In TNBC, cath-D is a tumor microenvironment antigen eligible for Fc-engineered antibody targeted therapy to trigger ADCC

ADCC initiation requires the presence of antibody-antigen complexes at the cell surface. Cath-D localizes at the cell surface through its binding to the M6P/IGF2 receptor via its mannose-6-phosphate (M6P) motif (BC cells and stromal fibroblasts)28 29 and to LRP1 (fibroblasts).23 We first examined the expression of cath-D, M6P/IGF2 receptor and LRP1 in a panel of TNBC cell lines (MDA-MB-231, SUM-159, MDA-MB-436, MDA-MB-453, BT-549) and hCAF1 cells (figure 1A). We detected the fully mature cellular lysosomal protease cath-D (34 kDa heavy chain) and the IGF2/M6P receptor in all TNBC cell lines and hCAF1s. We could not detect cath-D only in BT549 cells. Similarly, we did not detect LRP1 expression in TNBC cells, except for SUM159 and BT549 cells. Conversely, it was strongly expressed in hCAF1 cells. The 52-kDa cath-D precursor (pro-cath-D) was secreted in the culture medium of all TNBC cell lines and of hCAFs and was detectable in cell lysates, in agreement with its cell surface association after secretion (figure 1A). Moreover, we detected the 52-kDa cath-D precursor and IGF2/M6P receptor in the cytosol (ie, whole tumor lysate) of eight primary TNBC samples and in MDA-MB-231 and SUM-159 cell xenografts (figure 1B). Immunofluorescence analysis of cath-D in MDA-MB-231 cell xenografts showed intense, punctate staining at the tumor periphery and in cell membrane protrusions, potentially in areas of exocytosis (figure 1C), reflecting cath-D expression profile in TNBC biopsy samples (online supplemental figure 1). This is in agreement with recent IHC results showing membrane-associated cath-D expression in 85.7% of TNBC samples in a tissue microarray.27 These observations strongly suggest that cath-D is a tumor microenvironment antigen eligible for Fc-engineered antibody targeted therapy to trigger ADCC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Cath-D is a tumor microenvironment antigen eligible to trigger ADCC and relevant for Fc-engineered antibody-based targeted therapy in TNBC. (A) Cath-D expression and secretion, and M6P/IGF2 and LRP1 receptor expression in TNBC cell lines and breast hCAFs. Whole cell extracts (cell lysates, 20 µg proteins) and 24 hours conditioned media in the absence of serum (40 µL) were separated on 13.5% SDS-PAGE and analyzed by immunoblotting with the anti-cath-D mouse monoclonal antibody (#610801) (to detect mature cath-D), and anti-cath-D rabbit polyclonal antibody (E-7) (to detect pro-cath-D). Whole cell extracts were also immunoblotted with the 11H4 hybridoma against the LRP1 receptor, and with a mouse monoclonal antibody (clone MEM-238) against the M6P/IGF2 receptor (M6P/IGF2R). β-actin (#A2066) was used as loading control. Mr, relative molecular mass (kDa). (B) Cath-D expression in TNBC cytosols. Cath-D expression was determined in eight cytosols from primary TNBC samples. Whole cytosols (5 µg proteins) were immunoblotted with the mouse monoclonal (#610801) (to detect mature cath-D), and rabbit polyclonal (E-7) (to detect pro-cath-D) anti-cath-D antibodies. HSC70 (clone B-6) was used as loading control. MDA-MB-231 and SUM-159 TNBC cell xenografts were used as positive controls for cath-D expression. Mr, relative molecular mass (kDa). (C) Cath-D expression and localization in MDA-MB-231 cell xenografts. Representative image of MDA-MB-231 cell xenograft sections incubated with a monoclonal anti-human cath-D antibody (clone C-5) (green). Nuclei were stained with Hoechst 33 342 (blue). Scale bar, 10 µm (upper panel). Higher magnification of the boxed region showing cell-surface associated cath-D (lower panel). Arrows show the localization of cath-D at the cancer cell surface. (D) Binding of F1M1, F1M1-Fc⁻, and F1M1-Fc⁺ to human pro-cath-D secreted from MDA-MB-231 cells. Sandwich ELISA in which pro-cath-D from conditioned medium of MDA-MB-231 cells was added to wells precoated with the anti-human pro-cath-D M2E8 monoclonal antibody in the presence of increasing concentrations of F1M1, F1M1-Fc⁻, or F1M1-Fc⁺. Binding of F1M1, F1M1-Fc⁻ and F1M1-Fc⁺ to pro-cath-D was revealed with an anti-human Fc HRP-conjugated antibody. The EC₅₀ values are shown. (E) Binding of F1M1-Fc⁺ to human pro-cath-D secreted from MDA-MB-231 cells at acidic pH. Sandwich ELISA was performed as in (D), but at different pH values (7.2, 6.8, 6.2 and 5.6). (F) Binding of F1M1, F1M1-Fc⁻, and F1M1-Fc⁺ to human CD16a 158V and CD16a 158F by surface plasmon resonance. Increasing concentrations of F1M1, F1M1-Fc⁻, and F1M1-Fc⁺ (3.7, 11, 33, 100, 300 nM) were injected into the sensor chip on which hCD16a 158V (left panel) or hCD16a 158F (right panel) had been captured. The K_D values are shown. (G) Binding of F1M1, F1M1-Fc⁻, and F1M1-Fc⁺ to human CD16a 158V and CD16a 158F by flow cytometry. NK92 cells transduced to express hCD16a 158V (left panel) or hCD16a 158F (right panel) were incubated with increasing concentrations (0.6 to 6666 nM) of F1M1-Fc⁻, F1M1 or F1M1-Fc⁺ at 4°C for 30 min, followed by incubation with the APC-AF750 conjugated anti-CD16 3G8 antibody (positive control of human CD16 engagement) at 4°C for 30 min. Results were expressed as percentage of fluorescent 3G8 binding: (MFI without anti-cath-D antibody−MFI with anti-cath-D antibody) ×100/(MFI without anti-cath-D antibody). This allowed the standardization and comparison of results between NK92 hCD16a 158F and NK92 hCD16a 158V. ADCC, antibody-dependent cellular cytotoxicity; hCAFs, human breast cancer-associated fibroblasts; MFI, mean fluorescence intensity; TNBC, triple-negative breast cancer,

Generation of anti-cath-D F1M1 Fc-variant antibodies with different CD16a binding properties

We next evaluated whether manipulation of the Fc portion of F1, our lead anti-cath-D antibody, by protein-engineering to increase ADCC could potentiate its antitumor effectiveness in TNBC. We first improved F1 format by introducing two-point mutations (one in VL CDR3 and one in VH FR3) to abrogate N-glycosylation at these two sites (F1M1 antibody) that may lead to the production of glycoforms associated with the risk of immunogenic responses in patients, as described for cetuximab.30 The migration profiles of F1 and F1M1 in the IgG1 format showed that the light and heavy chains molecular weights were lower in F1M1 than F1, as expected due to reduced glycosylation (online supplemental figure 2A). Sandwich ELISA using human cath-D secreted from MDA-MB-231 cells showed that F1M1 (online supplemental figure 2B) had good binding capacities (0.04 nM), as previously shown for F1 (0.2 nM).27 Sandwich ELISA using cath-D secreted from mouse E0771 TNBC cells showed that F1M1 and F1 (online supplemental figure 2C) retained good binding capacities also toward mouse cath-D (EC₅₀=0.9 nM and 26.3 nM, respectively). The lower EC₅₀ of F1M1 indicated that the two-point mutations leading to Fab aglycosylation enhanced F1M1 capacity to recognize human and mouse cath-D, presumably because the VL-CDR3 glycosylation decreased F1 affinity due to steric hindrance. To compare the antitumor efficacy of F1M1 and F1, we xenografted athymic Foxn1^nu nude mice with MDA-MB-231 cells. When tumor volume reached 50 mm³, we treated mice with F1M1, F1, or negative isotype control (rituximab) (15 mg/kg) by ip three times per week for 35 days (days 21–56 postgraft), and then sacrificed all mice at treatment end. Both F1M1 and F1 significantly delayed tumor growth compared with negative control (online supplemental figure 2D; p=0.003 for F1M1, p<0.001 for F1). Thus, the VH and VL aglycosylated F1M1 antibody appears to be a suitable candidate for future clinical development.

We then designed an Fc-optimized F1M1 (F1M1-Fc⁺), in which the S239D, H268F, S324T, I332E mutations should increase the binding affinity to CD16a (FcγRIIIA) to enhance NK cell activation, and a F1M1 Fc-silent (F1M1-Fc⁻) in which the L234A, L235A and P329G mutations prevented binding to FcγRs8 and ADCC induction. F1M1-Fc⁻ activities should rely exclusively on Fab-mediated effector functions. Antibody preparations were highly pure, and after SDS-PAGE in reducing conditions, the heavy and light chains were of the expected molecular mass (50 and 25 kDa respectively) (online supplemental figure 3A). The three F1M1 antibody variants showed almost identical, concentration-dependent binding to secreted cath-D with EC₅₀ values in the low nanomolar range (EC₅₀=0.07 nM, 0.051 nM, and 0.048 nM for F1M1, F1M1-Fc⁻ and F1M1-Fc⁺, respectively) (figure 1D). Moreover, F1M1-Fc⁺ binding to pro-cath-D was comparable at pH values from 7.2 to 5.6 (figure 1E). All F1M1 variant antibodies immunoprecipitated secreted cath-D at pH values from 7.5 to 5.5 (online supplemental filgure 3B), suggesting that the highly acidic tumor microenvironment does not inhibit cath-D recognition and binding by antibodies. Then, we investigated by surface plasmon resonance F1M1-Fc⁺, F1M1 and F1M1-Fc⁻ binding to hCD16a (158V or 158F allotype). F1M1 binding to hCD16a 158V was 3.9-fold higher than to hCD16a 158F (K_D=42 nM and K_D 164 nM, respectively) (figure 1F). Compared with F1M1, F1M1-Fc⁺ binding to hCD16a 158V was increased by 21.2-fold (K_D=2 nM for F1M1-Fc⁺ and 42.4 nM for F1M1), and even more to hCD16a 158F (by 54.6-fold) (K_D=3 nM for F1M1-Fc⁺, and 164 nM F1M1). As F1M1-Fc⁻ showed no detectable binding to both hCD16a allotypes, we used it as control devoid of effector function. Finally, we investigated by flow cytometry F1M1-Fc⁺, F1M1 and F1M1-Fc⁻ binding to hCD16a (158V or 158F allotype) expressed on the surface of human NK92 cells. It has been shown that the hCD16a 158 V/F polymorphism in NK cells influences its binding to human IgG1.31 Specifically, hCD16a V158 displays increased affinity for human IgG1, resulting in increased NK cell activation. F1M1 binding to hCD16a 158V was approximately fivefold higher than to hCD16a 158F at all tested concentrations (figure 1G). Compared with F1M1, F1M1-Fc⁺ binding to hCD16a 158V was increased (2–3 fold), and even more to hCD16a 158F (7–20 fold).

F1M1-Fc⁺ increases NK cell activation in vitro, and ADCC of TNBC cells and CAFs

We next compared F1M1-Fc⁺, F1M1 and F1M1-Fc⁻ in vitro capacity to activate NK cells by analyzing cell surface CD107a expression, as a functional marker of NK cell degranulation, and intracellular IFNγ, a cytokine secreted following NK cell activation,32 in hCD16a-transduced human NK92 cell line. The percentages of CD107a⁺ NK, IFNγ⁺ NK and CD107a⁺ IFNγ⁺ NK subsets were increased similarly by 2–3 fold following incubation with F1M1-Fc⁺, compared with F1M1, in hCD16a 158V-expressing or 158F-expressing NK cells (figure 2A). Conversely, F1M1-Fc⁻ did not activate NK effector functions. Moreover, the percentages of CD107a⁺ NK, IFNγ⁺ NK and CD107a⁺ IFNγ⁺ NK cells were increased in a dose-dependent manner in the presence of increasing concentrations of cath-D bound to F1M1-Fc⁺ (figure 2B). Overall, F1M1-Fc⁺ was the most potent antibody to activate NK cell response. Next, we determined whether the F1M1-Fc⁺-enhanced binding to hCD16a-expressing NK cells translated into increased cytotoxic activity. First, incubation of MDA-MB-231 cells with F1M1-Fc⁺ significantly increased ADCC by 5-fold (p<0.0001) and 18-fold (p=0.0002), compared with F1M1, in the presence of NK92-hCD16a 158 V cells and NK92-hCD16a 158 F cells, respectively (figure 2C). F1M1-Fc⁻ did not induce ADCC. Moreover, F1M1-Fc⁺ triggered ADCC in a dose-dependent manner in the presence of NK92-hCD16a 158V cells (figure 2D). F1M1-Fc⁺-induced ADCC was exclusively dependent on Fc binding to the hCD16 receptor on NK92-hCD16a 158V cells, because addition of a Fc block completely inhibited ADCC, to a level similar to that observed with F1M1-Fc⁻ (figure 2E). We then evaluated ADCC against MDA-MB-231 cells using primary human NK cells (158V/F allotype) (online supplemental figure 4A). ADCC was significantly increased by 3.5-fold on incubation with F1M1-Fc⁺, compared with F1M1 (p<0.0001) (online supplemental figure 4B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Fc-engineered F1M1-Fc⁺ increases NK cell activation in vitro, and ADCC in TNBC cells and CAFs. (A) Functional responses of NK92 hCD16a 158V and NK92 hCD16a 158 F cells induced by F1M1-Fc⁻, F1M1, and F1M1-Fc⁺. For in vitro stimulation and functional responses of hCD16a-transduced NK92 cells, 96-well plates were sensitized with a saturating concentration (5 µg/mL; 33.3 nM) of F1M1-Fc⁻, F1M1, F1M1-Fc⁺, or anti-hCD16a (3G8) monoclonal antibodies at 4°C for 12 hours. NK92 hCD16a 158V and NK92 hCD16a 158 F cells were incubated on non-sensitized or sensitized plates with/without the anti-CD107a-PC5 antibody (1:20) and 0.1 µg/mL BD GolgiPlug containing brefeldin A at 37°C for 4 hours. Cells were then fixed and permeabilized and stained with an anti-IFNγ-PE antibody (1:20) at 4°C for 30 min to detect intracellular IFNγ. The percentages of degranulating NK92 cells (CD107a⁺IFNγ⁻; white bars), IFNγ-producing NK92 cells (CD107a⁻IFNγ⁺; black bars), and NK92 cells exhibiting both responses (CD107a⁺IFNγ⁺; gray bars) were quantified by flow cytometry after 4 hours of stimulation. Mean (%) ±SD (n=3). 3G8 (anti-hCD16a antibody) was used as positive control of CD16 engagement (Ctrl). (B) Functional responses of NK92-hCD16a 158 V cells induced by F1M1-Fc⁺ in the presence of increasing concentrations of cath-D. 96-well plates were coated with the M2E8 anti-cath-D antibody (2500 ng/well) at 4°C overnight, incubated or not with increasing concentrations of cath-D (3-50 nM) and then sensitized with F1M1-Fc⁺ (5 µg/mL). The percentage of degranulating (CD107a⁺IFNγ-; white bars), IFNγ-producing (CD107a⁻IFNγ⁺; black bars), and NK92-hCD16a 158 V cells exhibiting both responses (CD107a⁺IFNγ⁺; gray bars) was evaluated as in (A) after 4 hours of stimulation. (C) ADCC activity against MDA-MB-231 cells in the presence of NK92-hCD16a cells in response to F1M1-Fc⁻, F1M1 or F1M1-Fc⁺. MDA-MB-231 cells (target) were preincubated with F1M1-Fc⁻, F1M1, F1M1-Fc⁺, or with cetuximab at 100 µg/mL (666 nM) at 37°C for 30 min, followed by incubation with NK92-hCD16a 158V (left) or NK92-hCD16a 158F (right) cells (effector) at an effector:target ratio of 10:1 for 24 hours. MDA-MB-231 cell lysis was evaluated by measuring LDH release. Cetuximab (anti-EGFR antibody) was used as positive control (Ctrl). For hCD16a 158 V cells (left): ****p<0.0001 for F1M1-Fc⁺ vs F1M1, ****p<0.0001 for F1M1-Fc⁺ vs F1M1-Fc⁻, ***p=0.0003 for F1M1 vs F1M1-Fc⁻ (one-way ANOVA). For hCD16a 158 F cells (right): ***p=0.0002 for F1M1-Fc⁺ vs F1M1, ****p<0.0001 for F1M1-Fc⁺ vs F1M1-Fc⁻, *p=0.015 for F1M1 vs F1M1-Fc⁻ (one-way ANOVA). (D) Dose-dependent induction of ADCC against MDA-MB-231 cells in the presence of hCD16a 158V-expressing NK92 cells in response to F1M1-Fc⁺. ADCC was evaluated in MDA-MB-231 cells incubated with hCD16a 158V-expressing NK92 cells at an effector:target ratio of 10:1, as described in (C), after incubation with increasing concentrations of F1M1-Fc⁺, from 0.1 to 666 nM (0.015–100 µg/mL). (E) Effect of blocking Fc binding sites in NK92-hCD16a 158V cells on F1M1-Fc⁺-induced ADCC in MDA-MB-231 cells. NK92-hCD16a 158 V cells were preincubated or not with Fc block (to saturate the Fc binding sites in NK92-hCD16a 158 V cells) for 30 min. MDA-MB-231 cells, preincubated with F1M1-Fc⁺ at 100 µg/mL (666 nM) for 30 min, were incubated with NK92-hCD16a 158 V cells preincubated or not with Fc block. ADCC was analyzed as described in (C). Cetuximab (100 µg/mL), positive control (Ctrl); F1M1-Fc⁻, negative control (100 µg/mL). **p=0.0023 for F1M1-Fc⁺ vs F1M1-Fc⁺+Fc block, **p=0.0012 for F1M1-Fc⁺ vs F1M1-Fc⁻ (one-way ANOVA). (F) Blocking cath-D binding to M6P receptors in MDA-MB-231 cells affects F1M1-Fc⁺-induced ADCC in the presence of NK92-hCD16a 158 V cells. MDA-MB-231 cells (target) were preincubated with M6P (10 mM) or G6P (10 mM) for 24 hours. Then, ADCC was assessed in these cells with NK92-hCD16a 158 V cells (effector) at an effector:target ratio of 10:1, as described in C, after incubation with F1M1-Fc⁺ at 100 µg/mL (666 nM) in the presence of M6P (10 mM) or G6P (10 mM). M6P, mannose-6-phosphate; G6P, glucose-6-phosphate (negative control for M6P). Cetuximab (100 µg/mL), positive control (Ctrl). ****p<0.0001 for F1M1-Fc⁺ vs F1M1-Fc⁺+M6P, ***p=0.0002 for F1M1-Fc⁺+G6P vs F1M1-Fc⁺+M6P (one-way ANOVA). (G) Blocking cath-D binding to M6P receptors in hCAFs affects F1M1-Fc⁺-induced ADCC in the presence of NK92-hCD16a 158 V cells. Breast hCAF1 cells (target) were pre-incubated with M6P (10 mM) or G6P (10 mM) for 24 hours. ADCC was assessed in these hCAF1 cells in the presence of NK92-hCD16a 158 V cells (effector) at an effector:target ratio of 10:1, as described in F, after incubation with F1M1-Fc⁺ at 100 µg/mL (666 nM) and M6P (10 mM) or G6P (10 mM). Cetuximab (100 µg/mL), positive control (Ctrl); F1M1-Fc⁻, negative control (100 µg/mL). ****p<0.0001 for F1M1-Fc⁺ vs F1M1-Fc⁺+M6P, ****p<0.0001 for F1M1-Fc⁺+G6P vs F1M1-Fc⁺+M6P (one-way ANOVA). (H) ADCC against MDA-MB-231 cell spheroids in the presence of NK92-hCD16a 158 V cells in response to F1M1-Fc⁺. MDA-MB-231 cell spheroids (target) were first incubated with F1M1-Fc⁺, F1M1-Fc⁻, or cetuximab (100 µg/mL) positive control (Ctrl) at 37°C for 30 min. Then, spheroids were incubated with NK92-hCD16a 158 V cells (effector) at an effector:target ratio of 20:1 for 24 hours, and ADCC was assessed as described in (C). ****p<0.0001 for F1M1-Fc⁺ vs F1M1-Fc⁻ (one-way ANOVA). ADCC, antibody-dependent cellular cytotoxicity; ANOVA, analysis of variance; CAF, cancer-associated fibroblasts; LDH, lactate dehydrogenase; TNBC, triple-negative breast cancer.

As cath-D localizes at the surface of BC cells and stromal fibroblasts by binding to the M6P/IGF2 receptor,28 we assessed whether M6P/IGF2 receptor-bound cath-D was involved in ADCC induction by F1M1-Fc⁺. Thus, we evaluated ADCC against MDA-MB-231 cells in the presence of NK92-hCD16a 158 V cells and of excess M6P to compete with cath-D binding to the M6P/IGF2 receptor (figure 2F). Addition of M6P, but not glucose-6-phosphate (negative control for M6P), significantly inhibited ADCC induction by F1M1-Fc⁺ by 51% (p<0.0001), indicating that cath-D bound to its M6P/IGF2 receptor at the cell surface was involved in ADCC. Similarly, using primary human NK cells (158 V/V allotype) (online supplemental figure 4C), ADCC of MDA-MB-231 cells was inhibited by 70% (p<0.0001) in the presence of M6P (online supplemental figure 4D). Moreover, F1M1-Fc⁺ efficiently triggered ADCC against hCAF1 cells in the presence of NK92-hCD16a 158 V cells, and excess M6P significantly inhibited ADCC by 62% (p<0.0001) (figure 2G). Lastly, to better recapitulate the in vivo cancer cell environment, we prepared spheroids of MDA-MB-231 cells co-cultured with NK92 hCD16a 158V cells. F1M1-Fc⁺ also induced ADCC of MDA-MB-231 cell spheroids (figure 2H). Live imaging of ADCC in calcein-labeled MDA-MB-231 cell spheroids showed their time-dependent lysis in response to F1M1-Fc⁺ exposure (online supplemental figure 5A,B). At 24 hours, the fluorescence signal of calcein-labeled MDA-MB-231 spheroids was significantly decreased by 29.5% in response to F1M1-Fc⁺ compared with F1M1-Fc⁻ (p=0.005) (online supplemental figure 5C). Overall, our results demonstrated that F1M1-Fc⁺ is the most potent anti-cath-D antibody to trigger ADCC in TNBC cells and CAFs.

F1M1-Fc⁺ is the best candidate to reduce MDA-MB-231 cell xenograft growth and improve mouse survival

To compare the in vivo antitumor efficacy of F1M1-Fc⁺, F1M1, and F1M1-Fc⁻, we used athymic Foxn1^nu nude mice subcutaneously xenografted with MDA-MB-231 cells. When MDA-MB-231 tumors reached 50 mm³, we treated mice with F1M1-Fc⁺, F1M1, F1M1-Fc⁻, or the anti-human CD20 IgG1 rituximab, as negative isotype control (Ctrl) (15 mg/kg) by ip three times per week for 35 days (days 13–48 postgraft), and sacrificed mice when tumor volume reached 2000 mm³. Both F1M1-Fc⁺ and F1M1 significantly delayed tumor growth compared with negative control (p=0.001 for F1M1-Fc⁺, p=0.011 for F1M1). F1M1-Fc⁻ slightly delayed tumor growth (p=0.057) (figure 3A). F1M1-Fc⁺ tended to be more effective than F1M1-Fc⁻ (p=0.055). At day 48 (treatment end), tumor volume was significantly reduced by 77% (p=0.0002) in the F1M1-Fc⁺ group, by 60% (p=0.0007) in the F1M1 group, and by 40.5% (p=0.0289) in the F1M1-Fc⁻ group compared with control (figure 3B). Tumor volume was significantly smaller in the F1M1-Fc⁺ group than in the F1M1-Fc⁻ group (p=0.0041). The overall survival rate, reflected by a tumor volume <2000 mm³, was significantly longer in the F1M1-Fc⁺, F1M1, and F1M1-Fc⁻ groups than in controls. The median survival was 59, 57, 54.5, and 50 days in the F1M1-Fc⁺, F1M1, F1M1-Fc⁻, and control groups, respectively (figure 3C; Kaplan-Meier survival analysis, p=0.0003 for F1M1-Fc⁺, p=0.0011 for F1M1, p=0.0275 for F1M1-Fc⁻ compared with control). The overall survival rate was significantly longer in mice treated with F1M1-Fc⁺ than with F1M1-Fc⁻ (p=0.0253). Moreover, all treated mice gained weight (figure 3D) and displayed normal activities, suggesting no apparent toxicity. Overall, the results suggest that F1M1-Fc⁺ is the anti-cath-D antibody with the best antitumor activity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

F1M1-Fc⁺ is the best candidate to reduce growth of MDA-MB-231 cell xenografts and improve mouse survival. (A) Tumor growth. MDA-MB-231 cells were subcutaneously injected in nude mice. When tumor volume reached 50 mm³, mice were treated with F1M1 (n=10), F1M1-Fc⁻ (n=8), F1M1-Fc⁺ (n=9) (15 mg/kg), or rituximab (Ctrl; n=7) (15 mg/kg), three times per week for 35 days. Mice were sacrificed when tumor volume reached 2000 mm³. Tumor volume (in mm³) is shown as the mean±SEM. p=0.057 for F1M1-Fc⁻ vs Ctrl, **p=0.011 for F1M1 vs Ctrl, ***p=0.001 for F1M1-Fc⁺ vs Ctrl, p=0.055 for F1M1-Fc⁺ vs F1M1-Fc⁻, p=0.416 for F1M1-Fc⁺ vs F1M1, p=0.066 for F1M1 vs F1M1-Fc⁻ (mixed-effects multiple linear regression test). (B) Mean tumor volume at day 48. n=7 for Ctrl (rituximab); n=10 for F1M1; n=8 for F1M1-Fc⁻; n=9 for F1M1-Fc⁺. ***p=0.0003 for all groups (Kruskal-Wallis). *p=0.0289 for F1M1-Fc⁻ vs Ctrl, ***p=0.0007 for F1M1 vs Ctrl, ***p=0.0002 for F1M1-Fc⁺ vs Ctrl, **p=0.0041 for F1M1-Fc⁺ vs F1M1-Fc⁻, p=0.0947 for F1M1-Fc⁺ vs F1M1, p=0.2114 for F1M1 vs F1M1-Fc⁻ (Mann-Whitney t-test). Data are the mean±SEM. (C) Kaplan-Meier survival analysis. n=7 for Ctrl (rituximab); n=10 for F1M1; n=8 for F1M1-Fc⁻; n=9 for F1M1-Fc⁺. *p=0.0275 for F1M1-Fc⁻ vs Ctrl, **p=0.0011 for F1M1 vs Ctrl, ***p=0.0003 for F1M1-Fc⁺ vs Ctrl, *p=0.0253 for F1M1-Fc⁺ vs F1M1-Fc⁻, p=0.3093 for F1M1 vs F1M1-Fc⁻, p=0.0953 for F1M1-Fc⁺ vs F1M1 (log-rank Mantel-Cox test). (D) Mouse weight monitoring. n=7 for Ctrl (rituximab); n=10 for F1M1; n=8 for F1M1-Fc⁻; n=9 for F1M1-Fc⁺. Data are the mean±SEM.

F1M1-Fc⁺ triggers in vivo recruitment, activation, and cytotoxic activity of NK cells to induce ADCC in MDA-MB-231 TNBC cell xenografts

To investigate the in vivo mechanisms underlying the antitumor effect of F1M1-Fc⁺, F1M1, and F1M1-Fc⁻, we treated nude mice xenografted with MDA-MB-231 cells with F1M1-Fc⁺, F1M1, F1M1-Fc⁻, or rituximab as negative isotype control (same schedule as before), and then sacrificed all mice at treatment end (day 48). F1M1-Fc⁺, F1M1, and F1M1-Fc⁻ inhibited tumor growth compared with rituximab, but only F1M1-Fc⁺ resulted in significant tumor growth inhibition compared with rituximab (p=0.003) (figure 4A). At day 48, tumor volume was significantly reduced by 60.3% (p=0.0005) in the F1M1-Fc⁺ group, by 42.9% (p=0.02) in the F1M1 group, and by 36.6% (p=0.0368) in the F1M1-Fc⁻ group compared with control group (figure 4B,C). Tumor volume was significantly more reduced in the F1M1-Fc⁺ group than in the F1M1-Fc⁻ group (p=0.0368). Human cath-D expression level in tumors was not affected by F1M1-Fc⁺, F1M1, or F1M1-Fc⁻ treatment, as shown by IHC and western blot analyses (online supplemental figure 6A–C). To study the impact of the Fc part of F1M1 on NK cell recruitment and activation, we identified and quantified by flow cytometry tumor-infiltrating NK cells in MDA-MB-231 xenografts after treatment end (day 48). The overall percentage of living cells was not different between untreated (control) and treated groups (online supplemental figure 7). The percentage of NK (CD45⁺ F4/80^- CD3^- CD19^- CD11c^- NKp46⁺) cells within the living immune CD45⁺ cell population was significantly increased by 207.2% in the F1M1-Fc⁺ group compared with control (p=0.0053), but not in the F1M1 and F1M1-Fc⁻ groups (figure 4D). Linear regression analysis showed that the NK cell percentage was inversely correlated with tumor volume in all animals (four treatment groups together) (R²=0.2139, p=0.0062) (figure 4E). This suggested that in the F1M1-Fc⁺ group, NK cell recruitment in the tumor contributed to the antitumor response.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

F1M1-Fc⁺ triggers recruitment and activation of NK cells to induce ADCC in MDA-MB-231 cell xenografts. (A) Tumor growth. MDA-MB-231 cells were subcutaneously injected in nude mice. When tumor volume reached 50 mm³, mice were treated with F1M1 (n=11), F1M1-Fc⁻ (n=10), F1M1-Fc⁺ (n=11), or rituximab (15 mg/kg) (Ctrl; n=9), three times per week for 30 days. At day 48, mice were sacrificed. Tumor volume (in mm³) is shown as the mean±SEM. **p=0.003 for F1M1-Fc⁺ vs Ctrl, p=0.066 for F1M1 vs Ctrl, p=0.353 for F1M1-Fc⁻ vs Ctrl, p=0.061 for F1M1-Fc⁺ vs F1M1-Fc⁻, p=0.413 for F1M1 vs F1M1-Fc⁻, p=0.995 for F1M1-Fc⁺ vs F1M1 (mixed-effects multiple linear regression test). (B) Mean tumor volume at day 48. n=9 for Ctrl (rituximab); n=11 for F1M1; n=10 for F1M1-Fc⁻; n=11 for F1M1-Fc⁺. **p=0.0044 for all groups (Kruskal-Wallis), *p=0.0368 for F1M1-Fc⁻ vs Ctrl, *p=0.020 for F1M1 vs Ctrl, ***p=0.0005 for F1M1-Fc⁺ vs Ctrl, *p=0.0368 for F1M1-Fc⁺ vs F1M1-Fc⁻, p=0.6047 for F1M1 vs F1M1-Fc⁻, p=0.2234 for F1M1 vs F1M1-Fc⁺ (Mann-Whitney t-test). Mean±SEM. (C) Representative images of three tumors per treatment group. (D) NK cell recruitment at day 48. The percentage of NKp46⁺ NK cells was quantified by FACS and expressed relative to all living cells (n=7 for F1M1-Fc⁻; n=7 for F1M1; n=9 for F1M1-Fc⁺; n=8 for Ctrl). **p=0.0075 for all groups (Kruskal-Wallis), **p=0.0053 for F1M1-Fc⁺ vs Ctrl, **p=0.0033 for F1M1-Fc⁺ vs F1M1-Fc⁻, *p=0.0295 for F1M1-Fc⁺ vs F1M1, p=0.4359 for F1M1-Fc⁻ vs F1M1 (Mann-Whitney t-test). (E) Linear regression analysis of NK cells and tumor volumes at day 48. R² =0.2139; **p=0.0062; n=31 mice (all treatment groups). (F) NK cell maturation at day 48. NK cell maturation is divided into four stages [immature, mature (stage 1), mature (stage 2) and terminally differentiated] based on the expression of the CD27 and CD11b cell surface maturation markers, and is associated with NK cell-mediated cytotoxicity (upper panels). In the lower panels, the percentage of NKp46⁺ NK cells at the different maturation stages was quantified by FACS and expressed relative to all NKp46⁺ NK cells: immature CD27⁻CD11b⁻ NKp46⁺ NK cells [(n=7 for F1M1-Fc⁻; n=7 for F1M1; n=9 for F1M1-Fc⁺; n=8 for Ctrl). **p=0.0049 for all groups (Kruskal-Wallis); **p=0.0016 for F1M1-Fc⁺ vs Ctrl; **p=0.0033 for F1M1-Fc⁺ vs F1M1-Fc⁻; *p=0.0404 for F1M1-Fc⁺ vs F1M1; p=0.2086 for F1M1 vs F1M1-Fc⁻ (Mann-Whitney t-test)], mature (stage 1) CD27⁺CD11b⁻ NKp46⁺ NK cells [(n=7 for F1M1-Fc⁻; n=7 for F1M1; n=9 for F1M1-Fc⁺, n=8 for Ctrl). p=0.1248 for all groups (Kruskal-Wallis); *p=0.0360 for F1M1-Fc⁺ vs Ctrl; p=0.1142 for F1M1-Fc⁺ vs F1M1-Fc⁻; p=0.3510 for F1M1-Fc⁺ vs F1M1; p=0.62 for F1M1 vs F1M1-Fc⁻ (Mann-Whitney t-test)], mature (stage 2) CD27⁺CD11b⁺ NKp46⁺ NK cells [(n=7 for F1M1-Fc⁻; n=7 for F1M1; n=9 for F1M1-Fc⁺; n=8 for Ctrl). p=0.5473 for all groups (Kruskal-Wallis), p=0.2086 for F1M1-Fc⁺ vs Ctrl; p=0.4698 for F1M1-Fc⁺ vs F1M1-Fc⁻; p=0.3120 for F1M1-Fc⁺ vs F1M1; p>0.9999 for F1M1 vs F1M1-Fc⁻ (Mann-Whitney t-test)], and terminally differentiated CD27^-CD11b⁺ NKp46⁺ NK cells [(n=7 for F1M1-Fc⁻; n=7 for F1M1; n=9 for F1M1-Fc⁺; n=8 for Ctrl). p=0.8272 for all groups (Kruskal-Wallis); p=0.4234 for F1M1-Fc⁺ vs Ctrl; p=0.4698 for F1M1-Fc⁺ vs F1M1-Fc⁻; p=0.8371 for F1M1-Fc⁺ vs F1M1; p=0.8048 for F1M1 vs F1M1-Fc⁻ (Mann-Whitney t-test)]. (G) Quantification of Eomes mRNA expression at day 48. Total RNA was extracted from MDA-MB-231 tumor cell xenografts at treatment end, and Eomes expression level was analyzed by RT-qPCR. Relative fold change was normalized to Rps9 expression and expressed as fold change relative to control. Data are the mean±SEM (n=8 for F1M1-Fc⁻; n=8 for F1M1; n=8 for F1M1-Fc⁺; n=7 for Ctrl). **p=0.0016 for all groups (Kruskal-Wallis). ***p=0.0006 for F1M1-Fc⁺ vs Ctrl, ***p=0.0006 for F1M1-Fc⁺ vs F1M1-Fc⁻, *p=0.0281 for F1M1 vs F1M1-Fc⁺, p=0.0650 for F1M1 vs F1M1-Fc⁻ (Mann-Whitney t-test). (H) NK cell degranulation at day 48. The percentage of CD107a⁺ degranulating NK cells was quantified by FACS and expressed relative to all NKp46⁺ NK cells (n=7 for F1M1-Fc⁻; n=7 for F1M1; n=9 for F1M1-Fc⁺; n=8 for Ctrl). *p=0.0107 for all groups (Kruskal-Wallis), *p=0.0464 for F1M1-Fc⁺ vs Ctrl, **p=0.0079 for F1M1-Fc⁺ vs F1M1-Fc⁻, **p=0.0021 for F1M1-Fc⁺ vs F1M1, p=0.3176 for F1M1 vs F1M1-Fc⁻ (Mann-Whitney t-test). (I) Granzyme B⁺ NK cells at day 48. The percentage of granzyme-positive (GZMB⁺) activated NK cells was quantified by FACS and expressed relative to all NKp46⁺ NK cells (n=7 for F1M1-Fc⁻; n=7 for F1M1; n=9 for F1M1-Fc⁺; n=8 for Ctrl). *p=0.0119 for all groups (Kruskal-Wallis), **p=0.0079 for F1M1-Fc⁺ vs Ctrl, *p=0.0289 for F1M1 vs Ctrl, *p=0.0418 for F1M1-Fc⁺ vs F1M1-Fc⁻, *p=0.0262 F1M1 vs F1M1-Fc⁻, p>0.9999 for F1M1-Fc⁺ vs F1M1 (Mann-Whitney t-test). ADCC, antibody-dependent cellular cytotoxicity; NK, natural killer.

We also determined the NK cell phenotype based on the expression of the NK cell surface maturation markers CD11b and CD27.33 The immature CD27⁻CD11b⁻ NK cell population was significantly increased by 268% in F1M1-Fc⁺-treated animals (p=0.0016 compared with the control group), but not in the F1M1-treated and F1M1-Fc⁻-treated groups (figure 4F). The CD27⁺CD11b⁻ NK cell population (mature stage 1) also was increased in the F1M1-Fc⁺ group by 206.9% (p=0.0360 compared with the control group) (figure 4F). This suggests that F1M1-Fc⁺ treatment promoted the recruitment of immature NK cells and at an early stage of the maturation process. NK cell development and maturation is orchestrated by a network of transcription factors, including Eomes.34 Eomes is mainly expressed in immature NK cells, promotes survival of maturing NK cells, and plays a major role in the induction of genes associated with NK cell cytotoxicity, such as Prf1.35 RT-qPCR analysis of tumor samples showed that Eomes expression was significantly upregulated by 479.3% in the F1M1-Fc⁺ group (p=0.0006 compared with control), but not in the F1M1 and F1M1-Fc⁻ groups (figure 4G). This strongly suggested that F1M1-Fc⁺-based therapy triggers the recruitment of immature NK cells that might become cytotoxic in the MDA-MB-231 TNBC model.

Next, we analyzed cell surface expression of CD107a, as a functional marker of NK cell degranulation, by flow cytometry in tumor-infiltrating NKp46⁺ cells in MDA-MB-231 cell xenografts at day 48. The percentage of CD107a⁺ NK cells within the NK cell population was significantly increased by 142% in F1M1-Fc⁺-treated animals (p=0.0464 compared with control), but not in the F1M1 and F1M1-Fc⁻ groups (figure 4H). The percentage of CD107a⁺ NK cells was higher (by 142.6%) in the F1M1-Fc⁺ group compared with the F1M1-Fc⁻ group (p=0.0079). Linear regression analysis showed that the percentage of CD107a⁺ NK cells was inversely correlated with tumor volume in all animals (four treatment groups together) (R²=0.1703, p=0.0211) (online supplemental figure 8A). This suggested that in the F1M1-Fc⁺ group, in vivo tumor infiltration by activated CD107a⁺ NK cells may contribute to the antitumor response.

Lastly, we analyzed intracellular granzyme B (GZMB) expression as a marker of NK cell cytotoxic activity by flow cytometry in tumor-infiltrating NKp46⁺ cells in MDA-MB-231 cell xenografts at day 48. The percentage of GZMB⁺ NK cells was significantly increased by 173.8% (p=0.0079) and by 185.4% (p=0.0289) in the F1M1-Fc⁺ and F1M1 groups, respectively, compared with control, but not in the F1M1-Fc⁻ group (figure 4I). It also was higher (by 193%) in the F1M1-Fc⁺ group than in the F1M1-Fc⁻ group (p=0.0418). Linear regression analysis showed that the percentage of GZMB⁺ NK cells was inversely correlated with tumor volume in all animals (four treatment groups together) (R²=0.3085, p=0.0012) (online supplemental figure 8B). This suggested that in the F1M1-Fc⁺ and F1M1 groups, tumor infiltration by NK cells with cytotoxic activity contributed to the antitumor response.

In agreement, granzyme B (Gzmb) and perforin (Prf1) mRNA levels, as a read-out of NK cell cytotoxic activity, were upregulated (up to 166.3% and 173.8%, respectively) in the F1M1-Fc⁺ group compared with control group (p=0.3969 and p=0.2319), and also compared with the F1M1-Fc⁻ group (up to 256.3% and 243.4%; p=0.0207 and p=0.0379, respectively) (online supplemental figure 9A,B). The antitumor cytokine Tnf, secreted by activated NK cells, was also upregulated by 171.5% (p=0.0059) in the F1M1-Fc⁺ group and by 160.3% (p=0.0541) in the F1M1 group compared with control (online supplemental figure 9C). Tnf was upregulated (by 266.8%) (p=0.003) in the F1M1-Fc⁺ group compared with the F1M1-Fc⁻ group. Altogether, these results demonstrated that F1M1-Fc⁺ triggers NK cell recruitment, activation and cytotoxic activity in MDA-MB-231 cell xenografts, and strongly suggest the key role of the Fc part of this anti-cath-D antibody in its antitumor activity via NK cells.

NK cell depletion impairs F1M1-Fc⁺ therapeutic efficacy in MDA-MB-231 cell xenografts

The significant antitumor effects of F1M1-Fc⁺ and the F1M1-Fc⁺-mediated induction of NK cell recruitment, activation and cytotoxic activity strongly suggested that NK cells are crucial for F1M1-Fc⁺-antitumor activity. To confirm this hypothesis, we depleted NK cells in MDA-MB-231-bearing nude mice by intraperitoneal injection of an anti-asialo-GM1 antibody (αGM1). First, to determine the efficiency of NK cell depletion, we treated non-grafted nude mice with control (saline solution) or αGM1 at day 0 and day 3 (online supplemental figure 10A). Analysis of blood samples at day 4 and day 7 confirmed the efficacy of NK cell depletion, even 4 days (ie, day 7) after the last αGM1 injection (online supplemental figure 10B,C). Other immune cells (B cells, neutrophils, dendritic cells and macrophages) were not affected by αGM1 treatment, as indicated by their quantification in blood and spleen (online supplemental figure 10D,E).

Then, to investigate NK cell role in the antitumor effect of F1M1-Fc⁺, we treated nude mice harboring MDA-MB-231 cell xenografts with F1M1-Fc⁺ or rituximab (negative isotype control) in the presence or absence of αGM1, and then sacrificed all mice at treatment end (day 48), according to the schedule described in figure 5A. As previously observed, F1M1-Fc⁺ significantly inhibited tumor growth compared with control (p<0.001) (figure 5B). In the presence of αGM1, F1M1-Fc⁺ antitumor activity (tumor growth reduction) was decreased (p=0.021 vs control, and p=0.015 vs F1M1-Fc⁺ alone), revealing the crucial role of NK cells (figure 5B). At day 48, compared with the control group, tumor volume was significantly reduced by 63.3% (p=0.0003) in the F1M1-Fc⁺ group, but only by 30.7% (p=0.0881) in the F1M1-Fc⁺+αGM1 group (figure 5C). Moreover, tumor volume was significantly smaller in the F1M1-Fc⁺ group than in the F1M1-Fc⁺+αGM1 group (p=0.0281) (figure 5C). The number of tumor-infiltrating NK cells (CD45⁺ CD3⁻ CD19⁻ Ly6G⁻ NKp46⁺) within the living immune CD45⁺ cell population in tumors was significantly increased by 248.7% in the F1M1-Fc⁺ group compared with control (p=0.0312). Conversely, very few NK cells were detected in the F1M1-Fc⁺+αGM1 group (figure 5D). Analysis of blood samples at day 45 and of the spleens at day 48 validated NK cell depletion in the F1M1-Fc⁺+αGM1 group (online supplemental figure 11). This demonstrated the key role of NK cells in F1M1-Fc⁺ therapeutic efficiency.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

NK cell depletion impairs F1M1-Fc⁺ therapeutic efficacy in MDA-MB-231 cell xenografts. (A) Treatment schedule for NK cell depletion. MDA-MB-231 cells were subcutaneously injected in nude mice. At day 9 after MDA-MB-231 injection, half of the mice started the treatment with anti-asialo GM1 antibodies (αGM1, 50 µL by ip injection twice per week) to deplete NK cells. When tumor volume reached 50 mm³ (day 15), mice were treated with F1M1-Fc⁺ (n=8), or control rituximab (Ctrl; n=8) (15 mg/kg), three times per week for 30 days, in the absence or presence of αGM1. At day 48, mice were sacrificed. (B) Tumor growth. Tumor growth (in mm³) is shown as the mean±SEM. (n=8 for F1M1-Fc⁺; n=8 for F1M-Fc⁺+ αGM1; n=8 for Ctrl; n=8 for Ctrl+αGM1). p=0.449 for Ctrl+αGM1 vs Ctrl, ***p<0.001 for F1M1-Fc⁺ vs Ctrl, ***p<0.001 for F1M1-Fc⁺ vs Ctrl+αGM1, *p=0.015 for F1M1-Fc⁺+αGM1 vs F1M1-Fc⁺, *p=0.021 for F1M1-Fc⁺+ αGM1 vs Ctrl + αGM1, p=0.127 for F1M1-Fc⁺+ αGM1 vs Ctrl (mixed-effects multiple linear regression test). (C) Mean tumor volume at day 48. Mean±SEM. (n=8 for F1M1-Fc⁺; n=8 for F1M-Fc⁺+αGM1; n=8 for Ctrl; n=8 for Ctrl + αGM1). **p=0.0018 for all groups (Kruskal-Wallis), ***p=0.0003 for F1M1-Fc⁺ vs Ctrl, p=0.8785 for Ctrl + αGM1 vs Ctrl, **p=0.0019 for F1M1-Fc⁺ vs Ctrl + αGM1, *p=0.0281 for F1M1-Fc⁺+ αGM1 vs F1M1-Fc⁺, p=0.0881 for F1M1-Fc⁺+ αGM1 vs Ctrl, p=0.1605 for F1M1-Fc⁺+ αGM1 vs Ctrl + αGM1 (Mann-Whitney t-test). (D) NK cell recruitment at day 48. The number of NKp46⁺ NK cells in tumors was quantified by FACS and normalized per mg of tumor with precision count beads (n=8 for F1M1-Fc⁺; n=8 for F1M-Fc⁺+ αGM1; n=7 for Ctrl; n=8 for Ctrl + αGM1). ****p<0.0001 for all groups (Kruskal-Wallis), *p=0.0312 for F1M1-Fc⁺ vs Ctrl, ***p=0.0006 for Ctrl + αGM1 vs Ctrl, ***p=0.0002 for F1M1-Fc⁺ vs Ctrl + αGM1, ***p=0.0002 for F1M1-Fc⁺ + αGM1 vs F1M1-Fc⁺, ***p=0.0008 for F1M1-Fc⁺+ αGM1 vs Ctrl, p=0.1621 for F1M1-Fc⁺+ αGM1 vs Ctrl + αGM1 (Mann-Whitney t-test). NK, natural killer.

Drug-induced neutropenia is a potentially serious and life-threatening adverse event that may occur after therapy with various agents, including antibodies (eg, immune checkpoint inhibitors). Analysis by flow cytometry of blood from mice from the control and F1M1-Fc⁺ treated groups (from figure 5) showed that neutrophil count at day 45 was not affected, indicating the absence of neutropenia (online supplemental figure 12A). Similarly, no sign of leukopenia, thrombopenia, and anemia was observed after F1M1-Fc⁺ treatment, suggesting the absence of toxicity (online supplemental figure 12B).

F1M1-Fc⁺ inhibits tumor growth of patient-derived TNBC xenografts and improves mouse survival

Next, we tested F1M1-Fc⁺ therapeutic effect in mice harboring TNBC-PDXs (B1995 and B3977).36 First, we analyzed the expression of cath-D and the M6P/IGF2 receptor in PDX B1995 and B3977 xenografts (the two fastest growing PDXs in nude mice among a PDX collection),36 and in MDA-MB-231 and SUM-159 cell xenografts by western blot analysis. Cath-D and the M6P/IGF2 receptor were similarly expressed in all samples (figure 6A). Immunostaining of the PDX B1995 and B3977 confirmed that cath-D was expressed in tumor cells and in the microenvironment (figure 6B), as previously observed in TNBC biopsy samples (online supplemental figure 1). These results indicated these PDXs are representative of the disease, at least concerning cath-D expression. We then xenografted Swiss nude mice with the PDX B1995 or B3977. Tumor growth was significantly reduced by F1M1-Fc⁺ (15 mg/kg by ip three times per week) compared with control (rituximab) in both models (p<0.001) (figure 6C,D, left panel). At treatment end (day 48 for the PDX B1995 and day 42 for the PDX B3977), tumor volume was significantly reduced by 61.8% (p=0.0002) and by 44.6% (p=0.0018), respectively, in the F1M1-Fc⁺ group compared with the control group (figure 6C,D, middle panels). The overall survival rate, reflected by a tumor volume <1500 mm³, was significantly longer in mice treated with F1M1-Fc⁺ than in control group (median survival of 64 and 49 days for the F1M1-Fc⁺ group and for control animals in mice xenografted with PDX B1995, and of 57 and 46 days for the F1M1-Fc⁺ group and for control animals in mice xenografted with PDX B3977) (figure 6C,D, right panel; Kaplan-Meier survival analysis, p=0.0002 and p=0.0024, respectively). These results showed that F1M1-Fc⁺ monotherapy very efficiently delayed tumor growth in nude mice xenografted with TNBC PDXs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

The lead F1M1-Fc⁺ anti-cath-D antibody inhibits growth of TNBC-PDXs and improves mouse survival. (A) Cath-D expression and secretion, and expression of the M6P/IGF2 in TNBC-PDX tumors. Whole tumor cytosols (5 µg proteins) from PDX B1995 and PDX B3977 tumor xenografts, and from SUM159 and MDA-MB-231 cell xenografts were separated on 13.5% SDS-PAGE and analyzed by immunoblotting with the anti-cath-D mouse monoclonal antibody (#610801) (to detect mature cath-D) and anti-cath-D rabbit polyclonal antibody (E-7) (to detect pro-cath-D). Whole tumor cytosols were also immunoblotted with a mouse monoclonal antibody (clone MEM-238) against the M6P/IGF2 receptor (M6P/IGF2R). HSC70 was used as loading control. Mr, relative molecular mass (kDa). (B) Cath-D expression and localization in TNBC-PDX tumors. PDX B1995 (upper panels) and PDX B3977 (lower panels) tumor sections were incubated with a monoclonal anti-human cath-D antibody (clone C-5) (green). Nuclei were stained with Hoechst 33 342 (blue). Scale bar, 10 µm (left panels). Higher magnification of the boxed region shows cell surface-associated cath-D (right panels). Arrows show the localization of cath-D at the cancer cell surface. (C) Therapeutic effects of F1M1-Fc⁺ in mice xenografted with PDX B1995. Mice were xenografted with PDX B1995 and when tumor volume reached 100 mm³ (day 28), mice were treated with F1M1-Fc⁺ (15 mg/kg) or rituximab (Ctrl) (15 mg/kg) three times per week. Mice were sacrificed when tumor volume reached 1500 mm³, and the corresponding tumor growth curves were stopped (left panel). Tumor volume (in mm³) is shown as the mean±SEM; n=12 for Ctrl (rituximab); n=12 for F1M1-Fc⁺. ***p<0.001 for F1M1-Fc⁺ (mixed-effects multiple linear regression test). Mean tumor volume at day 48 (middle panel). Mean±SEM; n=12 for Ctrl (rituximab); n=12 for F1M1-Fc⁺. ***p=0.0001 for F1M1-Fc⁺ vs Ctrl (Mann-Whitney t-test). Kaplan-Meier survival analysis (right panel). n=12 for Ctrl (rituximab); n=12 for F1M1-Fc⁺. ***p=0.0002 for F1M1-Fc⁺ vs Ctrl (Log-rank Mantel-Cox test). (D) Therapeutic effects of F1M1-Fc⁺ in mice xenografted with PDX B3977. Mice were xenografted with PDX B3977 and when tumor volumes reached 100 mm³ (day 25), mice were treated with F1M1-Fc⁺ (15 mg/kg) or rituximab (Ctrl) (15 mg/kg) three times per week. Mice were sacrificed when tumor volume reached 1500 mm³, and the corresponding tumor growth curves were stopped (left panel). Tumor volume (in mm³) is shown as the mean±SEM; n=11 for Ctrl (rituximab); n=11 for F1M1-Fc⁺. ***p<0.001 for F1M1-Fc⁺ (mixed-effects multiple linear regression test). Mean tumor volume at day 42 (middle panel). Mean±SEM; n=11 for Ctrl (rituximab); n=11 for F1M1-Fc⁺. **p=0.0018 for F1M1-Fc⁺ vs Ctrl (Mann-Whitney t-test). Kaplan-Meier survival analysis (right panel). n=11 for Ctrl (rituximab); n=11 for F1M1-Fc⁺. **p=0.0024 for F1M1-Fc⁺ vs Ctrl (log-rank Mantel-Cox test). PDXs, patient-derived xenografts; TNBC, triple-negative breast cancer.

F1M1-Fc⁺ improves paclitaxel therapeutic efficacy

Next, we investigated the therapeutic benefit of combining F1M1-Fc⁺ with paclitaxel. First, we confirmed paclitaxel cytotoxicity in MDA-MB-231 cells in vitro (IC₅₀=53.4 nM) (online supplemental figure 13A). Then, we showed that incubation of MDA-MB-231 cells with 5 nM paclitaxel for 48 hours (dose equivalent to the IC₂₀; online supplemental figure 13A) did not affect cath-D expression (cell lysates) and secretion (conditioned media) (online supplemental figure 13B). As paclitaxel is frequently administered to patients with a weekly schedule, we treated nude mice bearing MDA-MB-231 cell xenografts with 1 mg/kg, 4 mg/kg, or 7 mg/kg paclitaxel (PTX) weekly. Higher PTX doses resulted in excessive toxicity (data not shown). PTX induced tumor growth inhibition in a dose-dependent manner (online supplemental figure 13C) that was significant at 4 mg/kg (p=0.007) and 7 mg/kg (p<0.001) compared with control (NaCl).

We then tested the combination therapy (paclitaxel and F1M1-Fc⁺) in nude mice bearing MDA-MB-231 cell xenografts using a low dose (PTX LD, 1 mg/kg) and a medium dose of paclitaxel (PTX MD, 4 mg/kg) that inhibited tumor growth on its own; online supplemental figure 13C). When tumor volume reached 50 mm³ at day 15 postgraft, we treated mice with F1M1-Fc⁺ (15 mg/kg; three times per week), PTX LD (1 mg/kg; once per week), F1M1-Fc⁺ (15 mg/kg; three times per week)+PTX LD (1 mg/kg; once per week), or control (rituximab, three times per week+saline, once per week; all ip) for 37 days. Both F1M1-Fc⁺ and PTX LD+F1M1-Fc⁺ significantly delayed tumor growth compared with control (p=0.039 for F1M1-Fc⁺, p=0.032 for PTX LD+F1M1-Fc⁺) (figure 7A). At day 52 (treatment end), tumor volume was reduced by 37.4% (p=0.07) in the F1M1-Fc⁺ group, by 18.7% (p=0.0005) in the PTX LD group, and by 50.5% (p=0.002) in the PTX LD+F1M1-Fc⁺ group compared with control group (figure 7B). Tumor volume was significantly smaller in the PTX LD+F1M1-Fc⁺ group than in the PTX LD group (p=0.0293). The overall survival rate, reflected by a tumor volume <2000 mm³, was significantly longer in mice treated with F1M1-Fc⁺ and with PTX LD+F1M1-Fc⁺ than in controls (median survival of 59 and 66 days for the F1M1-Fc⁺ and PTX LD+F1M1-Fc⁺ groups, respectively, compared with 55 days for control animals) (figure 7C; Kaplan-Meier survival analysis, p=0.0313 for F1M1-Fc⁺, p=0.0009 for PTX LD+F1M1-Fc⁺). The overall survival rate was significantly longer in mice treated with PTX LD+F1M1-Fc⁺ (66 days) than in mice treated with PTX LD (57 days) (p=0.0219). Overall, these results demonstrated that in combination therapy, F1M1-Fc⁺ improved the therapeutic efficacy of a suboptimal dose of paclitaxel that was poorly effective in vivo (online supplemental figure 13C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Therapeutic efficacy of the anti-cath-D F1M1-Fc⁺ antibody in combination with paclitaxel in mice xenografted with MDA-MB-231 cells. (A) Tumor growth in mice treated with F1M1-Fc⁺ and/or paclitaxel (1 mg/kg; PTX LD). MDA-MB-231 cells were subcutaneously injected in nude mice. When tumor volume reached 50 mm³ (day 15 postgraft), mice were treated with F1M1-Fc⁺ (15 mg/kg; three times per week) (n=9), PTX LD (1 mg/kg; once per week) (n=8), F1M1-Fc⁺ (15 mg/kg; three times per week)+PTX LD (1 mg/kg; once per week) (n=9), or rituximab (three times per week)+saline (once per week; all ip) (Ctrl; n=7) for 37 days. Mice were sacrificed when tumor volume reached 2000 mm³. Tumor volume (in mm³) is shown as the mean±SEM. *p=0.039 for F1M1-Fc⁺ vs Ctrl, p=0.689 for PTX LD vs Ctrl, *p=0.032 for PTX LD+F1M1-Fc⁺ vs Ctrl, p=0.092 for PTX LD+F1M1-Fc⁺ vs PTX LD, p=0.88 for PTX LD+F1M1-Fc⁺ vs F1M1-Fc⁺ (mixed-effects multiple linear regression test). PTX LD; low dose of paclitaxel (1 mg/kg). (B) Mean tumor volume at day 52 in mice receiving F1M1-Fc⁺ and/or paclitaxel (1 mg/kg; PTX LD). *p=0.0189 for all groups (Kruskal-Wallis); p=0.0712 for F1M1-Fc⁺ vs Ctrl; p=0.3211 for PTX LD vs Ctrl; **p=0.0020 for PTX LD+F1M1-Fc⁺ vs Ctrl; *p=0.0293 for PTX LD+F1M1-Fc⁺ vs PTX LD; p=0.5302 for PTX LD+F1M1-Fc⁺ vs F1M1-Fc⁺ (Mann-Whitney t-test); Data are the mean±SEM. (C) Kaplan-Meier survival analysis in mice receiving F1M1-Fc⁺ and/or paclitaxel (1 mg/kg; PTX LD). *p=0.0313 for F1M1-Fc⁺ vs Ctrl, p=0.0875 for PTX LD vs Ctrl, ***p=0.0009 for PTX LD+F1M1-Fc⁺ vs Ctrl, *p=0.0216 for PTX LD+F1M1-Fc⁺ vs PTX LD, p=0.3714 for PTX LD+F1M1-Fc⁺ vs F1M1-Fc⁺ (log-rank Mantel-Cox test). (D) Tumor growth in mice receiving F1M1-Fc⁺ and/or paclitaxel (4 mg/kg; PTX MD). MDA-MB-231 cells were subcutaneously injected in nude mice. When tumor volume reached 50 mm³ (day 15 postgraft), mice were treated with F1M1-Fc⁺ (15 mg/kg; three times per week) (n=9), PTX MD (4 mg/kg; once per week) (n=8), F1M1-Fc⁺ (15 mg/kg; three times per week)+PTX MD (4 mg/kg; once per week) (n=7), or rituximab (three times per week)+saline (once per week) (Ctrl; n=7) for 37 days. Mice were sacrificed when tumor volume reached 2000 mm³. Tumor volume (in mm³) is shown as the mean±SEM. *p=0.039 for F1M1-Fc⁺ vs Ctrl, *p=0.042 for PTX MD vs Ctrl, *p=0.013 for PTX MD+F1M1-Fc⁺ vs Ctrl, p=0.107 for PTX MD+F1M1-Fc⁺ vs PTX MD, p=0.857 for PTX MD+F1M1-Fc⁺ vs F1M1-Fc⁺ (mixed-effects multiple linear regression test). PTX MD; medium dose of paclitaxel (4 mg/kg). (E) Mean tumor volume at day 52 in mice receiving F1M1-Fc⁺ and/or paclitaxel (4 mg/kg; PTX MD). *p=0.0200 for all groups (Kruskal-Wallis); p=0.0712 for F1M1-Fc⁺ vs Ctrl; *p=0.0289 for PTX MD vs Ctrl; **p=0.0012 for PTX MD+F1M1-Fc⁺ vs Ctrl; p=0.1520 for PTX MD+F1M1-Fc⁺ vs PTX MD; p=0.4698 for PTX MD+F1M1-Fc⁺ vs F1M1-Fc⁺ (Mann-Whitney t-test). Data are the mean±SEM. (F) Kaplan-Meier survival analysis in mice receiving F1M1-Fc⁺ and/or paclitaxel (4 mg/kg; PTX MD). *p=0.0313 for F1M1-Fc⁺ vs Ctrl, *p=0.0312 for PTX MD vs Ctrl, ***p=0.001 for PTX MD+F1M1-Fc⁺ vs Ctrl, *p=0.0410 for PTX MD+F1M1-Fc⁺ vs PTX MD, p=0.2010 for PTX MD+F1M1-Fc⁺ vs F1M1-Fc⁺ (log-rank Mantel-Cox test). (G) Weight monitoring in mice receiving F1M1-Fc⁺ and/or paclitaxel. Mean±SEM. Ctrl (n=7); F1M1-Fc⁺ (n=9); PTX LD (n=8); F1M1-Fc⁺+PTX LD (n=9); PTX MD (n=8); F1M1-Fc⁺+PTX MD (n=7).

Similarly, F1M1-Fc⁺, PTX MD, and PTX MD+F1M1-Fc⁺ significantly delayed tumor growth compared with control (p=0.039 for F1M1-Fc⁺, p=0.042 for PTX MD, p=0.013 for PTX MD+F1M1-Fc⁺) (figure 7D). At day 52 (treatment end), tumor volume was reduced by 37.4% (p=0.07) in the F1M1-Fc⁺ group, by 32.9% (p=0.0289) in the PTX MD group, and by 55.5% (p=0.0012) in the PTX MD+F1M1-Fc⁺ group compared with control (figure 7E). The overall survival rate was significantly longer in mice treated with F1M1-Fc⁺, PTX MD and PTX MD+F1M1-Fc⁺ than in control (median survival of 59, 59 and 69 days for the F1M1-Fc⁺, PTX MD, and PTX MD+F1M1-Fc⁺ groups, respectively, compared with 55 days for control animals) (figure 7F; Kaplan-Meier survival analysis, p=0.0313 for F1M1-Fc⁺, p=0.0312 for PTX MD, p=0.001 for PTX MD+F1M1-Fc⁺). The overall survival rate was significantly longer in mice treated with PTX MD+F1M1-Fc⁺ (69 days) than with PTX MD (59 days) (p=0.041). All mice treated with the combination therapy gained weight (figure 7G) and displayed normal activities, suggesting no apparent toxicity. Overall, these results demonstrated that F1M1-Fc⁺ improved PTX therapeutic efficacy without increasing toxicity.

F1M1-Fc⁺ improves the therapeutic efficacy of the antiandrogen enzalutamide

We previously found that coexpression of cath-D and androgen receptor (AR) in non-metastatic TNBC is an independent prognostic factor of worse overall survival, suggesting the possibility of adjuvant combination therapy with anti-androgen drugs and anti-cath-D antibodies.16 We first showed that in the AR-expressing SUM159 TNBC cell line (online supplemental figure 14A), incubation with enzalutamide, an antiandrogen drug, did not decrease cath-D expression and secretion (online supplemental figure 14B), and incubation with F1M1-Fc⁺ did not reduce AR expression (online supplemental figure 14C).

Then, we treated mice harboring SUM159 cell xenografts with F1M1-Fc⁺ alone. Tumor volume increase was significantly slowed down in mice treated with F1M1-Fc⁺ (15 mg/kg by ip three times per week) compared with control (rituximab) (p<0.001) (figure 8A). Tumor volume at day 48 (treatment end) was significantly reduced by 71.9% in the F1M1-Fc⁺ compared with control group (p<0.0001) (figure 8B). The overall survival rate, reflected by a tumor volume <1000 mm³, was significantly longer in mice treated with F1M1-Fc⁺ than in the control group (median survival of 62 days and 55 days, respectively) (figure 8C; Kaplan-Meier survival analysis, p=0.0024), without any apparent toxicity (figure 8D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8

Therapeutic efficacy of the anti-cath-D F1M1-Fc⁺ antibody alone and in combination with enzalutamide in mice xenografted with SUM159 cells. (A) Tumor growth in mice receiving F1M1-Fc⁺ monotherapy. SUM159 cells were subcutaneously injected in nude mice. When tumor volume reached 50 mm³, mice were treated with F1M1-Fc⁺ (n=9) (15 mg/kg), or rituximab (Ctrl; n=9) (15 mg/kg), three times per week for 30 days. Mice were sacrificed when tumor volume reached 1000 mm³. Tumor volume (in mm³) is the mean±SEM. ***p<0.001 (mixed-effects multiple linear regression test). (B) Mean tumor volume at day 48 in mice receiving F1M1-Fc⁺ monotherapy. Mean±SEM. ****p<0.0001 for F1M1-Fc⁺ vs Ctrl (Mann-Whitney t-test). (C) Kaplan-Meier survival analysis in mice receiving F1M1-Fc⁺ monotherapy.*p=0.0024 for F1M1-Fc⁺ vs Ctrl (log-rank Mantel-Cox test). (D) Weight monitoring in mice receiving F1M1-Fc⁺ monotherapy. Mean±SEM. Rituximab (Ctrl; n=9), F1M1-Fc⁺ (n=9). (E) Tumor growth in mice receiving the F1M1-Fc⁺+enzalutamide combination therapy. SUM159 cells were subcutaneously injected in nude mice. When tumor volume reached 50 mm³, mice were treated with rituximab+corn oil (Ctrl; n=8), F1M1-Fc⁺ (n=8) (15 mg/kg, twice per week), enzalutamide (Enza; 30 mg/kg, five times per week, per os) (n=8), or F1M1-Fc⁺ (15 mg/kg, twice per week, ip) + enzalutamide (Enza; 30 mg/kg, five times per week, per os) (n=8) for 32 days. Mice were sacrificed when tumor volume reached 1500 mm³. Tumor volume (in mm³) is shown as the mean±SEM. ***p=0.001 for F1M1-Fc⁺ vs Ctrl, p=0.5 for Enza vs Ctrl, *p=0.019 for Enza vs F1M1-Fc⁺, **p=0.003 for Enza+F1M1-Fc⁺ vs Ctrl, **p=0.008 for Enza+F1M1-Fc⁺ vs Enza, p=0.887 for Enza+F1M1-Fc⁺ vs F1M1-Fc⁺ (mixed-effects multiple linear regression test). (F) Mean tumor volume at day 50 in mice receiving the F1M1-Fc⁺+ enzalutamide combination. Mean±SEM. ***p=0.0005 for all groups (Kruskal-Wallis), **p=0.0019 for F1M1-Fc⁺ vs Ctrl, *p=0.02 for Enza vs Ctrl, ***p=0.0002 for Enza+F1M1-Fc⁺ vs Ctrl, **p=0.0092 for Enza+F1M1-Fc⁺ vs Enza, p=0.0985 for Enza+F1M1-Fc⁺ vs F1M1-Fc⁺ (Mann-Whitney t-test). (G) Kaplan-Meier survival analysis in mice receiving the F1M1-Fc⁺ + enzalutamide combination.*p=0.0126 for F1M1-Fc⁺ vs Ctrl, *p=0.0196 for Enza vs Ctrl, ****p<0.0001 for Enza+F1M1-Fc⁺ vs Ctrl, *p=0.017 for Enza+F1M1-Fc⁺ vs Enza, **p=0.0091 for Enza+F1M1-Fc⁺ vs F1M1-Fc⁺ (log-rank Mantel-Cox test). (H) Weight monitoring in mice receiving the F1M1-Fc⁺+ enzalutamide combination. Mean±SEM. Rituximab (Ctrl; n=8), F1M1-Fc⁺ (n=8); Enza (n=8), Enza+F1M1-Fc⁺ (n=8).

We then treated nude mice bearing SUM159 cell xenografts with F1M1-Fc⁺ (15 mg/kg, twice per week, ip) or/and enzalutamide (Enza, 30 mg/kg, five times per week, per os). F1M1-Fc⁺ and Enza+F1M1-Fc⁺ significantly delayed tumor growth compared with control (p=0.001 for F1M1-Fc⁺, p=0.003 for Enza+F1M1-Fc⁺) (figure 8E). Tumor growth was significantly more reduced in the Enza+F1M1-Fc⁺ group than in the Enza group (p=0.008). At day 50 (treatment end), tumor volume was reduced by 55.7% (p=0.0019) in the F1M1-Fc⁺ group, by 43.4% (p=0.02) in the Enza group, and by 71% (p=0.0002) in the Enza+F1M1-Fc⁺ group compared with control (figure 8F). Tumor volume was significantly smaller in the Enza+F1M1-Fc⁺ group than in the Enza group (p=0.0082). Overall survival was significantly longer in mice treated with F1M1-Fc⁺, Enza, and Enza+F1M1-Fc⁺ than in controls (median survival of 63, 61, 64, and 57 days, respectively) (figure 8G; Kaplan-Meier survival analysis, p=0.0126 for F1M1-Fc⁺, p=0.0196 for Enza, p<0.0001 for Enza+F1M1-Fc⁺), without apparent toxicity (figure 8H). Overall survival was significantly longer in the Enza+F1M1-Fc⁺ group than in the Enza group (64 vs 61 days, p=0.017) and in the F1M1-Fc⁺ group (64 vs 63 days) (p=0.0091) (figure 8G). Overall, these results demonstrated that F1M1-Fc⁺ improves enzalutamide therapeutic efficacy.