Dual Mechanisms of Novel CD73-Targeted Antibody and Antibody–Drug Conjugate in Inhibiting Lung Tumor Growth and Promoting Antitumor Immune-Effector Function

This article is featured in Highlights of This Issue, p. 2233

Non–small cell lung cancer (NSCLC) represents one of the most common human malignancy and remains the leading cause of cancer-related death worldwide (1). Although significant progress has been made in the clinical management of advanced NSCLC owing largely to the tumor-selective tyrosine kinase inhibitor (TKI) therapy and cancer immunotherapy (e.g., anti–CTLA-4, anti–PD-1/L1; refs. 2, 3), a limited durable response rate and acquired resistance to these therapies are increasingly being encountered (4–7). In the area of immunotherapy, efforts are directed to understand the immune profile correlates of clinical response. Treatment benefits have been linked to tumor preexisting CD8⁺ T cells, tumor-intrinsic PD-L1 expression, and tumor mutational burden (8–11). Nevertheless, these biomarkers do not always provide prognostic role and predictive significance. It is likely that the immune landscape of NSCLC reflects a complex and dynamic cross-talk between the tumor and immune system (6, 7) and an altered cancer metabolism reflecting tumor heterogeneity and plasticity will collectively influence therapy response (12).

CD73 is a glycosylphosphatidylinositol (GPI)-anchored ecto-5′-nucleotidase (NT5E) that together with CD39 catalyzes the conversion from ATP/AMP to adenosine. Elevated CD73 expression occurs frequently in human cancers and often correlates with a worse prognosis observed in tumors of lung, colorectal, breast, ovarian, renal, and glioblastoma (13–18). The levels of CD73 can be induced by oncogenes, inflammatory cytokines, hypoxia, and anticancer drugs (19–21). CD73 is positively linked to the epithelial–mesenchymal transition (EMT) process (22) and resistance to anthracycline in triple-negative breast cancer (TNBC; ref. 15), resistance to BRAF-TKI in melanoma (23), and anti-HER2 monoclonal antibody in HER2/erbB2-driven breast cancer (24). These earlier studies suggested an important role of CD73 in regulating disease progression and poor treatment outcome.

CD73 has also emerged as a negative regulator of cancer immunity, which is thought to involve its enzyme product adenosine, an immunosuppressive molecule that can act on numerous immune-effector cells and suppressor cells (19). Gene knockout of tumor surface CD73 could improve antitumor efficacy in mouse models of cancer (25, 26). In TNBC, patients with high levels of CD73 expression on epithelial tumor cells show low tumor-infiltrating leukocytes (TIL) and worse clinical outcome (27). Such an inverse relationship for high CD73 level with low CD8⁺ TILs or low IFNγ signature also occurs in lung cancer patients, which correlates with reduced response to anti–PD-1/L1 (28, 29). In a metastatic melanoma trial with anti–PD-1, patients with high plasma soluble CD73 activity experience worse outcome (30). Additionally, dynamic induction of CD73 occurs in melanoma patients progressing under anti–PD-1/L1 treatment arguing for a role for CD73 in acquired resistance mechanism (31). In preclinical studies, CD73 inhibitor or CD73-targeted antibody could overcome AMP-mediated limitation of immunotherapy efficacy (32, 33). To date, several CD73-blocking antibodies have recently entered human cancer trials with results eagerly awaited (34–36).

Herein, we have generated highly potent and novel CD73-targeted antibodies and, for the first time, antibody–drug conjugate (CD73-ADC). We explored a rationale and pharmacologic feasibility for CD73-targeted therapy in preclinical models of advanced NSCLC. We found that CD73 is highly elevated in major NSCLC types and correlates with EMT status and an immunosuppressive tumor microenvironment. We describe the characterization of these new CD73-targeted anticancer agents, their promising efficacy and mechanism in a multifaceted remodeling of the tumor immune microenvironment.

Balb/c mice were immunized with an extracellular domain of human CD73 protein (CD73-ECD). Splenocytes were fused with SP2/0 myeloma cells, screened with ELISA and flow cytometry (FACS). Antibody variable regions were obtained using SMARTer RACE 5′/3′ Kit (Clontech) and were first constructed as mouse chimeric antibody Ab001 and Ab002. The CDRs were grafted to appropriate humanized templates generating Hu001 and Hu002 in human IgG1/kappa backbone (37). The CD73 ADCs were generated by bioconjugation of monomethyl auristatin E (MMAE) to the endogenous cysteine residues on Hu001 or hIgG1 via substituted maleimide linker with an average drug-to-antibody ratio of 4 (37).

Cell lines of NCI-H1975, NCI-H1993, NCI-H2228, NCI-H441, MDA-MB-231 (MDA231), and MDA-MB-453 (MDA453) were obtained from ATCC. Calu-1, Calu-6, NCI-H1299, NCI-H292, HCC827, U87MG, HCC1937, Hs578T, MDA-MB-436 (MDA436), MDA-MB-468 (MDA468), BT549, ZR75-1, and T47D were obtained from the Chinese Academy of Sciences (CAS, Shanghai). PC9 was obtained from European Collection of Cell Cultures (ECACC). Cells were cultured using standard methods and reagents (Invitrogen) and periodically tested for Mycoplasma contamination free. The pTRIPZ-based shRNA lentiviral vectors for human CD73/NT5E or nontargeting (Open Biosystems, cat. #V3LH_359876, cat. #RHS4346) were packaged in HEK293T cells. Tumor cells were stably infected with lentivirus and induced for CD73 depletion with 1 μg/mL doxycycline (Dox) for 6 days. Cancer lines were seeded in 96-well plates at predetermined density, treated with CD73-ADCs for 5 to 6 days before assayed for proliferation via MTS reagent (Promega). For immunoblotting, total cell lysates prepared from various caner lines were immunoblotted with anti-human CD73 (Cell Signaling Technology, cat. #13160) and β-actin (Bioworld).

For CD73-ELISA, CD73-ECD (1 μg/mL) was coated to ELISA plate, incubated with serially diluted CD73 antibodies and detected with HRP-labeled secondary antibody. Surface plasmon resonance was performed in a Biacore-T200 instrument (GE Healthcare). Anti-human IgG-Fc (GenWay) was immobilized on sensor chip to capture CD73 antibody. Serially diluted CD73-ECD was sequentially injected. KD values were obtained using the system software. For cell-surface CD73 binding, 1 × 10⁵ tumor cells were incubated with CD73 antibodies at 4°C for 1 hour, stained with R-PE-Goat anti-human IgG-Fc (Abcam), and analyzed by FACS (BD FACSAria II). For internalization, MDA-MB-231 or H441 cells pre-seeded in confocal chamber slides were incubated with CD73 antibody as indicated, fixed with 4% formaldehyde and permeabilized with 0.4% TritonX-100. The slides were incubated with anti-LAMP-2 (Abcam, cat. #ab125068) or lyso-Tracker-red (Life Technologies, cat. #L12492) before detected with Alexa-Fluor-488 or R-PE-labeled secondary antibody. Images were captured under confocal microscope (Zeiss, LSM710).

Reagents AMP, ATP, and APCP were purchased from Sigma-Aldrich. The CD73 enzyme assay principle was described (38). For soluble CD73, 25 μL CD73-ECD (0.1 μg/mL) was premixed with 50 μL serially diluted CD73 antibody in assay buffer (25 mmol/L Tris pH 7.5, 5 mmol/L MgCl₂, 0.005% Tween-20). The reaction initiated upon addition of 25 μL AMP/ATP mixture (final dose 0.3 mmol/L/0.1 mmol/L) and continued for 1 hour at 37°C. The remaining AMP was quantified using a CellTiter-Glo assay kit (Promega, cat. #G7571). For cell-surface CD73, cells were seeded in 96-well plates at appropriate density, washed, and pretreated with 50 μL serially diluted CD73 antibody for 30 minutes. The reaction initiated with 25 μL AMP (0.9 mmol/L) and continued for 3 hours. Reaction supernatant (25 μL) was mixed with 25 μL ATP (0.1 mmol/L) and assayed with 50 μL CellTiter-Glo reagent.

Frozen human PBMCs (Xidier Biotech) were thawed and cultured in RPMI-1640 supplemented with 10% FBS plus immobilized anti-CD3 (0.5 μg/mL), anti-CD28 (0.5 μg/mL; eBioscience), and 100 U/mL IL2 (PeproTech), selected using the CD3⁺ T-cell Isolation Kit (Stemcell, cat. #17951) and labeled with CFSE (Invitrogen). T cells were seeded in 96-well plates (2 × 10⁴/well), treated with CD73 antibody or CD73-ADC without or with AMP (final 250 μmol/L or as indicated). The T cells were grown for 4 to 5 days and analyzed for viability by FACS (Beckman CytoFLEX S). In some experiments, T cells were stained with anti-CD4 or anti-CD8 (Invitrogen, cat. #11004742, cat. #25008642) before FACS analysis. The culture supernatants were subjected to IFNγ ELISA using assay kit (Multi Sciences, cat. #EK180HS-48).

To generate conditioned medium (CM), H1299-ShNT and H1299-ShCD73 cells were preinduced with doxycycline for 6 days to deplete CD73. Cells were incubated with fresh medium plus 250 μmol/L AMP for 2 hours before collected as ShNT-CM, ShCD73-CM, and control-CM (mock). The CMs were added to 2 × 10⁴ CFSE-labeled T cells for 5 days and analyzed for growth by FACS. For tumor/T-cell coculture, 1 × 10⁴ H1299 cells were cocultured with 2 × 10⁴ CFSE-labeled T cells for 5 days without or with CD73 antibodies or APCP. Viable T-cell numbers were measured by FACS.

Frozen human PBMCs were thawed and seeded in 24-well plates at 1.5 × 10⁶/well. Cells were induced to differentiate in RPMI-1640 supplemented with 10% FBS plus GM-CSF (60 ng/mL) and IL4 (100 ng/mL; ref. 39). After 2 days, the nonadherent cells were gently removed, and the wells were replenished with fresh differentiation medium. Cells were exposed to 5 μg/mL CD73 antibody or CD73-ADC without or with AMP (on day 2) or 500 ng/mL LPS (on day 6). On day 7, the cells were costained with anti-CD11c and anti-CD86 (BioLegend, cat. #117306 and cat. #105008) then analyzed by FACS. On day 7, a separate set of identically activated DCs in 96-well plate were washed and cocultured with 2 × 10⁴ CFSE-labeled T cells for 5 days. Viable T-cell numbers were measured by FACS.

Tumor growth studies were performed under protocols approved by IACUC of the Fudan University School of Pharmacy. Female Balb/c nude mice (4–6 weeks old) were subcutaneously implanted with NCI-H1299 (10 × 10⁶), NCI-H441, U87MG, or NCI-H292 (all at 5 × 10⁶). For tumor/antibody coimplanting, tumor cells were mixed with 50 μg of antibody for 1 hour before coinjected. For therapeutic treatment, tumor-bearing mice were staged at 100 to 200 mm³ (growth) or 700 to 800 mm³ (regression) and randomized into treatment groups (n = 8). The indicated ADCs or docetaxel (BiochemPartner) were dosed i.v. 1×/weekly. Tumor volume was monitored using a caliper twice a week and calculated using the formula V = LW²/2 (where V = volume, L = length, and W = width).

For IHC, xenograft tumors were fixed in 4% formalin and embedded in paraffin. Slides were processed with standard protocol for staining with anti-CD31 (Servicebio, cat. #GB13036), CD73 (Cell Signaling Technology, cat. #13160), NCR1/NKp46 (Absci, cat. #AB37471), F4/80 (Biossci, cat. #BA9408), and CD206 (Abcam, cat. #ab64693). The lung cancer specimen IHC was performed on tissue microarrays (TMA; Outdo Biotech, cat. #HLugA180Su07) with anti-CD73. Staining results were graded by a pathologist on tumor cells and lung epithelial cells. The IHC combined score (CS) was quantified as staining intensity × percentage of positive tumor cells (cancer specimens) or lung epithelial cells (paracancer specimens). The CS values (qualified cancer = 86, paracancer = 81) were together grouped into CD73-negative (CS = 0; n = 66) and CD73-positive (CS>0; n = 101). The positive CS specimens were grouped into low positivity (</=median; n = 51). The remaining 50 specimens were grouped at a new median into medium positivity (n = 25) and high positivity (n = 25).

For IF, slides were costained with anti-CD31 and α-SMA (Abcam, cat. #ab119952, cat. #ab225883), costained with anti-CD86 and anti-CD11c (BioLegend, cat. #105008, cat. #117306), and costained with anti-CD11b (BioLegend, cat. #101207) and anti-Gr1 (Elabscience, cat. #E-AB-F1120C). Images were acquired using Leica microscope (model DMI4000D) and quantified by software Image pro plus 6.0.

All numerical data processing and statistical analysis were performed with Microsoft Excel and GraphPad Prism 6 software; values are mean ± standard deviation except for in vivo studies data, which are mean ± standard error. P values were calculated using an unpaired two-tailed Student t test, and values less than 0.05 indicate significant statistical difference.

We first performed CD73 IHC study on lung cancer TMA. Analysis of cancer (n = 86) and paracancer (n = 81) specimens detected an overall CD73 positivity in 86% cancer versus 33.3% paracancer, which included a combined medium and high positivity in 52.35% cancer (n = 45) versus 6.17% paracancer (n = 5; Fig. 1A and B). High CD73 staining was found in epithelial tumor cells and stromal cells in tumor microenvironment (Fig. 1C). We next analyzed NSCLC transcriptome data set from The Cancer Genome Atlas (TCGA, n = 503), which revealed a higher CD73 mRNA expression in the RAS-mutant tumors (n = 159; Log₂ value 10.05 ± 1.65 vs. 8.8 ± 1.63, P < 0.001) and RTK-mutant tumors (EGFR, EML4-ALK, n = 70; Log₂ value 10.23 ± 1.69 vs. 9.58 ± 1.75, P < 0.01; Fig. 1D). Analysis of the Cancer Cell Line Encyclopedia (CCLE) cell line data set (n = 186) also revealed an elevated CD73 mRNA in RAS-mutant lines (Log₂ values 7.86 ± 2.17 vs. 4.65 ± 1.58, P < 0.001) and RTK-mutant lines (Log₂ values 8.08 ± 1.45 vs. 5.97 ± 2.22, P < 0.001; Fig. 1E). Immunoblotting with NSCLC cell lines harboring oncogenic K/NRAS (H1299, Calu-1) or c-Met (H1993), EML4-ALK (H2228), or EGFR (HCC827, PC9, and H1975) showed an elevated CD73 protein in majority of the cell lines (Fig. 1F). We then clustered NSCLC cell lines (CCLE, n = 186) or tumors (TCGA, n = 336) into “CD73-high (CD73-H)” and “CD73-low (CD73-L)” based on CD73 mRNA expression. CD73 expression was enriched with the EMT genes vimentin (VIM), fibronectin 1 (FN1), TGFB1, MMP2, ZEB1 in cell lines (P < 0.001; Fig. 1G) and tumors (P < 0.001; Fig. 1H). CD73 expression was also linked to the biomarkers for tumor-associated macrophages (TAM) CD206, CD163, and myeloid-derived suppressor cells (MDSC) CD11b, CD33 (P < 0.001; Fig. 1I). Together, dysregulation of CD73 appears prevalent in major types of NSCLC patients, which likely contributes to tumor progression, metastasis, and immune escape.

Using modern discovery technology platform, we identified and characterized two new CD73 antibodies Ab001/Ab002 with novel complementarity-determining regions (CDR; Supplementary Fig. S1A and S1B) and obtained humanized versions Hu001/Hu002 (Supplementary Fig. S1C and S1D). Ab001/Ab002 demonstrated respective antigen binding affinity values 0.1 and 0.057 nmol/L (Biacore, Fig. 2A), 0.024 and 0.016 nmol/L (ELISA, Fig. 2B), or enzyme IC₅₀ values 0.025 and 0.031 nmol/L (soluble CD73, Fig. 2C). In FACS assays both antibodies strongly targeted cell-surface-CD73 with respective EC₅₀ values 1.04 and 0.39 nmol/L in H1299 (Fig. 2D) or 2.52 and 0.94 nmol/L in Calu-1 (Supplementary Fig. S2A). We confirmed that the cellular CD73 level was linearly proportional to AMP hydrolysis rates (Supplementary Fig. S2B). Ab001/Ab002 elicited a dose-dependent suppression of AMP hydrolysis when incubated with H1299 (IC₅₀<1 nmol/L, Fig. 2E) or Calu-1 (IC₅₀<1 nmol/L, Supplementary Fig. S2C). Both humanized Hu001 and Hu002 fully retained binding potency and exhibited no appreciable affinity with murine CD73 (Supplementary Fig. S2D). Nude mouse tumor studies showed that coimplantation of the RAS-mutant/CD73-high H1299 tumors (Fig. 2F) or H441 tumors (Fig. 2G) with Hu001 or Hu002 each substantially reduced tumor growth compared with that of isotype control hIgG1. These results overall indicate that both antibody series can effectively target the functional epitope(s) of human CD73 with high affinity leading to enzyme inhibition and antitumor efficacy in vivo.

CD73-targeted therapy is hoped to improve antitumor immune response. We assessed the effect by Hu001/Hu002 in blocking the immunosuppressive effect of adenosine/AMP on primary human T cells. Exogenous AMP (125–1000 μmol/L) strongly inhibited CD3⁺ T-cell proliferation (Fig. 3A). Remarkably, Hu001 and Hu002 each elicited a highly potent and dose-dependent rescue of T-cell growth from AMP-imposed toxicity achieving EC₅₀ values 0.08 ± 0.01 and 0.01 ± 0.001 nmol/L, respectively (Fig. 3B). ELISA results confirmed a restored IFNγ secretion in Hu001/Hu002 cotreated T cells (Fig. 3C). We collected CM from H1299-ShNT cells (ShNT-CM) or H1299-ShCD73 cells (ShCD73-CM). T-cell proliferation was inhibited when incubated with shNT-CM, but the toxicity was largely prevented when incubated with ShCD73-CM (Fig. 3D). We then developed an H1299 and T-cell coculture system as an in vitro model for tumor-T interaction. T-cell proliferation was inhibited when cocultured with H1299 cells, which was largely rescued by cotreatment with Hu001 or a CD73 inhibitor APCP resulting in restoration of T-cell growth (Fig. 3E, left) and IFNγ secretion (Fig. 3E, right). These results overall demonstrate that CD73-targeted antibody can efficiently protect T-cell growth and effector function in the in vitro model of the CD73-dysregulated tumor microenvironment.

Antitumor efficacy of therapeutic antibodies could involve antibody-dependent cell-mediated cytotoxicity (ADCC), which requires Fc effector binding to Fcγ receptors (FcγR) on nature killer (NK) cell (40). We generated triple mutations (L234F/L235E/P331S) in Hu001 Fc (Hu001/TM) to eliminate IgG effector function (41). In H1299 in vivo model, Hu001/TM significantly but partially reduced antitumor efficacy compared with that of Hu001 with respective tumor volumes 329.21 ± 28.04 vs. 133.27 ± 7.21 mm³ (Fig. 4A), even though both treatments equally reduced the CD73 level (Fig. 4B). There was a prominent increase in NKp46⁺ NK cell infiltration in the Hu001-treated tumors, which was significantly but partially diminished in the Hu001/TM-treated tumors (Fig. 4B). This result indicated an Hu001-stimulated ADCC in vivo as well as an enhanced NK cell infiltration in response to a diminished CD73/adenosine axis function in the absence of efficient ADCC. Next, because we found an association for CD73-high expression with M2-like TAMs and MDSCs in patients (Fig. 1I), we examined biomarkers in treated tumors. There were significantly less CD206⁺/F4/80⁺ macrophages in both the Hu001- and Hu001/TM-treated tumors compared with that of IgG1-control (P < 0.01) when we examined H1299 tumors (Fig. 4C) or H441 tumors (Fig. 4D). There was also a significantly reduced staining for CD11b⁺Gr1⁺ MDSCs in the Hu001- or Hu001/TM-treated H1299 tumors compared with that of IgG1 control (Fig. 4E). The reduction in MDSCs was also seen in treated H441 tumors (Fig. 4F). Because TAMs and MDSCs are well-established tumor-promoting immune-suppressive cells, these results are consistent with a conversion from an “immune-tolerant” to “cancer-immune” tumor microenvironment in response to treatment with anti-CD73 antibody. We then examined anti-CD73 effects on tumor blood vessel (42). Treatment with Hu001 but not hIgG1 compromised tumor angiogenesis as shown by a lower CD31 staining (P < 0.001) and a significant reduction in the vessel lumen area (P < 0.01) in both treated tumors of H1299 or H441 (Supplementary Fig. S3A and S3B). As mature vessels require pericytes and smooth muscle cells for stabilization, we examined pericyte marker α-SMA (43). Positive staining for α-SMA was absent or reduced in a significant percentage of CD31⁺ vessels with compressed lumen area in treated H1299 tumors (Fig. 4G), implying a diminished pericyte function. Collectively, these results indicate that the anti-CD73 Hu001 attenuates tumor growth, which involves an increased ADCC effect, tumor immuno-environment remodeling, and decreased angiogenesis.

Given the prevalence of CD73 overexpression in tumor cells and subsets of tumor-promoting immune cells, we wished to explore the therapeutic potential for CD73-ADC. In addition to the lung cancer TMA study (Fig. 1A and B), we analyzed the Affymetrix array data for representative tumor lines (CCLE) and normal human tissues (GTEX, http://gtexportal.org), which confirmed a higher normalized CD73 mRNA levels in tumor cells (Fig. 5A). CD73 mRNA was also much higher in NSCLC patients (n = 226) compared with normal lung (n = 20) by analysis of the Oncomine database (Fig. 5B). Analysis of breast cancer cell lines (CCLE) showed an elevated CD73 mRNA in high-invasive Basal/TNBC cells (Fig. 5C), which was verified at CD73 protein levels by immunoblotting (Fig. 5D). Upon incubation with MDA-MB-231 (Fig. 5E) or H441 cells (Supplementary Fig. S4), Hu001 was extensively internalized and colocalized with the lysozyme marker LAMP-2 or lyso-Tracker red, rendering it suitable for ADC development. Hu001 was bioconjugated with monomethyl auristatin E (MMAE) via a substituted maleimide linker (Supplementary Fig. S5), which is analogous to the approach used for Brentuximab Vedotin and expected to release MMAE and induce bystander killing. Hu001–MMAE elicited a highly potent cytotoxicity toward CD73-high cell lines (Calu-1, Calu-6, H292, H441, U87MG, and MDA-MB-231) with IC₅₀ <0.1 nmol/L but not the CD73-low cells (MDA-MB-453 and T47D) with IC₅₀ >100 nmol/L (Fig. 5F). The cytotoxicity was diminished in Calu-1 cells when CD73 was depleted, further supporting target specificity (Supplementary Fig. S6). Quantitative imaging analysis detected 52.3% of NSCLC specimens (45/86) showing high or medium level CD73-IHC gray values, which were comparable with the range of our cell line–derived xenograft tumors (Fig. 1A; Supplementary Fig. S7A and S7B). In nude mice bearing U87MG glioma (Fig. 5G), lung cancer H441 (Fig. 5H) and H292 (Fig. 5I) tumors, treatment with 3 or 1 mg/kg of Hu001–MMAE (i.v. 1× weekly) potently inhibited tumor growth in all three tumor models. The highest efficacy was observed with H441 and H292 lung tumors achieving a minimum effective dose (MED) of <1 mg/kg (Fig. 5G–I; Supplementary Table S1), whereas similar i.v. administration with naked antibody Hu001 did not achieve a robust efficacy (Supplementary Fig. S8). In the H1299 tumors, which express an elevated MDR1 gene (44), 5 mg/kg Hu001–MMAE was also more efficacious than 15 mg/kg docetaxel (Fig. 5J; Supplementary Table S1). These results overall demonstrate that Hu001–MMAE represents a class of new and highly effective CD73-ADC for specific targeting of multiple CD73-high expressing solid cancers including NSCLC.

In a tumor regression study, treatment of large-size H292 tumors (>700 mm³) with 5 mg/kg Hu001–MMAE led to a rapid and sustained tumor regression (P < 0.001), whereas treatment with 15 mg/kg docetaxel resulted mainly in tumor stasis (P < 0.05; Fig. 6A). Although MMAE and docetaxel are both microtubule-targeting agents, only the treatment with Hu001–MMAE triggered a dramatic remodeling of tumor immune environment. Specifically, IHC analysis of H292 (Fig. 6B) and H1299 (Fig. 6C) tumors showed that treatment with Hu001–MMAE, but not with docetaxel, induced an increased staining in the macrophage pan-marker F4/80 but a decreased staining in the M2-macrophage marker CD206, consistent with a robust accumulation of the inflammatory macrophages in the tumor environment. Similar results were also seen in H441-treated tumors (Supplementary Fig. S9). As DCs are important in antitumor immune response, we performed IF staining and found that treatment of H292, H1299, or H441 tumors with Hu001–MMAE each and all induced a robust expansion of mature DCs as shown by the increase in CD11c⁺CD86⁺ double-positive cell population, whereas no significant DC expansion occurred in docetaxel-treated tumors (Fig. 6D; Supplementary Fig. S10A–S10C). Taken together, these results reveal a new antitumor mechanism for CD73-targeted Hu001–MMAE that promotes an antitumor immune microenvironment in addition to direct cytotoxic effects on tumor cells.

To extend our in vivo tumor staining results, we used human PBMC-derived T cells and PBMC-differentiated DCs to examine the direct effect of Hu001–MMAE on these important antitumor immune-effector cells in vitro. Treatment of CFSE-labeled CD3⁺ T cells with Hu001–MMAE (10 nmol/L) up to 12 days did not significantly reduce proliferation rate or cell viability (Fig. 7A). Treatment of T cells with various doses of hIgG1–MMAE or Hu001–MMAE for 5 days also resulted in minimum overt toxicity in CD3⁺ T cells (IC₅₀>100 nmol/L) or CD4⁺/CD8⁺ T cells (Fig. 7B and C). Like the naked antibody Hu001, Hu001–MMAE dose dependently reversed the AMP-imposed suppression on T-cell growth and IFNγ production, achieving an EC₅₀ value 0.09 ± 0.01 nmol/L (Fig. 7D and E). We then adapted a primary human PBMC-DC-differentiation system (39). Treatment of differentiating PBMC-DCs with AMP reduced the level of CD11c⁺CD86⁺ mature DCs, which was largely abrogated by Hu001 or Hu001–MMAE (Fig. 7F). Hu001–MMAE induced significantly more mature DCs compared with that of Hu001, reaching a level even higher than the no-AMP control treatment (Fig. 7F). In the DC-differentiation culture without exogenous AMP, Hu001–MMAE also increased the level of mature DCs, which was comparable to the level obtained with LPS (Fig. 7G). A duplicate set of identically treated DCs (PBMC donor-1) were then washed and cocultured with CFSE-labeled T cells (PBMC donor-2). T-cell proliferation was significantly enhanced in the coculture of Hu001–MMAE- and LPS-activated DCs (52.25% and 51.25%, respectively) compared with that of hIgG1–MMAE or buffer control (41.85% and 41.90%, respectively; Fig. 7H). These results collectively demonstrate that Hu001–MMAE remains fully functional in inhibiting CD73 enzyme activity, protecting T-cell growth, and enhancing DC activation.

Cynomolgus monkeys were treated triweekly with Hu001–MMAE at 3 mg/kg on day −1 (cycle 1) and 6 mg/kg on day −22 (cycle 2) and observed for 42 days. There were no changes in food consumption or body weights. Results for blood chemistry, hematology, and immunology panel testing showed that Hu001–MMAE was well tolerated in nonhuman primates (Supplementary Fig. S11; Supplementary Table S2).

In both analyses of lung cancer TMA and transcriptome profile, we found an elevated CD73 expression in major subtypes of NSCLC, including those harboring mutated RAS or RTK oncogenes. CD73 is enriched in tumors with transcriptome signatures of EMT and immunosuppressive microenvironment, which are increasingly linked to cancer progression and therapy resistance. We next demonstrated antitumor efficacy by two series of novel CD73-targeted antibody in NSCLC tumor models with CD73-high expression. As expected, Hu001/Hu002 protected T-cell growth and effector function from AMP/adenosine- or tumor cell CD73-imposed toxicity. Hu001/Hu002 inhibited tumor growth in athymic mice via both an ADCC-dependent and independent actions as the ADCC-deficient Hu001/TM partially reduced antitumor efficacy. Because these antibodies do not target host CD73, we conclude that the retaining efficacy in Hu001/TM-treated tumors most likely reflects the inhibition of CD73/adenosine axis function in the tumor compartment, leading to tumor growth retardation accompanied by a multifaceted remodeling of the tumor microenvironment.

Our study highlighted a functional role for tumor-CD73 in promoting tumorigenic TAMs and MDSCs in the tumor environment. NSCLC tumors infiltrate predominately with M2-TAMs that negatively affect antitumor immunity and correlate with reduced overall survival in NSCLC patients (45). Using transcriptome analysis, we observed a positive association of CD73 and M2-TAM biomarker CD206/CD163. We further demonstrated that the CD73/adenosine blockade can substantially inhibit CD206⁺F4/80⁺ M2-like TAMs in two RAS-mutant H1299 and H441 NSCLC tumors in mice. These results strongly suggest that the CD73-targeted antitumor efficacy is mediated in part through the suppression of M2-like TAM function.

MDSCs are another negative regulator of cancer immunity and contribute to primary or acquired resistance to immunotherapy (7). We observed a positive association of CD73 and MDSCs biomarker CD11b/CD33 in NSCLC transcriptome. Recently, CD39⁺CD73⁺ MDSCs from NSCLC patient tumors are shown to suppress T-cell and NK cell and correlate with increased metastasis and resistance to chemotherapy (46). In our study, treatment with CD73 antibody significantly decreased tumor-infiltrating CD11b⁺Gr1⁺ MDSCs in two RAS-mutant H1299 and H441 NSCLC tumor models, further strengthening the functional role for CD73 in NSCLC immune escape. Together, our results uncovered a transformation from an “immune-tolerant” to “cancer-immune” microenvironment of NSCLC tumors in response to CD73-targeted antibody treatment. Although these results remain preliminary, they support the view that CD73-targeted therapy may provide a new strategy relevant for improving antitumor immunity.

ADC represents a powerful approach to eliminate tumor cells by specific transferring chemotherapeutic agents into target tumor cells while sparing healthy tissues (47). A major aspect of our study focused on the therapeutic potential of CD73-ADC. We noted a prominent cancer-selective differential expression window for CD73 compared with that of normal tissues. CD73 antibody could internalize rapidly and extensively to CD73-high human cancer cells, which supports CD73 as a feasible target for ADC development. Furthermore, our CD73 IHC and quantitative imaging studies confirmed a large percentage of clinical NSCLC tumors expressing an abnormally high level of surface CD73 that can be therapeutically targeted by CD73-ADC. Accordingly, Hu001–MMAE displayed an exquisitely potent toxicity in numerous CD73-high tumor cells while had no significant toxicity against normal proliferating human T cells or CD73-low tumor cells. Hu001–MMAE demonstrated promising efficacy in all four tested in vivo tumor models, including three NSCLC tumors with MED of </= 1 mg/kg. We conclude from these results that Hu001–MMAE represents a new class of highly potent and selective anticancer monotherapy.

CD73-ADC therapy can offer dual benefits in killing CD73-high tumors and improving cancer immunity response. Our results demonstrate that Hu001–MMAE remained fully functional in inhibiting CD73 enzyme activity and protecting effector T-cell function with the potency comparable to naked antibody. Hu001–MMAE induced tumor regression and accumulation of inflammatory macrophages and mature DCs in the tumor, which were not observed with hIgG1–MMAE or docetaxel, arguing for a CD73-ADC–specific remodeling of the tumor immuno-environment. These observations were further extended in a human PBMC-DC in vitro differentiation model where treatment with Hu001–MMAE led to an increase in the CD11c⁺CD86⁺ mature DCs that enhanced T-cell proliferation. In the previous studies with mouse bone marrow–derived DC (BMDC), CD73 expression is upregulated by TGFβ, which correlates with a tolerance-permissive DC phenotype (48). Silva-Vilches and colleagues recently reported that expression of CD73 in DCs triggers elevated levels of extracellular adenosine, which is a crucial mechanism for the induction of anergic T cells and tolerance (49). Our findings that Hu001–MMAE is a stronger inducer of DC maturation than Hu001 indicate an added benefit from MMAE portion of the ADC molecule. This observation is consistent with a previous report on microtubule-depolymerizing dolastatins exerting DC-maturation property (50).

In summary, our results have identified CD73-targeted antibodies and CD73-ADC as relatively safe and effective anticancer agents. Given the promising antitumor efficacy and multifaceted remodeling of the cancer immune environment, these results provide a strong rationale for CD73 as a therapeutic target for treating subsets of CD73-dysregulated lung cancer patients.

R. Jin, L. Liu, and K. Yu report a patent for WO2019/170131 A1 pending. No potential conflicts of interest were disclosed by the other authors.

R. Jin: Data curation, formal analysis, supervision, methodology, writing–original draft, writing–review, and editing. L. Liu: Data curation, formal analysis, supervision, methodology, and writing–original draft. Y. Xing: Data curation, formal analysis, methodology, and writing–original draft. T. Meng: Resources, data curation, and formal analysis. L. Ma: Data curation and formal analysis. J. Pei: Data curation, formal analysis, and methodology. Y. Cong: Data curation and methodology. X. Zhang: Methodology. Z. Ren: Data curation. X. Wang: Methodology. J. Shen: Resources and project administration. K. Yu: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review, and editing.

This work was funded by Fudan University (EZF301002), National Science Foundation of China (81373442), National Science & Technology Major Project of China (2018ZX09711002-008), and National Basic Research 973 Program of China (2013CB932500). The authors thank Fudan University School of Pharmacy Animal Facility, Instrument Center, and the CDSER/SIMM facility for study support.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.