Introduction

Metastatic lung cancer and breast cancer are among the most common human malignancy and cancer-related death worldwide. Clinical management of metastatic non-small cell lung cancer (NSCLC) and triple negative breast cancer (TNBC) remain a major unmet need [1, 2]. In these patients, chemotherapy, targeted therapy (e.g., EGFR, ALK inhibitor) and immunotherapy (e.g., anti-PD-1/L1) are established treatment paradigms [3, 4]. However, relapse is typical due to a relatively low response rate or development of drug resistance. Although combination of chemotherapy with immunotherapy or combination of different targeted drugs have improved outcome in some patients [5, 6], a significant patient population remain unresponsive or develop resistance to treatments [7, 8].

Recent studies have revealed a strong connection between AXL, a member of TAM (TYRO3, AXL, MERTK) family receptor tyrosine kinase, with disease recurrence in numerous cancer types [9]. GAS6 is the ligand of AXL, which can bind the extracellular domain of AXL leading to AXL phosphorylation [10]. The AXL-GAS6 signaling pathway mediates multiple downstream signaling pathway, including PI3K/AKT, ERK, JAK/STAT, and is further involved in cell migration, proliferation and poor survival [11,12,13]. AXL is positively linked to the drug resistance to EGFR-TKI in NSCLC [14], the recurrence and metastasis in aggressive TNBC [15], pancreatic ductal adenocarcinoma [16], and low response rate to anti-PD-1 in melanoma [17]. These earlier studies suggested an important role of AXL in regulating disease progression and poor prognosis.

The mechanisms associated with drug resistance are diverse and include genomic alterations, epithelial mesenchymal transformation (EMT) and modulation in tumor microenvironment (TME) [18, 19]. It was found that AXL, as an important regulator of EMT, has a regenerative feedback loop with a variety of EMT related markers [20, 21]. Patients with NSCLC treated with EGFR first and third generation inhibitors showed abnormal activation of AXL and EMT, and inhibition of AXL activity could reverse the drug resistance [22]. Targeting of AXL in EMT-high tumor cells is predicted to have a combinatorial benefit to enhance current immunotherapies [23].

The role of AXL in TME is much less understood. AXL is known to be developmentally expressed in myeloid lineage [24, 25], and we speculate that the dysregulated AXL can potentially influence the function of myeloid-derived immune cells including those of macrophages and dendritic cells in TME. Tumor-associated macrophages and dendritic cells are well established components of NSCLC and TNBC, which are increasingly recognized for impacting cancer treatment outcome [26, 27]. It is possible that AXL may act to promote the pro-tumor macrophages and/or negatively regulate the anti-tumor dendritic cells. These are emerging questions for dissecting how AXL might contribute to immune escape and disease progression.

Herein, we have performed transcriptome analysis, in vitro experiments, and preclinical mouse models to further establish a functional relationship between AXL and EMT-related immunosuppressive TME in NSCLC and TNBC. A combined targeting of AXL in tumor cells and myeloid-derived immune cells can efficiently inhibit metastasis and induce TME remodeling. We also describe the discovery of a novel and highly potent AXL antibody-drug conjugate (AXL-ADC), its promising efficacy in multiple AXL-high tumor models including those of drug-resistant NSCLC.

Materials and methods

AXL transcriptome analysis of TCGA and CCLE cohorts

For mRNA expression analysis, patient data of NSCLC (n = 966) including LUAD (n = 500, accessed in September 2020) and LUSC (n = 466, accessed in August 2022) or breast cancer (n = 900, accessed in March 2021) were downloaded from TCGA (http://www.cbioportal.org/). The method to distinguish Basal-like/TNBC in breast cancer was described previously [28]. Two subpopulations for NSCLC tumors expressing AXL-high (AXL-H, n = 150) and AXL-low (AXL-L, n = 150) or TNBC (AXL-H, n = 50; AXL-L, n = 50) were subjected to analysis on AXL-related prognosis and various biomarkers. Survival analysis was conducted in GraphPad Prism 8.0, and log-tank test was used to compare the statistical differences in overall survival (OS) between groups.

The immune cell fraction of all samples from TCGA was analyzed by CIBERSORT (https://cibersortx.stanford.edu/). The TCGA dataset was sorted by case and gene name and saved as TXT format. The data were then uploaded to CIBERSORT website. LM22 signature matrix were selected for further analysis, the relative proportion and corresponding genes of each immune cell typing were obtained. Monte Carlo sampling was performed to obtain the corresponding P value of CIBERSORT samples, and only the samples with P < 0.05 are considered qualified for further analysis [29]. The T cells and DCs were evaluated by mRNA relative ratio of GZMB/CD8A and CD103/CD11c.

The mRNA expression data of NSCLC cancer cell lines (n = 75) and TNBC cell lines (n = 58) were obtained from Cancer Cell Line Encyclopedia (CCLE, https://sites.broadinstitute.org/ccle), and were stratified according to the expression level of AXL.

The mRNA expression data and corresponding clinical information of NSCLC with different disease stages (GSE120622, n = 80), NSCLC received chemotherapy (GSE39345, n = 46) and Non-squamous NSCLC patients after anti-PD-L1 therapy (GSE93175, n = 20) were obtained from Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) [30, 31].

Generation of AXL antibody and antibody-drug conjugate

The previously reported anti-human/mouse AXL antibody YW327.6S2 [32] (referred in this paper as “Anti-AXL”) was produced in hIgG1 background, purified from 293 T cells, and was confirmed for target activity (supplementary Fig. S1a, S1b).

The newly discovered AXL02 was generated as follows. Balb/c mice were immunized with an extracellular domain of human AXL protein (AXL-ECD). Splenocytes were fused with SP2/0 myeloma cells, screened with ELISA and flow cytometry (FACS). Antibody variable regions were obtained, constructed as mouse chimeric antibody and subsequently humanized as AXL02 [33]. The amino acid sequence of AXL02, antibody production and purification are shown in Supplementary Fig. S1c, S1d, S1e.

We developed a new disulfide bridging linker platform (BL20-MMAE, US10987430B2), which generated a uniform DAR4 ADC. AXL02-MMAE and hIgG1-MMAE were generated based on methods described in US20210214447A1, including reagents, linker-payload (BL20-MMAE), and conjugation procedures. Briefly, AXL02 antibody interchain disulfide bonds were fully reduced by excessive (10 molar equivalent per mAb) tris(2-carboxyethyl)phosphine hydrochloride. BL20-MMAE (provided by Mabwell (Shanghai) Bioscience Co., Ltd.) was added to the reduced AXL02 with a 6 molar equivalent linker-payload/mAb ratio and incubated for 2 h. The resulting mixture was purified by preparative hydrophobic interaction chromatography (HIC) to give AXL02-MMAE with a DAR of 4. The relevant ADC AXL02-vc-MMAE and AXL107-vc-MMAE (EnaV) were prepared by partial reduction of AXL02, and conjugated with 6.0 equivalent of mc-VC-PAB-MMAE (Chemexpress, Cat#HY-126681, Shanghai, China) [34].

Cell culture, cell surface AXL levels and gene knockdown

Cell lines NCI-H1299, Calu-1, HCC827, NCI-H460, U87MG and Hs578t were obtained from the Chinese Academy of Sciences (CAS, Shanghai, China). MDA-MB-453 (MDA-453) and MDA-MB-231 (MDA-231) were obtained from ATCC. PC9 was obtained from ECACC. LCLC-103H was obtained from Cobioer Biotechnology (Nanjing, China). 4T1 was a gift from Prof. Qi-zhi Zhang of Fudan University School of Pharmacy [35]. All these cell lines were maintained in standard culture medium in a humidified CO2 incubator at 37 °C.

To measure cell surface expression of AXL, 1 × 105 tumor cells were incubated with serially diluted anti-human AXL (AXL02) or anti-human/mouse AXL (YW327.6S2) [32] at 4 °C for 1 h, stained with R-PE-Goat anti-human IgG-Fc (Abcam, Cat#98596, Cambridge, UK), quantified by FACS (BD, #FACS Aria II, NJ, USA).

pTRIPZ-based shRNA lentiviral constructs for human AXL or non-targeting (Open Biosystems, Cat#V2LHS_202535, Cat#RHS4346, AL, USA) were packaged in HEK293T cells. Tumor cells were stably infected with the lentivirus and induced for AXL depletion with 1 μg/mL doxycycline (Dox) for 5–7 days prior to experiments.

Immunocytochemistry and immunoblotting of cultured cells

Cells were grown on glass coverslips, fixed with 4% paraformaldehyde and subjected to immunocytochemistry (ICC) with primary antibodies to human AXL (CST, Cat#4566), Vimentin (Absci, Cat#21488, WA, USA) and E-cadherin (CST, Cat#14472, MA, USA), then detected with Alexa Fluor 488- or R-PE labelled secondary antibody. For immunoblotting, total cell lysates prepared from various tumor cells were immunoblotted with antibodies to mouse AXL (R&D, #AF854, CA, USA), human AXL (CST, Cat#4566), β-Actin (Bioworld, Cat#AP0060, MN, USA) and GAPDH (Bioworld, Cat#AP0063).

AXL antibody binding affinity, cell internalization and AXL degradation

For AXL-ELISA, AXL-ECD (1 μg/mL) was coated to ELISA plate, incubated with serially diluted (1:3) of AXL antibody and detected with HRP-labeled secondary antibody. For cell internalization, LCLC-103H cells pre-seeded in confocal chambers were incubated with 1 μg/mL AXL02 at 4 oC for 1 h or 37 oC for 4 h, fixed with 4% formaldehyde and permeabilized with 0.4% TritonX-100. The slides were incubated with anti-LAMP-2 (Abcam, Cat#ab125068) before detected with Alexa-Fluor-488 or R-PE-labelled secondary antibody. Images were captured under a confocal microscope (Zeiss, #LSM710, Oberkochen, Germany). For antibody-mediated AXL degradation, pre-plated tumor cells with a confluence of about 50% were incubated with 1 μg/mL AXL02 or hIgG1 for 24 h. Cell lysates were analyzed by immunoblotting.

Cell migration and lung metastasis

Appropriate number of tumor cells in 200 μL serum-free medium containing AXL antibody or IgG1 were added to the upper chamber of the 8 μm pore size Transwell system (Corning, Cat#3422, NY, USA), while the lower chamber contained 600 μL complete medium with 10% serum and the same treatment as the upper chamber. After 6–12 h incubation, the migrated tumor cells were stained with 0.2% crystal violet and counted under a microscope. Cells of each treatment were counted in 5 representative viewing fields at 200× magnification and analyzed using ImageJ software. Five-week-old female BALB/c mice were purchased from Charles River Laboratory (Beijing, China) and housed in an approved SPF animal care facility at Fudan University. For lung metastasis, 5 × 105 4T1 cells were injected in Balb/c mice via tail vein. The antibodies were dosed i.v. at 10 mg/kg on day −1, 1 and 8 (n = 4 per group). Mice were dissected on day 15 and lung tissues were fixed and subjected to histological analysis.

Proliferation assay, xenograft tumor efficacy and ADC in vivo distribution

Tumor cells were plated in 96-well plates at predetermined density, treated with AXL02-MMAE or hIgG1-MMAE for 5–8 days to ensure that the doubling of the cells is sufficient. Then MTS reagent (Promega, Cat#G111A, WI, USA) solution was added with replacing fresh medium, and cells were incubated for an appropriate time to ensure that the maximum net absorbance was between 0.5 and 1 at 490 nm.

Five-week-old female BALB/c nude mice were purchased from SIPPR-BK Lab animal Co., Ltd. (Shanghai, China) and housed in an approved SPF animal care facility at Fudan University. All mouse xenograft tumor experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of the Fudan University School of Pharmacy. U87MG (10 × 106), LCLC-103H (5 × 106), NCI-H1299 (10 × 106), and PC9 (5 × 106) cells were suspended in PBS then injected subcutaneously to the flank of Balb/c female nude mice (4–6 weeks old). Tumors were measured by caliper every 2–3 days. Before therapeutic treatment, tumor-bearing mice were staged at initial tumor volume of 100 to 300 mm3 (growth) or 1800 mm3 (regression) and randomized into treatment groups (n = 8). The ADCs or docetaxel (BiochemPartner, Cat#114977-28-5, Shanghai, China) were dosed i.v. once weekly, or total twice during entire study.

In vivo distribution of AXL02-MMAE and AXL02-vc-MMAE was evaluated in U87MG tumor-bearing mice (female BALB/c nu/nu) using a method described in the literature [36]. After a single i.v. injection of AXL02-MMAE and AXL02-vc-MMAE at 5 mg/kg, the mouse serum and tumor mass were harvested at 5 min, 24 h and 72 h after injection for analysis of free MMAE. The free MMAE concentration was determined by LC/MS/MS (Agilent, #1290-G6460A, CA, USA).

In vivo tumor immunohistochemistry and clinical tumor microarray

Mouse lung tissues (4T1 model) and xenograft human tumors were fixed in 4% formalin, embedded in paraffin and sectioned. Slides were stained with hematoxylin-eosin (H&E), TUNEL reagent or subjected to various immunohistochemistry (IHC) following standard protocols or methods specified by manufacturer. In some assays multiplexed IHC was performed with tyramide signal amplification (TSA method) and dyed with Tyr-488 and Tyr-CY3 (Record Biological Technology, Shanghai, China). Images were acquired in a slide scanner (Olympus, #VS200, Tokyo, Japan). The sections of each treatment (n = 3 tumors, 3 image view fields per section) were quantified by ImageJ software. The results were normalized to control group then plotted. For the proportion of AXL positive in macrophages, the counts of AXL+ macrophage marker+ cells were compared with macrophage marker+ cells, then the corresponding ratios were plotted. The IHC and TSA primary antibodies included mouse AXL (R&D, Cat#AF854), human AXL (CST, Cat#4566), p-AXL (CST, Cat#5724), Vimentin (Abcam, Cat#ab92547), p-Erk1/2 (CST, Cat#4370), p-Akt substrate (CST, Cat#9611 S), Ki67 (Abcam, Cat#279653), CD11b (Abcam, Cat#ab133357), CD11c (Servicebio, Cat#GB11059, Wuhan, China), CD86 (CST, Cat#19589), CD103 (Abcam, Cat#ab224202), CD8A (Sino Biological, Cat#50389-T26, Beijing, China), GZMB (Abcam, Cat#255598), CD19 (Abcam, Cat#ab245235), F4/80 (CST, Cat#70076), CD206 (Abcam, Cat#ab64693), CSF-1 (Abcam, Cat#233387), TNF alpha (Arigo, Cat#ARG10158, Taiwan, China).

NSCLC patient tumor tissue microarray (TMA, Outdo Biotech, Cat#HLugA180Su07, Shanghai, China) was stained with anti-human AXL (CST, Cat#4566), and CD206 (Abcam, Cat#ab64693).

Monocyte isolation and the mouse bone marrow-derived macrophage (BMDM) differentiation

Bone marrow cells were collected from six-week-old C57BL/6 mice, then cultured in RPMI-1640 supplemented with 10% heat-inactivated FBS and 20 ng/mL M-CSF (Peprotech, Cat#315-02, NJ, USA) for 8 days with medium replenished every 3 days. For M2 polarization, BMDMs were treated with hIgG1 or anti-AXL (6 μg/mL) and induced with 20 ng/mL IL-4 (Peprotech, Cat#214-14) and IL-13 (Peprotech, Cat#210-13) for 48 h. For M1 macrophages, BMDMs were stimulated sequentially 24 h each with 20 ng/mL IFN-γ (Peprotech, Cat#300-02) and 200 ng/mL LPS (Sigma, Cat#L2880, MO, USA). The phenotypes of mature macrophages were analyzed by flow cytometry.

Conditioned medium (CM) preparation and treatment

4T1 or LCLC-103H cells were cultured to reach 10% confluence, and then treated with indicated antibodies for 72 h. The CMs were collected, centrifuged at 5000 rpm for 10 min and filtered through 0.22 μm nylon syringe filters. RAW264.7 macrophages were treated with tumor cell CM for 48 h.

Statistical analysis

All numerical data were performed using GraphPad Prism 8 software and Microsoft Excel. Values of in vitro data are mean ± standard deviation, and in vivo research data are mean ± standard error. P values were calculated by an unpaired two-tailed Student t test to determine whether the difference between the two groups was significant. P values <0.05 represented significant statistical difference.

Results

AXL is elevated in NSCLC and TNBC and correlates with poor survival, epithelial-mesenchymal transition and immunosuppressive tumor microenvironment

We first analyzed TCGA datasets to investigate the relationship between AXL expression and prognosis of cancer patients. In both NSCLC and TNBC cohorts, tumors with AXL high expression (AXL-H) indicated lower overall survival (Figs. 1a, 1b). NSCLC patients with advanced stage or who received chemotherapy tended to show higher AXL expression (Figs. 1c, 1d). In NSCLC patients that received anti-PD-1 therapy [37], individuals with progressive disease (PD) tended to show higher AXL expression and M2 macrophage infiltration (Figs. 1e, 1f). In both NSCLC and TNBC, AXL-H was strongly associated with epithelial-mesenchymal transition (EMT) markers Vimentin (VIM), ZEB1 and FN1 (Fig. 1g, 1h) and may be linked to an immunosuppressive tumor microenvironment (TME) and therapy resistance [38]. Employing CIBERSORT analysis, we found that the AXL-H tumors contained significantly lower fractions of infiltrated dendritic cells (DCs), T lymphocytes, B lymphocytes and natural killer (NK) cells, while tumor-associated macrophages (TAMs) were enriched (Figs. 1i, 1j). Collectively, these results suggest that high AXL expression in NSCLC and TNBC may indicate high EMT state and immunosuppressive tumor microenvironment, which likely impact disease progression and therapy resistance.

Fig. 1: AXL is elevated in NSCLC and TNBC and correlates with poor survival, EMT and immunosuppressive tumor environment.
figure 1

Kaplan–Meier analysis of OS in groups of highest AXL expression (AXL-H) and lowest AXL expression (AXL-L) in NSCLC (a) and TNBC (b) patients. c, AXL mRNA expression level of NSCLC with stage IA/B (n = 36), stage IIA/B (n = 19) and stage III-IV (n = 25). d AXL mRNA expression level of peripheral blood mononuclear cells (PBMC) from NSCLC patients before chemotherapy (before chemo, n = 32) and after chemotherapy (after chemo, n = 14). AXL mRNA expression level (e) and macrophage subsets (f) of progressive disease (PD, n = 8) versus nonprogressive disease (NPD, n = 12) patients with anti-PD-1 therapy. Transcriptome signature of EMT in AXL-H vs. AXL-L group in NSCLC (g) or TNBC (h) patients. Fractions of innate immune cells (DCs and TAMs) and adaptive immune cells (T cells, B cells and NK cells) in the NSCLC (i) or TNBC (j) TCGA datasets were obtained by CIBERSORT analysis. *P < 0.05; **P < 0.01; ***P < 0.001.

Full size image

AXL inhibition attenuates migration and EMT in AXL-high expressing tumor cells

To extend the tumor transcriptome results and identify suitable cell models for AXL-targeted therapy, we next analyzed CCLE cell line cohorts. NSCLC and TNBC cell lines were grouped into AXL-H or AXL-L based on AXL mRNA expression. Consistently, cell lines with elevated AXL expression correlated with high EMT marker VIM, ZEB1 and FN1 (Figs. 2a, 2b). Flow cytometry (FACS) detected elevated levels of surface AXL in the NSCLC lines LCLC-103H, H1299, Calu-1, PC9 and TNBC lines 4T1 and MDA-231, while H460 and MDA-453 showed low or negligible AXL (Figs. 2c, 2d). Treatment of representative AXL-H and AXL-L cells with an anti-human/mouse AXL antibody YW327.6S2 [32] (termed “anti-AXL”) or AXL gene knockdown (AXL-shRNA; AXL-KD) all reduced AXL expression level and cell migration rate in the AXL-H human tumor cells LCLC-103H, MDA-231 and mouse tumor cell 4T1 but not in the AXL-L H460 and MDA-453 cells (Figs. 2e, 2f, Supplementary Fig. S2a, S2b). Likewise, treatment with anti-AXL or AXL-KD also reduced mesenchymal index as shown by the decrease in Vimentin and increase in E-cadherin (Fig. 2g, Supplementary Fig. S2c). These results collectively demonstrate that AXL plays an important role in high EMT and invasiveness, which can be reverted by AXL-targeted antibody or AXL-KD.

Fig. 2: Characterization of AXL-high tumor models and response to anti-AXL treatment in vitro.
figure 2

Analysis of transcriptome data of AXL and EMT in NSCLC cell lines (a AXL-H, n = 15; AXL-L, n = 12) and TNBC cell lines (b AXL-H, n = 16; AXL-L, n = 15). Cell surface AXL level for the indicated human and mouse NSCLC (c) or TNBC (d) cell lines were measured by FACS binding and plotted. Cell migration of the indicated NSCLC (e) and TNBC (f) cells treated with 40 nmol/L hIgG1 or anti-AXL antibody. g ICC staining for AXL, Vimentin and E-cadherin in LCLC-103H with AXL depletion and 4T1 treated with 40 nmol/L hIgG1 or anti-AXL antibody. *P < 0.05; **P < 0.01; ***P < 0.001.

Full size image

AXL inhibition reduces lung metastasis and improves antitumor immunity in a mouse model of TNBC

To study the effects of AXL inhibition in vivo, we established a syngeneic mouse model via tail-vein injected 4T1 TNBC cells. Once weekly i.v. treatment of mice with anti-AXL resulted in significant inhibition in the overall lung metastasis burden accompanied by the inhibition of AXL downstream signaling function. We observed diminished levels of p-Erk, p-Akt substrate (reflecting Akt activity) and Vimentin expression within the tumor nodules (P < 0.01) (Fig. 3a). The reduction of tumor burden in anti-AXL-treated group was highly associated with the increase of tumor necrotic area, activation of tumor cell apoptosis (P < 0.05) and inhibition of tumor cell proliferation (P < 0.001) (Fig. 3b). We then examined effects of anti-AXL on both the innate and adaptive immune cell infiltration in tumor environment. We found that CD11b+ myeloid cells were significantly reduced, while the activated DCs (CD11c+CD86+, CD11c+CD103+) were substantially increased in the anti-AXL treated group (P < 0.01) (Fig. 3c). We next quantified adaptive immunity response. Both the CD8+ T cells (P < 0.001) and CD19+ B cells (P < 0.01) were enriched in treated tumors, with a significant increase in cytotoxic T cells (CD8+GzmB+) (P < 0.01) (Fig. 3d). Taken together, these results demonstrated that anti-AXL antibody can effectively reduce metastatic tumor burden in the lung, which is linked to the downregulation of AXL signaling function, EMT state and activation of anti-tumor immunity.

Fig. 3: AXL inhibition reduces lung metastasis and induces antitumor immunity in a syngeneic mouse model of TNBC.
figure 3

Lung tissue sections were subjected to IHC or multiplexed-IHC (TSA method) analysis. a Lung metastatic nodules, staining level of the p-Erk, p-Akt substrate, AXL and Vimentin in tumor core area of 4T1 metastasis. b Staining of necrosis, apoptosis (TUNEL) and cell proliferation marker. c Dual-staining with DCs biomarkers (CD11b, CD11c/CD86 and CD11c/CD103). d Dual-staining with T cell marker CD8A and GzmB, and staining with B cell marker CD19. Representative views and statistics of relative areas were shown. *P < 0.05; **P < 0.01; ***P < 0.001.

Full size image

AXL-targeted inhibition of tumor-associated macrophages (TAMs) in tumor microenvironment

Tumor-associated macrophages (TAMs) resemble those of M2-polarized macrophages that play an important role in tumor progression and therapy resistance [39, 40]. As TAMs are elevated in the AXL-H patient tumors (Fig. 1f, 1g), we wished to study the potential relationship between AXL and TAMs. In analysis of TCGA-TNBC cohort, AXL expression is correlated with the M2-macrophage marker CD206 (P < 0.001) (Fig. 4a) and M2-derived cytokine CSF-1, CCL2 and TGFβ (Fig. 4b). In experimental 4T1 TNBC model, activated AXL was detected frequently as dual-positive staining of AXL+ and p-AXL+ in tumor cells and in tumor stroma (Fig. 4c). AXL was present in approximately 20% of macrophages (AXL+F4/80+) and was greatly enriched up to 80% of M2-polarized population (AXL+CD206+, AXL+CSF-1+) (Fig. 4c). Such a strong induction of AXL and its association with M2-TAMs was only observed in the tumor nodule but not in surrounding normal lung tissue of 4T1 model. This result was further confirmed by multiplex-IHC staining of human NSCLC tumor tissue, in which AXL was also expressed in M2-TAMs (Supplementary Fig. S3a, S3b). We next speculated that AXL activation could promote M2-TAMs function. Remarkably, treatment with anti-AXL antibody resulted in a complete suppression of p-AXL signaling leading to a nearly complete depletion of M2-TAMs (p-AXL+CD206+, CD206+F4/80+) and M2-derived cytokine (CD206+CSF-1+, p-AXL+CSF-1+) (Fig. 4d, Supplementary Fig. S3c). In contrast, the M1-like macrophage (CD86+F4/80+) and TNFα, a major tumor-killing cytokine secreted by M1-like macrophage, were markedly increased in the anti-AXL group (P < 0.05) (Fig. 4e). Together these results indicate that targeted inhibition of AXL on M2-TAMs likely contributes to its polarization towards M1 type and activation of antitumor immunity of macrophages.

Fig. 4: AXL-targeted inhibition of M2 macrophage polarization and function in tumor environment.
figure 4

a TCGA transcriptome dataset shows correlation of AXL with CD206 in TNBC patients. b Transcriptome data sets show M2-like macrophages related cytokines in AXL-H vs. AXL-L group from TNBC patients (TCGA) and cancer cell lines (CCLE). c Multiplex-IHC staining of AXL, p-AXL and macrophage biomarker (F4/80, CD206 and CSF-1) in metastatic nodules. The AXL+ fractions were quantified by positive cell counting. d Features of M2-TAMs and AXL activation were measured using p-AXL/CD206, CSF-1/CD206 and p-AXL/CSF-1 staining. e M1-like macrophages were detected by staining of CD86/F4/80, TNFα/F4/80 and TNFα in metastases. Representative views and statistics of relative areas are shown.*P < 0.05; **P < 0.01; ***P < 0.001.

Full size image

AXL inhibition interferes with M2 polarization of macrophages in vitro

To further investigate the role of AXL in macrophage M2 polarization, we examined the effect of anti-AXL in vitro. Mouse bone marrow derived macrophages (BMDMs) were induced for M1- or M2-polarization with or without anti-AXL treatment. We observed that anti-AXL significantly reduced the percentage of CD206+ M2-polarized BMDMs (Fig. 5a) correlating a reduction in cell surface expression of AXL level (Fig. 5b). To mimic the interaction of AXL-H tumor cell with macrophage in tumor environment, we prepared 4T1 conditioned medium (CM) after pre-treatment with hIgG1 (CM_4T1 + hIgG1) or anti-AXL (CM_4T1 + anti-AXL) (Fig. 5c). Coincubation of RAW264.7 macrophages with CM_4T1 + hIgG1 resulted in a profound increase in p-AXL and CD206 levels indicating enhanced AXL signaling and M2 polarization. Upon coincubation of RAW264.7 with CM_4T1 + anti-AXL, the M2-promoting effect was completely blocked and M1-marker CD86 was significantly induced (Fig. 5d). These results clearly demonstrate that AXL is induced and functionally present in M2 macrophages, and direct inhibition of macrophage-borne AXL in addition to tumor cell-AXL can efficiently interfere with M2 polarization and its tumor-promoting activity.

Fig. 5: AXL inhibition impairs the M2 polarization of macrophages in vitro.
figure 5

a Flow cytometry analysis of the M2 marker F4/80/CD206 and M1 marker F4/80/CD86. b AXL expression in macrophages of Fig. 5a and 5b were analyzed by flow cytometry. c Schematic diagram of the tumor cell conditioned medium (CM) collection process. d Immunoblotting of the indicated RAW264.7 treated with 4T1-CM 48 h.

Full size image

Discovery and characterization of novel humanized anti-AXL antibody and AXL-ADC

As we have revealed a role of AXL in EMT and tumor environment, we attempted to develop a novel AXL antibody and AXL-ADC. We analyzed the Affymetrix array data for representative tumor lines (CCLE) and normal human tissues (GTEX), which confirmed a higher normalized AXL mRNA levels in tumor cells (Fig. 6a). We generated a new and humanized AXL antibody AXL02 that contain the novel complementarity-determining regions (CDRs) (Supplementary Fig. S1c). AXL02 demonstrated antigen binding affinity value 0.055 nmol/L (ELISA) (Fig. 6b). In FACS assays, AXL02 strongly targeted cell surface-AXL with EC50 value 0.37 nmol/L in LCLC-103H (Fig. 6c). We confirmed that AXL02 selectively inhibited cell migration in the AXL-H cells (H1299, LCLC-103H) but not in AXL-L (H460) (Fig. 6d). Upon incubation with LCLC-103H, AXL02 was internalized into cytoplasm and co-localized with lysosome marker LAMP-2 (Fig. 6e), and ultimately resulted a significant AXL degradation (Supplementary Fig. S4). AXL02 was linked with monomethyl auristatin E (MMAE) by a site-specific conjugation platform (AXL02-MMAE) (Fig. 6f) and a traditional mc-VC-PAB conjugation (AXL02-vc-MMAE), respectively. Hydrophobic interaction chromatography (HIC) results confirmed that AXL02-MMAE has the advantage of high uniformity in drug antibody ratio of 4 (DAR4) compared with the variable DARs from that of AXL02-vc-MMAE or Enapotamab Vedotin (EnaV) [41] (Fig. 6g). The cytotoxicity of AXL02-MMAE was diminished in Calu-1 cells when AXL was depleted, further supporting target specificity (Fig. 6h). Isotype matched hIgG1-MMAE showed no specific killing of Calu-1 cells (Fig. 6i). AXL02-MMAE elicited a broad and highly potent cytotoxicity toward AXL-high cell lines (LCLC-103H, Hs578T, PC9, U87MG, and MDA-MB-231, Calu-1) with IC50 < 0.1 nmol/L but not in the AXL-low cells (MDA-MB-453 and H460) with IC50 > 100 nmol/L (Figs. 6i, 6j). These results overall demonstrate that AXL02-MMAE represents a novel and highly effective AXL-ADC for specific targeting of multiple AXL-high expressing solid cancers.

Fig. 6: Discovery of highly potent and humanized anti-AXL antibody and AXL-ADC.
figure 6

a Normalized mRNA levels for AXL from cancer cells (CCLE) or normal human tissues (GTEX, https://gtexportal.org/home/) are plotted. Affinities of AXL02 to human AXL-ECD via ELISA (b) and LCLC-103H cell-surface AXL (c). d Cell migration of H1299, LCLC-103H, and H460 cells were assayed with 40 nmol/L AXL02 treatment. e LCLC-103H cells were incubated with 5 μg/mL AXL02 as indicated and subjected to immunofluorescence. f Schematic diagram of AXL02-MMAE. g Hydrophobic interaction chromatography (HIC) showed the DAR composition of AXL02-MMAE, AXL02-vc-MMAE and EnaV. h Calu-1 cells transfected with shNT or shAXL were assayed for AXL02-MMAE cytotoxicity. i Calu-1 were treated with serial doses of AXL02-MMAE and hIgG1-MMAE, measured for proliferation. j The indicated cell lines were treated with serial doses of AXL02-MMAE, measured for proliferation. ***P < 0.001.

Full size image

AXL-ADC demonstrates promising anti-tumor efficacy in multiple AXL-high solid cancer models in vivo

To evaluate the utility of AXL-ADC as single agent therapy, we performed AXL IHC on lung cancer tumor microarray (TMA) and classified four representative AXL expression levels, from which the Medium- and High expression in TMA specimen were comparable with the expression range of our cell line-derived xenograft tumors, e.g., U87MG, NCI-H1299 and LCLC-103H (Fig. 7a). Treatment of staged U87MG tumors with AXL02-MMAE, AXL02-vc-MMAE or EnaV resulted in a complete suppression of tumor growth, which clearly lasted longer duration in the AXL02-MMAE-treated group compared with that of AXL02-vc-MMAE or EnaV (P < 0.001) (Figs. 7b, 7c). In LCLC-103H tumor expressing highest level of AXL, AXL02-MMAE at 1 mg/kg achieved a complete tumor suppression which was more sustained than that by EnaV (P < 0.001) (Fig. 7d). AXL02-MMAE was also effective against tumors PC9 (Fig. 7e) and NCI-H1299 that expresses elevated MDR [41] (Fig. 7f). Overall, the naked AXL02 did not inhibit tumor growth at dose up to 10 mg/kg (Supplementary Fig. S5a) and did not strongly suppress AXL phosphorylation in vitro (Supplementary Fig. S5b) but AXL02-MMAE exhibited antitumor activity in multiple tumor models with a minimum efficacy dose (MED) of ≤1 mg/kg (Supplementary Table S2). In vivo distribution study showed that AXL02-MMAE offered a higher total MMAE distribution to the tumor compared with that of AXL02-vc-MMAE, which is consistent with its superior antitumor efficacy (Figs. 7g, 7h). Together, these results identify AXL02-MMAE as a novel, AXL-targeted single agent therapy for potential treatment of AXL-high advanced cancers.

Fig. 7: Potent in vivo antitumor efficacy of novel AXL-ADC in athymic mice.
figure 7

a IHC was performed to detect the expression level of AXL in NSCLC TMA and experimentally constructed subcutaneous xenografts, and gray value was quantified by ImageJ. U87MG tumors were treated with 5 mg/kg (b) or 3 mg/kg (c) hIgG1-MMAE, AXL02-MMAE or AXL02-vc-MMAE as indicated. Tumor growth curves are shown. Tumors of LCLC-103H (d), PC9 (e) and H1299 (f) were treated with various doses of hIgG1-MMAE, AXL02-MMAE or EnaV as indicated. Tumor growth curves are shown. U87MG tumor-bearing mice were i.v dosed with 5 mg/kg AXL02-MMAE or AXL02-vc-MMAE, plasma and tumor tissue were collected at the indicated times and analyzed for total MMAE concentration as described in Methods. MMAE concentration in plasma (g) and tumor tissue (h) are plotted. Rx, treatment; *P < 0.05; ***P < 0.001.

Full size image

AXL-ADC induces a dramatic tumor regression and remodels TME differentiated from docetaxel treatment

To assess whether AXL02-MMAE can penetrate a large tumor, LCLC-103H tumors were grown to ~1800 mm3 before treatment. A single dose of 10 mg/kg AXL02-MMAE caused a rapid and nearly complete tumor regression, which was more profound than the efficacy with 15 mg/kg docetaxel (P < 0.01) and better tolerated without significant body weight loss (Fig. 8a). The dramatic tumor regression was accompanied by the loss of AXL and induction of apoptosis as shown by the TUNEL staining (P < 0.001) (Fig. 8b). Although MMAE and docetaxel are both microtubule-targeting agents, only the treatment with AXL02-MMAE triggered a substantial remodeling of tumor immune environment. Specifically, treatment with AXL02-MMAE, but not with docetaxel, induced an increased staining in M1-like macrophages (CD86+F4/80+, P < 0.05) but a decreased staining in the M2-TAMs (CD206+F4/80+, P < 0.001) and M2-related CSF-1 (P < 0.001) in TME (Fig. 8c), while activated DCs (CD86+CD11c+, CD103+CD11c+) were enriched in AXL02-MMAE treatment group (P < 0.05) (Supplementary Fig. S6). Because AXL02-MMAE does not cross-react with mouse AXL, the observed change in myeloid cell profile may reflect an indirect effect of tumor cell killing. In vitro study showed that AXL-KD in LCLC-103H cells significantly reduced CSF-1, a major growth factor of M2-macrophage (Fig. 8d). The CM with AXL-KD or anti-AXL-treated LCLC-103H cells each blocked CD206 expression in RAW264.7 (Fig. 8e). Although AXL02-derived CM only partially reduced macrophage CD206 possibly due to its less effectiveness in suppression of p-AXL in tumor cells (Fig. 8e, Supplementary Fig. S5b), AXL02-MMAE is expected to potently eliminate LCLC-103H cells, which could in turn impair paracrine signaling of tumor cell to macrophage resulting in a loss of AXL-dependent M2-polarization and/or infiltration in TME. Taken together, these results reveal a new antitumor mechanism for AXL-targeted AXL02-MMAE that promotes an antitumor immune microenvironment in addition to direct cytotoxic effects on tumor cells.

Fig. 8: AXL-ADC induces a dramatic tumor regression and suppression of tumor associated M2-like macrophages in athymic mice.
figure 8

a LCLC-103H xenografts constructed in nude mice were administered with 10 mg/kg hIgG1-MMAE, AXL02-MMAE or 15 mg/kg docetaxel (total 1 dose). Tumor growth curves and mice body weight are plotted. b IHC or multiplex-IHC staining were performed to detect apoptosis (TUNEL), proliferation (Ki67) and AXL protein. c Staining of M1-like macrophages (CD86/F4/80), M2-TAMs (CD206/F4/80 and CSF-1/CD206) in LCLC-103H tumor section. Representative views and statistics of relative areas are shown. d Immunoblotting of AXL knockdown LCLC-103H cells. e Immunoblotting of the indicated RAW264.7 treated with LCLC-103H-CM 48 h. Rx, treatment; *P < 0.05; **P < 0.01; ***P < 0.001.

Full size image

Discussion

In this study we have investigated the importance of AXL in the establishment of an EMT-related and immunosuppressive tumor architecture. Analysis of NSCLC and TNBC transcriptome profile showed that elevated AXL expression is associated with poor prognosis, EMT, and immunosuppressive tumor environment. EMT is increasingly recognized to orchestrate a large variety of complementary cancer features, and the regenerative feedback loop of AXL and EMT aggravated tumor plasticity, drug resistance and immune evasion [20, 23, 42, 43]. In our experimental NSCLC and TNBC cells with elevated AXL expression, inhibition of AXL by AXL-shRNA or two independent anti-AXL antibodies all significantly decreased the Vimentin/E-cadherin mesenchymal index and AXL-dependent cell migration.

TAMs participate in tumor growth and therapy resistance [39, 40, 44]. Very little is known about AXL’s involvement in cancer-promoting TAMs. We set out to investigate whether AXL plays a role in this setting and revealed a major relationship between AXL and tumorigenic macrophages in the tumor environment. Transcriptome analysis of patient tumors identified the positive link between AXL and elevated macrophage infiltration. In our mouse studies, p-AXL/AXL were present in the macrophages (p-AXL+/AXL+F4/80+) infiltrated into the lung tumor nodules but not in the surrounding normal lung tissue. We further discovered that the activated AXL was greatly enriched in the M2-TAMs population (p-AXL+CD206+), and was associated with the expression of M2-promoting cytokine CSF-1 (p-AXL+CSF-1+). Co-expression of p-AXL/AXL in M2-TAMs was further confirmed in the human NSCLC tumor specimen. Importantly, dual-blockade of AXL activity in tumor cells and M2-TAMs by an AXL antibody (YW327.6S2) resulted in a decrease in CSF-1 level and depletion of M2-TAMs in tumor microenvironment. Additional in vitro studies with conditioned medium suggested an important mechanism for paracrine signaling of AXL-H tumor to macrophage through secreted M2-promoting cytokines such as CSF-1 in initiating M2-polarization.

The current results support the notion that targeting of AXL-dependent TAMs function can confer antitumor benefits. In mouse model of TNBC, AXL-targeted loss of M2-TAMs was associated with significant increases in the M1-like macrophages (CD86+F4/80+, TNFα+F4/80+) and activated dendritic cells (CD86+CD11c+, CD103+CD11c+), all of which can induce production in tumor-killing cytokines such as TNFα while also favor infiltration of the adaptive antitumor effector cells [45]. Indeed, inhibition of AXL resulted in a robust transformation from immunosuppressive tumor environment to a potent anticancer microenvironment as demonstrated by the coordinated enhancements in cytotoxic T cells (CD8+GzmB+) and B cells (CD19+). These results strongly suggest that AXL-targeted antitumor efficacy is mediated in part through the activation of innate and adaptive immunity. Consistent with our results, a new report by Tirado-Gonzalez et al. showed that ablation of the AXL specifically in macrophages in the environment can stimulate anti-leukemic immunity [46]. Together these results highlighted a critical role for AXL-driven macrophages in tumor progression and provides a rationale for AXL-targeted therapy.

Antibody-drug conjugate (ADC) represents a new emerging class of highly potent pharmaceutical drugs to eliminate tumor cells by specific transferring chemotherapeutic agents into target tumor cells while sparing healthy tissues [47]. One aspect of our study focused on the discovery and therapeutic evaluation of a novel AXL-ADC. Transcriptome analysis confirmed a prominent cancer-selective differential expression window for AXL compared with that of normal tissues. Our newly discovered humanized antibody AXL02 can specifically suppress AXL-dependent cell migration and EMT in AXL-high tumor cells and could internalize rapidly and extensively to intracellular compartments. These features rendered AXL as a feasible target for ADC development.

We employed a novel disulfide rebridging linker technology (BL20-MMAE) to obtain AXL02-MMAE. This linker system was designed to form site-specific disulfide bonds through cross-linking to the reduced cysteines in the Fab and hinge regions of the antibody rendering a homogeneous DAR4 in AXL02-MMAE rather than an average DAR4 seen from the traditional vc-MMAE conjugated ADCs (e.g. AXL02-vc-MMAE and EnaV). Hamblett et al [48] showed that a drug-to-antibody ratio of 4 (DAR4) resulted in an optimal potency and safety profile. Therefore, a site-specific DAR4 conjugate may enable the best achievable therapeutic index. Besides, the N-(3,5-difluorophenyl)maleimide moiety accelerated the process of maleimide hydrolysis, which is monitored by LC-MS. Complete hydrolysis of maleimide could prevent the deconjugation by the retro-Michael reaction and stabilize antibody conjugates [49]. Together, AXL02-MMAE achieved a higher total MMAE distribution to the tumor tissue and a significantly longer-lasting antitumor efficacy. In multiple NSCLC tumor models expressing comparable AXL levels as that of clinical NSCLC specimen, AXL02-MMAE elicited potent and more sustained antitumor efficacy compared with that of the vc-MMAE-conjugated AXL02-vc-MMAE or EnaV. AXL02-MMAE achieved a superior efficacy in regressing large-sized tumors compared with that of chemotherapy. Mechanistically, AXL02-MMAE eliminated AXL-H tumor cells and AXL-induced M2-macrophage infiltration leading to a robust accumulation of inflammatory macrophages and mature dendritic cells, and a greatly enhanced apoptotic tumor death. We conclude from these results that AXL02-MMAE represents a new class of highly potent and selective antitumor monotherapy.

In summary, our results have provided a strong rationale for targeting AXL as new cancer therapy. The promising efficacy and mechanisms exhibited by AXL-targeted antibodies and AXL-ADC highlight a therapeutic opportunity for subsets of AXL-dysregulated solid cancer.