked decrease in pleural effusion. However, the patient developed neutropenic sepsis and died within 30 d of diagnosis.

Development of the UM-PEL-1 in Vivo Model.

PEL cells (2 × 10⁷) obtained from the diagnostic tap of the aforementioned patient were injected into the peritoneal cavities of immunodeficient NOD/SCID mice immediately after retrieval. Injected mice developed peritoneal swelling within 5–7 d, whereas mock-injected mice did not (Fig. 1A). Mice injected with PEL cells and then untreated became moribund because of disease within 2–3 wk. The in vivo PEL model could be propagated via collection of PEL cells from ascites (referred to as “UM-PEL-1 cells”) and injection into peritoneal cavities of naïve mice with a 100% engraftment rate (72 of 72 injected mice). Cell-surface antigen expression was monitored in sequential xenografts to ensure maintenance of PEL phenotype.

Histological Findings.

Peritoneal effusions as well as solid lymphomas were evident in mice at postmortem analysis, and tumor nodules were found on peritoneal or serosal surfaces or in association with the omentum (Fig. 1B). Some tumors infiltrated the parenchyma of the pancreas or bowel wall smooth muscle (Fig. 1C). No involvement of the spleen, bone marrow, lymph node, lung, liver, or kidney was seen. Tumor cells were medium to large in size with pleomorphic nuclear outlines, prominent nucleoli, moderate to abundant cytoplasm, and an immunophenotype identical to the primary PEL cells (Fig. 1 D–J). PCR analysis showed a monoclonal V_H4–34 IgG heavy chain with 83% homology to the germ line and evidence of antigen selection. The UM-PEL-1 karyotype depicted in Fig. S1B was identical to the karyotype of the primary cells and demonstrated 51, XY, dup(1)(q22q31), +4, +5, +7, +19, add(21)(p13). UM-PEL-1 cells derived from mice ascites also could grow and be propagated in vitro; these cells were designated “UM-PEL-1c.”

UM-PEL-1 Cells Are Infected with KSHV.

Immunofluorescence antibody analysis for KSHV LANA showed that UM-PEL-1 cells freshly isolated from mouse ascites or solid tumors exhibited the classic punctate pattern indicative of episomal KSHV (Figs. 2A and 3A). In addition, cells and tumors uniformly expressed the latent gene Kaposin (ORF K12). A fraction of UM-PEL-1 cells expressed the late lytic viral gene K8.1, which generally is the consequence of spontaneous lytic reactivation (Fig. 3A). Similar to other PEL cell lines, UM-PEL-1 can be induced lytically by the addition of the histone deacetylase inhibitor trichostatin A (TSA), causing a significant increase in the expression of the lytic transactivator gene RTA and K8.1 (Fig. 2 B and C). Cell-free TSA-induced supernatants were able to transfer KSHV to Vero cells, indicating that UM-PEL-1 cells bear infectious KSHV virions (Fig. 2 D and E). Comparative analysis of UM-PEL-1c cultured cells and UM-PEL-1 ascites and solid tumors carried out by real-time quantitative PCR (qPCR) showed variations in KSHV latent and lytic gene expression. UM-PEL-1c cells express more latent and lytic genes in vitro, whereas ascites and tumors show intermediate levels of expression (Fig. 3B). Relative viral load measurements showed the same trend (Fig. 3C). Taken together, these data further suggest that different microenvironments (in vitro versus in vivo) affect KSHV copy number and expression patterns.

Bortezomib Induces Apoptosis of UM-PEL-1 Cells.

We set out to use the UM-PEL-1 cells to compare the efficacies of established and proposed PEL therapies. We therefore evaluated the anti-PEL activity of bortezomib and doxorubicin. Direct comparisons of these therapies were conducted on UM-PEL-1c cells that were collected from mice and propagated in vitro in standard culture conditions. These cells were treated with the chemotherapies for up to 72 h, and cell viability was assessed by flow cytometric analysis. Although treatment with doxorubicin as well as bortezomib reduced UM-PEL-1c cell viability, the latter induced higher levels of apoptosis at earlier time points at clinically achievable concentrations (Fig. 4A and Fig. S3). Induction of apoptosis with either treatment was dose and time dependent, and a majority of UM-PEL-1c cells were sensitive to either treatment.

Bortezomib Enhances Survival of Mice Bearing PEL Xenografts.

Although bortezomib induced apoptosis more potently than doxorubicin in UM-PEL-1c cells in vitro, in vivo studies were necessary to compare the efficacies of the two chemotherapies because of the substantial changes in the KSHV and host transcriptome during ex vivo culture. To this end, UM-PEL-1 cells were used to establish lymphomatous effusions in the peritoneal cavities of NOD/SCID mice. On the third day after injection of UM-PEL-1 cells, mice began receiving twice-weekly i.p. injections of PBS, bortezomib, or doxorubicin for 3 wk. Mice were monitored daily. In vivo treatment of the UM-PEL-1–bearing mice with bortezomib (0.3 mg/kg) was well tolerated and improved mouse overall survival compared with control animals treated with PBS (P < 0.01; median survival of bortezomib-treated mice was 32 d, compared with 15 d in controls) (Fig. 4B). I.p. doxorubicin injections (1.25 mg/kg) at the same treatment intervals resulted in a median survival of 24 d, significantly inferior to median survival with bortezomib (P = 0.035). Because viral biology is intricately linked to PEL survival, we interrogated the effects of the therapies on KSHV gene expression. We found that in vivo both bortezomib and doxorubicin led to an increase in the overall KSHV gene expression, including up-regulation of the lytic transactivator RTA, early lytic genes ORF49, ORF57, and vGPCR and late lytic genes K8.1 and glycoprotein B (gB) (Fig. 4 C and D). Although both drug treatments led to an induction of KSHV lytic replication, they did so in a temporally distinct manner. Specifically, lytic reactivation was evident at 1 d after bortezomib treatment and 7 d after doxorubicin treatment.

Bortezomib Does Not Affect Constitutive Activation of NF-κB in UM-PEL-1 Cells.

The potent anti-PEL activity of bortezomib prompted us to examine several potential mechanisms for this effect. Bortezomib has been shown to induce apoptosis by several mechanisms, one being the inhibition of the NF-κB prosurvival transcription factor (21–24). Previous reports have shown that PELs may exhibit constitutive activation of NF-κB that is necessary for their survival (17, 25). We therefore examined the status of NF-κB activation in UM-PEL-1 cells freshly obtained from mice ascites. NF-κB EMSA showed intranuclear p50 and p65 proteins indicative of constitutive activation of NF-κB. The activation of this transcription factor could be inhibited by in vitro treatment with the selective NF-κB inhibitor Bay-11 (Fig. 5A). Treatment with bortezomib (20 μM), however, was not effective in inhibiting constitutive NF-κB activity, as evidenced by the continued presence of intranuclear p50 and p65 proteins.

Bortezomib Treatment Does Not Induce a Terminal Unfolded Protein Response.

PELs, which exhibit a plasma cell-like gene expression signature, have been shown to constitutively exhibit a partially activated unfolded protein response (UPR) (26). The source of the activated UPR is unknown, because these lymphomas, unlike plasma cells, do not express Ig protein (27). It has been shown that apoptosis and cell death can be induced in cells exhibiting an activated UPR by inhibiting the 26S proteasome (21). We therefore examined the expression of several UPR-associated proteins in UM-PEL-1 cells after treatment with bortezomib. GRP78 is an endoplasmic reticulum (ER) molecular chaperone that is up-regulated during a physiological, or nonterminal, UPR (28). We observed constitutive expression of GRP78 in UM-PEL-1 cells, indicative of an activated UPR, but no further induction of protein levels after bortezomib treatment (Fig. 5B).

Cells undergoing extensive ER stress frequently block de novo protein synthesis via phosphorylation of eukaryotic initiation factor 2α (eIF-2α) (29). Bortezomib treatment did not result in enhanced phosphorylation of eIF-2α (Fig. 5B), suggesting that inhibiting the 26S proteasome in PEL did not significantly enhance ER stress. We did observe an increase in total eIF-2α protein levels in cells treated with 20 nM bortezomib for 12 h; this transient effect may have been caused by inhibition of eIF-2α degradation by the 26S proteasome.

Activating transcription factor 4 (ATF4) frequently is associated with terminal UPR induction which up-regulates expression of proapoptotic GADD153/CHOP (30). ATF4 levels remained constant during bortezomib treatment (Fig. 5B), further suggesting that this drug does not induce apoptosis by activating a terminal UPR in UM-PEL-1 cells.

In Vivo Bortezomib Treatment Down-Regulates Expression of Genes Regulated by c-myc and E2F3 as well as Cell-Cycle Progression and DNA Replication Genes.

To examine further the mechanism of the bortezomib-induced apoptosis, we carried out gene-expression profiling of UM-PEL-1 cells. The microarray analysis was conducted on mice bearing UM-PEL-1 xenografts after treatment for 24 h with PBS or bortezomib. We used Gene Set Enrichment Analysis (GSEA) to identify gene sets that were concordantly induced or repressed by treatment with bortezomib, compared with treatment with PBS (Fig. 5C and Table S2). Bortezomib significantly repressed genes involved in cell-cycle progression, DNA replication, and response to DNA damage stimuli. [All false discovery rates (FDRs) < 0.001.] Furthermore, genes induced by Myc or E2F3 forced expression, and genes that are bound by Myc in human B cells as determined by ChIP paired-end ditags (ChIP-PET) (31, 32) also were repressed by bortezomib. A total of 1,248 individual known genes were repressed at least 2-fold (mean, 3.5-fold; maximum, 41-fold) in bortezomib- versus PBS-treated cells, whereas 502 were up-regulated (mean, 2.8-fold; maximum, 34-fold). Many of these genes were modulated similarly by bortezomib treatment of Adult T-cell Leukemia (ATL) cells (Fig. S4) (33). The correlation between expression levels of these genes in our experiment compared with bortezomib-treated ATL cells was 0.61 (P < 0.0001). Analysis of the direction of changes in gene expression in the experiments also showed strong concordance (P < 0.0001 by χ²-test).

As mentioned, targets of the Myc transcription factor were shown to be down-regulated after 24 h of bortezomib treatment in the gene-expression analysis. In agreement with the array results, we observed down-regulation of Myc protein levels after 24 h of bortezomib treatment (Fig. 5B). Interestingly, bortezomib treatment led to a transient induction in Myc levels at 12 h before subsequent down-regulation.

We also examined the effects of bortezomib on the phosphorylation status of Akt and JNK. No activation of JNK and Akt was observed in untreated or bortezomib-treated UM-PEL-1 cells (Fig. S5).

Reagents.

Bortezomib was obtained from Millennium Pharmaceuticals, and doxorubicin was purchased from Sigma-Aldrich. P-eIF2α, eIF2α, GRP78, pAkt, and Myc antibodies were purchased from Cell Signaling Technologies. ATF4, pJNK, and actin antibodies were purchased from Santa Cruz Biotechnology. CD3, CD19, CD20, CD30, CD38, CD45, CD79a, CD138, and HLA-DR antibodies were purchased from Beckman Coulter.

UM-PEL-1 Cells.

Pleural effusion collected during the diagnostic tap of the patient with PEL was centrifuged over Ficoll/Paque gradient following the manufacturer's instructions (GE Healthcare) and was resuspended in effusion fluid. PEL cells (2 × 10⁷) were injected i.p. into 6- to 8-wk-old female NOD/SCID mice (Jackson Laboratory). Mice were monitored daily for tumor growth. After 5–7 d, mice developed peritoneal swelling and were killed. A portion of the ascites was removed immediately postmortem via paracentesis; these cells are referred to as “UM-PEL-1” cells. For propagation, UM-PEL-1 cells were injected into the peritoneal cavities of additional NOD/SCID mice without ex vivo culture.

Cell Lines.

For in vitro studies, UM-PEL-1 cells collected from mice were cultured in RPMI medium 1640 (Mediatech,) supplemented with 10% FBS (Mediatech), 2 nM glutamine (Gibco BRL), and penicillin/streptomycin (Gibco BRL). Cells cultured in vitro are referred to as “UM-PEL-1c” cells. The KSHV-positive, EBV-negative BC-3 PEL cell line derived from an HIV-negative patient (45) also was cultured in the supplemented RPMI medium 1640. 293T cells were cultured in DMEM supplemented with 10% FBS, 2 nM glutamine, and penicillin/streptomycin.

NF-κB EMSA and Apoptosis Studies.

UM-PEL-1 cells were isolated from peritoneal cavities of mice harboring lymphomatous effusions by paracentesis, were placed into culture, and were treated with either Bay-11 or bortezomib. Cells (5 × 10⁶) were collected at the indicated time points and used for nuclear protein extraction and EMSA supershift analyses as described previously (46) and summarized briefly in SI Materials and Methods.

For apoptosis studies, 10⁵ cells/mL were incubated with or without bortezomib or doxorubicin for specified time periods, collected, washed with 1× PBS, and stained with propidium iodide (PI) (Invitrogen) and YO-PRO (Invitrogen) following the manufacturer's instructions. Flow cytometric analysis was performed on a Becton-Dickinson LSR analyzer (BD Biosciences).

In Vivo Tumor Studies.

UM-PEL-1 cells (2 × 10⁷) isolated from ascites of tumor-bearing mice were resuspended in 250 μL ascites fluid and injected i.p. into additional NOD/SCID mice. Mice were assigned randomly to a treatment group and on day 3 began receiving i.p. injections of PBS, bortezomib (0.3 mg/kg), or doxorubicin (1.25 mg/kg). Mice were monitored daily and were killed when moribund or exhibiting signs of discomfort in accordance with institutional guidelines. All procedures with animals were conducted in conformity with an approved institutional animal protocol.

RNA Isolation and Real-Time Quantitative PCR.

RNA was isolated with the RNeasy Plus kit (QIAGEN) and DNase treatment on columns. We transcribed 500 ng of RNA into cDNA using the Reverse Transcription System (Promega) and Random Primers (Promega) according to the manufacturer's instructions. Real-time quantitative PCR (qPCR) was performed using an ABI Prism 7000 Sequence Detection System (Applied Biosystems) with Sybr Green PCR Master Mix (New England Biolabs). Sequences of primer sets are given in SI Materials and Methods. Melting curve analysis was performed to verify specificity of products as well as water and RT controls in every run. Data were analyzed using the △C_t method as in Mutlu et al. (47). Target gene expression was normalized to GAPDH by taking the difference between Ct values for target genes and GAPDH (△△ Ct value). These values then were calibrated to that of the control sample to give the Ct value. The fold target gene expression is given by the formula: 2^–△△C_t.

Analysis of IgG V_H gene rearrangement was done as previously reported (48).

Vero Cell Infections.

UM-PEL-1 cells were induced with TSA at 40 nM for 48 h and then at 400 nM for 24 h. Supernatants were centrifuged three times at 500 × g for 10 min at 4 °C. One milliliter of supernatant was added to 2 × 10⁵ Vero cells and incubated for 24 h. The medium was removed, and cells were washed three times with Dulbecco's PBS (Gibco BRL) and incubated for 72 h. LANA staining was performed as described below.

Immunofluorescence Staining, Western Blot Analysis, and Microarray Hybridization.

Immunostaining was performed as previously described by Mutlu et al. (47). Whole-cell extracts were prepared by lysing 5 × 10⁶ cells, and Western blot analysis was performed as previously described (49).

Microarray hybridization was performed using the Affymetrix Human Gene 1.0ST arrays. A brief description of the procedure and analysis is provided in SI Materials and Methods.

Statistical Analysis.

Mice survival curves were estimated using the Kaplan–Meier product-limit method and were compared using the log-rank test. P values <0.05 were considered statistically significant.