Research ArticleCancer

AZD3759, a BBB-penetrating EGFR inhibitor for the treatment of EGFR mutant NSCLC with CNS metastases

See allHide authors and affiliations

Science Translational Medicine  07 Dec 2016:
Vol. 8, Issue 368, pp. 368ra172
DOI: 10.1126/scitranslmed.aag0976

Crossing the BBB to pursue tumors

Non–small-cell lung cancer remains difficult to treat despite recent advances in targeted therapy. One reason for this is metastasis to the central nervous system. Drugs that inhibit the epidermal growth factor receptor (EGFR), a common target in this cancer, do not effectively penetrate the blood-brain barrier, which means that metastatic tumors can grow unchecked once they spread to the brain or spinal cord. Yang et al. have now developed a drug that not only can inhibit EGFR as effectively as clinically approved therapeutics but also can cross the blood-brain barrier to target metastases. This drug shows promising effectiveness in multiple different mouse models, as well as signs of antitumor activity in human patients.

Abstract

Non–small-cell lung cancer patients with activating mutations in epidermal growth factor receptor (EGFR) respond to EGFR tyrosine kinase inhibitor (TKI) treatment. Nevertheless, patients often develop central nervous system (CNS) metastases during treatment, even when their extracranial tumors are still under control. In the absence of effective options, much higher doses of EGFR TKIs have been attempted clinically, with the goal of achieving high enough drug concentrations within the CNS. Although limited tumor responses have been observed with this approach, the toxicities outside the CNS have been too high to tolerate. We report the discovery and early clinical development of AZD3759, a selective EGFR inhibitor that can fully penetrate the blood-brain barrier (BBB), with equal free concentrations in the blood, cerebrospinal fluid, and brain tissue. Treatment with AZD3759 causes tumor regression in subcutaneous xenograft, leptomeningeal metastasis (LM), and brain metastasis (BM) lung cancer models and prevents the development of BM in nude mice. An early clinical study in patients with BM and LM treated with AZD3759 confirms its BBB-penetrant properties and antitumor activities. Our data demonstrate the potential of AZD3759 for the treatment of BM and LM and support its further clinical evaluation in larger trials.

INTRODUCTION

Patients with central nervous system (CNS) metastases, such as leptomeningeal metastasis (LM) and brain metastasis (BM), have poor prognosis and low quality of life. Non–small-cell lung cancer (NSCLC) patients with epidermal growth factor receptor–activating mutations (EGFRm+) have a much higher risk of developing CNS metastases. The cumulative incidence of BM and LM in these patients is more than 50% (13). The median overall survival (OS) for patients with BM and those with LM is about 16 months (4) and 4.5 to 11 months (5, 6), respectively.

Whole-brain radiation therapy (WBRT) is a standard of care for BM. The toxicities associated with WBRT are severe, whereas its benefit for patient survival is questionable (7, 8). EGFR tyrosine kinase inhibitors (TKIs), such as gefitinib and erlotinib, have been evaluated in patients with BM. Some benefits have been reported, but they are usually variable and do not last, probably because of their poor capability to penetrate the blood-brain barrier (BBB) (9, 10). Consistent with this hypothesis, most patients with BM experienced CNS metastases as the first site of treatment failure after the initial response (1113). No drug has been approved to treat LM from lung cancer. The lack of effectiveness of available anticancer drugs is likely due to pharmacokinetics (PK) because only a few drugs can achieve effective concentrations in the cerebrospinal fluid (CSF) with standard doses. When patients were given doses that were up to 10-fold higher than the approved doses for gefitinib or erlotinib, drug concentration in the CSF could transiently reach the predicted efficacious level, and modest palliative effect was indeed observed. However, patients could not tolerate such a treatment because of serious adverse effects (13, 14). Other measures, such as the intrathecal injection of cytotoxic agents, TKIs, or antibodies, have been attempted with the same goal of achieving a high enough exposure in the CSF (15). However, this approach has not been approved as a standard of care for solid tumors with LM because of a lack of clinical evidence with large sample size. These observations suggest that a drug with much better CNS penetration is needed to effectively treat patients with BM and LM. To this end, we searched for such an agent and thus designed AZD3759, an oral, potent EGFR TKI with full BBB penetration and improved selectivity between mutant and wild-type (WT) EGFRs (16). This investigational product is currently being evaluated in an ongoing phase 1 study [a phase 1, open-label, multicenter study to assess the safety, tolerability, pharmacokinetics, and preliminary antitumor activity of AZD3759 in patients with EGFR mutation–positive advanced stage NSCLC (BLOOM); clinical trial identifier, NCT02228369]. Here, we report our assessment of its antitumor activities in a variety of preclinical animal models, as well as preliminary clinical data in patients with LM and BM.

RESULTS

AZD3759 demonstrated potent activity in EGFRm+ lung cancer cell lines in vitro and excellent CNS penetration in vivo

As shown in Table 1, AZD3759 demonstrated similar potency to erlotinib in its activity against phosphorylated EGFR (pEGFR) and inhibition of cell growth (GI) in tumor cells driven by EGFR L858R (H3255) and exon 19 deletion (PC-9). A greater selectivity between EGFRm+ and WT cells was observed for AZD3759 than for erlotinib, as measured by the inhibition of pEGFR. Additional tumor cell lines with overexpression of WT EGFR, activating mutations, and T790M mutation, which causes resistance to gefitinib and erlotinib, were tested for their sensitivity to AZD3759 (table S1). The EGFR protein expression of lung cancer cell lines is shown in fig. S1.

Table 1. Comparison of in vitro activity between AZD3759 and erlotinib.

pEGFR IC50 and GI50 were assessed as described previously (16). IC50, drug concentration causing 50% inhibition of pEGFR; GI50, drug concentration for 50% of maximal inhibition of cell proliferation; n, technical replicates.

View this table:

AZD3759 demonstrated excellent CNS penetration in mice. Kpuu,brain (the ratio of the unbound brain concentration to the unbound plasma concentration) and Kpuu,CSF (the ratio of the unbound CSF concentration to the unbound plasma concentration) were 0.65 and 0.42, respectively, which were significantly higher than those of erlotinib (Kpuu,brain, 0.13; Kpuu,CSF, 0.14) (P < 0.001 and P < 0.01 for Kpuu,brain and Kpuu,CSF, respectively) (Table 2). In addition, AZD3759 has other drug metabolism and pharmacokinetics (DMPK) properties needed for an excellent oral drug (16).

Table 2. Pharmacokinetic parameters of erlotinib and AZD3759 in nude mice.

Median Kpuu,brain and Kpuu,CSF of AZD3759 were 0.65 and 0.42, respectively. Median Kpuu,brain and Kpuu,CSF of erlotinib were 0.13 and 0.14, respectively. AUC, area under the curve.

View this table:

AZD3759 demonstrated profound antitumor activities in LM and BM models in nude mice

The LM model was established by injecting PC-9_Luc cells through the cisterna magna. When luciferin signals reached >5 × 106, animals were randomly assigned to different treatment groups. Vehicle, erlotinib, and AZD3759 were administered by oral gavage. On the basis of PK modeling, a dosage of 15 mg/kg once daily of erlotinib in nude mice gave exposure comparable to the 150-mg once-daily dosage in humans [150 mg once daily is the maximum tolerated dose (MTD) in humans (17)] (table S2). On the basis of the same PK modeling, AZD3759 at 7.5 and 15 mg/kg once daily would give comparable exposure to the predicted human efficacious dosage of 100 and 200 mg twice daily (18). At week 2.6 (the time point when >50% of the animals were alive and the measurement of luciferin signal was most reliable with minimal interference due to animal performance status), signal intensity in the animals treated with AZD3759 at 15 mg/kg was minimal (Fig. 1, A and B, and fig. S2), and animals gained body weight (Fig. 1C). Most animals (seven of eight) in the AZD3759 group survived longer than 4 weeks, significantly better than those in the erlotinib group (P = 0.0113, log-rank test; Fig. 1D). At the end of the experiment, no tumor was found in 50% of the animals in the AZD3759 group (table S3), and only small tumor areas on the leptomeninges were identified in the rest (fig. S3). Erlotinib somewhat decreased luciferin signal intensity (Fig. 1, A and B), but it neither prevented body weight loss (Fig. 1C) nor showed benefit in animal survival (Fig. 1D).

Fig. 1. Antitumor activity of AZD3759 in the PC-9 LM model.

PC-9_Luc cells were injected through the cisterna magna. After luciferin signals reached >5 × 106 photons/s, animals were randomly divided into three treatment groups: vehicle, erlotinib at 15 mg/kg, and AZD3759 at 15 mg/kg. (A) At week 2.6, significantly lower luciferin signals were detected in the erlotinib group than in the vehicle group [#P = 0.0083, two-way analysis of variance (ANOVA)]. Even lower luciferin signals were detected in the AZD3759 group than in the erlotinib group (**P < 0.0001, two-way ANOVA). (B) Representative images of luciferin signals at week 2.6 after treatment. (C) Body weight gain was observed in AZD3759-treated animals. (D) More than 80% of the animals in the AZD3759 group continued to survive 4 weeks after treatment, significantly better than the erlotinib group (**P = 0.0113, log-rank test).

The BM model was established by injecting PC-9_Luc cells through the carotid artery. Similar to the LM model, tumor growth in the brain was continuously monitored by luminescence intensity. Tumor growth was halted during the 8-week treatment period by AZD3759 at 7.5 mg/kg, and more profound tumor regression was observed at 15 mg/kg. Both 7.5 and 15 mg/kg doses of AZD3759 demonstrated significantly superior antitumor activity than erlotinib (P = 0.003 and P < 0.0001, respectively, two-way ANOVA) (Fig. 2, A and B, and fig. S4), with no obvious loss of body weight (Fig. 2C). There was a significant improvement in animal survival in the AZD3759 groups compared with the erlotinib group (P = 0.035 and P = 0.006 for AZD3759 at 7.5 and 15 mg/kg, respectively, log-rank test; Fig. 2D). At the end of the experiment, the histological assessment showed that the AZD3759 15 mg/kg group had significantly smaller tumor areas than the erlotinib group (P = 0.0312, one-way ANOVA; fig. S5), with 43% of the animals achieving tumor eradication in the brain (table S4).

Fig. 2. Antitumor activity of AZD3759 in the PC-9 BM model.

PC-9_Luc cells were injected through the carotid artery. After luciferin signals reached >6 × 108 photons/s, animals were randomly divided into four treatment groups: vehicle, erlotinib at 15 mg/kg, and AZD3759 at 7.5 and 15 mg/kg. (A) At week 4, significantly lower luciferin signals were detected in animals treated with AZD3759 at 7.5 and 15 mg/kg, compared with the erlotinib group (*P = 0.003 and **P < 0.0001 for AZD3759 at 7.5 and 15 mg/kg, respectively, two-way ANOVA). (B) Representative images of luciferin signals at week 4 after treatment. (C) Less than 5% body weight loss was observed in AZD3759-treated animals. (D) More than 70% of the animals continued to survive at week 8 with AZD3759 treatment, significantly better than the erlotinib group (*P = 0.035 and **P = 0.006 for AZD3759 at 7.5 and 15 mg/kg, respectively, log-rank test).

AZD3759 prevented the formation of tumor in the brain

Animals were treated with the vehicle, erlotinib (15 mg/kg), or AZD3759 (7.5 and 15 mg/kg) for 1 week before PC-9_Luc cells were implanted into the brain through intracerebral injection. In the vehicle group, luciferin signals reached a peak at week 2 (1 week after cell inoculation) (Fig. 3) and remained at the peak level during the 4-week observation period. There was a transient drop of the luciferin signal in the erlotinib-treated group at week 3, but the signal rebounded at weeks 4 and 5. Compared to erlotinib, AZD3759 significantly lowered luciferin signals in a dose-dependent manner [P < 0.0001 for both doses of AZD3759 (7.5 and 15 mg/kg), two-way ANOVA; Fig. 3 and fig. S6]. At the end of the study, animals were randomly selected for a histological assessment. All animals in the vehicle group and 83% of the animals in the erlotinib group had tumor formation in the brain, but 67 and 100% of the animals in the AZD3759 7.5 and 15 mg/kg groups remained tumor-free (P = 0.0152, AZD3759 at 15 mg/kg versus erlotinib, Fisher’s exact test; table S5).

Fig. 3. The effect of AZD3759 in the PC-9 BM prevention model.

Nude mice were treated with vehicle, erlotinib at 15 mg/kg, and AZD3759 at 7.5 and 15 mg/kg for 1 week before PC-9_Luc cells were implanted through intracerebral (ICB) injection. Mice continuously received the above treatment for another 4 weeks after cell inoculation. (A) At week 5 after the start of treatment, significantly lower luciferin signals were detected in AZD3759-treated animals (*P < 0.0001 and **P < 0.0001 for AZD3759 at 7.5 and 15 mg/kg, respectively, compared with the erlotinib group, two-way ANOVA). (B) Representative images of luciferin signals at weeks 2 and 5 after the start of compound treatment.

AZD3759 induced tumor regression in subcutaneous xenograft models

Although the properties of AZD3759 were suited for treating CNS metastases, to be a most useful drug, the agent should also control extracranial tumors. Two EGFRm+ cell lines (H3255 and PC-9) and one EGFR WT cell line (A549) were implanted subcutaneously. When tumor volume reached about 200 mm3, mice were randomly assigned to different groups with a comparable tumor volume at the baseline. Erlotinib (15 mg/kg) induced tumor regression in the H3255 (Fig. 4A) and PC-9 (Fig. 4B) subcutaneous xenograft (SC) models and about 28% inhibition of tumor growth in the A549 model (Fig. 4C) (P < 0.0001 in the H3255 and PC-9 models, P = 0.0065 in the A549 model, compared with the vehicle group, two-way ANOVA). AZD3759 behaved similarly in these SC models. There was no apparent body weight loss with different treatments (Fig. 4, D to F). Figures S7 to S9 show the tumor growth curves of individual animals in these three models.

Fig. 4. Antitumor efficacy of AZD3759 in the SC models.

H3255, PC-9, and A549 cells were implanted subcutaneously into nude mice. When tumor volumes reached about 200 mm3, mice were randomly divided into the following groups: vehicle, erlotinib at 15 mg/kg, and AZD3759 at 3.75, 7.5, and 15 mg/kg. (A) Tumor growth curve in the H3255 SC model. At week 2.5, both erlotinib and AZD3759 (all doses) resulted in smaller tumor volumes than the vehicle. There was no significant difference in tumor volume between the erlotinib and AZD3759 groups (#P < 0.0001, erlotinib versus vehicle, two-way ANOVA). (B) Tumor growth curve in the PC-9 SC model. At week 2, smaller tumor volumes were observed in both erlotinib- and AZD3759-treated animals (all doses) (#P < 0.0001, erlotinib versus vehicle; **P = 0.004, AZD3759 at 15 mg/kg versus erlotinib, two-way ANOVA). (C) Tumor growth curve in the A549 SC model. At week 4, tumor volumes were 72 and 50% of that in the vehicle group with erlotinib and AZD3759 15 mg/kg treatment, respectively (#P = 0.0065, erlotinib versus vehicle, two-way ANOVA). (D to F) Body weight changes in the H3255, PC-9, and A549 models. No body weight loss was observed in all groups.

PK/PD correlation was assessed in the BM, LM, and SC models

In the BM model, AZD3759 at 3.75 and 15 mg/kg showed a dose-dependent increase of free brain exposure after single oral dosing. Both doses reached their maximum concentration (Cmax) at 1 hour, and their free brain concentrations were above the pEGFR IC50 for about 5 and 10 hours, respectively. At these two doses, there was 50% reduction of pEGFR in the brain tumor tissue, which lasted for about 4 and 7 hours, respectively (Fig. 5A). Erlotinib (15 mg/kg) reached its free brain Cmax at 1 hour after dosing. Its free brain concentrations were above the pEGFR IC50 for about 5 hours. About 30% decrease of pEGFR in the brain tumor tissue was observed only at the Cmax time point (Fig. 5B). Figure 5 (C and D) shows representative images of pEGFR in the tumor tissues at 1 hour after dosing.

Fig. 5. Correlation between PK and PD in the PC-9 BM model.

(A) AZD3759 at 3.75 and 15 mg/kg reached its free Cmax in the brain at 0.5 hours after a single dose and declined to an undetectable level at about 16 hours. The free brain concentrations of 3.75 and 15 mg/kg were above the pEGFR IC50 of PC-9 cells for about 5 and 10 hours, respectively. At these two doses, more than 50% reduction of pEGFR in the brain tumor tissue lasted for about 4 and 7 hours, respectively. (B) Erlotinib reached its free brain Cmax at 1 hour after a single dose. Its free brain concentrations were above the pEGFR IC50 for about 4 hours. About 30% reduction of pEGFR was only detected at the Cmax time point. (C and D) Representative images of pEGFR in the brain tumor tissues at 1 hour after dosing. The inset is a magnification (×10) of the view field designated by the black box. Scale bars, 200 μm. Cu, free drug concentration.

In the LM model, AZD3759 at 15 mg/kg reached its Cmax in the CSF at 0.5 hours after a single oral dose. Its CSF concentrations were above the pEGFR IC50 for longer than 7 hours. More than 50% pEGFR inhibition was detected in the tumor cells on the leptomeninges, which lasted longer than 10 hours (Fig. 6A). Erlotinib reached its Cmax in the CSF at 1 hour after dosing. Its CSF concentration dropped below the pEGFR IC50 in less than 6 hours. At the Cmax time point, only about 30% inhibition of pEGFR was detected in the tumor cells (Fig. 6B). Figure 6 (C and D) shows representative images of pEGFR in the tumor cells on the leptomeninges. In addition to the tumor cells on the leptomeninges, an 89% decrease of pEGFR was also detected in tumor cells in the CSF after AZD3759 treatment (Fig. 6, E and F).

Fig. 6. Correlation between PK and PD in the PC-9 LM model.

(A) AZD3759 at 15 mg/kg reached its Cmax in the CSF at 0.5 hours after dosing. Its CSF concentrations were above the pEGFR IC50 for longer than 7 hours. About 50% decreased pEGFR in leptomeningeal tumor cells was sustained for longer than 10 hours. (B) Erlotinib achieved its Cmax in the CSF at 1 hour after dosing. Its CSF concentrations were above the pEGFR IC50 for about 6 hours. About 30% reduction of pEGFR was detected at its Cmax time point. (C and D) Representative images of pEGFR in the leptomeningeal tumor cells. The inset is a magnification (×10) of the view field designated by the black box. Scale bars, 200 μm. (E and F) Representative images and quantification of pEGFR in CSF tumor cells [pooled samples (n = 5 per group)]. Scale bars, 20 μm.

In the PC-9 SC model, AZD3759 at 3.75 and 15 mg/kg demonstrated a dose-dependent increase of free blood exposure after single dosing. Both doses reached their Cmax at 0.5 hours. The free blood concentrations of 3.75 and 15 mg/kg were above the pEGFR IC50 for about 8 and 14 hours, respectively. There was a more than 50% reduction of pEGFR in the tumor tissue achieved by the two doses, which sustained for longer than 7 and 14 hours, respectively (Fig. 7A). Erlotinib reached its free blood Cmax at 1 hour after dosing. Its free blood concentrations were above the pEGFR IC50 for about 6 hours. There was an about 50% decrease of pEGFR in the tumor tissue, which lasted for about 7 hours (Fig. 7B). Figure 7 (C and D) shows representative images of pEGFR in the tumor tissue.

Fig. 7. Correlation between PK and PD in the PC-9 SC model.

(A) After a single dose of AZD3759 at 3.75 or 15 mg/kg, free blood concentrations were above the pEGFR IC50 for about 8 and 14 hours, respectively. At these two doses, more than 50% reduction of pEGFR was sustained for longer than 7 and 14 hours, respectively. (B) After a single dose, free blood concentration of erlotinib was above the pEGFR IC50 for about 6 hours, with about 50% inhibition of pEGFR lasting for about 6 hours. (C and D) Representative images of pEGFR in the tumor tissue. Scale bars, 201.3 μm.

pEGFR target inhibition by AZD3759 and antitumor activity were noted in patients

AZD3759 is being evaluated in a phase 1 clinical study (BLOOM) (19). Modified Response Evaluation Criteria in Solid Tumor (mRECIST) were applied to assess tumor response to AZD3759 treatment. At doses of ≥100 mg twice daily, antitumor activity in the brain was observed in a number of patients (19). Here, we report one case of LM and one case of BM in this ongoing clinical study, as evidence of target inhibition by AZD3759 and its antitumor activity.

The first patient was a 56-year-old East Asian female from South Korea who was diagnosed with stage IIA EGFRm+ (L858R) NSCLC in 2012. After surgery, the patient received five cycles of cisplatin plus paclitaxel and two cycles of paclitaxel alone as an adjuvant therapy. The patient had a disease relapse in 2015 and received gefitinib treatment for 4 months, during which time she developed LM. The patient was enrolled in the dose escalation cohort of the BLOOM study and received AZD3759 treatment (200 mg twice daily). Plasma and CSF samples were collected before and 1 week after treatment to assess PK, pEGFR expression, and tumor cell numbers in the CSF. Only L858R (that is, no T790M) was detected in the CSF sample before treatment. The free plasma concentration of AZD3759 at the Ctrough time point was 23.5 nM, with a corresponding CSF concentration of 25.2 nM, resulting in Kpuu,CSF = 1.1. At the same time point, 71% pEGFR inhibition and 64% fewer tumor cells were found in the CSF (Fig. 8, A to C).

Fig. 8. Target inhibition by AZD3759 in a patient with LM and antitumor activity in a patient with BM.

(A to C) Quantification of the pEGFR signal in CSF tumor cells (A), a representative image of pEGFR (B), and CSF tumor cell number (C) in a patient with LM. After 1-week repeated dosing of 200 mg of AZD3759 twice daily, a 71% reduction of pEGFR and a 64% decrease in tumor cell number were detected. Scale bars, 20 μm. (D) Brain magnetic resonance imaging (MRI) scan of a patient with multiple BM. This patient had two measurable BM lesions (white nodules) in the occipital and parietal lobes before treatment, with longest diameters of 4.9 and 1.4 cm, respectively. After twice daily treatment of 100 mg of AZD3759 for 6 weeks, the tumor diameters reduced to 2.8 and 0.6 cm.

The second patient was a 65-year-old East Asian female from South Korea who was diagnosed with stage IV EGFRm+ (L858R) NSCLC in 2012. She received gefitinib treatment for 11 months and initially had a partial response. After disease progression, she was treated with six cycles of pemetrexed and four cycles of paclitaxel until she developed multiple BMs and was enrolled in the dose escalation cohort of the BLOOM study. She was treated with 100 mg of AZD3759 twice daily for a week before blood and CSF samples were taken for PK analysis. The free plasma and CSF concentrations of AZD3759 at the Ctrough time point were 11.8 and 16.7 nM, respectively, with a calculated Kpuu,CSF of 1.4. The brain MRI scan at week 6 showed 46% tumor shrinkage, compared with its size before treatment (Fig. 8D).

DISCUSSION

The present study focused on evaluating the biological activities of AZD3759 in a range of preclinical animal models (BM, LM, SC, and BM prevention models), which were designed to mimic the different disease settings in the clinic. With these models, the biological activities of AZD3759 were evaluated from a variety of angles: systemic (SC model), micrometastases [intracarotid (ICA) BM model], extensive BBB destruction (ICB BM model), prevention (BM prevention model), and tumor cells spreading throughout the subarachnoid space via CSF circulation (LM model). The resulting data would therefore potentially support clinical positioning of AZD3759 in different disease indications as well as disease stages.

Before EGFR TKI treatment, CNS metastases have the same EGFR mutations as primary tumors in the lung (20, 21). During EGFR TKI treatment, 63% of the patients acquire T790M mutation in extracranial lesions as a resistance mechanism (22). For patients with T790M mutation outside the CNS, less than 17% have detectable T790M mutation within CNS metastases (2325), suggesting that, in most cases, tumor metastasis to the CNS is likely an early event, before T790M mutation appears extracranially. This also suggests that there is no strong selection pressure for T790M mutation within the CNS, consistent with our finding that current TKIs cannot achieve and maintain efficacious concentrations within the CNS. For our LM patient, with L858R mutation only, gefitinib was effective in treating her extracranial lesions but could not stop intracranial tumor growth, whereas AZD3759 could, which further underscores the importance of BBB penetration for treating CNS metastases. Our data support the view that BBB remains largely functional in patients with CNS metastases (2628), and that to achieve optimal therapeutic effect, a drug with good BBB penetration is required.

In EGFRm+ SC models, AZD3759 demonstrated comparable activity with erlotinib, suggesting that AZD3759 could be as effective as erlotinib in controlling systemic disease. This is critical for AZD3759’s overall clinical value (5, 9, 29, 30). Erlotinib was chosen as a comparator based on its proven clinical activity as well as higher Kpuu,brain and Kpuu,CSF values than those of other approved EGFR TKIs (16). The selected dosage of 15 mg/kg once daily, which is equivalent to the human MTD (17), should reflect the highest achievable effect of erlotinib on CNS metastases. In the SC model, for both erlotinib and AZD3759, free drug concentration in the blood above the pEGFR IC50 for longer than 6 hours correlated with more than 50% reduction of pEGFR in the tumor tissue as well as tumor regression. The same was also true for AZD3759 in the BM and LM models, where tumor regression and improvement of animal survival were associated with free brain and CSF drug concentrations above the pEGFR IC50 for longer than 7 hours and greater than 50% reduction of pEGFR. In contrast, free brain and CSF drug concentrations of erlotinib were above the pEGFR IC50 for less than 6 hours, with a mild decrease of pEGFR detected only at Cmax time point, which may explain the weak efficacy of erlotinib observed in the BM and LM models. For the two patients treated with AZD3759, at the Ctrough time point, their CSF drug concentrations were still at least twofold higher than the pEGFR IC50. At the Ctrough time point, 71% reduction of pEGFR was observed in the CSF tumor cells, demonstrating good PK/PD (pharmacokinetics/pharmacodynamics) correlation in these patients. The fact that AZD3759 treatment resulted in shrinkage of brain tumors in a patient with BM who had progressed after multiple lines of treatment, including EGFR TKIs, further confirmed the clinical benefit of AZD3759.

Our data validated Kpuu,brain and Kpuu,CSF as reliable parameters for evaluating an agent’s ability to penetrate the BBB. It has been reported that a Kpuu,brain of >0.4 in animals is needed to have good CNS penetration in humans (31). Twice as much AZD3759 was needed to achieve the same extent of tumor regression in the BM and LM models as that in the SC models, predicted by its Kpuu,brain of 0.65 and Kpuu,CSF of 0.42 in mice. The Kpuu,CSF values for AZD3759 from the two patients were 1.1 and 1.4, which gives us confidence that AZD3759 should be similarly efficacious both intracranially and extracranially.

Over the last two decades, we have learned a lot about EGFR mutation–driven NSCLC from EGFR inhibitors preclinically and clinically. CNS metastases, however, are an emerging medical challenge. Our knowledge about their biology and treatment is still limited. This is especially true for LM, where there is no clear clinical diagnostic standard and OS is likely the only reliable end point for efficacy. The CNS metastases models used in this study are not widely used. Although we are very encouraged by the efficacy and PK data from the two patients, more patient data are needed to confirm the predictive value of these models and the effectiveness of the drug.

MATERIALS AND METHODS

Study design

This study was designed to assess the BBB penetration and biological activity of AZD3759 in preclinical animal models and preliminary antitumor activities in patients. The in vitro activity of AZD3759 was evaluated in cellular phosphorylation and proliferation assays. In vivo BBB penetration and antitumor efficacy were assessed in the LM, BM, BM prevention, and SC models. Sample size was estimated based on 80% power (one-sided). Randomization was applied for all efficacy studies. Drug concentrations in the blood, brain, and CSF samples were measured by liquid chromatography–tandem mass spectrometry (LC–MS/MS). EGFR phosphorylation in tumor tissue and CSF tumor cells was determined by immunohistochemistry (IHC) assay. In the clinical study, BBB penetration, target inhibition, and antitumor activity were assessed by LC–MS/MS, IHC assay, and mRECIST.

Cell lines

The human lung cancer cell lines NCI-H3255 (L858R), HCC827 (Exon19del), NCI-H1975 (L858R/T790M), NCI-H838 (EGFR WT), and A549 (EGFR WT) were obtained from the American Type Culture Collection. The PC-9 (Exon19del) cell line was obtained in 2011 from A. Hiraide at Preclinical Sciences R&D, AstraZeneca, Japan, and tested and authenticated by short-tandem repeat analysis in May 2013. The H3255 cells were maintained in bronchial epithelial basal medium (Lonza; CC-3171), containing 10% fetal bovine serum (FBS) (Gibco; 10099), supplemented with a BEGM kit (Lonza; CC-4175). The HCC827, H1975, H838, A549, and PC-9 cells were maintained in RPMI 1640 (Gibco; 22400), containing 10% FBS. Cells were grown in a humidified incubator at 37°C with 5% CO2.

Establishment of PC-9_Luc cell line

The PC-9 cells were transfected with pGL4.50 [luc2/CMV/Hygro] vector containing luciferase gene using Lipofectamine LTX. The stable cell clone was selected with hygromycin B (300 μg/ml) by serial dilution. The luciferase intensity was measured using the Bright-Glo Luciferase Assay System in vitro.

Cellular pEGFR assays, cellular proliferation assays, and Western blots

The assessment of cellular EGFR phosphorylation using enzyme-linked immunosorbent assay and cell proliferation was performed as described previously (16). Cellular pEGFR assay using the Meso Scale Discovery method was performed following the manufacturer’s instruction (Meso Scale; N45ZB-1). Western blot analysis was performed according to a published method (32).

Mice

Six- to 8-week-old specific pathogen–free immunodeficient nude mice were purchased from Vital River and housed as described previously (16). All the animal studies were approved by the Institutional Animal Care and Use Committee of AstraZeneca.

Establishment of SC model

H3255, PC-9, and A549 cells were harvested and suspended in phosphate-buffered saline. Single-cell suspensions of greater than 90% viability were used for injection. Tumor cells were injected subcutaneously in the left flank of a nude mouse in a volume of 0.1 ml.

Establishment of LM and BM models

The LM model was established by implanting 10 μl of cell suspension containing 1 × 104 PC-9_Luc cells through the cisterna magna (33, 34). The BM model was established by ICA artery injection (35, 36) of 70 μl of cell suspension containing 3.5 × 105 PC-9_Luc cells. The BM prevention model was established by ICB injection (37, 38) of 4 μl of cell suspension containing 3 × 105 PC-9_Luc cells. Model characterization was achieved by measuring luciferin signal and assessing tumor areas through histological examination at the same time points. A good correlation was identified between luciferin signal and tumor areas in the brain (fig. S10).

The formulations for the drugs were 10% dimethyl sulfoxide (Sigma; D2650) and 0.1% Lutrol F 68 (Poloxamer 188) (BASF; 9003-11-6) in saline for erlotinib and 1% methyl cellulose (Sigma; 9004-67-5) in water for AZD3759; 0.1% Lutrol F 68 in saline was used as the vehicle control.

In vivo DMPK assay to evaluate CNS penetration

A short oral absorption assay (16) was used to assess CNS penetration in mice. Drug concentrations in the blood, brain, and CSF were analyzed by LC–MS/MS. After determining the protein binding of compounds in the blood and brain tissue by equilibrium dialysis (39), Kpuu,brain and Kpuu,CSF values were calculated by the following equations:

Kp,brain = AUCbrain/AUCblood. Kp,CSF was the average of the CSF-to-blood ratios at the evaluated time points. Kpuu,brain = Kp,brain × fu,brain/fu,blood; Kpuu,CSF = Kp,CSF/fu,blood.

IHC assay in tumor tissue and CSF tumor cells

Tumor tissue preparation. Tumor-bearing brain tissue and SC tissues were collected from mice after a single dose of vehicle, erlotinib, or AZD3759. Tissues were fixed in 10% buffered formalin for 24 hours following standard procedure for processing (40), paraffin-embedding, and sectioning to assess pEGFR by IHC assay.

CSF tumor cell preparation. A total of 10 μl of CSF was collected from mice after a single dose of vehicle or AZD3759. Cytology slides were prepared using a Thermo Scientific Shandon Cytospin 4 cytocentrifuge (Thermo Fisher Scientific; A78300005) at 300 rpm for ~3 to 5 min. The slides were fixed in 10% buffered formalin for 5 min followed by pEGFR IHC assay.

A total of 4 ml of CSF was obtained from patients with LM, of which 3 ml was used for the assessment of pEGFR. Another 1 ml was used for the assessment of tumor cell numbers by Wright-Giemsa staining. Cytology slides were prepared as described above.

pEGFR IHC assay. A rabbit phospho-EGFR (Tyr1068) antibody (diluted to 0.175 μg/ml) (Cell Signaling Technology; 2234) was used as a primary antibody on a Ventana automated immunostainer (Discovery XT; Ventana Medical Systems). The staining procedure was as follows: primary antibody incubation at 37°C for 20 min, secondary antibody Ventana UltraMap anti-Rb HRP (Roche; 05269717001) at 37°C for 16 min, ChromoMap DAB (Roche; 5266645001) development for 5 min, counterstaining with hematoxylin II (Roche; 05277965001) for 4 min, and then counterstaining with Bluing reagent (Roche; 05266769001) for 8 min.

Determination of CSF tumor cells by Wright-Giemsa staining

A total of 500 μl of Wright-Giemsa reagent (Sigma; WG16) was added on the cytology slides and reacted for 1 min at room temperature. An equal volume of phosphate buffer (pH 7.2) (Sigma; P3288-12VL) was added and gently shaken for 2 min. The slides were then thoroughly rinsed with water followed by air-drying.

Quantification of CSF tumor cells and pEGFR signals. The CSF tumor cell number was counted manually. The total tumor cell number was determined by adding cell numbers on all slides prepared from the 4 ml of CSF.

The pEGFR signals were quantified by the Aperio image analysis system and the blinded pathologist’s own visual observation and expressed as an “H” score. Both staining intensity and positive percentage were used to generate the H score. For the CSF sample, ≥10 tumor cells were required to determine the H score.Embedded Image

Detection of EGFR mutations in the CSF sample

CSF samples (0.7 ml) were collected from patients with LM before treatment. Circulating tumor DNA was extracted from the CSF using the QIAamp Circulating Nucleic Acid Kit (Qiagen; 55114). EGFR mutations were detected by using the droplet digital polymerase chain reaction as previously described (41).

Determination of plasma and CSF PK in patients

Venous blood (about 4 ml) and CSF samples (about 0.3 ml) were collected from patients with LM and BM, at the predose time point (Ctrough) after 1 week of repeated dosing of AZD3759 to determine its concentration. The analysis was carried out using LC–MS/MS (Sciex API 4000).

Statistical analysis

The comparison of luciferin signals in the BM and LM models and tumor volume in the SC models in different treatment groups was performed by applying two-way ANOVA. The comparisons of tumor areas and tumor formation in the brain were performed by using one-way ANOVA and Fisher’s exact test, respectively. Log-rank test was used to compare the difference in animal survival in the BM and LM models. The correlation between the luciferin signals and tumor areas was performed using Pearson’s coefficient analysis. All analyses were performed using GraphPad Prism software. A P value of <0.05 was considered statistically significant.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/8/368/368ra172/DC1

Fig. S1. EGFR protein expression in lung cancer cell lines.

Fig. S2. Tumor growth curves of individual animals in the PC-9 LM model.

Fig. S3. Comparison of tumor areas on the leptomeninges in the PC-9 LM model.

Fig. S4. Tumor growth curves of individual animals in the PC-9 BM model.

Fig. S5. Comparison of tumor areas in the brain in the PC-9 BM model.

Fig. S6. Tumor growth curves of individual animals in the PC-9 BM prevention model.

Fig. S7. Tumor growth curves of individual animals in the H3255 SC model.

Fig. S8. Tumor growth curves of individual animals in the PC-9 SC model.

Fig. S9. Tumor growth curves of individual animals in the A549 SC model.

Fig. S10. Characterization of the PC-9 BM model.

Table S1. In vitro activity of AZD3759 in lung cancer cell lines.

Table S2. Estimation of human equivalent dose of erlotinib in nude mice.

Table S3. Tumor formation on the leptomeninges in the PC-9 LM model.

Table S4. Tumor formation in the brain in the PC-9 BM model.

Table S5. Tumor formation in the brain in the PC-9 BM prevention model.

REFERENCES AND NOTES

  1. Acknowledgments: We thank A. Hiraide for providing the PC-9 cell line, N. Prokoph for formatting the figures, and A. Bilal for proofreading the revised manuscript. Funding: AstraZeneca sponsored the preclinical and clinical studies of AZD3759. Author contributions: Z.Y. contributed to idea conception, designed the preclinical and clinical studies, conducted the clinical study, interpreted the data, and wrote the manuscript. Y.B., Q.G., Y.W., L.Z., Y.X., and K.C. conducted the preclinical in vitro, in vivo, and DMPK experiments and contributed to data interpretation and manuscript writing. Z.C. and X.Y. contributed to data interpretation and manuscript writing. S.R. conducted PK analysis of human plasma and CSF samples. D.-W.K., M.-J.A., and J.C.-H.Y. conducted the clinical studies and contributed to manuscript writing. X.Z. contributed to idea conception, overall planning of the study, and critical review of the manuscript. Competing interests: Z.Y., Q.G., Y.W., K.C., L.Z., Z.C., Y.X., X.Y., Y.B., S.R., and X.Z. are AstraZeneca employees. Z.Y. and X.Z. have equity interest in AstraZeneca, which develops AZD3759. AstraZeneca holds a patent on AZD3759. Z.Y. and X.Z. are inventors on patent application WO 2014/135876 A1, held by AstraZeneca AB, which covers AZD3759. D.-W.K., M.-J.A., and J.C.-H.Y. are investigators of the ongoing phase 1 study.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article