Inhibition of Hsp90 Suppresses PI3K/AKT/mTOR Signaling and Has Antitumor Activity in Burkitt Lymphoma

Burkitt lymphoma (BL) is a germinal center B-cell-derived malignancy that is molecularly characterized by the translocation of the MYC proto-oncogene to either the immunoglobulin heavy or the κ or λ light chains. Beyond the MYC translocation, BL is heterogeneous from a genomic standpoint. Recurrent somatic alterations have recently been described in TP53 (approximately 50% of cases); TCF3 or its negative regular ID3 (34–70% of cases); CCND3 (38% of cases); and the SWI/SNF family of chromatin remodeling genes (17–43% of cases; refs. 1–4). Novel therapies are needed in BL, where the survival for patients with relapsed disease is <20% (5, 6). One approach to therapeutic targeting in BL given the molecular heterogeneity is to focus on global mechanisms of lymphomagenesis, which are independent of genetic subtype.

Hsp90 has emerged as a desirable target in cancer due to its role in the regulation of pathways required for malignant growth (7). Hsp90 is an ATP-dependent molecular chaperone that protects proteins from proteolytic degradation including oncogenic signaling complexes. Inhibitors that broadly target Hsp90, however, have been limited in their clinical development due to suboptimal target inhibition and off-target toxicities. More recent studies have elucidated a biochemically distinct Hsp90 complex in cancer cells that is unique from the fraction of Hsp90 engaged in housekeeping functions (8). This “tumor-enriched” Hsp90 is amenable to small molecule inhibition. PU-H71 is a purine scaffold Hsp90 inhibitor with preferential uptake in tumor tissue and selective binding to the tumor-enriched fraction of Hsp90 (9). PU-H71 specifically targets oncogenic Hsp90 clients, for example: Bcr-Abl but not cellular Abl; mutated, but not WT mB-Raf; and the aberrantly activated signaling complex containing Bcl6 (8, 10). This agent, as well as other inhibitors of Hsp90, are currently being evaluated in phase I and phase II clinical trials (clinicaltrials.gov ID: NCT01393509, NCT02261805, NCT02474173).

More recently, a chaperome network (coined the epichaperome) was identified in a subset of cancers, and found to be specifically targeted by PU-H71 (11). This epichaperome appears to rely on dysregulated MYC, and is present in tumors that overexpress the MYC oncoprotein. Although BL is characterized by MYC translocation and deregulation, the Hsp90 chaperome and the effects of PU-H71 have not been studied in BL. In this study we evaluated Hsp90 as a therapeutic target in BL. We hypothesized that Hsp90 inhibition would disrupt oncogenic client proteins required for BL survival. We found that Hsp90 inhibition has anti-lymphoma effects in vitro and in vivo, including activity in a patient-derived xenograft (PDX) model of BL. In order to determine the relevant Hsp90 clients in BL and thereby define the Hsp90 oncoproteome, we used PU-H71 affinity capture and proteomics (8, 12). We identified multiple components in the PI3K/AKT/mTOR signaling pathway as Hsp90 clients in BL. In addition, we found that Hsp90 inhibition was synergistic with PI3K/mTOR inhibition. Collectively, these results support Hsp90 as a rational therapeutic target in BL and identify a novel strategy for combination therapy.

Ramos, Raji, Namalwa, Daudi, CA-46, Ly10, and Ly18 cell lines were purchased from ATCC in 2013. Jiyoye and DG-75 were purchased from ATCC in 2014. Raji 2R and Raji 4RH were provided by Roswell Park Cancer Institute (M.J.B.). All cells were used for experiments within 2 months of thawing. Cell line characterization was performed by Idex Bioresearch using short tandem repeat profiling and multiplex PCR to report cell line characterization and interspecies contamination. Testing was last performed in February 2014. Cells were cultured in RPMI1640 media (Invitrogen) supplemented with 10% heat inactivated FBS and gentamicin 50 μg/mL (Sigma-Aldrich). PU-H71, PU-H71-beads, and control beads were synthesized and prepared at Memorial Sloan Kettering Cancer Center as previously reported (13, 14). Idelalisib (CAL-101), IPI-145, BKM120 (15), BEZ-235 (16), ganetespib (17), and BIIB-021 (18) were purchased from Selleck Chemicals. MPC-3100 (19) and CUDC-305 (20) were purchased from Santa Cruz and ChemieTek, respectively. The chemical structures for PU-H71 and BEZ-235 are presented in Supplementary Fig. S1.

Cell lysis and immunoblot were performed by standard procedure using the following antibodies: GAPDH (Genetex); p110δ and p110α (Santa Cruz); p110β and p85α (Abcam); PARP (BD Pharmingen); AKT, p110δ, p110β, eIF4E (Cell Signaling Technology); PARP (Santa Cruz); RB and pRB (Cell Signaling Technology). IHC was performed using the following antibodies: Hsp90 (Enzo Life Sciences) p-4EBP1 (BD Pharmingen), BCL2 (Cell Marque), BCL6 (Dako), Ki67 (Dako), TUNEL (prepared as previously described; ref. 21). A TMA was constructed using 0.6 mm cores obtained from formalin-fixed paraffin embedded BL tumor samples. Each case was represented on the TMA in duplicate. Quantification of IHC was performed using the Halo image analysis software program (Indica Labs).

Cell viability was determined using trypan blue staining and an ATP-based luminescent assay (CellTiter-Glo, Promega). Luminescence was measured using the GloMax Multi+ Detection System (Promega). IC₅₀ values were calculated using Prism 6 software. For combination treatments, cells were exposed to increasing doses of each drug alone or the combination at a constant ratio in 96-well plates. Cell viability was measured and normalized to vehicle control. Compusyn software (Biosoft) was used to plot dose effect curves and calculate combination index (CI) value. To evaluate cell viability in germinal center (GC) B cells, lymphocytes were extracted from de-identified primary human tonsil specimens which were obtained after routine tonsillectomy. Total lymphocytes were treated with PU-H71 in RPMI containing 20% FBS and 100 μg/mL Primocin and cultured for 24 hours on gamma-irradiated mouse L CDw32 feeder cells. At indicated times posttreatment lymphocyte cultures were incubated with fixable viability stain (FVS, BD Catalog No. 564406). Cells were pelleted and resuspended in 100 μL cold PBS without calcium and magnesium containing 5% FBS, and 0.1% Sodium Azide (FACS Block) and incubated on ice for 15 minutes after which 100 μL cold PBS containing 0.5% FBS and 0.1% Sodium Azide (FACS Wash) was added. Cells were pelleted and resuspended in FACS Wash containing B cell phenotype panel as follows: Ig Lambda CD19-PE (BD Catalog No. 240720), CD38-PECy7 (BD Catalog No. 560667), IgD-PerCP Cy5.5 (BD Catalog No. 561315), and CD27-APC H7 (BD Catalog No. 560222). Parallel cultures of Ramos treated with PU-H71 and stained only with FVS were included as positive control. Data were acquired on a BD LSR2 Flow Cytometer and analyzed using FlowJo software. Data reported for primary tonsil specimens are the percentage of CD19⁺ lymphocytes, which were both germinal center cells (CD38^hi, IgD^low) and viable (FVS-).

Cells were treated as indicated and 5 × 10⁵ cells were harvested by centrifugation at indicated timepoints, washed once in PBS and resuspended in 1× AnnexinV staining buffer (BD Catalog No. 556454) containing 3 μL/test AnnexinV-Alexa Fluor 647 (ThermoFisher A23204) and 5 μL/test 7-AAD (BD Catalog No. 559925) and incubated at room temperature for 15 minutes. Data were acquired with a BD LSR2 analytical flow cytometer and analyzed using FlowJo software. Necrotic cells were defined as annexin V+/7AAD+ or annexinV−/AAD+ and early apoptotic cells were defined as annexin V+/7AAD−.

Cells were treated as indicated and 5 × 10⁵ cells were harvested by centrifugation, washed once in PBS, and resuspended in 1× Transcription Factor Fix/Perm buffer and incubated for 50 minutes on ice. An equal volume of 1× Perm/Wash buffer was added and cells were pelleted and washed once more in 1× Perm/Wash. Buffers for this procedure were purchased from BD (Catalog No. 562725). Ki67 staining was performed overnight at 4°C in 1× Perm/Wash using 0.5 μL Ki67-BV510/test (BD Catalog No. 563462). After staining cells were washed twice with 1× Perm/Wash and resuspended in PBS containing 5 μL/test propidium iodide (BD Catalog No. 556463). Data were acquired with a BD LSR2 analytical flow cytometer, analyzed using FlowJo software, and plotted using the ggplot2 package in R.

Affinity capture and proteomics analysis were performed as previously described (8). Briefly, cells were lysed by collecting them in Felts buffer [20 mmol/L HEPES, 50 mmol/L KCl, 5 mmol/L MgCl₂, 0.01% (w/v) NP-40, freshly prepared 20 mmol/L Na₂MoO₄ (pH 7.2–7.3)] with added 1 μg/μL protease inhibitors (leupeptin and aprotinin), followed by three successive freeze (in dry ice) and thaw steps. Total protein concentration was determined using the BCA Kit (Pierce) according to the manufacturer's instructions. The PU-H71 conjugated beads were added at a volume of 80 μL to the cell lysate (1 mg), and the mixture incubated at 4°C for 4 hours. The beads were washed five times with Felts lysis buffer and separated by SDS-PAGE, followed by a standard Western blotting procedure. Proteins identification was then performed by LC/MS-MS as follows: purified PU-H71 interacting protein complexes were resolved using SDS-PAGE, followed by brief staining with Simply Blue (Life Technologies) and excision of the separated protein bands. In situ trypsin digestion of polypeptides in each gel slice was performed and the tryptic peptides were resolved on a nano-capillary reverse phase column. Peptides were directly infused into a linear ion-trap mass spectrometer (LTQ Orbitrap XL, Thermo Electron Corp.). The mass spectrometer was set to collect one survey scan (MS1), followed by MS/MS spectra on the nine most intense ions observed in the MS1 scan. Proteins were identified by searching the tandem mass spectra against a human segment of Uniprot protein database (20,273 sequences; European Bioinformatics Institute, Swiss Institute of Bioinformatics and Protein Information Resource) using Mascot search engine (Matrix Science; version 2.5.0; http://www.matrixscience.com/) Decoy database search was always activated and, in general, for merged LC/MS-MS analysis of a gel lane with P < 0.01, FDR averaged around less than 1%. Scaffold (Proteome Software, Inc.) version 4_4_1 was used to further validate and cross-tabulate tandem mass spectrometry (MS-MS) based protein and peptide identifications.

Six- to eight-week-old male NOD-SCID mice obtained from Jackson Laboratory were injected in the subcutaneous flank with 1 × 10⁷ Namalwa or Ramos cells in PBS. Mice were evaluated daily for weight and tumor measurement. A 7.5 mg/mL PU-H71 solution was prepared in a 10 mmol/L PBS and delivered via intraperitoneal injection. A 5 mg/mL BEZ-235 solution was prepared in 1:9 N-methylpyrrolidone:polyethylene glycol and delivered via oral gavage. Tumor volumes were measured daily with calipers and tumor volume was calculated as 0.5 length × width². AUC (22) was used to measure tumor volume over time (GraphPad Prism software). Mice were sacrificed when tumors reached a volume of 3,000 mm³ or at the completion of the observation period (15 days). Tumors were harvested after animals were sacrificed and sectioned for IHC staining.

PDXs were established by implanting fresh tumor fragments in the subcutaneous flanks of NOD-SCID interleukin-2R gamma null (NSG) mice (n = 3). Tumor formation was monitored weekly by inspection and measurement using calipers. When the tumors reached a volume of 1,000 to 1,500 mm³, the PDX tumors were collected after humane euthanasia. The tumors were sectioned into 2 mm fragments and implanted into additional animals for subsequent drug testing.

Two-tailed unpaired t test was used unless otherwise specified. Survival estimates were calculated using Kaplan–Meier analysis. All statistical analyses were carried out using Prism software (GraphPad).

De-identified patient samples were obtained under approval from the Institutional Review Board (23) of Weill Cornell Medical College. The Research Animal Resource Center of Weill Cornell Medical College approved cell line xenograft studies. PDX studies were conducted at Columbia University under the approval of the Columbia University Institutional Animal Care and Use Committee.

Hsp90 is ubiquitously expressed in human cells and is upregulated in conditions of chronic cellular stress including disease states and cancer (24). Hsp90 is overexpressed in a variety of cancer subtypes and overexpression has been linked to more aggressive disease and resistance to chemotherapy (25, 26). To assess the potential for Hsp90 as a therapeutic target in BL, we first evaluated Hsp90 expression in primary human BL tumors. IHC for Hsp90 was performed on a BL tissue microarray containing 59 primary tumors and tonsil tissue for control. Fifty-three of 59 cases (90%) demonstrated high levels of Hsp90 expression defined as >90% tumor cell positivity. Quantification of Hsp90 by mean optical density demonstrated a significant increase in expression over normal tonsil (0.673 vs. 0.275, P = 0.001; Fig. 1A and B). In Hsp90 + tumors, nonmalignant infiltrating cells were negative for Hsp90 (representative images shown in Fig 1B). Because the cell of origin for BL is the germinal center B-cell (27), we also examined Hsp90 expression in normal GC cells in tonsil and lymph node specimens where GCs could be visualized and staining of both tumor and GC cells was performed on the same slide. GC B cells demonstrated moderate Hsp90 expression relative to high expression seen in BL tumors (Supplementary Fig. S2).

Although BL tumors were found to express high levels of Hsp90, total amount of this chaperone protein does not necessarily reflect cellular dependency on this protein or sensitivity to Hsp90 inhibitors (11). To determine the sensitivity of BL to Hsp90 inhibition, we performed in vitro viability assays with the ATP-based CellTiter-Glo assay in nine human BL cell lines (Ramos, DG-75, Raji, Namalwa, Daudi, Jiyoye, CA-46, Raji 2R, and Raji 4RH) and two DLBCL cell lines known to be sensitive to PU-H71 (Ly10, Ly18; Fig. 1C and Supplementary Fig. S3A). Cells were exposed to increasing concentrations of PU-H71 and viability was assessed at 24, 48, and 72 hours. All cell lines demonstrated time and dose-dependent growth inhibition. Results were confirmed by evaluating cell viability with trypan blue (Supplementary Fig. S3B). The concentration of PU-H71 that inhibited viability by 50% (IC₅₀) was in the nanomolar range (171–337 nmol/L) and was consistent across cell lines. The Raji 2R and Raji 4RH cell lines which have acquired resistance to rituximab and multiple chemotherapeutic agents (28) were equally sensitive to PU-H71 (IC₅₀ 175–181 nmol/L). To determine if PU-H71 was broadly toxic to GC B cells, we exposed primary human tonsil tissue to PU-H71 and evaluated viability in GC B cells using flow cytometry. Ramos cells but not GC B cells demonstrated substantial reduction in viability upon exposure to PU-H71 (Supplementary Fig. S3C). To determine if the effect on BL growth was unique to PU-H71, we tested four additional small molecule Hsp90 inhibitors in BL cells: BIIB-021, CUDC-305, ganetespib, and MPC-3100 (refs. 17–20; Fig. 1D). Cells were sensitive to all four Hsp90 inhibitors with IC₅₀ ranging from 3.8 to 149.5 nmol/L. This work demonstrates that BL is broadly sensitive to Hsp90 inhibition.

Given data from our group that Hsp90 inhibition may preferentially target gamma herpesvirus-associated malignancies (29), we included both EBV-positive (n = 4) and EBV-negative (n = 3) subtypes of BL in our PU-H71 screen including both the sporadic and endemic forms of BL (Fig. 1C). There was no difference in sensitivity in EBV-positive and EBV-negative cells (IC₅₀ 290 and 217 nmol/L, respectively, P = 0.239). This suggests that the mechanism of action of PU-H71 is independent of EBV-mediated oncogenesis in BL. This may be explained by the restricted EBV latency observed in BL where few viral proteins are expressed and are not major oncogenic drivers (30).

To determine the mechanism of growth inhibition, we evaluated the effect of treatment on cell cycle using flow cytometry and PI/Ki67 staining (Fig. 2A). Cell-cycle analysis revealed an accumulation of cells in G₀ upon treatment with PU-H71, suggesting that PU-H71 inhibits the growth of BL primarily through cell-cycle arrest. An evaluation of apoptosis using flow cytometry and Annexin V/7AAD staining revealed an increase in necrotic cells (annexin V+/7AAD+ or annexinV−/7AAD+) by 72 hours of treatment, but no corresponding increase in early apoptotic cells (annexinV+/7AAD−) at any timepoint (Fig. 2B). Moreover, we did not observe PARP cleavage as measured by Western blot analysis (Fig. 2C) or increases in caspase 3 or 7 activity as measured by the fluorescent assay Caspase-Glo (Fig. 2D) consistent with a nonapoptotic mechanism of cell death. This is also supported by a lack of increase in the sub-G₀ population in the cell-cycle analysis, which reflects DNA fragmentation and is characteristic of apoptotic cell death. Based on this we conclude that PU-H71 induces cell-cycle arrest in BL followed by nonapoptotic cell death.

To determine the anti-lymphoma effect of Hsp90 inhibition in vivo, we established Namalwa xenografts in NOD-SCID mice by subcutaneous injection. After tumors reached 150 to 200 mm³, mice were randomized to receive PU-H71 or vehicle control (4 mice/cohort). PU-H71 was administered daily by intraperitoneal injection at a dose of 50 mg/kg for 10 days. This dose was chosen based on a pilot experiment where we evaluated BL xenografts treated at 75 mg/kg i.p. daily, as previously reported (31, 32), and 50 mg/kg i.p. daily and found similar efficacy with both doses. Mice were monitored daily for evidence of toxicity and measurement of tumor volume. Mice were sacrificed once tumors reached 3,000 mm³. To evaluate tumor growth over time, we calculated both tumor volume and tumor AUC as previously described (31).

PU-H71 significantly decreased tumor growth relative to vehicle control in Namalwa bearing xenografts (P = 0.0117; Fig. 3A and B). Similar results were seen in a Ramos xenograft model (P = 0.0057; Supplementary Fig. S4). There was no evidence of toxicity from PU-H71 based on behavioral evaluation and serial weight measurements. More extensive toxicity profiling has been performed in previous studies including complete blood count, blood chemistry, liver function tests, and thyroid hormone testing with no significant toxicity noted (10, 33). This safety profile finding is concordant with that seen in the phase I clinical study of PU-H71 in patients with previously treated solid tumors, lymphoma, and myeloproliferative neoplasms (34).

Because cell line xenografts may not fully recapitulate disease characteristics found in patients, we also evaluated the effect of Hsp90 inhibition in a BL PDX (Fig. 3C and D). Tumor for PDX studies was obtained from a pediatric patient with an abdominal presentation of BL who underwent a biopsy prior to the initiation of therapy. After written informed consent for PDX generation was obtained, surplus primary BL tumor sample was obtained fresh. The tumor was implanted into the subcutaneous flank space of NSG mice. Once tumors reached an approximate volume of 1,500 mm³, the animals (P = 0) were euthanized and the tumor removed. This tumor was then separated into 2 mm fragments and implanted into (P = 1) animals for subsequent drug testing.

We first evaluated PDX tumors to determine if they were similar to the patient tumor phenotype. PDX BL tumors resembled primary BL by histology, immunophenotype, and cytogenetic analysis. The PDX tumors demonstrated the classic starry sky appearance on H&E, had a Ki67 proliferative index >90%, and were positive for CD20 and BCL6 and showed only very weak staining for BCL2, consistent with a BL immunophenotype. In addition, both the PDX and the primary tumor harbored the characteristic MYC translocation (Supplementary Fig. S5).

To study the effect of Hsp90 inhibition, PDX mice were treated with PU-H71 (n = 6) or vehicle (n = 6) once tumors reached 150 to 200 mm³. PU-H71 was given at a dose of 50 mg/kg daily ×15 days. Mice were monitored daily for weight, tumor measurements, and toxicity. Treatment with PU-H71 resulted in a significant decrease in tumor volume compared to control (P = 0.0074; Fig. 3C). PDX mice did not demonstrate any evidence of PU-H71 induced toxicity. The antitumor effect of PU-H71 was more pronounced in the PDX model than in either cell line xenograft model, highlighting the differences between cell lines and patient-derived tumors.

Given the promising anti-lymphoma activity of PU-H71 in BL models, we wanted to further explore the mechanism of action to better understand Hsp90-dependent signaling networks in BL and to develop rational therapeutic combinations. To evaluate the targets of PU-H71 in BL, we performed high-affinity capture followed by proteomic analysis (8). Cellular lysates from two BL cell lines (Ramos, Daudi) were incubated with PU-H71 conjugated agarose beads to chemically precipitate Hsp90 complexes. The protein cargo was then subject to SDS-PAGE and proteins were analyzed using LC/MS-MS. A total of 752 and 1,783 Hsp90 client proteins were identified in Ramos and Daudi, respectively (Supplementary Table S1). A total of 701 proteins were common to both cell lines.

Ingenuity Pathway Analysis was used to determine the signaling networks represented by Hsp90 clients in BL. In both cell lines, the PI3K/AKT/mTOR pathway was identified among the most highly represented canonical pathways including “PI3K/AKT signaling”; “p70S6K signaling”; “regulation of eIF4 and p70S6K signaling”; and “mTOR signaling” (Supplementary Table S2). Other top hits which are relevant to lymphocyte biology, included “MYC mediated apoptosis,” “EIF2 signaling,” and “protein ubiquitination pathway.” To validate the IPA findings, we performed an independent analysis of the common proteins identified in both Ramos and Daudi using the Search Tool for the Retrieval of Interacting Genes/proteins (STRING) with a CI of 0.99. PI3K proteins formed one of the primary clusters identified in this analysis (Fig. 4A). Other clusters included the proteasome, the ribosome, replication machinery, ribonucleoproteins, translation machinery, and the COP9 signalosome. Both the STRING and the IPA analyses demonstrate that PU-H71 is targeting PI3K signaling among other oncogenic pathways in BL.

Recent genomic and functional studies in BL indicate that PI3K signaling contributes to lymphomagenesis (4, 35). Given that PI3K, AKT, and mTOR signaling were among the top networks identified in our proteomic screen, we explored the relationship between these pathways and Hsp90. We interrogated our proteomic data and found significant enrichment for PI3K proteins including: multiple isoforms of the PI3K catalytic subunit (PIK3CB, PIK3CD, PIK3CG), the PI3K regulatory subunit (PIK3R1, PIK3R4), and downstream elements (MTOR, S6K1, and EIF4E; Fig. 4B). These findings were validated using chemical precipitation with PU-H71-conjugated beads (Fig. 4C). Cellular lysates from Ramos and Daudi were incubated with PU-H71-conjugated beads. The protein cargo was then evaluated using immunoblot analysis probing for PI3K pathway proteins. In both Ramos and Daudi, p110β, p110γ, p110δ, p85α, AKT, and eIF4E were precipitated with PU-H71 beads, confirming their role as Hsp90 clients. In contrast, MCL1, which was not identified as an Hsp90 client in our proteomic screen, was not precipitated by the PU-H71 beads.

To determine if PI3K pathway proteins require Hsp90 chaperoning to maintain expression, we exposed Ramos and Namalwa cells to increasing doses of PU-H71 and then analyzed protein abundance by immunoblot (Fig. 5A). P-mTOR, p70S6K, and AKT demonstrated both time- and dose-dependent decrease upon exposure to PU-H71. To determine if this effect was unique to PU-H71, we also tested other small molecule Hsp90 inhibitors. Treatment with CUDC-305 or ganetespib resulted in similar dose-dependent downregulation of PI3K signaling, confirming the dependence of these proteins on Hsp90 (Fig. 5B).

To determine if Hsp90 inhibition impacts PI3K/AKT/mTOR signaling in vivo, we performed IHC on Namalwa xenograft tumors after exposure to PU-H71. NOD-SCID mice were injected with 1 × 10⁷ Namalwa cells and monitored for tumor progression (n = 9). Once tumors reached 500 mm³, mice were randomized 1:2 to vehicle or a single dose of PU-H71 at 100 mg/kg. Tumors were harvested 8 or 24 hours after treatment and evaluated by IHC. IHC for p-mTOR and p-p70S6K were negative in all cases including control, reflecting the transient nature of phosphorylation of these proteins. 4EBP1, however, which is downstream in PI3K signaling was moderately decreased in the PU-H71-treated mice at the 8-hour time point and markedly decreased by 24 hours (Fig. 5C). Of note, we also evaluated Ki67 and TUNEL in this model to determine if the mechanisms of cell death in vivo were consistent with what we observed in vitro. Similar to our in vitro experiments, we observed a decrease in proliferation as evidenced by Ki67 upon treatment with PU-H71, which was statistically significant at 8 hours (Supplementary Fig. S6A) and no increase in TUNEL (Supplementary Fig. S6B) consistent with a nonapoptotic mechanism for cell death.

Our data suggest that Hsp90 inhibition is effective in BL in part due to the dependence of PI3K pathway proteins on Hsp90 chaperoning. Given this, we were interested in evaluating the effect of isolated PI3K inhibition on BL. PI3K inhibitors can broadly be categorized into class I PI3K isoform-specific inhibitors, pan-class I PI3K inhibitors, and dual class I PI3K/mTOR inhibitors (36). We tested all three classes of PI3K inhibitors in a panel of BL cell lines. Cells were broadly resistant to isoform specific PI3K inhibition with idelalisib/CAL101 (p110δ inhibitor) and IPI-145 (p110δ/γ inhibitor). Cells demonstrated moderate sensitivity to pan-PI3K inhibition with the pan-class I PI3K inhibitor BKM-120 (15). Cells were most sensitive to BEZ235, which targets both mTOR and all class I isoforms of PI3K (ref. 16; Table 1).

We suspected that PI3K inhibition would be most effective in BL when multiple components of the pathway are targeted, which can be achieved by PU-H71 or BEZ235. Based on this, we hypothesized that PU-H71 might synergize with PI3K inhibitors through more robust dual targeting of PI3K signaling and inhibition of possible feedback mechanisms. To test this we evaluated the combination of PU-H71 and PI3K inhibitors in four BL cell lines (Ramos, Namalwa, Daudi, DG-75). Cells were exposed to increasing doses of each drug alone and the combination. The combined response at 48 and 72 hours was evaluated using the Chou-Talalay method (37). A combination index (CI) of <1 was considered synergistic. CI of 1 was considered additive and >1 infra-additive. Because cells were resistant to idelalisib and IPI-145, CI could not be calculated for these combinations. The combination of PU-H71 and BKM-120 was additive or synergistic in three of four cell lines. The combination of PU-H71 and BEZ 235 was additive or synergistic in all four cell lines (Fig. 6A).

To further evaluate combination therapy with PU-H71 and BEZ235, we established Namalwa bearing xenografts. Once tumors reached 150 to 200 mm³, mice were randomized to receive PU-H71, BEZ235, the combination, or vehicle (6 mice/cohort). Mice were treated daily with PU-H71 at 50 mg/kg delivered by intraperitoneal injection, BEZ235 at 50 mg/kg delivered by oral gavage, the combination, or vehicle for 15 days. As single agents both PU-H71 and BEZ235 reduced tumor growth compared to vehicle (P = 0.0196 and 0.0004, respectively). The combination of PU-H71 and BEZ235 resulted in marked tumor growth suppression with minimal or no growth in most cases. The combination of PU-H71 and BEZ235 suppressed BL tumor growth more than vehicle (P ≤ 0.001), PU-H71 alone (P = 0.0008), or BEZ235 alone (P = 0.0386; Fig. 6B and C).

Our work has identified Hsp90 as a novel therapeutic target in BL and has expanded our understanding of the function of Hsp90 in BL. We found that BL tumors overexpress Hsp90 and that BL cells are sensitive to inhibition with small molecule Hsp90 inhibitors. Hsp90 plays a crucial role in normal cellular function by supporting the proteome machinery, which is involved in a number of housekeeping activities. In states of cellular stress Hsp90 exists in an active binding conformation, which is distinct from the “housekeeping” Hsp90 species and is uniquely sensitive to small molecule inhibition (8, 38). This unique sensitivity allows for specific targeting of oncogenic signaling pathways with minimal disruption of “housekeeping” Hsp90. Immunoprecipitation with PU-H71 conjugated beads allows specific pulldown of Hsp90-dependent oncogenic protein clients which are required by tumor cells to maintain growth (8, 10, 29, 31). This pulldown likely includes not only proteins that bind directly to Hsp90 but also those in the epichaperome network (co-chaperones, adaptors, and folding enzymes) that are dependent on Hsp90 to maintain the malignant phenotype (11). In this work we utilized the specificity of PU-H71 as an unbiased approach to characterize the Hsp90-dependent chaperome in BL. We identified multiple components of the PI3K/AKT/mTOR pathway as Hsp90 clients. We also found other essential pathways in BL to interact with Hsp90 including MYC-mediated apoptosis, B-cell receptor signaling, and protein ubiquitination. Our findings highlight the importance of these pathways in BL and provide a unique opportunity to target multiple oncogenic pathways with a single agent.

The PI3K pathway is thought to play a role in BL pathogenesis based on recent genomic and functional studies. BL tumors harbor recurrent alterations in TCF3 and its negative regulator ID3 which result in tonic B-cell receptor signaling and activation of PI3K (2–4). In addition, a mouse model in which MYC and PI3K are constitutively active in germinal center B cells recapitulates the histomorphology and clinical phenotype (35). Despite the apparent biologic dependence on PI3K, we found isolated PI3K inhibition to have only a modest effect in BL. We suspect that the improved anti-lymphoma effect of PU-H71 over isolated PI3K inhibition is due to targeting of not just PI3K but multiple cellular pathways in BL that depend on Hsp90. It is also likely that isolated PI3K inhibition in BL results in activation of feedback loops and alternative survival pathways, which facilitate escape. In contrast, Hsp90 inhibition prevents escape by also inhibiting these feedback and alternative pathways, which are likely to also be Hsp90 dependent.

In aggressive lymphomas such as BL, single-agent therapy is unlikely to be curative. Therefore, the use of targeted therapies will require rational combinatorial treatment. Here we demonstrate synergy between PU-H71 and dual PI3K/mTOR inhibition in BL. This treatment combination is particularly well suited for clinical application as both categories of drugs are being evaluated in clinical trials in lymphoma and the drugs are not expected to have overlapping toxicities. A phase I study of PU-H71 is nearing completion and phase II studies are being developed. In addition, other next-generation Hsp90 inhibitors, which we also found to be effective in BL, are moving toward clinical implementation with minimal toxicity reported to date (39). Dual PI3K/mTOR agents are also in phase I/II trials in lymphoma (NCT02669511, NCT02249429, NCT01991938).

Although the majority of patients with BL in resource rich settings are cured of their disease (>85%), there remain distinct patient subsets for whom outcomes are poor and targeted therapies are needed (40). Specifically, patients with relapsed disease have a poor response to therapy and an expected overall survival of <20%. In addition, patients with endemic BL have an OS ranging from 25% to 62% due to the inability to safely deliver intensive chemotherapy in resource-limited settings (41). Our work suggests that Hsp90 inhibition-based treatment might be effective in both settings. The sensitivity of chemotherapy and immunotherapy resistant BL cells to Hsp90 inhibition (Raji 2R, Raji 4RH) suggests that this approach may be effective in relapsed disease. Because we were only able to evaluate two resistant cell lines, additional work on other models of resistant BL (such as PDX mice from refractory cases) will be needed to determine the role of PU-H71 in chemotherapy-resistant disease. With regard to use of this agent in the endemic subtype of BL, we observed an anti-tumor effect in endemic BL cell lines in-vitro (Raji, Daudi, Namalwa, Jiyoye) and in-vivo (Namalwa) suggesting that Hsp90 inhibition may be a rational approach in sub-Saharan Africa.

In summary, our work demonstrates that small molecule inhibition of Hsp90 suppresses PI3K/AKT/mTOR signaling and has robust antitumor activity in BL. We describe, for the first time, the Hsp90-dependent oncoproteome in BL and identify multiple components of PI3K/AKT/mTOR signaling as Hsp90 clients in this malignancy. Finally, we identify a novel treatment strategy in BL through dual targeting of Hsp90 and PI3K/mTOR. This approach may offer improved efficacy and reduced toxicity over standard chemo-immunotherapy for patients with BL.

A. Rodina is a consultant/advisory board member for Samus Therapeutics. T. Taldone is a consultant and a consultant/advisory board member for Samus Therapeutics. G. Chiosis is a Director, board member at Samus Therapeutics; has ownership interest (including patents) of Samus Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Giulino-Roth, E. Cesarman

Development of methodology: L. Giulino-Roth, H.J. van Besien

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Giulino-Roth, H.J. van Besien, T. Dalton, J.E. Totonchy, A. Rodina, H. Erdjument-Bromage, J. Sadek, A. Chadburn, F.S.D. Cruz, A. Rainey, A.L. Kung

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Giulino-Roth, H.J. van Besien, T. Dalton, J.E. Totonchy, H. Erdjument-Bromage, J. Sadek, F.S.D. Cruz, A.L. Kung

Writing, review, and/or revision of the manuscript: T. Dalton, J. Sadek, A. Chadburn, M.J. Barth, E. Cesarman, Giulino-Roth

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Giulino-Roth, H.J. van Besien, A. Bolaender, M.J. Barth

Study supervision: E. Cesarman

Other (prepared and provided reagents): T. Taldone

Other (provided reagents and methods, and know-how for their use): G. Chiosis

We thank Susan Mathew, MD, for performing FISH analysis on PDX tumor samples for the MYC translocation.

This work was funded by grants from the NIH (R01-CA103646 and R01-CA154288; to E. Cesarman), Lymphoma Research Foundation (to L. Giulino-Roth), Hyundai Hope on Wheels (to L. Giulino-Roth), St. Baldrick's Foundation (to L. Giulino-Roth), and the Leukemia and Lymphoma Society (SCOR 7012-16; to E. Cesarman). G. Chiosis is funded by R01 CA172546 and R01 CA155226.

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.