Preclinical and clinical evaluation of Buparlisib (BKM120) In Recurrent/Refractory Central Nervous System Lymphoma

Abstract

Central Nervous System (CNS) Lymphomas are aggressive brain tumors with limited treatment options. Targeting the phosphoinositide 3-kinase (PI3K) pathway yields promising responses across B-cell malignancies, but its therapeutic potential in CNS lymphomas remains unexplored. We present pre-clinical and clinical data on the pan-PI3K inhibitor Buparlisib in CNS lymphomas. In a primary CNS lymphoma-patient derived cell line, we define the EC50. Four patients with recurrent CNS lymphoma were enrolled in a prospective trial. We evaluated Buparlisib plasma and cerebrospinal fluid pharmacokinetics, clinical outcomes, and adverse events. Treatment was well tolerated. Common toxicities include hyperglycemia, thrombocytopenia, and lymphopenia. Presence of Buparlisib in plasma and CSF was confirmed 2h post-treatment with a median CSF concentration below the EC50 defined in the cell line All four patients were evaluated for response and median time to progression was 39 days. Buparlisib monotherapy did not lead to meaningful responses and the trial was prematurely stopped.

Clinical Trial Registration:

NCT02301364

Introduction

Primary Central Nervous System Lymphoma (PCNSL) is an aggressive subtype of diffuse large B-cell lymphoma (DLBCL) that manifests exclusively in the central nervous system (CNS)[1]. Standard of care for PCNSL includes radiation, methotrexate therapy, and anti-CD20 antibody therapy (rituximab), all of which are associated with limited efficacy, substantial morbidity and treatment recurrence [2], as well as poorer outcomes than non-CNS DLBCL [3].

Tumor cells from systemic non-Hodgkin lymphoma can invade the CNS (including the brain parenchyma and leptomeninges), concomitant to systemic lymphoma progression, leading to Secondary Central Nervous System Lymphoma (SCNSL). SCNSL has poor prognosis with a median survival of 1-5 months [4]. As more first-line therapy approaches have been developed for systemic DLBCL, the incidence of SCNSL has subsequently increased over time as a direct result of disease recurrence and progression. Similar to PCNSL, SCNSL therapy is challenging and usually involves methotrexate-based chemotherapy. For patients failing chemotherapy, palliative radiation remains the only option. Patients who relapse after initial therapy have an especially poor prognosis with a swift progression and aggressive clinical course [5]. Based on published long-term follow-up data, the relapse rates after high-dose methotrexate-based therapy with or without consolidation whole-brain radiation is 44% [6].

Gene expression profiling has enabled the classification of DLBCL into two major subgroups: germinal center B-cell (GCB) and activated B cell (ABC) [7]. The ABC subtype has poorer clinical prognosis than GCB; ABC lymphoma is also associated with B-cell receptor (BCR) activation, and tumor cells present with frequent mutations in genes from the BCR signaling axis such as MYD88 and CD79B [8, 9]. ABC DLBCL cell lines respond remarkably well to targeted inhibition of the phosphatidylinositol-3-kinase (PI3K) pathway in vitro [10]. More than 95% of PCNSL cases are of the ABC/non-germinal center subtype [11]. Moreover, the mammalian target of rapamycin (mTOR) inhibitor temsirolimus was associated with responses in recurrent PCNSL even though the progression-free survival was limited[12]. Therefore, the use of PI3K inhibitors could be promising in PCNSL, particularly in patients that have failed conventional methotrexate-containing chemotherapy regimens.

The maximum-tolerated dose (MTD) of Buparlisib has been defined in as 100 mg daily dosing achieving mean Cmax serum level of 1010 ng/mL at 1 hour after oral administration[13]. Pre-clinical assessments suggest blood-brain barrier penetration[14-16]. In a multicenter single-agent phase II trial in glioblastomas, 65 patients were dosed with 100mg daily. In resected tumors, the mean buparlisib concentration was 612ng/g [17]. There was limited clinical efficacy with a median progression-free survival (PFS) of 1.7 months and no responders. In another multicenter trial, 33 glioblastoma patients received a Buparlisib, dosed at 50 and 80 mg daily, in combination with the MET inhibitor INC280 (capmatinib). The median plasma Cmax ranged from 429-680 ng/mL in the 50 mg cohort and 779-853 ng/mL in the 80mg cohort, depending on the INC280 concentration, indicative of a drug-drug interaction. None of the patients responded to the combination[18].

Here we report the preclinical and clinical evaluations of the efficacy of the pan-PI3K inhibitor Buparlisib (BKM120) in patients with r/r PCNSL and SCNSL. We define the EC50 in a PCNSL patient-derived cell line. We show that at the 100mg daily dosing of Buparlisib was well tolerated with expected adverse events. Plasma and cerebrospinal fluid (CSF) PK analysis demonstrated expected plasma biodistribution and established CNS penetration. Even though, CNS penetration was seen, the concentrations reached remained below the EC50 observed in the PCNSL cell line. We also performed correlative genomic studies on tumor biopsies through MSK-IMPACT panel sequencing, and catalogued the mutations present at recurrence. Due to lack of substantial clinical benefit from Buparlisib in r/r PCNSL or SCNSL the study was prematurely stopped. Overall, our data highlights the urgent need of conducting prospective studies in this patient population to identify and validate the efficacy of novel therapeutic approaches to treat this aggressive disease.

Methods

Generating Orthotopic xenografts and stable cell line:

Fresh samples were collected from patients undergoing diagnostic biopsy at MSKCC. A single cell suspension was generated from the sample by mechanical dissociation as well as enzymatic treatment with Accumax (Innovative Cell Technologies). Cells were counted and checked for viability prior to injection into the striatum of SCID mice (Taconic Farms) (200,000 cells / injection; 5 mice per tumor biopsy). At time of development of neurologic deficits, the mice were sacrificed, the brains were removed in a sterile fashion and mechanically and biochemically dissociated to obtain a single cell suspension which was again injected into the brain of SCID mice (200,000 cells / injection: 5 mice per passage). These xenografts were maintained for at least 3 passages. Two patient-derived xenografts were generated: PCNS#6 and PCNS#11; both generated from a non-germinal center tumor sample. Molecular characterization with the MSK-IMPACT panel[19] identified mutations in CD79B in PCNS#6 and CD79B and MYD88 in PCNS#11. As previously described, both models have an activated PI3Km/TOR pathway as demonstrated by immunohistochemical staining[20].

The cell line PCNSL-MSK was generated from the orthotopic xenograft model PCNS#11[20]. Patient-derived tumor cells were injected into mice brains and propagated through serial re-injections. PCNS#11 cells were also injected subcutaneously. In 3/5 animals subcutaneous (SC) tumors developed. These tumors were resected under sterile conditions. Part of the SC tumors was manually and enzymatically disrupted to generate a single cell suspension that was placed and initially maintained in Roswell Park Memorial Institute (RPMI) media supplemented with 20% fetal bovine serum (FBS; Omega scientific, FB-11), 100U/mL penicillin and 100 U/streptomycin (Gemini Bio-Products) in 5% CO₂ atmosphere at 37°C. After the establishment of a stable cell line media was change and cells maintained in RPMI/10% FBS. The cell line was generated from a non-germinal center PCNSL. Sequencing identified a CD79B Y196D and MYD88 L265P mutation.

Viability/cell death assay:

For viability/growth assessment, 100,000 PCNSL#11 cells were seeded in 6cm dishes in RPMI/10% FBS. Cells were treated with Buparlisib (BKM120 (NVP-BKM120, Buparlisib; S2247; Selleck Chemicals; diluted in dimethylsulfoxide (DMSO)) at 1, 10, 100, 500, and 100 nM concentrations for 5 days. DMSO served as vehicle control (0.1 % of final concentration) and each condition was seeded in triplicate. Following treatment, cells were harvested and counted on a ViCell Cell Viability analyzer. Viability is displayed as fold change between day 5 and a day 1 control. The instrument also uses trypan blue to assess cell death. Cell death was expressed as the fraction of trypan-blue-positive cells over the total number of cells on day 5. Statistical significance was assessed by one-way ANOVA testing (** = p ≤ 0.01, **** = p ≤ 0.0001).

Immunoblotting:

For western blotting, 1,000,000 PCNSL#11 cells were seeded in 6cm dishes in RPMI/10% FBS. Cells were treated with Buparlisib (BKM120 (NVP-BKM120, Buparlisib; S2247; Selleck Chemicals; diluted in dimethylsulfoxide (DMSO)) at 31.25, 62.5, 125, 250, 500, 1000, 5000, and 10,000 nM concentrations for 24 hours. DMSO served as vehicle control (0.1 % of final concentration). Cells were then harvested, lysed and homogenized in 1% triton lysis buffer (#9803, Cell Signaling) containing fresh protease and phosphatase inhibitors by standard procedures. Protein concentrations were quantified with the BCA Protein Assay kit (Pierce Chemical Co.), and proteins were separated in a gradient (4-15%) SDSPAGE gel, transferred to nitrocellular membranes, and hybridized with antibodies to the indicated antigens by standard procedures. Signals were detected by chemoluminescence using ECL detection reagents (Amersham Pharmacia Biotech). Primary antibodies to the following antigens were used: phospho-Akt (Ser473/587F11, #4051, Cell Signaling), phospho-Akt (Thr308/C31E5E, #2965, Cell Signaling), phospho-S6 (Ser235/236, #2211, Cell Signaling), phospho-4EBP1 (Thr37/46 236B4, #2855, Cell Signaling), CD20 (clone L26, M0755, Dako), cleaved PARP (Asp214/D64E10 XP, #5625, Cell Signaling), cleaved caspase-3 (Asp175, #9661, Cell Signaling) and Vinculin (clone hVIN-1, V9131, Sigma).

Patients:

The intended target accrual for this study was twenty-one patients. The therapy would have been considered worth pursuing if at least 12 out of 21 participants (original enrollment target) had remained progression-free at 6 months post-treatment. However, the trial was prematurely closed to enrollment because the drug manufacturer withdrew support, after only a single partial response was observed among the first four patients of the study. This trial was conducted following approval by the Institutional Review Board at Memorial Sloan Kettering Cancer Center, and in accordance with an assurance filed with and approved by the U.S. Department of Health and Human Services. Informed consent was obtained from each subject or subject's guardian. The clinical trial registration number is NCT02301364).

Main inclusion criteria included histologically confirmed PCNSL or SCNSL (in patients with SCNSL, CNS involvement was confirmed either through tissue biopsy or CSF cytology), relapsed/refractory PCNSL or SCNSL with at least one prior CNS-directed therapy (with no restrictions in the number of recurrences), unequivocal evidence of disease progression on MRI or new CSF cytology finding, Karnofsky Performance Status (KPS) ≥ 50, adequate bone marrow and organ function, recovery from grade 1 toxicity from prior lines of therapy, submission of 20 unstained slides from the initial tissue diagnosis for confirmation of diagnosis and correlative studies, and age≥18. Prior autologous stem cell transplant as well as radiation to the CNS was allowed; allogenic stem cell transplant recipients were excluded. Exclusion criteria included: active disease outside the CNS, prior treatment with a PI3K inhibitor, AKT inhibitor, or mTOR inhibitor, > 8 mg of dexamethasone daily at enrollment, warfarin or any other Coumadin-derivative anticoagulant, history of invasive malignancy, known human immunodeficiency virus (HIV) infection, any severe psychiatric disease (≥ Common Terminology Criteria for Adverse Events (CTCAE) (v 4.03)grade 3 anxiety, depression grade ≥ grade 2 (CTCAE (v 4.03)), history of or active major depressive episode, bipolar disorder (I or II), obsessive-compulsive disorder, schizophrenia, history of suicidal attempt or ideation, or homicidal ideation), active cardiac disease or cardiac dysfunction, poorly controlled diabetes mellitus or steroid-induced diabetes (glycosylated hemoglobin >8%), recent major systemic surgery (≤ 2 weeks), and pregnant or nursing women.

The clinical trial was pre-maturely stopped after only 1 out of the first 4 patients had a clinical response to therapy.

Treatment and clinical assessment:

Buparlisib was provided by Novartis. All participants received daily oral Buparlisib treatment at a fixed dose of 100 mg which is the maximal tolerated dose (MTD) established in prior single agent trials: NCT01068483 and NCT01283503). Drug treatment was administered on a continuous schedule, once daily. Response was assessed with MRI brain scans every 8 weeks following the IPCG criteria. Adverse events were assess using CTCAE v 4.03.

Pharmacokinetic analysis:

Fifteen days after treatment initiation, patients underwent steady state CSF and plasma pharmacokinetic analysis, to evaluate the drug distribution in plasma and CSF. In patients with an Ommaya reservoir (n=3), blood and CSF were collected 1h, 2h and 4h after drug administration. In patients without an Ommaya reservoir (n=1), blood and CSF were collected 2h after study drug administration. Each sample was collected at +/− 30 minutes of each time point. CSF and plasma concentrations of the drug was measured by liquid chromatography with tandem mass spectrometry (LC system: Shimadzu LC-20AD system; Mass spectrometer: API 4000, Applied Biosystems/Sciex; at WuXi AppTec, Shanghai, China)).

MSK-IMPACT Heme panel sequencing:

DNA was isolated from formalin-fixed paraffin-embedded (FFPE) tumor tissue and from normal matched blood samples, according to standard procedures[19]. DNA sequencing library preparation and sequencing on a HiSeq 2500 was performed as described [19]. Hybridization capture was performed using a modified MSK-IMPACT (MSK-IMPACT Heme) targeted sequencing panel including 589 key cancer genes (out of which 480 genes are related to hematological malignancies)[21]. Read alignment and processing were performed using BWA and the Picard using standard parameters. GATK tool was use for local realignment and quality score recalibration as of GATK best practices (https://www.broadinstitute.org/gatk/guide/best-practices). Mutation analysis was performed using Mutect v 1.1.7 [22] and GATK Haplotype Caller (https://www.broadinstitute.org/gatk/guide/tooldocs/org_broadinstitute_gatk_tools_walkers_haplotypecaller_HaplotypeCaller.php).

Statistical analysis:

The primary endpoint of the study was to assess the progression free survival at 24 weeks (PFS24w) where follow up was calculated from treatment start date until first documented progression or death due to any cause. Secondary endpoints included the exploration of safety and tolerability of Buparlisib in patients with PCNSL and SCNSL, by assessing the frequency and severity of adverse events, overall response rate (ORR) on MRI, assessment of progression free survival at 12 weeks (PFS12w) and 48 weeks (PFS48w), duration of response (DoR) and overall survival (OS). Prior trials in patients with recurrent PCNSL have reported a PFS6m of 35-45% with conventional chemotherapy regimens.

Results

Pre-clinical efficacy of PI3K inhibition

Using a PCNSL patient-derived cell line [20], we established an in-vitro EC50 between 100-500nM confirming a prior observation in two PCNSL patient-derived xenograft models[20]. Cell growth was significantly reduced at 500nM of buparlisib (Fig. 1A). The induction of cell death was observed at concentrations ≥500nM (Fig. 1B/C). Moreover, pathway members were inhibited at concentrations ≥500nM (Fig.1C).

Figure 1: Preclinical efficacity of buparlisib.

Buparlisib reduces growth (A), induces cell death (B/C) and inhibits PI3k pathway signaling (C) in a PCNSL patient-derived cell line (trypan blue) dose dependently with an IC50 of <500nM.

Patient characteristics

A total of four patients were enrolled (Fig. 2A). Three of them were men, their median age was 64 (range: 55-79), and the median KPS at enrollment was 90 (range 60-100). Two patients had PCNSL and two SCNSL. All patients had intraparenchymal lesions without additional involvement of the CSF. None of the patients had prior radiation to the brain and only one had received prior stem cell transplant (auto-transplant). The median prior CNS-directed treatments were 1.75 (range: 1-3).

Figure 2: Genomic Alterations of Trial Patients.

A) Baseline clinical parameters. B) Summary of genomic alteration identified by using the hybridization capture-based next-generation sequencing assay MSK-HemePACT for targeted deep sequencing of tumor samples from patients enrolled in the trial. Members of the B-cell receptor-signaling (BCR) pathway (MYD88, CD79B, CARD11) were frequently mutated. B) Frequencies of BCR axis alterations. MYD88 and CD79B were most frequently mutated (75%). C) Mutations are found in known hotspots in MYD88 and CD79B.

Molecular characterization of the patients in the study

All four patients original tumor tissue was assessed for cell-of-origin based on the Hans classification [23] using the immunohistochemical markers CD10, BCL6, and MUM-1. In the four patients, tumor cells were of the activated B-cell (ABC) type (Fig 2B). Additional genomic alterations were identified by using the targeting sequencing. We identified frequent mutations affecting the BCR signaling pathway. MYD88 and CD79B were mutated in 75% of cases (Fig 2C/D) and were found in known hotspots (MYD88 L265P; CD79B Y196) (Fig 2B). These data confirm the presence of BCR signaling pathway gene alterations in the four patients enrolled in the study.

Treatment and toxicity

One patient discontinued Buparlisib because of psychiatric symptoms including hallucinations (grade 3), encephalopathy (grade 2) and cognitive disturbance/ (grade 3). There were two drug-related grade 4 toxicities (lymphopenia and neutropenia) in one patient that resolved after the drug was held (drug was restarted after resolution of toxicity). Grade 3 toxicities observed included hyperglycemia, lymphopenia, cognitive disturbance and hallucinations. The most common toxicities were hyperglycemia and thrombocytopenia. Overall, buparlisib treatment was generally well tolerated (Table 1) and followed the established toxicity profiles of this drug.

Table 1:

Buparlisib treatment related adverse events

Toxicity	1	2	3	4	All
Hyperglycemia	3	-	1	-	4
Platelet count decreased	1	1	-	-	2
Alanine aminotransferase increased	1	-	-	-	1
Alkaline phosphatase increased	1	-	-	-	1
Aspartate aminotransferase increased	1	-	-	-	1
Cholesterol high	1	-	-	-	1
Cognitive disturbance	-	-	1	-	1
Depression	-	1	-	-	1
Dry skin	1	-	-	-	1
Encephalopathy		1	-	-	1
Fatigue	1	-	-	-	1
Hallucinations	-	-	1	-	1
Lymphocyte count decreased	-	-	-	1	1
Memory impairment	1	-	-		1
Neutrophil count decreased		-	-	1	1
Pruritus	1	-	-	-	1
Rash maculo-papular	1	-	-	-	1
Seizure	-	1	-	-	1
White blood cell decreased	-	1	-	-	1

Efficacy

The overall response rate to treatment was 25% (Fig. 3A/C). One patient showed a partial response. However, this patient rapidly developed psychiatric symptoms which prompted drug discontinuation. Due to rapid clinical deterioration, this patient was shortly directed to hospice care, and thus the response to treatment was not confirmed in a follow-up scan. The three other patients developed neurologic symptoms prior to the scheduled treatment response assessment, requiring unscheduled imaging at a median of 37 days after trial drug initiation. The development of neurologic symptoms was associated with increased tumor size and progression of disease on imaging (Fig 3A). The PFS at 12, 24, and 48 weeks was 0% with a median PFS of 39 days (Fig. 5B/C). Overall, the median progression free survival was 39 days with a median overall survival of 196 days (Fig. 3B/C).

Figure 3: Clinical Response to Buparlisib.

A) Magnetic resonance imaging (MRI) T1+contrast sequences from baseline and after 4 weeks of Buparlisib treatment in each patient. B) Kaplan-Meier curves demonstrating progression-free survival (PFS) and overall survival (OS) in days. C) Table summarizing clinical efficacy. ORR: overall response rate; PFS: progression-free survival; OS: overall survival.

Pharmacokinetic studies

Buparlisib concentrations were assessed 2h after drug treatment at steady state (15 days after therapy initiation) in plasma (n=4) and CSF (n=3) patients. CSF could not be collected in one of the four patients due to technical difficulties during lumbar puncture. The mean plasma concentration at 2h was 1114ng/mL (range: 844-1630) with associated mean CSF concentration of 81.3ng/mL (54.9-110) (Fig. 4). The mean CSF concentration of 81.3ng/mL correlated with 198.1nM which is below the EC50 identified in our pre-clinical model. Only one patient achieved a CSF concentration above the EC50 at 205ng/mL (equal to 500nM). This patient had a partial response on imaging assessment, but eventually developed therapy-related CNS toxicity.

Figure 4: Plasma and CSF pharmacokinetics of Buparlisib.

Plasma and CSF drug concentrations 2h after oral Buparlisib on day 15 of treatment. The dashed reference lines compare molar and ng/mL concentrations of Buparlisib.

Discussion

PCNSL is an aggressive subtype of DLBCL that manifests exclusively in the CNS. Therapeutic options for PCNSL are scanty, and usually associated with substantial morbidity and treatment recurrence [2]. PCNSL outcomes are poorer than other DLBCL subtypes that manifest outside the CNS [3]. SCNSL is defined as metastases to the brain parenchyma or CSF [24]. Patients with either PCNSL or SCNSL who relapse after initial therapy have a particularly dismal prognosis [6, 25]. Only a few prospective clinical trials have studied novel therapies for this patient population, that yielded objective response rates of 30-60% with a PFS of only 2-6 months in the best cases for most chemotherapy agents [26-31].

Our study shows that the pan-PI3K inhibitor Buparlisib shows manageable toxicity in patients with recurrent/refractory CNS lymphoma, but we were not able to demonstrate single-agent activity in this patient population exceeding currently available treatment regimens.

Our pharmacokinetic analysis of plasma samples demonstrated that Buparlisib concentration reached comparable levels to published data[13, 17, 18]. We observed a mean plasma concentration of 1114ng/mL compared to 1010ng/mL in the original phase I study[13] which was higher than those observed in glioblastoma patients (612ng/mL)[17]. CSF samples collected from trial participants showed that the drug is penetrating the blood-brain barrier but only at levels that did not reach a concentration sufficient to match the EC50 experimentally determined in vitro using our PCNSL patient-derived cell line. These sub-optimal CSF concentrations could be the underlying reason for the lack of clinical efficacy of Buparlisib. Our study underlines the importance of pharmacokinetic studies using CSF to understand the lack of clinical efficacy. The value of targeting the PI3K pathways in CNS lymphoma remains to be defined. As a next step, agents with an EC50 in the lower nanomolar range, like copanlisib[20] or GDC-0084 (paxalisib)[32] might be more efficacious and clinical trials are ongoing (NCT03581942 and NCT04906096, respectively). Additionally, Buparlisib might show clinical activity when used in combination with other chemotherapeutic or targeted agents in this disease population.

There are limitations to this study. The preclinical data is based on the results in two PCNSL-derived xenograft models (previously published[20]) and one PCNSL-derived cell line. PCNSL-derived xenografts and cell lines represent the more appropriate pre-clinical models over ABC-subtype cell lines frequently used (OCY-LY3 SUDHL-2, HBL-1). Like the majority of PCNSL, our models are all non-germinal subtype harboring mutations in MYD88 and CD79B which are frequently found in PCNSL. Our models still might not be representative of the entire disease population. Unfortunately, the trial was prematurely terminated and only 4 patients were enrolled. In a larger population, we might have identified a clinical efficacy signal. Of note, all patients enrolled had non-germinal center subtype tumors.

Overall, we identified an EC50 for Buparlisib in PCNSL models and demonstrated that Buparlisib penetrates the CNS but does not reach appropriate level to be effective. Even though treatment was well tolerated clinical efficacy was lacking.