Osimertinib plus Bevacizumab Improve the Outcome of Leptomeningeal Metastasis Patients with EGFR Mutant Non-Small-Cell Lung Cancer

EGFR-mutant non-small cell lung cancer (NSCLC) is prone to leptomeningeal metastasis (LM) after Tyrosine kinase inhibitors (TKIs) treatment. Our previous study suggested that osimertinib plus bevacizumab was safe and effective in LM from EGFR-mutant lung cancer. This study aimed to compare the ecacy of osimertinib plus bevacizumab with osimertinib in EGFR-mutant NSCLC patients with LM. Methods We retrospectively reviewed data from 27 LM patients with EGFR-mutant lung cancer who received osimertinib with or without bevacizumab at the Second Aliated Hospital of Nanchang University. Next, we investigated the antitumor ecacy of osimertinib plus bevacizumab in an LM xenograft model using the H1975 (EGFR exon20 T790M and exon21 L858R) cell line. We examined the ability of osimertinib plus bevacizumab compared with osimertinib to penetrate the blood-brain barrier (BBB) and explored the potential mechanism. downstream signaling pathways including p-AKT and reduced tumor microvessel density (TMD), indicated that combined osimertinib with bevacizumab play a synergistic effect in EGFR-mutant LM model by modulating the level of E-cadherin.


Introduction
Non-small cell lung cancer (NSCLC) is one of the leading causes of cancer death worldwide [1]. EGFRmutant lung adenocarcinoma is prone to leptomeningeal metastasis (LM) after rst-generation TKI treatment [2,3]. Patients with LM have a poor prognosis and low quality of life [4,5]. Approximately 3-5% of patients with NSCLC will develop LM, and the median OS is approximately 4. 5-11 months[6, 7].
Osimertinib, a third-generation EGFR TKI, is currently the standard therapy for lung cancer metastases in the central nervous system (CNS)(including brain/leptomeningeal metastases) [8], which extended the OS to 18.8 months in LM patients with NSCLC [4,9]. The incidence of LM in patients with lung cancer, especially in patients with EGFR mutations, is increasing with the emergence of new targeted drugs [1].
TKIs combined with antiangiogenic drugs such as bevacizumab have shown e cacy in lung cancer [10,11]. A phase II clinical study showed that osimertinib plus bevacizumab is bene cial in the treatment of lung cancer brain metastases [12], and our previous study indicated that osimertinib plus bevacizumab was safe and effective for the treatment of LM in EGFR-mutant lung cancer [NCT04148898] [13]. LM is different from brain parenchymal metastasis, and its mechanism and treatment are complicated issues in current clinical treatment. There are currently no research data on the e cacy of osimertinib plus bevacizumab compared with osimertinib in LM. This is the rst study to compare the e cacy of osimertinib plus bevacizumab with osimertinib in LM with EGFR-mutant NSCLC through pre-clinical experiment. Aims to investigate the e cacy and potential mechanisms of osimertinib plus bevacizumab in LM with EGFR-mutant NSCLC.

Patients
We retrospectively reviewed the charts of 27 patients diagnosed with LM from EGFR-mutant NSCLC who received osimertinib with or without bevacizumab at the Second A liated Hospital of Nanchang. The date of LM diagnosis was de ned as the date of rst CSF cytology examination revealing malignant cells or the date of rst MRI (brain or spine) demonstrating LM. The assessment of LM response was based on the modi ed RANO LM radiological criteria, and the CNS and extra-CNS response was evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1: comprehensive evaluation according to the improvement of clinical symptoms, the performance of MRI and the clearance of CSF tumor cells [14]. Given that lumbar puncture is an invasive procedure, the response criteria of LM were judged according to the improvement of clinical symptoms and the performance of MRI in our study. All LM patients underwent 1.5T whole brain and spinal cord enhanced MRI scan at baseline, and the MRI scan thickness was 1mm, bravo and cube sequence having a high sensitivity and speci city in the diagnosis of LM. The Overall survival (OS) was de ned as the time between the initiation of diagnosis to the date of death or last follow-up by November 5, 2020. The intracranial Progression-free Survival (iPFS) was de ned as the time from the diagnosis of LM to the disease progression or death. Four weeks after the initiation of Osimertinib and bevacizumab, neurological evaluations, brain MRI and chest/abdominal computed tomography were routinely performed and were then performed every 1 months. The main endpoint of this study was iPFS.

Page 4/19
The eligibility criteria were as follows: (i) age 18-80 years; (ii) histologically or cytologically con rmed NSCLC; (iii) the detection of an EGFR mutation, with EGFR status identi ed from primary lung tumors using the ampli cation refractory mutation system (ARMS) or next-generation sequencing (NGS) analysis; (iv) LM de ned by CSF positivity for malignant cells and/or focal or diffuse enhancement of leptomeninges, nerve roots or the ependymal surface diagnosed by magnetic resonance imaging (MRI) with gadolinium contrast; (v) Patients receiving osimertinib plus bevacizumab or osimertinib. (vi) Patients without history of treatment with osimertinib before a diagnosis of LM. Patients who did not meet these criteria were excluded. Clinical outcomes were compared with the Kaplan-Meier log-rank test. This study was approved by the Institutional Review Board of the Second A liated Hospital of Nanchang University.

Cell Culture
H1975 cells (EGFR exon20 T790M and exon21 L858R) were purchased from Pro-cell in 2020 (Wuhan, China, was identi ed by STR) and maintained in RPMI-1640 medium supplemented with 10% Fetal Bovine Serum and 1% antibiotic solution in a humidi ed incubator with 5% CO 2 . All reference compounds were purchased from Gibco.
H1975 cells were stably transfected with a GV260 (GeneChem, Shanghai, China) vector containing luciferase, and bioluminescence signals were measured by an in vivo imaging system. H1975-luc tumor cells were prepared for injection after trypsinization and washing with PBS. A viable cell count was performed with trypan blue to adjust the cell concentration to 2 × 10 6 cells in PBS for each injection.

Animal model of LM
BALB/C nude female mice (6-8 weeks) were obtained from institutional animal breeding services (the Nanchang Royo Biotech Co., Ltd.). The animals were group housed ve a cage in a temperaturecontrolled room on a 12-h/12-h light/dark cycle with unlimited access to food and water. All animal procedures in this study were approved by the Animal Care and Use Committee of Nanchang University, and ethical approval was obtained from the Institutional Review Board of the Second A liated Hospital of Nanchang University.
Mice were anesthetized with 4% iso urane. The skull was exposed with a skin incision (1 cm) under sterile conditions (ethanol skin wipe) to locate the bregma. A Hamilton syringe needle with H1975-luc human NSCLC cells was injected into the right lateral ventricle (anteroposterior 2.0 mm from the bregma; lateral 0.2 mm to the right; and dorsoventral 4 mm) at 2 µl per min. A total volume of 5 µl of cell suspension was injected [15,16]. The tumor burden of intracranial lesions and the tumor size in the pia mater were measured using a BLI technique with an in vivo imaging system, MRI and H&E staining.

In Vivo Pharmacodynamic Study
After con rming tumor formation, the mice were randomly divided into 4 groups (n=9 per group): the control(0.9% normal saline, daily, oral), osimertinib (25 mg/kg, daily, oral) [17], and bevacizumab (5 mg/kg twice a week, i.p.)[18], and osimertinib plus bevacizumab. Brain tissues were collected at 1, 6, 12, and 24 h after a single dose of treatment to determine the penetration of osimertinib in the brain with a validated liquid chromatography-tandem mass spectrometry method(n=3). Furthermore, tumor tissues were collected in formalin or frozen at the endpoint of experiments, EGFR downstream signals were evaluated by immunoblotting, and an angiogenesis marker (CD31) was evaluated by immuno uorescence. Once the nude mice lost more than 20% of their weight, the experiment reached the endpoint, and the mice were euthanized by intraperitoneal injection of pentobarbital sodium(150mg/kg). Each experimental group has at least 9 mice, and a total of 40 mice are included.

Formaldehyde perfusion of the brain
To keep the mouse brain tissue as intact as possible, this experiment used formaldehyde perfusion to x the brain. The mice were anesthetized by the intraperitoneal injection of 1% pentobarbital sodium(50 mg/kg), the chest skin was cut open with scissors, the heart was exposed, a syringe needle was inserted into the left ventricle from the left apex towards the aorta, the needle was xed with hemostatic forceps, physiological saline was quickly perfused until the liver tissue color turned gray, and then paraformaldehyde was slowly perfused. The mouse limbs twitched, and the whole body became stiff. After successful perfusion, the mouse brain tissue was removed and xed with paraformaldehyde for subsequent experiments.
2.6. Histology 2.6.1. H&E staining and immuno uorescence Mouse brain tissues were quickly excised, immersed in 10% paraformaldehyde and embedded in para n. Brain tissues were cut into 5-µm thick sections. Slides were stained with hematoxylin and eosin (H&E). Pictures were obtained with ×10 and ×40 objective respectively, the pathology diagnosis was completed by two independent pathology doctors. Tumor microvessel density (TMD) was measured by staining for CD31 (Proteintech, Wuhan, China), as previously reported [19]. Brie y, TMD was assessed in hot spots of brain tissue cross-sections identi ed by light microscopy. Five equal areas were then photographed with a 40x objective (400x magni cation). The staining was scored by two independent and experienced pathologists and calculated as the product of the staining intensity. The areas and integrated optical density (IOD) of the slides were analyzed by Image-Pro Plus 6.0 software.

IHC
Tumor-bearing brain tissue were collected from mice of each group. Tissues were xed in 10% buffered formalin for 24 hours following standard procedure for processing, para n-embedding, and sectioning to assess PCNA (1:200, Servicebio, GB11010), E-cadherin (1:200, Servicebio, GB14076) by IHC assay. The reaction was visualized using the Servicebio image analysis system, the staining was scored by two independent and experienced pathologists and calculated as the product of the staining intensity. rst divides the positive grade: negative without staining, score 0, weak positive light yellow, score 1, medium positive brown, score 2, strong positive brown, score 3 points. Then analyze and calculate the area of weak, medium, and strong positive in the measurement area, the tissue area of the measurement area,

Statistics
Preclinical data are presented as the mean ± SD. Data were analyzed by one-way ANOVA, followed by Dunnett's test or Student's t-test. Survival analysis of animal was performed using a Kaplan-Meier survival curve and a log-rank test. The Kaplan-Meier method was used to estimate and graphically present OS and iPFS. All analyses were performed using GraphPad Prism 7.0 (GraphPad Prism Software, Inc) or IBM SPSS statistics 22. Experiments were repeated independently at least three times. p <0.05 was considered statistically signi cant.

Osimertinib plus bevacizumab improved the e cacy of LM patients with EGFR-mutant NSCLC.
A total of 70 patients diagnosed with lung cancer and LM in at the Second A liated Hospital of Nanchang University from October 2017 to March 2020 were collected. 27 patients with EGFR mutationpositive NSCLC with LM were selected according to the inclusion criteria (Table 1). Enrolled patients received osimertinib 80 mg orally daily with or without bevacizumab 7.5 mg/kg intravenously every three weeks. Treatment continued until the disease progressed, unacceptable adverse events occurred, or the patient withdrew consent. The pathological type of all patients is lung adenocarcinoma. In these patients, patients had progressive disease (PD). In osimertinib plus bevacizumab group, Of the 16 total patients, 81.3% (13) had a clinical response, 6 of patients achieved PR (37.5%),7 of patients had SD, 3 of patients had PD. The assessments of response to osimertinib or osimertinib plus bevacizumab in 27 LM patients were shown in supplementary table 1 and 2. The Swimmer plot was accordance with treatment duration of 27 LM patients (Fig. 1A). There is an increasing of median OS of the patients who received osimertinib and bevacizumab (n = 16, 18.0 months) as compared with patients with Osimertinib treatment (n=11, 13.7 months) (log-rank test, p=0.046, HR=2.867, 95%CI:1.007-8.162). The median iPFS is 10.6 months versus 5.5 months (log-rank test, p=0.037, HR=3.401, 95%CI:1.079-10.720) (Fig. 1B and C). These data suggested that osimertinib plus bevacizumab improve the survival time of patients with LM from EGFR mutant NSCLC.

Osimertinib plus bevacizumab demonstrated impressive CNS penetration in vivo
To determine the e cacy of osimertinib plus bevacizumab in LM in EGFR-mutant NSCLC, we constructed a model of LM with lateral ventricle injection ( Fig. 2A), and the lung cancer LM xenograft model was con rmed by IVIS imaging, MRI and H&E (Fig. 2B, C and E). Mass spectrometry analysis showed that the average concentration of osimertinib in the brain tissue in combined group was higher than osimertinib group [ 803.2 vs 606.5 ng/ml (1 hour after treatments), 359.6 vs 193.8 ng/ml (6 hour), 91.8 vs 31.2 ng/ml (12 hour), 3.8 vs 1.0 ng/ml (24 hour) (n=3 per group, p<0.05)] (Fig. 2D). These results suggested that when osimertinib is combined with bevacizumab, the concentration of osimertinib in the mouse brain was increased.

Osimertinib plus bevacizumab effectively inhibits the growth of EGFR-mutant LM xenografts in nude mice
The effects of osimertinib plus bevacizumab in the LM xenograft model were validated. Within 2 weeks of treatment, the three groups had different degrees of tumor regression except for the progression of the control group compared to baseline ( Fig. 3A and B). The body weight change and survival curves implied that the tumor regression of the combined group was superior to that of the osimertinib group after 3 weeks of treatment. The mice in the bevacizumab group lost a signi cant amount of weight and died quickly, similar to mice in the control group. (n=6 per group, p<0.05) (Fig. 3C and F). Compared with that in the osimertinib group, the tumors in the osimertinib plus bevacizumab group were signi cantly regressed ( Fig. 3D and G). These results suggested that the combination effectively inhibits the growth of EGFR-mutant LM tumors in vivo.

Osimertinib plus bevacizumab suppresses the EGFR downstream signaling pathway and modulates E-cadherin levels in EGFR-mutant LM model mice
The EGFR downstream signaling pathway was examined in tumor tissues by Western blotting. The results showed that p-AKT and EGFR were signi cantly decreased in the osimertinib plus bevacizumab group compared with the other groups. (n=3, p<0.05) ( Fig. 4A and B (a-e)). Meanwhile, our data showed that TMD was signi cantly decreased in the osimertinib and bevacizumab group compared with that in the other groups (n=3, p<0.05) ( Fig. 4C and B (f)). These results further veri ed that osimertinib and bevacizumab could play a synergistic effect in EGFR-mutant LM model. To understand the potentially mechanism of the combination treatment, we assessed PCNA, E-cadherin, ADAM9 and HIF-1α levels in xenograft tumors receiving the treatments. The combination did not affect the levels of ADAM9 and HIF-1α, but the E-cad levels was more signi cantly decreased in combination group compared with osimertinib group (n=3, p<0.05) (Fig. 5). Hence, we demonstrated the in vivo modulation of E-cadherin by the combination of osimertinib and bevacizumab.

Discussion
Due to the unclear mechanism of LM and the existence of the blood-brain barrier, it is di cult for drugs to reach an effective intracranial concentration. The treatment of LM in lung cancer is still a complicated problem. This is the rst study to compare the effects of osimertinib with or without bevacizumab in LM of EGFR-mutant NSCLC. The e cacy is compared through a combination of basic research and clinical analysis and its possible mechanism is explored.
Studies have shown that EGFR-TKIs (erlotinib) and angiogenesis inhibitors that target endothelial growth factor receptor (anti-VEGFR) (bevacizumab) (A+T) achieved superior PFS and acceptable safety in NSCLC patients with intracranial metastasis [22]. Our study retrospectively analyzed 27 LM patients with EGFR-mutant lung cancer who received osimertinib with or without bevacizumab, the median OS of osimertinib plus bevacizumab group (n=16) compared osimertinib group (n=11) was 18.0 months versus 13.7 months (log-rank test, p=0.046, HR=2.867, 95%CI:1.007-8.162). The median iPFS is 10.6 months versus 5.5 months (log-rank test, p=0.037, HR=3.401, 95%CI:1.079-10.720). Then we investigated the antitumor effects of osimertinib and bevacizumab in an EGFR-mutant LM model. We found that osimertinib plus bevacizumab signi cantly improved the concentration of osimertinib in mouse brain tissue.
A series of studies have supported that A+ T therapy can result in improved survival bene ts [23]. The JO25567 recommended combined erlotinib and bevacizumab as a rst-line regimen in EGFR mutationpositive NSCLC [24,25]. The NEJ026 also suggested that TKIs combined with bevacizumab extended PFS 3.6 months compared to erlotinib in EGFR-mutant NSCLC [11]. The PFS of NSCLC patients with pleural or pericardial effusion is expected to be prolonged with the combined use of osimertinib plus bevacizumab and to demonstrate their safety [26]. Consistent with the above studies, our preclinical experiments and retrospective analysis indicated that osimertinib and bevacizumab improve the survival of LM patients with EGFR mutant NSCLC. In contrast, some reports showed that TKIs combined with angiogenesis inhibitors did not improve PFS in EGFR-mutant NSCLC [27]. Although the overall response rate (ORR) was better with osimertinib plus bevacizumab than osimertinib alone (68% vs 54%), median PFS was not longer with osimertinib plus bevacizumab (9.4 months vs 13.5 months) [28]. Dr. Toi reported that compared with osimertinib alone (n=40), the combination of osimertinib and bevacizumab (n=41) did not increase the PFS of NSCLC patients. Although the ORR of the combined group was higher, there was no difference in OS between the two groups [28]. Recent studies have found that 160 mg of osimertinib could have better e cacy than 80 mg in controlling LM[8, 29,30]. Further investigation to compare the e cacy of osimertinib 80 to 160 mg in patients with LM is warranted.
VEGF is a key regulator of angiogenesis and a validated target for NSCLC [31]. The biologically synergistic antitumor activity of EGFR inhibition in combination with VEGF/VEGFR pathway blockade have been demonstrated in preclinical studies [32]. In EGFR-mutant NSCLCs, up-regulated EGFR signaling increases VEGF through hypoxia-independent mechanisms, and elevated VEGF, in turn, contributes to the emergence of resistance to EGFR tyrosine kinase inhibitors (TKIs) [33,34], and EGFR, similar to VEGFR-2, can be expressed on tumor-associated endothelial cells [23,32]. The inhibitory effects of afatinib on EMT and tumorigenesis may be associated with the ERK-VEGF/MMP9 signaling pathway [35]. The TMD has been regarded as one important indicator for quantitatively analyzing tumor angiogenesis, which can clearly re ect the intra-tumoral blood vessels state and tumor-induced angiogenesis ability [36,37]. Tumor vascularization is critical to the pathogenesis of solid tumors, and TMD is related to tumor invasiveness and metastasis formation which could be used as a potential predictive marker for bevacizumab bene t[38]. Bevacizumab prunes vessels while normalizing those remaining [39]. Osimertinib is a small molecule targeted drug, which is easier to penetrate the blood-brain barrier than the rst and second-generation targeted drugs [17]. These studies suggest that dual blockade of the VEGF and EGFR pathways would be more effective. Our results showed that the combination group has a more signi cant reduction in TMD. Perhaps it is the reason that bevacizumab improved the concentrations of osimertinib in the mouse brain, which further indicated that osimertinib plus bevacizumab have a synergistic effect. The further mechanism still needs to be investigated. Considering, our study found that the combined treatment signi cantly increased the effective intracranial concentration of osimertinib, modulated the level of E-cadherin and downregulated the levels of EGFR and downstream signaling pathways including p-AKT and reduced TMD, indicated that combined osimertinib with bevacizumab could play a synergistic effect in EGFR-mutant LM model possibly by modulating the level of E-cadherin. In preclinically, osimertinib combined with bevacizumab is a more ideal and optimized treatment plan for EGFR mutant NSCLC with LM. It is di cult to assess the iPFS of LM, and it is currently believed that a comprehensive assessment should be based on patient neurological examination, C radiological evaluation, and cerebrospinal uid cytology. The main challenge is to de ne measurable and non-measurable (target) damage, and allow assessment of response changes, perform cerebrospinal uid cytology for assessment, due to lumbar puncture is an invasive test, most patients refuse to perform. Therefore, it is di cult and subjective to evaluate iPFS in patients with LM, and the present results must be interpreted cautiously. The data were obtained from medical les, and we cannot exclude the possibility of unde ned biases and/or confounding factors. Third, the interaction mechanism of osimertinib and bevacizumab needed to be further explored. The next step of our research, the phase II study of osimertinib plus bevacizumab for LM is already ongoing (NCT04425681). We are collecting the CSF and blood of NSCLC patients with LM in osimertinib group and the combination with bevacizumab group to further explore the mechanism and nd the biomarkers for prognostic.
In Availability of data and material All data generated or analyzed during this study are included in this published article.

Competing interests
Page 14/19 The authors declare they have no con icts of interest relevant to this submission.  Antitumor activity of osimertinib and bevacizumab in the EGFR-mutant LM model Figure 4