Polatuzumab vedotin in CNS lymphoma: proof-of-concept study for blood-brain barrier penetration Open Access

Polatuzumab vedotin (pola) is a CD79B antibody-drug conjugate (ADC) with significant clinical activity in systemic B-cell lymphomas,¹ and it is now a standard frontline treatment in combination with chemotherapy for diffuse large B-cell lymphoma (DLBCL).² Although pola is efficacious in systemic B-cell lymphomas, it is unclear whether it can cross the blood-brain barrier (BBB) or induce clinical responses in central nervous system lymphomas (CNSL). These are timely and relevant questions as the long-term survival of patients with primary CNSL (PCNSL) and secondary CNSL (SCNSL) is only 30% to 50%,^3-6 underscoring the importance of identifying improved treatments for these patients. As previous studies have demonstrated the potential efficacy of other antibody-based therapies in CNSL,^3,7-9 we hypothesized that pola would also penetrate the BBB, which could provide a rationale for its evaluation in future CNSL clinical trials. Thus, in this study we sought to provide preclinical support for the efficacy of pola in CNSL by using PCNSL and SCNSL xenograft models, and to demonstrate the ability of pola to penetrate the BBB using on-treatment samples from patients with CNSL treated with pola-based therapy.

In the OCI-Ly10 PCNSL model, 0.3 × 10⁶ luciferase-expressing OCI-Ly10 DLBCL cells were stereotactically implanted into the right forebrain of NOD scid gamma (NSG) mice to generate an isolated central nervous system (CNS) tumor. In the OCI-Ly10 SCNSL model, 2 × 10⁶ luciferase-expressing OCI-Ly10 cells were injected IV into NSG mice, resulting in systemic engraftment with secondary CNS dissemination within 19 days. For the DHL-4 PCNSL model, 0.2 × 10⁶ luciferase-expressing DHL-4 DLBCL cells were stereotactically implanted into the right forebrain of NSG mice. In both OCI-Ly10 PCNSL and OCI-Ly10 SCNSL models, a single dose of pola (1.8 mg/kg) or human immunoglobulin G1 isotype control antibody (1.8 mg/kg; BioXcell, catalog no. BE0297) was administered by tail vein injection once CNS tumor was detected by bioluminescence imaging (BLI). For the DHL-4 PCNSL model, 1.8 mg/kg of pola or isotype control antibody was administered weekly until mice reached their survival endpoint. Tumor burden, as determined by BLI, was balanced between the pola and isotype groups before treatment was initiated. Tumor burden was then tracked weekly by BLI and long-term survival was monitored. Mice were age- and sex-matched (6-10 weeks old, 50% each gender). Mice that did not have tumor engraftment were excluded from the analysis. The Kaplan-Meier method was used to estimate the survival of mice for each group. Survival comparisons were performed using a log-rank test. Hazard ratios and 95% confidence intervals were calculated using the Mantel-Haenszel method. Statistical analyses were performed using GraphPad Prism v10.05.

Cerebrospinal fluid (CSF) and plasma were collected from pola-treated patients at City of Hope. Antibody-conjugated monomethyl auristatin E (acMMAE) and unconjugated MMAE concentrations were determined using qualified assays that were similar to previously validated immunoaffinity, liquid chromatography with tandem mass spectroscopy methods.¹⁰ Further details regarding these methods can be found in the supplemental Data. All patients provided written informed consent. The study was approved by the City of Hope institutional review board and met requirements of the Declaration of Helsinki.

To assess the preclinical efficacy of pola in CNSL, we tested its antitumor effects in PCNSL and SCNSL xenograft models. In this study, we used the human DLBCL cell line, OCI-Ly10. We selected OCI-Ly10 because it exhibits biological and genetic hallmarks associated with CNSL, such as an activated B-cell-like cell-of-origin and the MYD88^L265 hot spot mutation.^11-13 OCI-Ly10 also demonstrates CNS tropism with secondary dissemination to the CNS after IV injection in immune-deficient NSG mice, which suggests it can serve as a robust model of SCNSL. Confirming these observations, we identified tumor engraftment within both CNS and non-CNS locations 19 days after IV injection as determined by CD20 immunohistochemistry and bioluminescent imaging (Figure 1A-B). CNS involvement occurred in both the leptomeninges and brain parenchyma (Figure 1A). Treatment of mice bearing OCI-Ly10 SCNSL tumors with a single dose of pola (1.8 mg/kg) on Day 20 significantly decreased tumor burden in both CNS and non-CNS compartments (Figure 1C; supplemental Figure 1A) and significantly prolonged the survival of mice (Figure 1D). Similar results were observed in an orthotopic PCNSL model, where OCI-Ly10 was directly implanted into the right forebrain by stereotactic injection (Figure 1E-F; supplemental Figure 1B), and in a second orthotopic PCNSL model using the DLBCL cell line, DHL-4 (supplemental Figure 1C-D). Thus, these data indicate pola exhibits significant preclinical activity against PCNSL and SCNSL models that accurately recapitulate key biological and mutational features of the disease.

Based on these encouraging preclinical findings, we sought to compile a clinical case series of patients with CNSL that were treated with pola-based therapy at our institution. The clinical characteristics of the 3 consecutively treated patients are provided in Table 1. All 3 patients had biopsy-proven DLBCL. One patient had PCNSL and 2 patients had SCNSL. Patients had relapsed/refractory disease with a median of 4 prior lines of therapy. Two patients presented with disease confined to the CNS at the time of pola-based treatment, and 1 patient had concomitant systemic and CNS disease. Sites of CNS involvement included the CNS parenchyma (n = 1) and concomitant parenchymal and leptomeningeal disease (n = 2).

Patients were treated with off-label pola using standard dosing at 1.8 mg/kg every 3 weeks. Pola was combined with intrathecal chemotherapy (n = 1), ibrutinib (n = 1), and mosunetuzumab (n = 1). The median number of pola cycles was 6 (range, 2-11). Transient grade 3 transaminitis occurred in 1 patient and 1 patient had grade 2 neutropenia. No significant neurologic side effects were noted from treatment.

On-treatment CSF samples were collected from 2 patients with CNSL to determine the extent to which pola crossed the BBB. CSF was also collected from 2 additional patients with systemic DLBCL who underwent lumbar punctures as part of standard CNS prophylaxis procedures performed during treatment with pola-R-CHP (pola, rituximab, cyclophosphamide, hydroxydaunorubicin, and prednisone). CSF drug concentrations were determined for antibody-conjugated pola (evaluated as acMMAE), as well as the unconjugated MMAE payload using qualified mass spectrometry-based assays. All CSF samples were collected by lumbar puncture under direct fluoroscopy. The timing of CSF collections in relation to pola dosing is provided in Table 1. Pola was detected in the CSF of all 4 patients (Figure 2A). The acMMAE concentrations were 881 pM and 398 pM in the 2 patients with CNSL, and 52 pM and 102 pM in the 2 systemic DLBCL patients. The acMMAE CSF concentrations in the patients with CNSL were notably higher than the 50% inhibitory concentration of pola for several lymphoma cell lines as established by a previous enzyme-linked immunosorbent assay-based assay (eg, Ramos cell line, 71 pM).¹⁴ Unconjugated MMAE CSF concentrations (4.93-36.4 pM) were lower than acMMAE CSF concentrations, which was expected given the known stability of the conjugated MMAE payload to polatuzumab.¹ Finally, synchronous plasma and CSF collections from 3 patients indicated acMMAE CSF concentrations that were 0.56% to 1.31% of those observed in plasma (Figure 2A).

In terms of clinical efficacy, patient 1 received pola in combination with intrathecal methotrexate/cytarabine. This patient had active systemic and CNS lymphoma, and the lymphoma involved multiple CNS locations, such as the cranial nerves, brainstem, spinal cord, leptomeninges, and CSF. After 2 cycles of pola, the patient exhibited a complete radiographic response in both systemic and CNS compartments (Figure 2B), with complete clearance of lymphoma cells from the CSF. The patient’s neurologic deficits (third nerve palsy and left facial/arm weakness) also significantly improved within days of treatment. After 2 doses of pola, the patient’s remission was consolidated with axicabtagene ciloleucel CAR T-cell therapy, and the patient remains in remission for >10 months since initiating pola. Although we are unable to definitively attribute the patient’s CNS response to pola because of the concomitant use of intrathecal chemotherapy, prior reports indicate intrathecal chemotherapy has poor penetration into the CNS parenchyma,^15-17 and would therefore unlikely be effective for the deep CNS parenchymal lesions in this patient. Patient 2 had SCNSL with leptomeningeal disease, positive CSF cytology, and multiple large enhancing CNS parenchymal lesions measuring up to 33 mm. The patient was started on treatment with pola and ibrutinib. A restaging brain MRI after 6 weeks of treatment revealed a complete remission with resolution of the large contrast enhancing brain lesions (Figure 2C). A lumbar puncture also confirmed complete clearance of lymphoma from the CSF. Ibrutinib was discontinued in this patient after 4 months because of a clinically significant gastrointestinal bleed. The patient was continued on pola treatment for a total of 11 doses and remains in ongoing complete remission for more than 13 months since starting treatment. Finally, patient 3 had refractory PCNSL with a large 60 mm left frontal lobe mass, and was treated with pola in combination with mosunetuzumab. Improvement in focal neurologic deficits (right-sided hemiplegia) was noted within the first week of treatment. A total of 6 cycles were administered with a best response of stable disease, at which time the patient electively discontinued treatment.

In this study, we provide preclinical support for the activity of pola in CNSL xenograft models and demonstrate pola can partially penetrate the BBB in patients with CNSL. Pola CSF drug levels in patients were 0.56% to 1.31% of those observed in plasma, with pola CSF drug concentrations being notably higher than its established 50% inhibitory concentration for several lymphoma cell lines.¹⁴ This degree of BBB penetration is comparable or perhaps slightly higher than what has been described with other therapeutic antibodies in CNSL including rituximab (0.1%) and glofitamab (0.1%-0.4%),^8,18 as well as small molecule inhibitors such as ibrutinib (∼1.4%).¹⁹ Thus, these data provide additional support for the ability of therapeutic antibodies to partially penetrate the BBB in patients with CNSL.

Although demonstrating the clinical efficacy of pola in CNSL was not the primary objective of the study, we did observe that 2 of 3 patients with CNSL treated with pola-based therapy achieved complete responses to treatment. The extent to which pola contributed to the clinical responses in these 2 patients cannot be fully established because both were treated with concomitant therapies. However, these preliminary clinical observations combined with the pola CSF drug levels and preclinical xenograft results, provide hypothesis-generating data to support the evaluation of pola-based therapies in future prospective CNSL clinical trials. Indeed, we hypothesize that pola could be uniquely positioned to benefit CNSL from a disease biology perspective, as the majority of CNSL exhibit an activated B-cell like cell-of-origin,²⁰ which is known to be associated with increased pola sensitivity.^2,21 The clinical activity of pola in CNSL, however, will need to be confirmed through future carefully designed clinical trials in which its safety and efficacy can be more formally assessed.

Although our data provide support for evaluating pola in CNSL there are potential concerns with this approach. For instance, there were no significant differences in the CNS relapse rates between the 2 treatment arms of the POLARIX study, which compared pola-R-CHP to R-CHOP in previously untreated DLBCL.² While these data suggest that pola could be ineffective as prophylaxis against the development of CNSL in patients with DLBCL, relevant retrospective studies have also indicated that the incorporation of either intrathecal or high-dose IV methotrexate as CNS prophylaxis during frontline DLBCL treatment also does not significantly reduce CNS relapse rates despite methotrexate being one of the most effective treatments for CNSL.^13,22,23 Therefore, these data suggest that a therapy does not need to be capable of preventing CNS relapse for it to be an effective treatment for CNSL. A second concern for evaluating pola in CNSL is the neurotoxic MMAE payload, as it is conceivable that delivering pola to CNS tumors could be associated with enhanced neurologic toxicities because of the diffusion of MMAE to surrounding brain tissue. Because neurologic toxicities from pola are related to unconjugated “free” MMAE, this is thought to be a low risk since we find unconjugated MMAE levels are low in the CSF of patients with CNSL. Supporting this notion, we did not observe neurologic toxicities in our clinical case series. Similarly, much larger clinical studies evaluating trastuzumab emtansine, a HER2 ADC with a similar neurotoxic microtubule inhibitor payload, have not identified enhanced neurologic toxicities in patients treated with brain metastases from HER2⁺ cancers.^24,25 Thus, these data suggest that treatment with microtubule-based ADCs can be safe in patients with brain malignancies. However, given the theoretical increased risk for neurologic toxicities in patients with CNSL treated with an MMAE-based ADC, the investigation of other ADCs with nonneurotoxic payloads such as the CD19 ADC, loncastuximab tesirine, should also be considered in future studies.

Finally, these data extend prior findings supporting the penetration of other antibody-based therapies (monoclonal, bispecific, and ADC) across the BBB for a variety of CNS malignancies including brain metastases from melanoma, lung, and breast cancer.^3,7-9,26-31 Importantly, many of these studies have also suggested that antibody-based therapies yield similar efficacy results in patients with and without brain metastases.^27,29 For example, recent data with the HER2 ADC, trastuzumab deruxtecan, indicate response rates of 54% vs 62% in patients with and without active brain metastases from HER2⁺ breast cancer, respectively.²⁷ Thus, there is a growing body of literature that suggests antibody-based therapies have significant potential to penetrate the BBB and improve the stubbornly poor outcomes of patients with a variety of CNS malignancies.

The authors thank the participants of their repository protocols for support of this research. Human specimens were collected from patients registered at City of Hope National Medical Center (COHNMC) who had consented to an institutional review board–approved protocol. Patient sample acquisition was approved by the institutional review boards at the COHNMC, in accordance with an assurance filed with and approved by the Department of Health and Human Services, and met all requirements of the Declaration of Helsinki.

Research reported in this publication included work performed in the Hematopoietic Tissue Biorepository Core supported by the National Cancer Institute of the National Institutes of Health under grant P30CA033572. The research was also supported by City of Hope Comprehensive Cancer Center grant 2K12CA001727-29 (J.K.G.).

The research content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Contribution: L.G. performed the in vivo studies and data analysis; G.S., A.K., and J.H.B. contributed to the clinical case series; S.P. analyzed the pharmacokinetic data; Y.Q. assisted with in vivo studies and created the graphical abstract; J.Y.S performed pathologic review of the clinical cases; G.M., L.Y.G., and C.G. oversaw the consenting and acquisition of biobanked patient samples; S.-J.H., O.M.S., H.D., and R.D. developed and qualified the bioanalytical assays and analyzed concentrations of antibody-conjugated monomethyl auristatin E (MMAE) and unconjugated MMAE in the cerebrospinal fluid and plasma samples; J.K.G. designed the study, analyzed the data, and wrote the manuscript; and all authors reviewed and edited the manuscript.

Conflict-of-interest disclosure: J.K.G. reports research support from Merck, Janssen, AbbVie, Genmab, and Genentech. L.E.B. reports research support from Merck, Amgen, Cellular Biomedicine Group, and AstraZeneca; and provides consultancy as part of an advisory board for ADC Therapeutics, AstraZeneca, AbbVie, Nurix, Kite Pharma, Bristol Myers Squibb, Janssen, Regeneron, Genentech, and Roche. G.S. provides consultancy as part of an advisory board for AstraZeneca, BeiGene, and Kite Pharma; and serves on a speaker’s bureau for BeiGene and Kite Pharma. A.K. is on the speaker’s bureau for Genmab. J.H.B. reports research support from Genentech, Janssen, Regeneron, and Kite Pharma; and provides consultancy as part of an advisory board for Kite Pharma. S.-J.H., O.M.S., H.D., R.D., C.L.B., and J.M. are employees of Genentech. The remaining authors declare no competing financial interests.

Correspondence: James K. Godfrey, Department of Hematology and Hematopoietic Cell Transplantation, City of Hope Comprehensive Cancer Center, 1500 East Duarte Rd, Duarte, CA 91010; email: jagodfrey@coh.org.