Pralatrexate Is Synergistic with the Proteasome Inhibitor Bortezomib in in Vitro and in Vivo Models of T-Cell Lymphoid Malignancies

Published OnlineFirst May 25, 2010; DOI: 10.1158/1078-0432.CCR-10-0671 Published OnlineFirst on July 13, 2010 as 10.1158/1078-0432.CCR-10-0671

Cancer Therapy: Preclinical Clinical Cancer Research Pralatrexate Is Synergistic with the Proteasome Inhibitor Bortezomib in In vitro and In vivo Models of T-Cell Lymphoid Malignancies

Enrica Marchi1, Luca Paoluzzi1, Luigi Scotto1, Venkatraman E. Seshan2, Jasmine M. Zain1, Pier Luigi Zinzani3, and Owen A. O'Connor1

Abstract Purpose: Pralatrexate (10-propargyl-10-deazaaminopterin) is an antifolate with improved cellular uptake and retention due to greater affinity for the reduced folate carrier (RFC-1) and folyl-polyglutamyl synthase. Based on the PROPEL data, pralatrexate was the first drug approved for patients with relapsed and refractory peripheral T-cell lymphoma. Bortezomib is a proteasome inhibitor that has shown some activity in patients with T-cell lymphoma. Experimental Design: Assays for cytotoxicity including mathematical analysis for synergism, flow cytometry, immunoblotting, and a xenograft severe combined immunodeficient-beige mouse model were used to explore the in vitro and in vivo activities of pralatrexate alone and in combination with bortezomib in T-cell lymphoid malignancies. Results: In vitro, pralatrexate and bortezomib exhibited concentration- and time-dependent cytotoxicity against a broad panel of T-lymphoma cell lines. Pralatrexate showed synergism when combined with bortezomib in all cell lines studied. Pralatrexate also induced potent apoptosis and caspase activation when combined with bortezomib across the panel. Cytotoxicity studies on normal peripheral blood mononuclear cells showed that the combination was not more toxic than the single agents. Western blot assays for proteins involved in broad growth and survival pathways showed that p27, NOXA, HH3, and RFC-1 were all significantly modulated by the combination. In a severe combined immunodeficient-beige mouse model of transformed cutaneous T-cell lymphoma, the addition of pralatrexate to bortezomib enhanced efficacy compared with either drug alone. Conclusion: Collectively, these data suggest that pralatrexate in combination with bortezomib repre- sents a novel and potentially important platform for the treatment of T-cell malignancies. Clin Cancer Res; 16(14); 3648–58. ©2010 AACR.

Malignancies derived from mature (post-thymic) T cells grams are still the most commonly used regimens despite and natural killer cells, collectively referred to as peripher- the suboptimal outcomes. Overall, CHOP-based chemo- al T-cell lymphomas (PTCL), encompass a variety of rare therapies achieve overall response rates of 30% to 60% and often challenging diseases. PTCLs represent 10% to with an overall survival at 5 years of approximately 15% 15% of all non–Hodgkin's lymphomas worldwide (1), to 20%. Efforts to improve on these approaches with more accounting for 6,000 to 9,000 cases annually in the United dose-intense combination chemotherapy regimens (2, 3) States. CHOP (cyclophosphamide, doxorubicin, vincris- have failed to show a benefit (62% versus 56%, respec- tine, and prednisone) and CHOP-like chemotherapy pro- tively). Given the often dismal results seen in patients with PTCL compared with other subtypes of non–Hodgkin's lymphoma, there is an urgent need to identify new agents Authors' Affiliations: 1New York University Cancer Institute, NYU Langone Medical Center; 2Department of Epidemiology and with demonstrable activity in these T-cell neoplasms. Over Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New the past few years, a number of promising new drugs have York; and 3Institute of Hematology and Medical Oncology “L. & A. emerged, which seem to have marked single-agent activity Serágnoli,” University of Bologna, Bologna, Italy in T-cell lymphomas, including gemcitabine (4–7), alemtu- Note: Supplementary data for this article are available at Clinical Cancer zumab (8, 9), and the histone deacetylase (HDAC) inhibi- Research Online (http://clincancerres.aacrjournals.org/). tors (10). More recently, pralatrexate, a novel antifol with Corresponding Author: Owen A. O'Connor, Division of Hematologic Malignancies and Medical Oncology, New York University Cancer Insti- improved cellular uptake and retention, has become the tute, NYU Langone Medical Center, New York, NY 10016. Phone: 212- first drug approved by the U.S. Food and Drug Administra- 263-9884; Fax: 212-263-9210; E-mail: owen.o'[email protected]. tion for the treatment of relapsed/refractory PTCL. doi: 10.1158/1078-0432.CCR-10-0671 The results of the PROPEL registration study based on ©2010 American Association for Cancer Research. 109 evaluable patients showed an overall response rate

3648 Clin Cancer Res; 16(14) July 15, 2010

Pralatrexate-Based Combination in T-Cell Malignancies

of 29% in patients with very heavily treated disease.4 and CEM (18) are T-acute lymphoblastic leukemia (T- Subgroup analyses have shown that the activity of pra- ALL) lines resistant to γ-secretase inhibitors, and KOKT- latrexate was equivalent irrespective of the amount of K1 (19), DND-41, and HPB-ALL (20) are T-ALL lines prior therapy, age, prior autologous stem cell transplant, sensitive to γ-secretase inhibitors. All leukemic cell lines or subtype of PTCL. Pralatrexate belongs to the class of were provided by the laboratory of Dr. Adolfo Ferrando folate analogues known as 10-deazaaminopterins, which (Columbia University, New York, NY). Peripheral blood have shown markedly superior antitumor effects com- mononuclear cells (PBMC) from healthy donors were pared with methotrexate in severe combined immunode- purchased from Allcells. All cell lines were grown as ficient (SCID)-beige xenograft models of lymphoma (11, previously described (21). 12). Pralatrexate was designed to have greater affinity than methotrexate for the natural folate and antifolate Materials principal transporters reduced folate carrier-1 (RFC-1; All reagents for Western blotting were obtained from an oncofetal protein) and folyl-polyglutamyl synthase, Bio-Rad Laboratories and Pierce Biotechnology, Inc.; leading to enhanced intracellular accumulation and DMSO was obtained from Sigma. Pralatrexate was from polyglutamylation in tumor cells. Allos Therapeutics, and bortezomib from Millennium. Recent preclinical studies by our group have established that pralatrexate synergizes with other agents active in Cytotoxicity assays PTCL, including gemcitabine. These preclinical data (13) For all in vitro assays, cells were counted and resus- have been confirmed in a phase I clinical trial exploring pended at an approximate concentration of 3 × 105 per the schedule-dependent activity of pralatrexate and gemci- well in a 96-well plate (Becton Dickinson Labware) and tabine in patients with non–Hodgkin's lymphoma. In ad- incubated at 37°C in a 5% CO2 humidified incubator dition to PTCL, pralatrexate seems to be very active in for up to 72 hours. Pralatrexate was added at concentra- patients with drug-resistant cutaneous T-cell lymphoma tions from 100 pmol/L up to 200 μmol/L, whereas borte- (CTCL), producing an overall response rate of 55% in pa- zomib was tested at concentrations from 1 to 100 nmol/L tients who received a median of 6 prior regimens (range, to determine growth inhibition curves for all cell lines. In 1-25; ref. 14). Collectively, these data suggest that prala- combination experiments, pralatrexate was added at con- trexate exhibits marked activity across different subtypes centrations of 2 to 5.5 nmol/L, and bortezomib at 3 to of T-cell neoplasms and could combine favorably with a 6 nmol/L. These concentrations were selected to approxi- host of other agents known to be active in T-cell lympho- mate the IC25-50 (inhibitory concentration of 25-50% cells ma. The observation that bortezomib has now been for each drug) for up to 72 hours. Following incubation at shown to exhibit significant activity across both CTCL 37°C in a 5% CO2 humidified incubator, 100 μLfrom and PTCL raises the interesting prospect that these agents each well were transferred to a 96-well opaque-walled could also be combined in an effort to define a new plat- plate; CellTiter-Glo reagent (Promega Corporation) was form for these diseases that is not CHOP dependent. For used according to the manufacturer's instruction. The example, Zinzani et al. (15) have reported that bortezo- plates were allowed to incubate at room temperature for mib produces an ORR of 67% in patients with relapsed 10 minutes before recording luminescence with a Synergy or refractory CTCL and PTCL, including a complete re- HT Multi-Detection Microplate Reader (Biotek Instru- sponse rate of 17%. ments, Inc.). Each experiment was done in triplicate and Given the relatively poor prognosis of patients with T- repeated at least twice. cell malignancies and the relatively poor outcomes asso- ciated with conventional chemotherapy regimens such as Flow cytometry for apoptosis CHOP, there is now a strong rationale to begin to com- H9, HH, P12, and PF382 cells were seeded at a density of bine and integrate these targeted agents into novel com- 3×105/mL and incubated with pralatrexate (2-5.5 nmol/L) bination treatment regimens for patients with T-cell and bortezomib (2-6 nmol/L) alone or in combination for malignancies that are not CHOP based. Based on this 48 or 72 hours. A minimum of 1 × 105 events were acquired rationale and the available clinical data, we sought to from each sample. To quantitate apoptosis, Yo-Pro-1 and explore the preclinical activity of pralatrexate and borte- propidium iodide (PI) were used (Vybrant apoptosis assay zomib in models of T-cell lymphoma. kit #4, Invitrogen) according to the manufacturer's instruction. The fluorescence signals acquired by a FACSCalibur Materials and Methods System were resolved by detection in the conventional FL1 and FL3 channels. Cells were considered early apopto- Cells and cell lines tic if Yo-Pro-1 positive but PI negative, late apoptotic if Yo- H9 (16) and HH (17) are CTCL cell lines, obtained Pro-1 and PI positive, and dead if only PI positive. from the American Type Culture Collection. P12, PF382, Caspase activation assays H9 cells were seeded at a density of 1 × 106/mL and in- 4 O'Connor OA, Pro B, Pinter-Brown L, et al. Pralatrexate in patients with relapsed or refractory peripheral T-cell lymphoma (PTCL). Results from the cubated with pralatrexate and bortezomib alone or in pivotal PROPEL study. JCO, submitted March 2010. combination at 4 to 6 nmol/L. After incubation for up

www.aacrjournals.org Clin Cancer Res; 16(14) July 15, 2010 3649

Marchi et al.

to 24 hours, aliquots of 300 μL were obtained from each Statistical analysis sample and transferred into Eppendorf tubes. One micro- For the different in vitro experimental groups, permuta- liter of FITC-IETD-FMK (CaspGLOW Fluorescin Active tion tests were done to determine whether any of the ex- Caspase-8, Biovision) and 1 μLofRed-LEHD-FMK perimental groups was superior to a control group. The (CaspaseGLOW red Active Caspase-9, Biovision) were analysis entailed comparing groups based on repetitions added to each tube. A negative control, the caspase inhi- (typically 3) using ANOVA after a normalizing transforma- bitor Z-VAD-FMK (1 μL/mL), and a positive control, eto- tion. Because multiple hypotheses were simultaneously poside (100 nmol/L), were also evaluated in each tested, all P values were adjusted using the Dunnett meth- experiment. The protocol followed the manufacturer's in- od (22). For each cell line, the IC50 (refs. 23, 24) and the struction. All data from flow cytometry were analyzed by drug-drug interaction in terms of synergism, additivity, or FlowJo software. Each experiment was done at least in du- antagonism were computed using the Calcusyn software plicate and repeated at least twice. Data are presented as (Biosoft; ref. 25) for the cytotoxicity assay. A combination average ± SD. index (CI) of <1 defines synergism, CI = 1 additivity, and CI > 1 antagonism. For the apoptosis data, the drug-drug Western blot analysis interactions were computed using the relative risk ratio Cells were incubated with the same concentrations of (RRR) analysis (GraphPad; http://www.graphpad.com) pralatrexate and/or bortezomib used in the apoptosis with RRR < 1 defining synergism, RRR = 1 additivity, and caspase assays under normal growth conditions for and RRR > 1 antagonism. For the mouse experiments, up to 24 hours. Proteins from total cell lysates were re- thetumorvolumesandareasunderthecurvewerelog solved on 4% to 20% SDS-PAGE and transferred onto ny- transformed and evaluated using ANOVA for four-way lon membranes. Membranes were blocked in PBS, 0.05% comparison and Wilcoxon test for pairwise comparisons. Tween 20 containing 5% skim milk powder and were then Median absolute deviation was used as a measurement of probed overnight with specific primary antibodies. Anti- variability. bodies were detected with the corresponding horseradish peroxidase–linked secondary antibodies. Blots were devel- Results oped using SuperSignal West Pico chemiluminescent sub- strate detection reagents. The membranes were exposed to Pralatrexate interacts synergistically with bortezomib X-ray films for various time intervals. The images were cap- in T-cell lymphoma lines tured with a GS-800 calibrated densitometer (Bio-Rad), The IC50 values for pralatrexate alone at 48 and 72 hours and the ratios were quantified by densitometric analyses were generally in the low nanomolar range. Specifically, within the linear range of each captured signal. The follow- these values (nmol/L) at 48 and 72 hours, respectively, ing monoclonal and polyclonal antibodies were used: are as follows: H9, 1.1 and 2.5; P12, 1.7 and 2.4; CEM, p27, p100/p52, and acetylated histone H3 (from Cell 3.2 and 4.2; PF-382, 5.5 and 2.7; KOPT-K1, 1 and 1.7; Signaling Technology), Noxa (Imgenex Corp.), RFC-1 DND-41, 97.4 and 1.2; and HPB-ALL, 247.8 and 0.77. (SLC19A1 antibody from abcam), and acetylated α-tubulin HH was relatively resistant after 48 hours of exposure, (from Sigma). with the IC50 at 72 hours being 2.8 nmol/L (Table 1; Fig. 1A). Two CTCL lines (H9 and HH) and two T-ALL cell Mouse xenograft models lines (P12 and PF-382) were selected for combination ex- In vivo experiments were done as follows: 5- to 7-week-old periments with bortezomib. Generally, the range of IC50s SCID-beige mice (Charles River Laboratories) were injected for bortezomib in these cells was more similar. For 6 s.c. with 20 × 10 HH cells on the flank. When tumor vo- example, the IC50s for bortezomib alone at 48 hours were lumes approached 50 mm3, mice were separated into treat- as follows: H9, 6 nmol/L; HH, 3.9 nmol/L (after 72 hours ment groups of 9 to 10 mice each. Tumors were assessed of incubation); P12, 4.7 nmol/L; and PF382, 2.2 nmol/L. using the two largest perpendicular axes (l, length; w, width) In the cytotoxicity assays, the combination of pralatrexate measured with standard calipers. Tumor volume was calcu- and bortezomib showed synergism in all cell lines: H9, lated using the formula (4/3)r3, where r =(l + w) / 4. Tumor- CI ≤ 0.38; P12, CI ≤ 0.513; PF382, CI ≤ 0.352; and HH, bearing mice were assessed for weight loss and tumor CI ≤ 0.4 (Fig. 1A and B). These data support the mathe- volume at least twice weekly. Animals were sacrificed matical synergy of these two agents in vitro. when one-dimensional tumor diameter exceeded 2.0 cm or after loss of >10% body weight in accordance with Pralatrexate plus bortezomib enhances apoptosis in institutional guidelines. Complete response (CR) was de- T-cell lymphoma fined as nonpalpable tumor. Both pralatrexate and borte- Treatment with the combination of pralatrexate and zomib were given by i.p. injection. In the combination bortezomib for 48 hours exhibited significant induction experiments, pralatrexate was administered at 15 mg/kg of apoptosis in H9, P12, PF382, and HH cell lines. For ex- (1/4 of the maximum tolerated dose) on days 1, 4, 8, ample, treatment of the CTCL line H9 with bortezomib and 11; bortezomib (B) was administered at 0.5 mg/kg alone at 5 nmol/L induced apoptosis in up to 28.6% of on days 1, 4, 8, and 11. Control groups were treated with the population. When H9 cells were treated with pralatrex- the vehicle solution alone. ate at 3 or 4 nmol/L (IC50-60), 46% and 58% of the cells,

3650 Clin Cancer Res; 16(14) July 15, 2010 Clinical Cancer Research

Pralatrexate-Based Combination in T-Cell Malignancies

Table 1. Cytotoxicity of pralatrexate and methotrexate in eight T-cell lymphoma and leukemia cell lines

Cell line Subtype Pralatrexate (nmol/L) Methotrexate (nmol/L)

IC50, 48 h (c.i.) IC50, 72 h (c.i.) IC50, 48 h (c.i.) IC50, 72 h (c.i.)

H9 CTCL 1.13 (0.1-8.9) 2.54 (0.5-12.9) 4.8 (2.5-9) 4.41 (2.9-6.7) HH N/A 2.8 (0.3-29.2) N/A 18.9 (0.09-3,538) P12 T-ALL (resistant to GSI) 1.73 (1.7-1.74) 2.43 (1.6-3.6) 4 (3.5-4.7) 16.8 (15.4-18.4) CEM 3.23 (1.6-6.5) 4.16 (3.8-4.6) 3.4 (2.3-5) 16.2 (13.1-20) PF382 5.47 (4.3-6 9) 2.72 (1.9-3 9) 21.6 (17-27 3) 19 (13.3-26.1) KOPT-K1 T-ALL (sensitive to GSI) 1 (0.8-1.3) 1.69 (1.3-2.2) 14.1 (13.2-14.5) 10 (7.3-13.8) DND-41 97.37 (25.8-368.2) 1.21 (1-1.3) 188.3 (77-458) 6.9 (2.4-19.7) HPB-ALL 247.78 (60.6-1,012 9) 0.77 (0.5-1.1) 156.7 (45.8-536) 11.1 (2.7-46.6)

NOTE: The table presents the IC50 values with the confidence interval for pralatrexate and methotrexate after 48 and 72 h of

incubation. In most of the cell lines, the IC50 values at 48 and 72 h of incubation are in the low nanomolar range. There is minimal cytotoxic activity appreciated at 24 h in all the cell lines studied. Abbreviations: c.i., confidence interval; GSI, γ-secretase inhibitor.

respectively, became apoptotic. The combination of prala- with 3 nmol/L bortezomib, 5.5 nmol/L pralatrexate, or the trexate (3 or 4 nmol/L) and bortezomib (5 nmol/L) pro- combination, 72%, 65%, and 94% of the cells became ap- duced more than 70% induction of apoptosis, and the optotic, respectively. The RRR of the combination was 0.6 RRR for the combination of pralatrexate (3 nmol/L) and (Fig. 3B; Table 2). These data support the observations bortezomib was ≤0.8. Analysis of various pralatrexate made in the cytotoxicity experiments, confirming the syn- and bortezomib concentrations revealed that the RRR ergistic interaction of pralatrexate and bortezomib in T-cell seemed to be more sensitive to the relative concentration lymphoid malignancies. of bortezomib. For example, 53% of H9 cells were apoptotic when treated with 6 nmol/L bortezomib, whereas 46% Pralatrexate enhances the activity of bortezomib in and 58% of H9 cells were apoptotic when treated with 3 inducing caspase-8 and caspase-9 activation and 4 nmol/L of pralatrexate, respectively. Treatment of To evaluate the effect of the combination on the extrinsic H9 with 6 nmol/L bortezomib and the highest concentra- and intrinsic arms of the apoptositc cascade, we explored tions of pralatrexate (4 nmol/L) produced nearly 90% in- the effect of pralatrexate and bortezomib on the activation duction of apoptosis. Furthermore, the RRR for the latter of caspase-8 and caspase-9 in the CTCL line H9. Incubation combination was <0.6. Importantly, all combinations of H9 with pralatrexate alone at 4 nmol/L, bortezomib studied were statistically significant, favoring the combina- alone at 6 nmol/L, and the combination for 24 hours re- tion over the control and any single-agent exposure (Fig. sulted in activation of caspase-8 in 14%, 10%, and 37% 2A; Table 2). In one of the more resistant cell lines, the of cells, respectively. Similarly, activation of caspase-9 CTCL HH line, the combination was superior to any single was seen in 13%, 5%, and 39% of H9 cells, respectively agent. For example, treatment of HH with the combination (Fig. 4A). These data are again concordant with the obser- of pralatrexate (3 nmol/L) and bortezomib (2 nmol/L) for vations made in the cytotoxicity and apoptosis assays de- 72 hours induced apoptosis in 53%, compared with 39% scribed above and confirmed that caspase activation with (pralatrexate) and 32% (bortezomib) of cells treated with the combination was also synergistic. either drug alone (Fig. 2B; Table 2). Again, the RRR analysis confirmed the synergistic interaction between the two Pralatrexate does not increase the toxicity of agents (RRR ≤0.66). Overall, these observations held up bortezomib in PMBCs from healthy donors over the entire panel of T-cell malignancies screened. When To assess the toxicity of the combination in normal the T-ALL line P12 was treated with 5 nmol/L bortezomib PBMNC, we also investigated the induction of apoptosis or with 2 or 3 nmol/L pralatrexate alone, 57%, 25%, and of PBMCs from healthy donors. We explored the effects 49% of the cells became apoptotic, respectively. The com- of bortezomib at 3 and 5 nmol/L, pralatrexate at 2, 3, bination of pralatrexate and bortezomib induced apopto- and 4 nmol/L, and the combination at all the possible per- sis in 75% and 87% of cells, with more apoptosis being mutations after 48 hours of exposure. Importantly, all observed at the higher concentration of pralatrexate. The combination groups did not show any significant increase RRRs of the combinations were ≤0.8 and ≤0.7, respectively, in apoptosis compared with the bortezomib-alone group and the P values were statistically significant relative to or between the combination group and pralatrexate (P > both the controls and the single-agent exposures (≤0.001; 0.8). Pralatrexate also did not exhibit any increased cyto- Fig. 3A; Table 2). When the T-ALL line PF382 was treated toxicity relative to the untreated controls (Fig. 4B).

www.aacrjournals.org Clin Cancer Res; 16(14) July 15, 2010 3651

Marchi et al.

Pralatrexate and bortezomib influence the expression volved in cell cycle control, apoptosis, chromatin remodel- of proteins belonging to different pathways in the ing (histone acetylation), and cellular transport. Using intracellular network precisely the same concentrations that effected synergy in In an effort to identify discrete pathways influenced by the in vitro assays, we identified significant changes in p27, the combination not markedly influenced by either drug NOXA, HH3, and RCF-1 in H9 after exposure to 5 nmol/L alone, broad probes of the cell cycle and apoptotic ma- bortezomib, 3 nmol/L pralatrexate, and the combination chinery were analyzed by Western blot. Treatment of the (after 16, 24, and 48 hours of exposure). These data (Sup- H9 and P12 cell lines with bortezomib and pralatrexate plementary Data A) showed that bortezomib increased the revealed changes in a host of proteins known to be in- accumulation of p27, NOXA, and HH3 acetylation in the

Fig. 1. A, cytotoxicity of pralatrexate in four T-cell lymphoma and leukemia cell lines. The plots graphically show the cytotoxicity curves for pralatrexate after 48 h of incubation in the cell lines selected for the combination studies. B, pralatrexate interacts synergistically with bortezomib in T-cell leukemia and lymphoma cell lines. Isobologram analysis of the interaction between pralatrexate and bortezomib in H9, HH, P12, and PF382. All data points below the red line define synergistic interaction between the two drugs (CI < 1).

3652 Clin Cancer Res; 16(14) July 15, 2010 Clinical Cancer Research

Pralatrexate-Based Combination in T-Cell Malignancies

Fig. 2. Enhanced apoptosis after treatment with pralatrexate plus bortezomib in CTCL (H9 and HH). A, H9 cells were incubated with pralatrexate (P), bortezomib (B), and their combination from 3 to 4 nmol/L and from 5 to 6 nmol/L, respectively, for 48 h (P ≤ 0.01). control Exp., expected. B, HH cells were incubated with 3 nmol/L pralatrexate, 2 nmol/L bortezomib, and their combination for 72 h. Etop, etoposide. The histograms show the results after the relative risk (RR) analysis. All the figures also show the “expected apoptosis value” if the interaction between pralatrexate and bortezomib were just additive.

CTCL cell line H9, as seen in the bortezomib-alone and intracellular accumulation. Supplementary Data B reveals the bortezomib plus pralatrexate combination groups. the expression of p27, NOXA, and HH3 in the T-ALL cell Pralatrexate alone had little to no effect on the accumula- line P12 after exposure to 5 nmol/L bortezomib, 2 and tion of NOXA and HH3 acetylation, while clearly increas- 3 nmol/L pralatrexate, and the possible combinations. ing p27 levels similar to that seen in the bortezomib-alone These observations were similar to those in the H9 cell group. Moreover, pralatrexate seemed to increase the accu- line, except that the increase in p27 seemed to be smaller. mulation of RFC-1 in these cells, suggesting that the expo- Whereas these drugs are likely to produce a number of ef- sure to pralatrexate may induce an increase in its own fects on these cell lines at both the transcriptional and the

Table 2. Range of apoptosis induction for the single drug compared with the combination of pralatrexate and bortezomib in the CTCL and T-ALL lines studied

Cell line Bortezomib (% apoptosis) Pralatrexate (% apoptosis) Pralatrexate + bortezomib RRR (% apoptosis − % expected)

H9 5 nmol/L (26-31%) 3 nmol/L (40-51%) 63-73% (58%) 0.7 6 nmol/L (43-63%) 3 nmol/L (40-51%) 73-92% (72%) 0.7 6 nmol/L (43-63%) 4 nmol/L (51-63%) 83-94% (78%) 0.6 HH 2 nmol/L (28-34%) 3 nmol/L (38-41%) 50-55% (46%) 0.6 P12 5 nmol/L (50-67%) 2 nmol/L (21-27%) 65-81% (58%) 0.8 5 nmol/L (50-67%) 3 nmol/L (42-55%) 84-92% (76%) 0.7 PF382 3 nmol/L (60-85%) 5.5 nmol/L (64-67%) 91-98% (89%) 0.6

NOTE: The table displays the range of apoptosis induction for the drugs alone or in combination and the RRR value after the analysis. RRR < 1, synergy.

www.aacrjournals.org Clin Cancer Res; 16(14) July 15, 2010 3653

Marchi et al.

Fig. 3. Pralatrexate and bortezomib synergystically induce apoptosis after 48 h of incubation in T-ALL. A, P12 (T-ALL) cells were incubated with pralatrexate at doses of 2 and 3 nmol/L, bortezomib at a dose of 5 nmol/L, and their combination for 48 h (P ≤ 0.001). B, PF382 (T-ALL) cells were incubated with pralatrexate at a dose of 5.5 nmol/L, bortezomib at a dose of 3 nmol/L, and their combination for 48 h (P ≤ 0.001). All the figures also show the expected apoptosis value if the interaction between pralatrexate and bortezomib were just additive.

translational levels, the survey of these proteins suggests ther significant weight loss nor death was observed in that the combination of bortezomib and pralatrexate like- any of the cohorts. These data support the in vitro experi- ly exerts its effect by influencing both cell cycle and apo- ments in establishing the superior efficacy of this combi- ptotic pathways. nation in T-cell malignancies.

Pralatrexate enhances the activity of bortezomib in a Discussion SCID-beige xenograft model The in vivo efficacy of pralatrexate combined with bor- PTCLs are, with few exceptions, considered to be aggres- tezomib was investigated in a xenograft model of CTCL sive diseases with a poor prognosis. During the last few (HH; Supplementary Data B). HH was selected because years, the proteasome inhibitor bortezomib has been ap- it was the most resistant line based on the in vitro assays. proved by the Food and Drug Administration for the treat- After 30 days from the beginning of the experiment, the ment of multiple myeloma and mantle cell lymphoma results in the combination group treated with pralatrex- and has been evaluated in a multicenter phase II clinical ate at a dose of 15 mg/kg (1/4 of the maximum tolera- trial in patients with CTCL. Albeit a small study, an ORR ted dose) and bortezomib given on days 1, 4, 8, and 11 of 67% including 2 CRs with a duration of response rang- at a dose of 0.5 mg/kg were statistically significant com- ing from 7 to 14 months was reported (15). In addition to pared with pralatrexate alone (P = 0.002), bortezomib its single-agent activity, bortezomib has been found to alone (P = 0.001), and the control (P = 0.001; Supple- synergize with innumerable agents, including HDAC mentary Data C). Interestingly, CRs were observed only inhibitors, Bcl-2–targeted agents (26), and conventional in the combination cohort, where 6 of 10 mice experi- cytotoxic agents (27). Similarly, pralatrexate has been enced CR in the combination cohort at day 18, with shown to exhibit marked activity across a panoply of B- two of those CRs being maintained beyond day 30. Nei- and T-cell malignancies with unique activity in a subset

3654 Clin Cancer Res; 16(14) July 15, 2010 Clinical Cancer Research

Pralatrexate-Based Combination in T-Cell Malignancies

of patients with T-cell lymphoma (28). While the precise The cytotoxicity assays suggested IC50 values in the na- mechanism for how these drugs complement each other is nomolar range for the cutaneous lymphoma lines (H9 and likely to be multifactorial, these agents seem to exhibit a HH) and the T-cell acute lymphoblastic leukemia lines synergistic interaction in T-cell–derived malignancies (P12, PF382, and CEM; KOKPT K1, DND-41, and HPB- based on these preclinical data. ALL) for both drugs. We also confirmed in this panel of The experiments presented here support the potent activity T-cell lines that pralatrexate is at least a log more potent of pralatrexate in a variety of T-cell lymphoma and leukemia than methotrexate (Table 1). This study showed that the lines. At present, there are no cell lines or xenograft models exposure of T-cell lines to even nanomolar concentrations representative of PTCL. As a result, our studies are restricted of pralatrexate and bortezomib produced potent induc- to the wider range possible of cell lines of T-cell lineage, in- tion of apoptosis in these lines. For all the cell lines stud- cluding T-ALL and transformed mycosis fungoides (MF). ied (H9, P12, PF382, and HH), the drug-drug interaction

Fig. 4. Caspase-8 and caspase-9 activation in CTCL and lack of enhanced toxicity in PBMCs from healthy donors after treatment with pralatrexate ± bortezomib. A, the histograms show the percentage of cells with caspase-8 and caspase-9 activation after treatment with bortezomib (6 nmol/L) and/or pralatrexate (4 nmol/L) after 24 h of incubation. The histogram called “expected” reveals the degree of caspase activation assuming additivity of pralatrexate and bortezomib. The percentage of caspase-8/caspase-9 activation was calculated based on the percent of the cells with caspase activation in the untretated cells (“0” limit) versus the percentage of cells with caspase activation treated with etoposide (100 nmol/L). B, at the same dilution that showed synergistic effect, PMBCs were treated with bortezomib at 3 or 5 nmol/L and/or pralatrexate at 2, 3, and 4 nmol/L for 48 h. Each combination group was not significantly more cytotoxic than bortezomib given alone (P > 0.87).

www.aacrjournals.org Clin Cancer Res; 16(14) July 15, 2010 3655

Marchi et al.

was described as synergistic based on the calculation of the (Bcl-2, Bcl-XL, and Bcl-w) mediated by the BH3-only mi- CI and the RRR. To determine if the two drugs affect their metic and marked increases in the BH3-only proteins action through the extrinsic or intrinsic apoptotic path- NOXA and PUMA mediated by bortezomib. It is likely ways, flow cytometry was done to analyze the activation that, whatever the ultimate mechanism of apoptosis of caspase-8– and/or caspase-9–mediated apoptosis. induced by pralatrexate, the lowering of the apoptotic Available published literature has shown that bortezomib threshold with bortezomib serves to sensitize the T-cell itself is able to activate both the extrinsic (caspase-8) and lines to the antifol. Despite the multitude of hypotheses the intrinsic (caspase-9) apoptotic pathways (29). Interest- about the mechanism of action of proteasome inhibitors, ingly, pralatrexate seems to activate the extrinsic and in- it is difficult to be dogmatic regarding the relative contri- trinsic pathways of apoptosis as well. The combination butions of one mechanism of action over another. of these agents was not only synergistic in conventional Furthermore, while pralatrexate has marked activity in cytotoxicty assays but also seems to synergistically activate patients with relatively drug-resistant T-cell lymphoma, both pathways of apoptosis. To establish that the synergy the biological basis for the activity seen in T- over B-cell appreciated in the tumor cell lines did not extend to nor- lymphomas remains unclear. In an effort to identify mal lymphocytes, we explored the combination in PBMCs unique biomarkers of activity that might account for the from healthy donors. The data showed that essentially the synergy between these two agents, a series of Western blot same and even higher concentrations of pralatrexate and analyses focused on proteins involved in cell cycle regula- bortezomib shown to be synergistic in tumor cell lines tion and apoptosis were evaluated. Although there were had no combined effect on normal PBMCs. The effect of no significant changes in most of these potential biomar- the combination was not statistically more toxic against kers, these data suggested that there were four proteins in- PBMCs than either drug alone (P > 0.8), although the cluding p27, NOXA, RFC-1, and HH3 that seemed to be concentration of bortezomib drove the toxicity against favorably modulated by these drugs used as single agents PBMCs. and in combination. Preliminary data have suggested that A xenograft experiment of CTCL (HH) in SCID-beige pralatrexate might actually lead to an increase in the ex- mice with bortezomib combined with pralatrexate was pression of RFC-1 on the target cells, potentially leading done and showed a statistically significant advantage for to an “autocrine” increase in the intracellular uptake of the combination compared with the single agents and pralatrexate. Although there was no evidence that bortezo- the control after 18 days (P ≤ 0.002). Bortezomib was giv- mib had any effect on RFC-1, these data support the ob- en at a dose of 0.5 mg/kg and pralatrexate was adminis- servation that pralatrexate, at very low concentrations, may tered at 15 mg/kg (1/4 of the maximum tolerated dose) be able to induce the expression of its own transporter. on days 1, 4, 8, and 11 to maximize activity while mini- Pralatrexate treatment also leads to an accumulation of mizing toxicity (13). These doses were selected based on p27 (likely cell cycle arrest effect) seen predominantly in preliminary animal toxicology studies which established H9 cells, whereas bortezomib increases p27, NOXA, and that higher doses of pralatrexate (30 and 60 mg/kg) in HH3. Thus, the combination of pralatrexate and bortezo- combination with bortezomib (0.5 mg/kg) were not toler- mib leads to augmented accumulation of RFC-1, p27, ated. Interestingly, CRs (6 of 10 animals) were only ob- NOXA, and, surprisingly, HH3. served in the combination cohort. No significant toxicity Increase in NOXA is emerging as an interesting potential (weight loss in excess of 10% or toxic death) was seen in biomarker of effect for proteasome inhibitors. Its accumu- any of the dose cohorts. The combination of bortezomib lation following exposure to combination therapy may ex- and pralatrexate at 1/4 of the maximum tolerated dose plain potential synergistic effects, although admittedly, it produced significant activity with minimal toxicity in the has not shed light on the important upstream pathways. xenograft model studied. NOXA belongs to the group of Bcl-2 homology domain Proteasome inhibition is known to affect a diverse array 3 (BH3) only proteins and seems to be crucial in fine- of intracellular signaling pathways, including effects on tuning cell death decisions by targeting the prosurvival NF-κB (impaired degradation of IB), cell cycle regulation molecule Mcl-1 for proteasomal degradation (33). This (accumulation of p21/p27), modulation of Bcl-2 family event seems to be critical for cell death induction in lym- members (upregulation of proapoptotic and BH3-only phoma cells treated with bortezomib and the combina- members), and accumulation of p53 (30–32). In two tion. Pralatrexate itself has been shown to selectively recent articles by our group (26, 27), AT-101, a small- induce apoptosis in human T-cell lymphotrophic virus- molecule inhibitor of Bcl-2, Bcl-XL, and Mcl-1, potently sy- 1–infected adult T-cell lymphoma/leukemia Pautrier mi- nergized with the conventional agents (cyclophosphamide croabcesses from a patient treated with a single dose of and rituximab) and the irreversible proteasome inhibitor drug (34). Biopsies of the patient's skin revealed selective carfilzomib in in vitro and in vivo models of mantle cell caspase activation and apoptosis only in the malignant lymphoma and diffuse large B-cell lymphoma. In addi- cells and not in the normal surrounding keratinocytes. Al- tion, the BH3-only mimetic ABT-737 synergized with the though NOXA was not specifically evaluated in the tissue proteasome inhibitor bortezomib, which was mechanisti- biopsies, these important observations confirm in a pa- cally attributed to an effect on both the relative balance tient that pralatrexate seems to induce apoptosis via cas- between BAX/BAK binding to antiapoptotic proteins pase activation selectively in malignant T cells.

3656 Clin Cancer Res; 16(14) July 15, 2010 Clinical Cancer Research

Pralatrexate-Based Combination in T-Cell Malignancies

The increase in HH3 acetylation is surprising, and sug- how to combine select classes of agents where other a gests that bortezomib may have some properties as a HDAC priori rationales are limited. A convergence of data around inhibitor. Previous data by Marcucci et al. have established the reproducible observation on the modulation of p21/ that bortezomib has an effect as a hypomethylating agent, p27, NOXA, Mcl-2, and, unique to pralatrexate, RFC-1 and there is emerging data that hypomethylating agents may begin to provide insights into potential biomarkers may also increase histone acetylation (35). The observation of activity and possibly response. Preclinical and clinical of course raises the suspicion that this doublet of agents studies are now under way to evaluate how best to com- could be combined with HDAC inhibitors, which also have bine these agents in an effort to develop radically new significant activity in T-cell lymphomas, but it does not pro- platforms that are not CHOP-centric for T-cell lymphoma. vide specific insights into how this contributes to the cell At the moment, these efforts are oriented toward under- kill noted. standing the drug:drug interactions between pralatrexate, An emerging number of studies are beginning to con- proteasome inhibitors, HDAC inhibitors, gemcitabine, verge on the idea that proteasome inhibitors like bortezo- forodesine, and Bcl-2–targeted agents. Phase I clinical trials mib may mediate their effects through NOXA and p27 in are now being designed to determine the dose-limiting T-cell malignancies. In one report by Ri et al. (36), bortezo- toxicity and maximum tolerated dose of pralatrexate given mib induced upregulation of NOXA at both the transcrip- in combination with proteasome and HDAC inhibitors. tional and protein levels in a p53-independent manner in Collectively, both these preclinical and available clinical mature T cells (CTCL and adult T-cell lymphoma/leukemia). data suggest that innovative and highly effective plat- Repression of NOXA by small interfering RNA rescued the forms can be developed for patients with aggressive T-cell CTCL cells. These data confirmed that the time-dependent lymphoma. binding of NOXA to Mcl-1 led to loss of the transmembrane mitochondrial potential, an observation we have similarly Disclosure of Potential Conflicts of Interest validated in mantle cell lymphoma as well. These data, cou- pled with the recent observation by Heider et al. (37) that O.A. O'Connor: commercial research grant, Allos Therapeutics, Inc. both HDAC inhibitor and bortezomib led to the upregulation of p21 and p27 in CTCL cells, which synergistically in- Acknowledgments duced apoptosis in these cells, are starting to suggest a We thank Adolfo Ferrando for providing the T-cell leukemia lines we possible mechanism for these drugs in T-cell malignancies. used in the in vitro experiments. Although the a priori rationale for combining drugs active in T-cell malignances is at the moment largely focused Grant Support on agents with clinically documented single-agent activity, it is becoming increasing clear that important themes in Food and Drug Administration Orphan Product R01 Award the underlying molecular pharmacology are emerging. In (FD0003498-01; O.A. O'Connor). fact, it is likely that these themes appreciated at the molec- Received 03/23/2010; revised 04/28/2010; accepted 05/19/2010; ular level and will become the immediate rationale for published OnlineFirst 05/25/2010.

References 1. Rudiger T, Weisenburger DD, Anderson JR, et al. Peripheral T-cell proach for advanced lymphoma using vinorelbine, gemcitabine, and lymphoma (excluding anaplastic large-cell lymphoma): results from filgrastim. Intern Med J 2007;37:760–6. non Hodgkin's Lymphoma Classification Project. Ann Oncol 2002; 8. Enblad G, Hagberg H, Erlanson M, et al. A pilot study of alemtuzu- 13:140–9. mab (anti-CD52 monoclonal antibody) therapy for patients with re- 2. Gressin R, Peoch M, Deconick E, et al. The VIP-ABVD regiment is not lapsed or chemotherapy-refractory peripheral T-cell lymphomas. superior to the CHOP 21 for the treatment of non epidermotropic pe- Blood 2004;103:2920–4. ripheral T cell lymphoma. Final results of the LTP95 protocol of the 9. Moon J, Kim J, Sohn D, et al. Alemtuzumab plus CHOP as front-line GOELAMS [Abstract]. Blood 2006;108:Abstract #2464. chemotherapy for patients with peripheral T-cell lymphomas 3. Escalón MP, Liu NS, Yang Y, et al. Prognostic factors and treatment [abstract]. J Clin Oncol 2007;25:Abstract #8069. of patients with T-cell non-Hodgkin lymphomas: the M.D. Anderson 10. Prince HM, Bishton MJ, Harrison SJ. Clinical studies of histone dea- Cancer Center experience. Cancer 2005;103:2091–8. cetylase inhibitors. Clin Cancer Res 2009;15:3958–69. 4. Marchi E, Alinari L, Tani M, et al. Gemcitabine as frontline treatment 11. Sirotnak FM, DeGraw JI, Schmid FA, Goutas LJ, Moccio DM. New for cutaneous T-cell lymphoma: phase II study of 32 patients. Cancer folate analogs of the 10-deaza-aminopterin series. Basis for the 2005;104:2437–41. structural design biochemical and pharmacologic properties. Cancer 5. Zinzani PL, Baliva G, Magagnoli M, et al. Gemcitabine treatment in Chemother Pharmacol 1984;12:18–25. pretreated cutaneous T-cell lymphoma: experience in 44 patients. 12. Wang ES, O'Connor OA, She Y, Zelenetz AD, Sirotnak FM, Moore MA. J Clin Oncol 2000;18:2603–6. Activity of a novel anti-folate (PDX, 10-propargyl10-deazaaminopterin) 6. Arkenau HT, Chong G, Cunningham D, et al. Gemcitabine, cisplatin against human lymphoma is superior to metotrexate and correlates and methylprednisolone for the treatment of patients with peripheral with tumor RFC-1 gene expression. Leuk Lymphoma 2003;44: T-cell lymphoma: the Royal Marsden Hospital experience. Haemato- 1027–35. logica 2007;92:271–2. 13. Toner LE, Vrhovac R, Smith EA, et al. The schedule-dependent 7. Spencer A, Reed K, Arthur C. Pilot study of an outpatient-based ap- effects of the novel antifolate pralatrexate and gemcitabine are

www.aacrjournals.org Clin Cancer Res; 16(14) July 15, 2010 3657

Marchi et al.

superior to methotrexate and cytarabine in models of human non- synergizes the antineoplastic activity of proteasome inhibitors in lym- Hodgkin's lymphoma. Clin Cancer Res 2006;12:924–32. phoid malignancies. Blood 2008;112:2906–16. Epub 2008 Jun 30. 14. Horwitz SM, Duvic M, et al. Pralatrexate (PDX) is active in cutaneous 27. Paoluzzi L, Gonen M, Gardner JR, et al. Targeting Bcl-2 family mem- T-cell lymphoma (CTCL): results of a multicenter dose finding trial. bers with the BH3 mimetic AT-101 markedly enhances the therapeu- ASH 2009. 2009;Abstract 919. tic effects of chemotherapeutic agents in in vitro and in vivo models 15. Zinzani PL, Musuraca G, Tani M, et al. Phase II trial of proteasome of B-cell lymphoma. Blood 111:5350–8. inhibitor bortezomib in patients with relapsed or refractory cutaneous 28. O'Connor OA, Horwitz S, Hamlin P, et al. Phase II-I-II study of two T-cell lymphoma. J Clin Oncol 2007;25:4293–7. different doses and schedules of pralatrexate, a high-affinity sub- 16. Gootenberg JE, Ruscetti FW, Mier JW, Gazdar A, Gallo RC. Human strate for the reduced folate carrier, in patients with relapsed or re- cutaneous T cell lymphoma and leukemia cell lines produce and fractory lymphoma reveals marked activity in T-cell malignancies. respond to T cell growth factor. J Exp Med 1981;154:1403–18. J Clin Oncol 2009;27:4357–64. Epub 2009 Aug 3. 17. Starkebaum G, Loughran TP, Jr., Waters CA, Ruscetti FW. Establish- 29. Zang QL, Wang L, Zhang YW, et al. The proteasome inhibitor borte- ment of an IL-2 independent, human T-cell line possessing only the zomib interacts synergistically with the histone deacetylase inhibitor p70 IL-2 receptor. Int J Cancer 1991;49:246–53. suberoylanilide hydroxamic acid to induce T-leukemia/lymphoma 18. Foley GE, Lazarus H, Farber S, et al. Continuous culture of human cells apoptosis. Leukemia 2009;23:1507–14. Epub 2009 Mar 12. lymphoblasts from peripheral blood of child with acute leukemia. 30. Palombella VJ, Rando OJ, Goldberg AL, Maniatis T. The ubiquitin- Cancer 1965;18:522–9. proteasome patway is required for processing the NF-κB1 pre- 19. Kojika S, Sugita K, Inukai T, et al. Mechanisms of glucocorticoid re- cursor protein and activation of NF-κB. Cell 1994;78:773–85. sistance in human leukemic cells: implication of abnormal 90 and 70 31. Adams J. The proteasome: a suitable antineoplastic target. Nat Rev kDa heat shock proteins. Leukemia 1996;10:994–9. Cancer 2004;4:349–60. 20. Sangster RN, Minowada J, Suciu-Foca N, et al. Rearrangement and 32. Paoluzzi L, O'Connor OA. Mechanistic rationale and clinical evidence expression of the alpha, beta, and gamma chain T-cell receptor for the efficacy of the proteasome inhibitors against indolent and genes in human thymic leukemia cells and functional T cells. J Exp mantle cell lymphomas. BioDrugs 2006;20:13–23. Med 1986;163:1491–508. 33. Ploner C, Kofler R, Villunger A. Noxa: at the tip of the balance be- 21. O'Connor OA, Smith EA, Toner LE, et al. The combination of the pro- tween life and death. Oncogene 2009;27:S84–92. teasome inhibitor bortezomib and the bcl-2 antisense molecule ob- 34. Marneros AG, Grossman ME, Silvers DN, et al. Pralatrexate-induced limersen sensitizes human B-cell lymphomas to cyclophosphamide. tumor cell apoptosis in the epidermis of a patient with HTLV-1 adult Clin Cancer Res 2006;12:2902–11. T-cell lymphoma/leukemia causing skin erosions. Blood 2009;113: 22. Hockberg Y, Tamhane AC. Multiple comparison procedures. New 6338–41. Epub 2009 Apr 23. York (NY): Wiley; 1987, p. 1–450. 35. Liu S, Liu Z, Xie Z, et al. Bortezomib induces DNA hypomethyla- 23. Chou TC. Relationship between inhibition constants and fractional tion and silenced gene transcription by interfering with Sp1/NF- inhibitions in enzyme-catalyzed reactions with different numbers of κB-dependent DNA methyltransferase activity in acute myeloid reactants, different reaction mechanisms, and different types of me- leukemia. Blood 2008;111:2364–73. chanisms of inhibition. Mol Pharmacol 1974;10:235–47. 36. Ri M, Iida S, Ishida T, et al. Bortezomib-induced apoptosis in mature 24. Chou TC. Derivation and properties of Michaelis-Menten type and Hill T-cell lymphoma cells partially depends on upregulation of NOXA type equations for reference ligands. J Theoret Biol 1976;39:253–76. and functional repression of Mcl-1. Cancer Sci 2008. 25. Zhao L, Wientjes G, Au JLS. Evaluation of combination chemotherapy: 37. Heider U, Rademacher J, Lamottke B, et al. Synergistic interaction of integration of nonlinear regression, curve shift, isobologram, and com- the histone deacetylase inhibitor SAHA with the proteasome inhibitor bination index analyses. Clin Cancer Res 2004;10:7994–8004. bortezomib in cutaneous T cell lymphoma. Eur J Haematol 2009;82: 26. Paoluzzi L, Gonen M, Bhagat G, et al. The BH3-only mimetic ABT-737 440–9. Epub 2009 Feb 10.

3658 Clin Cancer Res; 16(14) July 15, 2010 Clinical Cancer Research

Pralatrexate Is Synergistic with the Proteasome Inhibitor Bortezomib in In vitro and In vivo Models of T-Cell Lymphoid Malignancies

Enrica Marchi, Luca Paoluzzi, Luigi Scotto, et al.

Clin Cancer Res Published OnlineFirst May 25, 2010.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-10-0671

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2010/05/24/1078-0432.CCR-10-0671.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2010/07/08/1078-0432.CCR-10-0671. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.