CD40L and IL-4 suppress NK cell-mediated antibody-dependent cellular cytotoxicity through the HLA-E:NKG2A axis

Abstract

Background: Anti-CD20 antibodies are first-line treatments for B cell malignancies. Natural killer (NK) cells are important mediators of anti-CD20 antibody efficacy in humans through antibody-dependent cellular cytotoxicity (ADCC). In B cell malignancies, the lymph nodes are a critical site of pathology and the T cell-derived signals CD40L and IL-4 within the lymph node microenvironment can mediate tumour proliferation, survival and resistance to pro-apoptotic therapy. CD40L and IL-4 have recently been shown to inhibit NK cell activation against chronic lymphocytic leukaemia (CLL) cells via the HLA-E:NKG2A immune checkpoint axis. However, the effect of these signals on NK cell-mediated ADCC of malignant B cells is unclear.

Methods: Using a combination of clinical samples, murine models, flow cytometry, immunoblotting, immunohistochemistry, ELISA, bioinformatics and functional assays, we examined the impact of lymph node-mimicking conditions on NK cell-mediated ADCC against malignant B cells. Exogenous CD40L and IL-4 were used to mimic T-B cell interactions in 2D malignant B cell cultures, in addition to a 3D spheroid model of T cell-dependent CLL proliferation.

Results: CD40L and IL-4 increased HLA-E expression on the surface of primary CLL cells and non-Hodgkin's lymphoma (NHL) cell lines, and this decreased NK cell-mediated ADCC via ligation of the inhibitory receptor NKG2A. High HLA-E surface expression was observed in lymph node FFPE sections of CLL and NHL patients and in a 3D ex vivo lymph node-mimicking model of CLL. NKG2A blockade potentiated NK cell-mediated ADCC against malignant B cells treated with CD40L and IL-4 and improved anti-CD20 antibody therapy in a murine model of B cell lymphoma.

Conclusion: These results reveal a novel mechanism of resistance to anti-CD20 therapy in B cell malignancies and demonstrate that the combination of anti-NKG2A with anti-CD20 could improve the treatment of patients with CLL or NHL.

Keywords: ADCC; HLA-E; NK cells; NKG2A; lymph nodes; monalizumab; rituximab.

Graphical Abstract

Figure 1.

CD40L and IL-4 suppress NK cell-mediated lysis of B cell leukaemia and lymphoma cells. (a and b) Primary CLL cells (>90% tumour) were treated with 300 ng/ml CD40L and 10 ng/ml IL-4 or left untreated for 24 h at 37°C before co-culture with NK cells from healthy donors for a further 6 h at 1:1, 5:1, and 10:1 effector:target (E:T) ratios. Propidium iodide was used to measure the proportion of lysed CLL cells by flow cytometry. Representative FACS plots of raw lysis data are shown in (a) and summarized data of NK cell-specific lysis are shown in (b) (n = 5). (c–f) NHL cell lines Ramos, Raji, SU-DHL-6, SU-DHL-4, and JeKo-1 were treated with 300 ng/ml CD40L and 10 ng/ml IL-4 for 24 h. Representative FACS plots of HLA-E expression shown in (c) and data is represented as mean±SEM for HLA-E (d) and total HLA (e) (n = 6). (f) Abundance of HLA-E protein in primary CLL and Ramos cells as measured by immunoblotting (representative of three repeats). (g–l) Ramos, Raji, and SU-DHL-6 cells were treated as above before co-culture with NK cells from healthy donors at 1:1 E:T ratio for 24 h. Propidium iodide was used to measure the proportion of lysed NHL cells. Representative FACS plots of raw lysis data shown in (g), (i), and (k) and summarized data of NK cell-specific lysis shown in (h), (j), and (l) (n = 5–6). (m) Ramos cells were treated as above before co-culture with healthy donor PBMCs at a 5:1 E:T for 4 h. Degranulation (CD107a expression) of NKG2A+ and NKG2A− NK cells was assessed by flow cytometry and data are shown as CD107a expression (%) minus the no target control. Statistical significance was calculated by two-way ANOVA with Sidak’s correction for multiple comparisons or multiple paired T tests with the Holm-Sidak’s multiple comparisons correction. *P < .05, **P < .01, ***P < .001, ****P < .0001.

Figure 2.

CD40L and IL-4 suppress NK cell-mediated ADCC of leukaemia and lymphoma cells. (a–i) Target cells were treated with 300 ng/ml CD40L and 10 ng/ml IL-4 or left untreated for 24 h before incubation with rituximab or obinutuzumab at the indicated concentrations or isotype control (10 or 1 μg/ml) and co-culture with NK cells from healthy donors at a 1:1 (Raji, CLL) or 1:5 (Ramos, SU-DHL-6) E:T ratio. Representative FACS plots of raw lysis data of Ramos cells are shown in (a) and NK cell-specific lysis is summarized as mean±SEM in the presence of increasing concentrations of rituximab (b–e) and obinutuzumab (f–i) (n = 3–5). (j and k) Primary CLL cells (>90% tumour) were treated with CD40L and IL-4 as above before co-culture with healthy donor PBMCs at a 5:1 E:T for 4 h in the presence of rituximab (10 μg/ml). Degranulation of NKG2A+ and NKG2A− NK cells was assessed by flow cytometry (n = 6). (l) 721.174 cells were treated with CD40L and IL-4 as above and HLA-E expression was assessed by flow cytometry. (m) 721.174 and primary CLL cells were treated with CD40L and IL-4 as above and MCL-1 protein abundance was assessed by immunoblotting (values under blot represent densitometry). Data for primary CLL are shown as a positive control. Representative of four repeats for 721.174 cells and three different donors for CLL. (n and o) 721.174 cells were treated with CD40L and IL-4 as above before incubation with rituximab at the indicated concentrations and co-culture with NK cells from healthy donors at a 1:10 E:T for 4 h. Representative FACS plots shown in N and NK cell specific lysis is summarized as mean±SEM in the presence of increasing concentrations of rituximab shown in (o) (n = 5). Statistical significance was calculated using two-way ANOVA with Sidak’s correction for multiple comparisons. *P < .05, **P < .01, ***P < .001, ****P < .0001.

Figure 3.

Identification of HLA-E and NK cells in the lymph nodes of patients with B cell malignancies. (a) FFPE diagnostic lymph node biopsies from consented patients with indicated B cell malignancies were stained with antibodies for HLA-E or CD20 by immunohistochemistry. Images representative of ≥2 patients for each disease are shown (10× magnification, scale bar 50 μm). The insert in each image represents 1× magnification taken from the same section. Gene expression associated with resting (b) and activated (c) NK cells extracted from Cibersort X quantified as gene set variation analysis score for DLBCL patients in the GOYA trial (n = 552) treated with rituximab or obinutuzumab plus CHOP chemotherapy who survived <12 months, 12–36 months, and >36 months. Statistical significance was calculated using Student’s t-test in Python. *P < .05 and **P < .01.

Figure 4.

HLA-E expression increases on CLL cells in a patient-derived 3D spheroid model of the lymph node microenvironment. 3D spheroids were generated from CLL patient PBMCs (>10% CD4+ T cells). Spheroids were cultured in ULA plates for up to 7 days with CpG (1 μg/ml), IL-2, IL-15, and IL-21 (all 25 ng/ml). (a) Representative images of spheroids taken at X4 magnification over time (scale bar 1 mm). Expression of HLA-E and total HLA on CLL tumour cells (CD19+CD5+) of the spheroid culture was assessed over time by flow cytometry, alongside FSC-A. Representative FACS plots shown in (b) and data is summarized as fold change expression relative to day 0 as mean±SEM (n = 7–9) in (c). (d and e) Primary CLL samples were cultured in 2D with CpG, IL-2, IL-15, and IL-21 individually or left untreated for 24 h before HLA-E (d) and total HLA (e) expression was assessed by flow cytometry. Data are presented as mean±SEM (n = 8). (f) Spheroids were cultured in the presence or absence of CD40 antagonist (anti-CD40) for 4 days and HLA-E expression was measured on the CLL cells by flow cytometry. Data is representative of two donors. (g and h) Purified CLL cells (>95% tumour) were cultured as above to generate spheroids in the absence of T cells. HLA-E expression on CLL tumour cells (CD19+CD5+) of the spheroid culture was assessed over time by flow cytometry. Representative FACS plots shown in (g) and data are summarized as mean±SEM in (h) (n = 3). (i) Spheroids generated from CLL samples with total PBMCs or >92% tumour were co-cultured with NK cells from healthy donors in combination with rituximab, obinutuzumab, or isotype control for 48 h at a 1:1 E:T ratio. Data are summarized as mean±SEM. Statistical significance was calculated using two-way or one-way ANOVA with Dunnett’s or Sidak’s correction for multiple comparisons. *P < .05, **P < .01, ***P < .001, ****P < .0001.

Figure 5.

Suppression of NK cell-mediated ADCC by CD40L and IL-4 can be overcome by NKG2A blockade. (a–d) Target cells were treated with 300 ng/ml CD40L and 10 ng/ml IL-4 or left untreated for 24 h before incubation with rituximab (1 μg/ml) or obinutuzumab (0.1 μg/ml) or isotype control (1 μg/ml). NK cells from healthy donors were incubated with 10 μg/ml Z199 (anti-NKG2A) before co-culture with the target cells at a 1:5 (Ramos a and b) or 1:1 (Raji c and d) E:T ratio. Representative FACS plots of raw lysis data are shown in (a) and (c) and NK cell-specific lysis is summarized as mean±SEM in (b) and (d) (n = 6–7). (e) Secretion of IFNγ by NK cells co-cultured with Raji as above was assessed by ELISA and data are summarized as mean±SEM (n = 10). (f and g) 721.174 cells were treated with 300 ng/ml CD40L and 10 ng/ml IL-4 for 24 h before incubation with rituximab (1 μg/ml) or obinutuzumab (0.1 μg/ml) or isotype control (1 μg/ml). NK cells from healthy donors were incubated with 10 μg/ml Z199 (anti-NKG2A) before co-culture with the target cells at a 1:10 E:T ratio. Representative FACS plots of raw lysis data are shown in (f) and NK cell-specific lysis is summarized as mean±SEM in G (n = 5). Statistical significance was calculated using two-way ANOVA with Dunnett’s correction for multiple comparisons. *P < .05, **P < .01, ***P < .001.

Figure 6.

Clinically relevant anti-NKG2A monalizumab enhances NK cell-mediated ADCC in vitro and NKG2A blockade improves anti-CD20 mAb therapy in vivo. Raji cells were treated with 300 ng/ml CD40L and 10 ng/ml IL-4 or left untreated for 24 h before incubation with rituximab or isotype control (1 μg/ml). NK cells from healthy donors were incubated with 15 μg/ml monalizumab (anti-NKG2A) before co-culture with the treated Raji cells at a 1:1 E:T ratio. Representative FACS plots of raw lysis data are shown in (a) and NK cell-specific lysis is summarized in (b) (n = 6). (c) Secretion of IFNγ by NK cells co-cultured with Raji as above was assessed by ELISA (n = 9). Statistical significance was calculated using two-way ANOVA with Dunnett’s correction for multiple comparisons. (d) Expression of Qa-1^b on BCL₁ murine lymphoma cells (CD19+Idiotype+) was confirmed by flow cytometry. Open histogram indicates isotype control-stained cells and the shaded histogram indicates Qa-1^b stained cells. (e) 10⁴ BCL₁ cells were injected into mice i.v. before injection with anti-NKG2A on days 6 and 13 (200 μg, i.v.) and anti-CD20 on day 7 (200 μg, i.p.) or respective isotype controls. Tumour growth was monitored by palpation of spleens and mice were terminated once humane endpoint was reached. (f) Kaplan–Meier survival curve of BCL₁-bearing mice treated with isotype controls (control), anti-CD20, anti-NKG2A, or anti-CD20 and anti-NKG2A antibodies (combination). Statistical significance calculated with the log-rank (Mantel-Cox) test, n = 5 per group. (g) Blood was withdrawn from mice via tail lancing two days after the last injection of anti-NKG2A (day 15), and KLRG1 expression was assessed on murine NK cells (CD3-NKp46+CD49b+) by flow cytometry. Statistical significance calculated with one-way ANOVA with Tukey’s correction for multiple comparisons. *P < .05, **P < .01, ***P < .001, ****P < .0001.