Delivery versus potency in treating brain tumors: BI-907828, a MDM2-p53 antagonist with limited BBB penetration but significant in vivo efficacy in glioblastoma

Abstract

MDM2-p53 inhibition may be effective in glioblastoma. This study evaluates the PK/PD of BI-907828, a potent antagonist of MDM2, in GBM, and demonstrates a translational paradigm with a focus on a unified “Delivery – Potency – Efficacy” relationship in drug development for CNS tumors. BI-907828 was tested for cytotoxicity and MDM2-p53 pathway inhibition. Systemic pharmacokinetics and transport mechanisms controlling CNS distribution were evaluated in mice. BI-907828 free fractions in cell media, mouse and human specimens were measured to determine “active” unbound concentrations. Efficacy measures, including overall survival and target expression were assessed in mouse orthotopic GBM xenografts. BI-907828 exhibited potent inhibition of MDM2-p53 pathway and promoted cell death in GBM TP53 wild-type cells. MDM2-amplified cells are highly sensitive to BI-907828, with an effective unbound concentration of 0.1 nM. The CNS distribution of BI-907828 is limited by BBB efflux mediated by P-gp, resulting in a K_{p,uu_brain} of 0.002. Despite this seemingly “poor” BBB penetration, weekly administration of 10 mg/kg BI-907828 extended median survival of orthotopic GBM108 xenografts from 28 to 218 days (P < 0.0001). This excellent efficacy can be attributed to high potency, resulting in a limited, yet effective, exposure in the CNS. These studies show that efficacy of BI-907828 in orthotopic models is related to high potency even though its CNS distribution is limited by BBB efflux. Therefore, a comprehensive understanding of all aspects of the “Delivery – Potency – Efficacy” relationship is warranted in drug discovery and development, especially for treatment of CNS tumors.

Introduction

Discovery of effective compounds and subsequent delivery of “enough” active compound to the target site are critical steps in the drug development process. In oncology, medicinal chemists have made significant contributions in discovering potent drug candidates for oncogenic signaling pathways. However, adequate delivery of potent drugs to the tumor is a challenge, especially for central nervous system (CNS) tumors, such as glioblastoma (GBM). The blood-brain barrier (BBB) protects the CNS from both endogenous and exogenous compounds through its physical barrier properties (e.g., tight junctions) and biochemical barrier properties (e.g., efflux transporters), and is a major obstacle to delivery to brain (1,2). P-gp and BCRP, efflux systems expressed in the BBB, actively transport out of the brain and the restricted BBB penetration of many drugs can be attributed to these efflux transporters.

p53, a tumor suppressor, is a master regulator of response to genotoxic damage (Figure S1). In resting cells, p53 is maintained at low levels by a series of regulation factors including murine double minute 2 (MDM2), which functions as E3 ubiquitin ligase leading to ubiquitin-mediated p53 (3,4). When cells are exposed to genotoxic stress, p53 is activated by a series of post-translational modifications (e.g., phosphorylation) that free it from MDM2, resulting in a rapid rise of p53 levels (5–7). The activated p53 subsequently promotes cell cycle arrest or apoptosis to prevent tumor growth (6–8).

Given the significant role of MDM2 in the degradation of p53, blocking the protein-protein interactions between MDM2 and p53 with small-molecule antagonists has gained intense interest for treating tumors with wild-type p53 (5,9,10). BI-907828, a novel MDM2-p53 antagonist, has displayed encouraging clinical results in patients with liposarcoma and hepato-biliary tumors with a tolerable side effect profile (11,12). BI-907828 also has shown robust cytotoxicity and significant survival benefits in GBM patient-derived xenograft (PDX) models (13,14), that is currently in phase 0/1a evaluation (NCT05376800) in GBM patients at the Mayo Clinic.

The objective of this study was to evaluate the pharmacokinetics (PK), pharmacodynamics (PD), and survival benefit of BI-907828 in GBM, exploring the relationship between drug potency, drug delivery to site of action, and efficacy in GBM PDX mouse models. The results from this study demonstrate that despite limited BBB penetration, BI-907828 exhibits significant therapeutic efficacy against GBM due to its extremely high potency in inhibiting the MDM2-p53 pathway. Through this study, we affirm that a comprehensive evaluation of drug “Delivery – Potency – Efficacy” relationship is warranted in the drug discovery and development process, especially for the treatment of CNS tumors.

Materials and Methods

Drug and Cell Culture

BI-907828 was obtained from Boehringer Ingelheim (Ingelheim/Rhein, Germany). The chemical structure of BI-907828 was disclosed recently (Figure S2) (15). The synthesis of BI-907828 can be found in patent application WO 2017/060431 A1 (compound Ia-34) (16). GBM patient-derived xenograft (PDX) short-term explant cultures (GBM10, GBM14, GBM120, GBM108) were maintained in serum free stem cell media (StemPro NSC SFM, ThermoFisher) on plates coated with 1 μL/cm² laminin (Sigma) as previously described (17). GBM120 has two TP53 mutations – c.473G>A_p.Arg158His and c.747G>T_p.Arg249Ser. GBM10 and GBM14 have wild-type MDM2, while GBM108 has amplified MDM2. All experiments were performed within 1 – 2 passages. Cell lineage authentication was performed by short tandem repeat analysis in conjunction with each animal study.

Cell viability assays and in vitro drug treatment

Short term explant cultures (17) were plated at a density of 2,000 cells/well in 96 well plates and cell viability measured using Cell Titer-Glo 3D (Promega) 7 days after treatment with 0 – 100 nM BI-907828. For each PDX model, two (GBM120) or three (GBM10, GBM14, and GBM108) independent experiments were performed, and results are reflective of three independent experiments.

Western blot

Cultured tumor cells were collected 6 hr after treatment with 0, 3, or 10 nM BI-907828. Protein extraction and Western blot was performed as previously described (18). Briefly, Western blot was performed using primary antibodies directed against PUMA (Cell Signaling #12450, 1:500), p53 (Cell Signaling # 9282S, 1:2000), MDM2 (Cell Signaling #86934S, 1:1000), p21 (Cell Signaling # 2947, 1:1000), β-actin (Cell Signaling # 4967, 1:1000) and (HRP)-conjugated rabbit secondary antibody (Cell Signaling #7074S, 1:5000).

Animal models

All animal studies were conducted in accordance and under the auspices of an approved Institutional Animal Care and Use Committee protocol in the University of Minnesota or Mayo Clinic, Rochester. FVB mice (8 – 16 weeks) of different transporter-deficient genotypes (wild-type, Mdr1a/b^−/−, Bcrp1^−/−, Mdr1a/b^−/−Bcrp1^−/−) were used for pharmacokinetic studies (19). PDX models were established and maintained in female athymic nude mice (Hsd:athymic Nude-Foxn1^nu) at age 6 – 7 weeks (Charles River Laboratories).

Orthotopic xenografts were established in athymic nude mice after ~ 10–14 days by intracranial injection of 3 × 10⁵ GBM cells suspended in PBS as previously described (17). To enable accurate tumor localization, GBM108 cells were transduced with a modified pGIPZ lentiviral vector encoding a fusion of firefly luciferase (Luc2) with tandem tomato (tdT) red fluorescent protein or enhanced green fluorescent protein (eGFP) as described (20).

Drug binding studies

The unbound (free) fractions of BI-907828 in various matrices were determined by the rapid equilibrium dialysis (RED) method. Details are listed in the supplemental materials.

PK studies after single intravenous or oral administration

BI-907828 was administered to FVB wild-type, P-gp-knockout (Mdr1a/b^−/−), BCRP-knockout (Bcrp1^−/−), and triple-knockout (Mdr1a/b^−/−Bcrp1^−/−) mice intravenously or orally as a single dose of 10 mg/kg. Multiple blood, brain and spinal cord specimens were harvested (n = 4 in each time point), from 10 min to 32 hr postdose, following euthanasia by CO₂ inhalation. Drug formulation and sample collection and preparation details are listed in supplemental information.

Distribution of BI-907828 in brain regions and other major organs at steady state

Alzet^® micro-osmotic pumps (Model 1003D; DURECT Corporation, Cupertino, CA) were primed and then implanted into the intraperitoneal cavity of FVB wild-type and P-gp-knockout (Mdr1a/b^−/−) mice, and infused for 66 hr to achieve steady state at the constant infusion rate of 10 μg/hr (n = 6 for each genotype). The surgical procedure has been previously described (21). Mice were euthanized by CO₂, followed by the collection of plasma, major organs, and different brain anatomic regions.

Distribution of BI-907828 in orthotopic GBM tumor models

Athymic nude mice with established orthotopic GBM108-fLuc2/eGFP tumors were treated with a single oral dose of 2 mg/kg or 10 mg/kg BI-907828, followed by euthanasia at 6 hr or 24 hr postdose (n = 5 for each treatment). Plasma and tumor-bearing brain were collected and flash frozen prior to processing. A fluorescence-guided dissection sampled the intracranial tumor and normal brain regions from each brain (Figure S3).

In vivo pharmacodynamics

Athymic nude mice with established orthotopic GBM108-fLuc2/tdT tumors were treated with a single dose of vehicle or BI-907828 (2 or 10 mg/kg), euthanized, and tumor tissue collected after 48 hr. Whole brains were serially sectioned and frozen tissue blocks prepared in OCT compound (Sakura Tissue-Tek) as described (17). Serial 5 μm sections were cut for hematoxylin and eosin (H&E) staining and immunohistochemistry (IHC). IHC was performed at the Mayo Pathology Research Core using antibodies against Ki67 (Clone EP5; Cell Marque; 1:1000) and cleaved caspase 3 (polyclonal; Cell Signaling; 1:200) on a Leica Bond RX stainer (Leica).

Efficacy of BI-907828 in orthotopic PDX

Athymic nude mice with established orthotopic GBM108-fLuc2/tdT tumors were randomized 10 – 14 days after tumor implantation to treatment with vehicle, 2 mg/kg, or 10 mg/kg BI-907828 (n = 9 or 10 per group) by oral gavage once weekly until moribund. Mice with established GBM10 or GBM14 orthotopic tumors were randomized to treatment with vehicle or 10 mg/kg BI-907828 weekly until moribund (n = 9 or 10 per group).

LC-MS/MS analysis of BI-907828

BI-907828 concentrations in all specimens were determined by LC-MS/MS. Details are given in the supplemental materials.

PK calculations

BI-907828 concentration in the brain was corrected by estimating residual blood volume in the brain microvessels; determined to be 1.4 % of whole brain volume (22). Correction for spinal cord concentration was in the same method by assuming that vascular space in spinal cord was the same as brain. Pharmacokinetic data analysis was performed in Phoenix WinNonlin (Certara USA, Inc., Princeton, NJ) by using non-compartmental analysis. The variance of AUC_0−∞ was calculated using Yuan’s method with to the sparse sampling in pharmacokinetic studies (23). Partition coefficients (K_p) and unbound partition coefficients (K_p,uu) were calculated through following equations. The variances of K_p and K_p,uu were calculated through propagation of error formulae for division (24).

$$K_{p\_\mathit{\text{brain}}}=\frac{{AUC}_{0-\infty ,\mathit{\text{brain}}}}{{AUC}_{0-\infty ,\mathit{\text{plasma}}}}$$

$$K_{p,uu\_\mathit{\text{brain}}}=\frac{{AUC}_{0-\infty ,\mathit{\text{brain}}}}{{AUC}_{0-\infty ,\mathit{\text{plasma}}}}\times \frac{f_{u,\mathit{\text{brain}}}}{f_{u,\mathit{\text{plasma}}}}$$

Statistical analysis

All other data represent mean ± S.D., and P < 0.05 was considered statistically different for all comparisons. A two-tailed t-test was used to make comparison between two groups. One-way ANOVA was used to analyze the difference among groups. Animal survival was estimated by the Kaplan – Meier analysis. All statistical analyses were performed using GraphPad Prism 9.4.1 (GraphPad Software, La Jolla, CA).

Data availability

The summary data generated in this study are available within the article and the supplementary materials. Additional raw data will be provided upon request.

Results

In vitro efficacy of BI-907828 in GBM

The effects of BI-907828 on cell viability and p53 pathway activation were evaluated in vitro across a panel of GBM PDXs with varying MDM2 and p53 gene alteration status (Figure 1A). Across three p53 wild-type (WT) models, BI-907828 decreased cell viability over a concentration range 0 – 100 nM as measured by Cell Titer Glo 3D viability assay (Figure 1B). GBM108 (MDM2-amplified) showed the greatest sensitivity to total BI-907828 concentrations (IC50 = 0.53 nM, IC90 = 4.78 nM) relative to the p53-WT, MDM2 non-amplified GBM10 (IC50 = 5.33 nM, IC90 = 47.94 nM) and GBM14 (IC50 = 0.86 nM, IC90 = 7.75 nM) models. In contrast, GBM120, which harbors TP53 p.R158H and p.R249S mutations, was resistant to BI-907828 across this concentration range.

Figure 1.

Effects of BI-907828 on viability and p53 pathway modulation in vitro. (A) Summary of p53 and MDM2 status of PDX lines. (B) Effects of BI-907828 on viability. Short-term explant cultures were treated with varying concentration of BI-907828 (0–100 nM) and cell viability measured after 7 days using the Cell Titer Glo 3D assay. Values represent mean ± S.D. of 2–3 independent experiments. (C) Expression of p53 and target genes evaluated by Western blot 6 hours after BI-907828 treatment.

Modulation of the p53 pathway by BI-907828 in vitro was assessed by Western blot (Figure 1C, Figure S4). As shown in Figure 1C, cells treated with 0, 3, or 10 nM BI-907828 showed a concentration-dependent modulation in expression of the direct p53 transcriptional targets PUMA and p21 at the protein level. Expression of MDM2 was elevated in GBM108, consistent with the presence of MDM2 amplification, and MDM2 expression was increased by BI-907828 across all models. MDM2 is transcriptionally activated by p53, providing a negative feedback loop to control p53 expression (7). Overall, the results above demonstrate that BI-907828 is a highly potent MDM2-p53 antagonist that suppresses the in vitro proliferation of GBM tumors.

Unbound fraction of BI-907828

The in vitro and in vivo exposure of unbound (free) BI-907828 can be assessed by measuring the total concentration and determining the unbound fraction (fu) in a specific matrix. The unbound fractions of BI-907828 in various matrices; including cell culture media, and mouse and human tissue specimens, were measured by rapid equilibrium dialysis, and are shown in Table 1. BI-907828 was highly bound in all matrices. The percent free fraction (% fu) of BI-907828 in stem cell media, utilized in all in vitro studies, was 2.24 ± 0.17 %. Based on the in vitro cell viability results in Figure 1B, this unbound fraction resulted in a target unbound concentration of 0.1 nM (unbound IC90, 2.24 % × 4.78 nM) BI-907828 in GBM108 cells.

Table 1.

Unbound fraction of BI-907828 in cell media, mouse and human specimens

Letters for each GBM patient represent distinct tumor regions.

Data represent mean ± S.D.

^####P < 0.0001 compared with human brain homogenate

^****P < 0.0001 comparison among 13 GBM tumors

The unbound fractions of BI-907828 in mouse plasma and CNS tissues are 0.15±0.01% and 0.03±0.01%, respectively, that were used to estimate the free drug partition coefficient, K_p,uu, in brain and spinal cord. The unbound fractions of BI-907828 among 13 separate tumor regions derived from 4 GBM patients were statistically greater (P < 0.0001) when compared to normal human brain tissue (Table 1). The average unbound fraction in GBM tumors was 0.17 %, similar to plasma and higher than normal brain (~ 3-fold), suggesting more free drug in tumor tissues than normal brain when the total drug exposures are the same in tumor and surrounding normal brain.

Distribution of BI-907828 in mouse after systemic administration

The systemic PK and CNS distribution of BI-907828 were evaluated in FVB wild-type mice following either intravenous or oral administration. BI-907828 was orally bioavailable (F = 0.79) with a long half-life of 10 hr (Table 2). When intravenously dosed to a wild-type mouse, BI-907828 distributed with a volume of distribution of 643 mL/kg and was cleared from the systemic circulation at a plasma clearance of 48.7 mL/hr/kg (Table 2).

Table 2.

Pharmacokinetic parameters of BI-907828 in FVB mice after a single dose of 10 mg/kg

Parameter (unit)	Wild-type	P-gp-knockout	Wild-type
	Intravenous	Intravenous	Oral
t1/2 (hr)	10.2	9.4	10.6
AUC0−∞, plasma (hr*ug/mL)	205 ± 16	310 ± 16	162 ±19
AUC0−∞, brain (hr*ug/mL)	1.91 ± 0.25	81.6 ± 5.3	1.22 ± 0.15
CL or CL/F (mL/hr/kg)	48.7	32.2	61.8
Vd or Vd/F (mL/kg)	643	443	948
K p	0.009 ± 0.012	0.263 ± 0.227	0.008 ± 0.016
K p,uu	0.002 ± 0.003	0.059 ± 0.051	0.002 ± 0.004
Distribution Advantage	1	28.3	-

Data represent mean ± S.D. for AUC_0−∞ (n=4 at each time point).

Abbreviations: t_1/2, half life; AUC_{0- ∞}, area under the curve from zero to infinite; CL, CL/F, clearance; Vd, Vd/F, volume of distribution; Kp, the ratio of AUC_{(0−∞,brain)} to AUC_{(0−∞,plasma)} using total drug concentrations; K_p,uu, (AUC ratio), the ratio of AUC_{(0−∞,brain)} to AUC_{(0−∞,plasma)} using unbound drug concentrations; Distribution Advantage, the ratio of Kp_knockout to Kp_WT.

Drug exposure in the CNS tissues (Figure 2A), as measured by the area under the curve (AUC), was significantly lower than in plasma, with a total partition coefficient (K_p) of 0.009 in brain (Table 2) and 0.021 in spinal cord (Table S1). The values of K_{p_brain} and K_{p_spinal cord} indicated that after a single injection, only 1–2 % BI-907828 accumulated in the CNS, demonstrating limited BBB penetration. Similarly, after single oral dose, BI-907828 had limited distribution into brain and spinal cord with K_p values of 0.008 and 0.026, respectively (Table 2, Table S1).

Figure 2.

In vivo distribution of BI-907828 in plasma, brain and spinal cord. (A, C) Plasma, brain, spinal cord concentration-time profiles of BI-907828 in FVB wild-type (A) and P-gp-knockout mice (C) after a single intravenous administration of 10 mg/kg. (B) Concentration-time profiles of BI-907828 after a single oral dose of 10 mg/kg in FVB wild-type mice. (D) Total drug distribution of BI-907828 within CNS tissues, including different anatomic brain regions, in wild-type and P-gp-knockout mice after an intraperitoneal infusion to steady state. “T” - thalamus; “HT” - hypothalamus.

Data represent mean ± S.D., n=4 at each time point in (A-C); n=5 at each time point in (D) except medulla which was pooled for concentration analysis.

Efflux transporters P-gp (coded by Mdr1a/b) and BCRP (Bcrp1) function as biochemical barriers at the BBB to actively efflux drugs out of the brain. Transport mechanisms controlling limited CNS distribution of BI-907828 were examined using three genotypes of transgenic mice, lacking either or both of the efflux transporters (Mdr1a/b^−/−, Bcrp1^−/−, Mdr1a/b^−/−Bcrp1^−/−). After a single intravenous dose of BI-907828, limited brain distribution was observed in BCRP-knockout (Bcrp1^−/−) mice (K_{p_brain} = 0.006) (Table S2), similar to wild-type mice (K_{p_brain} = 0.009). These data indicate that BCRP is not a significant contributor to the limited BBB penetration in wild-type mice. However, when P-gp was genetically deleted, in P-gp-knockout (Mdr1a/b^−/−) or triple-knockout (Mdr1a/b^−/−Bcrp1^−/−)), BI-907828 distribution into brain and spinal cord was significantly enhanced by approximately 20-fold (wild-type K_{p_brain} = 0.009; P-gp-knockout K_{p_brain} = 0.263; triple-knockout K_{p_brain} = 0.192) (Table 2, Table S2). We can conclude that P-gp was the primary efflux transporter preventing the extent of BI-907828 distribution into the CNS. Moreover, the P-gp-knockout and triple-knockout mice, lacking P-gp-mediated efflux, require a longer time to reach a distributional equilibrium of BI-907828 between the CNS tissues and plasma (Figure S5C–S5D) (25).

The total exposures of BI-907828 in anatomic brain regions (cortex, cerebellum, pons, medulla, thalamus-hypothalamus, and midbrain) and spinal cord were also investigated in wild-type and P-gp-knockout mice after an intraperitoneal infusion to steady state (Figure 2D). Consistent with single dose results, limited BBB penetration was observed in all brain regions and spinal cord in wild-type mice, and the differences among brain regions were negligible within the same genotype (P = 0.58 for wild-type; P = 0.21 for P-gp-knockout).

According to the free drug hypothesis, only free (unbound) drug will interact with the pharmacological targets, meaning that the unbound drug concentration is actually the critical concentration leading to pharmacological responses (26). The unbound brain-to-plasma partition coefficient, K_p,uu, determined by the ratio of free drug AUC in brain to the free drug AUC in plasma, was used to measure the extent of “active” drug penetration across the BBB. The K_{p,uu_brain} of BI-907828 was 0.002 in wild-type mice (Table 2), indicating a significant limitation of drug delivery to normal brain by the BBB. When the primary efflux transporter for BI-907828, i.e., P-gp, was genetically deleted, the K_{p,uu_brain} of BI-907828 increased to 0.059 in P-gp-knockout mice (Table 2) and 0.043 in triple-knockout mice (Table S2), a 28-fold and 21-fold enhancement of free drug delivery, respectively. Due to similar transporter-mediated efflux in spinal cord, K_{p,uu_spinal cord} (0.004) of BI-907828 was similar to K_{p,uu_brain}, and lack of P-gp significantly enhanced delivery by 19-fold in the spinal cord (Table S1).

BI-907828 distribution into other organs was also evaluated in wild-type and P-gp-knockout mice after an intraperitoneal infusion to steady state (Figure S6). The distribution difference between the two genotypes was observed only in the CNS tissues (P < 0.0001); BI-907828 distribution into other organs was similar between wild-type and P-gp-deficient mice. The liver had the highest K_p, measured by tissue-to-plasma concentration ratio at steady state, followed by small intestine, kidneys, heart, lungs, bone marrow, and spleen (Figure S6B). The K_p of BI-907828 in the bone marrow was 0.39 in wild-type mice, a target organ for hematologic toxicity in clinical studies (27,28).

In vivo efficacy of BI-907828

The in vivo efficacy of BI-907828 was evaluated in three orthotopic GBM PDX models (Figure 3 and S7). Drug exposure of BI-907828 in orthotopic GBM108 tumors was quantitatively evaluated at 6 hr and 24 hr after single oral administration of 2 mg/kg and 10 mg/kg BI-907828 (Figure 3A). The total drug concentrations in plasma, intracranial tumors, and surrounding normal brain were measured by LC-MS/MS, and the free drug concentrations were determined. The intra-tumoral free BI-907828 concentrations were 1.00 nM after the dose of 10 mg/kg (0.11 nM with 2 mg/kg dose) after 6 hr and 0.66 nM (0.03 nM with 2 mg/kg dose) after 24 hr, respectively (Figure 3A). BI-907828 concentrations in tumors at 6 and 24 hr were significantly higher than in adjacent normal brain (6 hr, P = 0.0055 for 2 mg/kg, P = 0.0031 for 10 mg/kg; 24 hr, P = 0.0064 for 2 mg/kg, P = 0.0010 for 10 mg/kg), under the same driving force of the plasma concentrations, demonstrating BI-907828 had a longer residence time in tumor than in normal brain (Figure 3A). The drug concentration in tumor was less after 24 hr than after 6 hr, a result from the decline of the driving-force concentration in the plasma. The BI-907828 tumor concentration relative to plasma at 24 hr, represented by the tumor-to-plasma concentration ratio, was higher than at 6 hr, and these tumor-to-plasma ratios were significantly higher than normal brain-to-plasma ratios at either time point (Figure S8). These data could be explained by a decrease in the clearance process from the tumor back to the plasma, possibly through a loss in efflux, when compared to normal brain. Importantly, in addition to the longer residence time in the tumor, the free BI-907828 concentrations in tumors were higher than 0.1 nM (10-fold higher at 6 hr postdose in the 10 mg/kg group), the “active” free-concentration that was associated with the IC90 in GBM108 cells, with the exception of the 2 mg/kg after 24 hr (Figure 3A).

Figure 3.

PK-PD-efficacy of BI-907828 in orthotopic GBM108 tumor models. (A) Unbound concentration of BI-907828 in plasma, normal brain and tumor after a single oral dose of 2 mg/kg or 10 mg/kg at 6 hr or 24 hr. The unbound IC90 in GBM108, 0.1 nM, is defined as a target unbound concentration in GBM108 cells. *P < 0.05, **P < 0.01, ***P < 0.001, comparison between brain and tumor in each group by paired t-test. (B) Expression of cleaved caspase-3 (CC-3) and Ki67 in a representative GBM108 orthotopic tumor at 48 hr after treatment with vehicle, 2 or 10 mg/kg BI-907828 (scale bar, 100 μm). (C) Survival of animals with orthotopic GBM108 xenografts orally treated with BI-907828 at 0, 2, 10 mg/kg once weekly until moribund. Data represent mean ± S.D. (n=5).

We have also evaluated the impact of BI-907828 on apoptosis and tumor proliferation in orthotopic GBM108 tumor-bearing mice (Figure 3B). IHC for cleaved caspase-3 and Ki67 was performed on serial sections from the same GBM108 orthotopic tumors. The cleaved caspase-3 was markedly elevated and Ki67 proliferation index decreased at 48 hr postdose in mice treated with 2 or 10 mg/kg BI-907828, in a dose-dependent modulation (Figure 3B). Increased apoptosis and diminished tumor cell proliferation after the treatment of BI-907828 were consistent with an extension of survival in GBM models. When dosed at 2 mg/kg or 10 mg/kg weekly until moribund, BI-907828 extended median survival in GBM108 (MDM2-amplified) from 28 days (vehicle) to 57 days (2 mg/kg) (2.0-fold greater, P < 0.0001) and 218 days (10 mg/kg) (7.8-fold greater, P < 0.0001) (Figure 3C). Survival was also extended in GBM10 and GBM14 (TP53-WT, MDM2 non-amplified) intracranial tumor models (Figure S7). BI-907828 dosed at 10 mg/kg extended survival of GBM14 approximately 3.3-fold relative to vehicle treated animals (82 days vs. 25 days, respectively) and approximately 1.6-fold in GBM10 (40.5 days vs. 25 days, respectively) (P < 0.0001). These in vivo efficacy differences in different PDXs mirror the relative in vitro sensitivities of these tumor models to BI-907828 (Figure 1).

Discussion

In this study, we evaluated the in vitro drug potency, in vivo drug delivery, and survival efficacy of BI-907828 in orthotopic GBM PDX xenografts. Our data suggest that, despite limited BBB penetration, BI-907828 is highly efficacious in GBM pre-clinical models due to its extremely high potency in inhibiting the MDM2-p53 pathway. The results reaffirm that drug potency and BBB penetration are critical partners for efficacious treatment of brain tumors, especially invasive tumors like GBM (Figure 4) and need to be evaluated as a linked pair of parameters when making decisions for further development.

Figure 4.

“Delivery – Potency – Efficacy” relationship considerations for CNS drug discovery and development.

BI-907828 had a minimum effective concentration of 4.78 nM in GBM108 (p53-WT, MDM2-amplified) models in vitro as defined by IC90. When corrected for the unbound fraction of BI-907828 in stem cell media (2.24 %), the active (free) concentration was 0.1 nM in p53-WT, MDM2-amplified tumors, i.e., GBM108. In contrast, the TP53-mutant GBM120 model was resistant to BI-907828, consistent with efficacy of BI-907828 being linked to p53-dependent on-target effects (Figure 1B). Significantly higher free drug exposures were observed in tumor than in the surrounding normal brain in orthotopic GBM108 tumor models (Figure 3A). This is likely related to the more disrupted BBB in regions of tumor core (1,29), and higher free fraction in GBM tumors (Table 1). The free concentrations in tumor are above the effective free concentration (0.1 nM) observed for GBM108 in vitro (Figure 3A) and correlate with a remarkable survival extension (Figure 3C). Based upon these data, we can predict that 0.1 nM is an effective minimum intra-tumoral free concentration of BI-907828 targeting MDM2-amplified tumors. These preclinical data inform an ongoing phase 0 clinical trial in GBM (NCT05376800) where the primary outcome measures include the determination of free and total concentrations of BI-907828 in contrast enhancing and noncontrast enhancing tumor regions.

In addition to drug exposure, BI-907828 efficacy was determined by intrinsic sensitivity of tumor cells to the drug. We observed varying sensitivity of three TP53-WT GBM PDXs (GBM10, GBM14, and GBM108) to BI-907828 both in vitro (Figure 1B) and in vivo (Figure 3C, Figure S7). The MDM2-amplified GBM108 model showed profound sensitivity to BI-907828 and significant survival extension (Figure 3C), that was more limited in GBM14 and GBM10 models (Figure S7). This finding is in keeping with our prior studies, which demonstrate that MDM2-amplified tumors are highly sensitive to MDM2 inhibition (30,31). Ongoing studies are aimed at identifying biomarkers predictive of response in TP53-WT MDM2 tumors as well as the mechanisms by which MDM2 enhances sensitivity to MDM2 inhibition.

In addition to the positive influence of high potency, adequate exposure of drug to the tumor sites in the brain is critical for efficacy. In this regard, an unbound brain-to-plasma partition coefficient (K_p,uu) has been proposed to quantitatively describe “how the BBB handles the drug regarding passive transport and active influx/efflux” (32). This concept has been widely applied in both academia and the pharmaceutical industry to estimate drug delivery to the CNS in drug discovery programs, and make decisions about future development of compounds for CNS diseases (26). However, the K_p,uu values of currently known CNS-active drugs can vary by up to 150-fold (0.02 – 3) (32). This wide range raises an important, yet often ignored, question for CNS drug discovery and development: how much drug is enough for effective treatment? Are drugs with a higher K_p,uu necessarily associated with better efficacy? Our findings with BI-907828 indicate that compounds with high potency at the target can be effective even with “limited” drug delivery. BI-907828 had a limited distribution to brain and spinal cord in wild-type mice (Figure 2), with K_{p,uu_brain} of 0.002 following intravenous or oral administration (Table 2). The remarkable difference of K_{p,uu_brain} between wild-type and P-gp deficient mice suggests BI-907828 is a substrate of the efflux transporter P-gp. Our current study demonstrated potent efficacy of BI-907828 in orthotopic GBM tumors, regardless of the limited CNS distribution, i.e., a low K_{p,uu_brain}. In contrast, SAR405838, shown to be a relatively potent MDM2 inhibitor in vitro with a IC50 higher than 100 nM in GBM108, had a K_{p,uu_brain} of 0.007 but was ineffective in orthotopic tumors in the same GBM108 models, even with a daily dose of 50 mg/kg (31,33). Although similarly low free drug exposure (K_{p,uu_brain} < 0.01) was observed in BI-907828 and SAR405838, BI-907828 was significantly more effective following weekly dose of 2 and 10 mg/kg with much higher potency (31,33). SAR405838 was only efficacious when the BBB was disrupted with enhancing CNS exposure (31). The findings above imply that effective treatment depends on both drug potency and enough drug exposure at the target site. While this may seem obvious, but is not applied in current practice, drug potency and delivery must be considered in unison when developing treatments for CNS tumors. In screening compounds for CNS tumors, a predominance of BBB penetration in compound selection will result in the possible loss of potentially useful, high-potency drugs. A comprehensive, holistic understanding of each step in the drug “Delivery – Potency – Efficacy” relationship is highly warranted in drug discovery and development process for CNS tumors.

Many questions need answers in order to connect the “Delivery – Potency – Efficacy” relationships in this comprehensive drug development paradigm (Figure 4). These discoveries are key in determining the necessary connections in a “chain of drug development” where the success of this chain is only as likely as the strength of the weakest connection (link). For instance, once a pharmacological target is identified, compounds are screened to narrow the scope of drug candidates to those with adequate delivery (BBB penetration) and potency (in vitro efficacy). Drug potency and delivery properties should be evaluated together before candidate selection, since a highly potent compound may have adequate delivery even if the absolute exposure to the tumor is low. Drug potency properties are critical, e.g., total and unbound drug concentration ranges for in vitro efficacy, as well as reliable biomarkers to predict on-target effects. The discovery of predictive biomarkers has been challenging, but they are indispensable in predicting efficacy from phase 0/1 clinical studies. Other questions, such as the time course of effect after drug treatment are also essential. As for drug delivery evaluation, the systemic PK properties and CNS distribution of the drug, including total and free drug exposure in plasma and the target sites of drug action, PK parameters and metrics (e.g., half-life, K_p,uu), are necessary for dosing schedule (how much and how often) design for preclinical efficacy studies, informing phase 0/1 trials.

In essence, drug potency and delivery cannot be isolated in their contributions to efficacy. Does the drug show enough exposure (concentration and time) to the site of action following a designated dosage regimen? How long does it take the drug to activate the targets after drug administration? All of these questions are important and the answers should support understanding the relationships of a drug “Delivery – Potency – Efficacy” paradigm. In this relationship, the contribution of both drug potency and drug delivery to result in an efficacious treatment is important for clinical prediction. In the case of BI-907828 for GBM, the limited BBB penetration does not rule out eventual efficacy, given the remarkably high potency of the compound. BI-907828 would have been missed as a potentially useful GBM therapeutic if the initial evaluation was only based on the BBB penetration, without considering its high potency. Given our struggles over the years to develop an effective therapy for GBM, any practices in our discovery and development paradigms that may result in the loss of an effective treatment should be carefully reconsidered. In conclusion, this study supports that the efficacy of BI-907828 in orthotopic GBM models is related to high potency even though distribution of this drug into the normal CNS is limited. The findings from this study affirm that a comprehensive understanding of drug “Delivery – Potency – Efficacy” relationship is warranted in drug discovery development, especially for treatment of CNS tumors.