Tumor-infiltrating mast cells are associated with resistance to anti-PD-1 therapy

Abstract

Anti-PD-1 therapy is used as a front-line treatment for many cancers, but mechanistic insight into this therapy resistance is still lacking. Here we generate a humanized (Hu)-mouse melanoma model by injecting fetal liver-derived CD34⁺ cells and implanting autologous thymus in immune-deficient NOD-scid IL2Rγ^null (NSG) mice. Reconstituted Hu-mice are challenged with HLA-matched melanomas and treated with anti-PD-1, which results in restricted tumor growth but not complete regression. Tumor RNA-seq, multiplexed imaging and immunohistology staining show high expression of chemokines, as well as recruitment of FOXP3⁺ Treg and mast cells, in selective tumor regions. Reduced HLA-class I expression and CD8⁺/Granz B⁺ T cells homeostasis are observed in tumor regions where FOXP3⁺ Treg and mast cells co-localize, with such features associated with resistance to anti-PD-1 treatment. Combining anti-PD-1 with sunitinib or imatinib results in the depletion of mast cells and complete regression of tumors. Our results thus implicate mast cell depletion for improving the efficacy of anti-PD-1 therapy.

Fig. 1. Generation of Hu-mice.

a Schematic of Hu-mice reconstitution. Six-week-old female NSG mice are all treated with the myelo-depleting drug (busulfan [30 mg/kg]; i.p.) 24 h before they receive purified fetal liver-derived CD34⁺ cells (1 × 10⁵; i.v.) and autologous thymus grafts (~2 mm) under the renal capsule. After day 50, mice are periodically bled (100 μl) and characterized for human immune cells by standard flow cytometry assay using fluorochrome-conjugated anti-mouse or anti-human antibodies. b Repopulation of human CD45⁺ cells in circulating blood of reconstituted mice. A representative example of mice (n = 45) after 8–12 weeks of human CD34⁺ cell injection showed increased levels of human CD45⁺ cells (brown squares; p = 0.00025) in circulating blood when compared to control non-reconstituted female NSG mice (blue squares; n = 5). c Enhanced repopulation of human lymphocytes after AAV8-hu-cytokine transgenes delivery. Significant increase in circulating human CD45⁺ cells (p = 0.022 for days 48 and 72 [closed blue circles and brown squares] and p = 0.0094 for day 112 [closed black triangles]) in mice (n = 7) that received AAV8 hu-cytokines (IL3, IL-7, and GM-CSF; 2 × 10⁹ GC/ml; i.v.; 5 days after CD34 injection; right panel) when compared to mice (n = 5) that did not receive hu-cytokines (left panel). d Myeloid lineage cells after administration of hu-cytokines. CD33⁺, CD15⁺, CD11b⁺, and CD14⁺ cells are also seen in circulating blood after week 12 of CD34⁺ cells administration and when mice (n = 10) receive AAV8 hu-cytokines (see above) plus DNA-hu-cytokines (SCF, FLT-3, THPO; 50 μg; i.m.; multiple sites). An independent batch of mice was used for this experiment. e Reconstituted mice show the presence of CD56⁺ NK (innate immune) cells. Mice (n = 16) bled at 10 weeks showed increased CD56⁺ cells that decrease significantly to physiological levels by week 16 (p = 0.002). An independent batch of mice was used for this experiment. From d, all mice received AAV8 or DNA plasmid-encoded human cytokines as described in the method section. f–i Repopulation of human T- and B-cells. Generally, by 12–14 weeks, physiological levels of human T-and B-cells (f, left panel) and human CD4⁺ and CD8⁺ T cells are observed in circulating blood (f, right panel). Each point in the scatter plot represents blood drawn from an individual mouse (n = 16). g–i Repopulation of lymphoid organs with human immune cells. In H&E staining, there is dense repopulation of human lymphocytes in reconstituted mouse thymus (g; right panel; scale bar: 200 μm) and spleen (h; right panel; scale bar: 250 μm) when compared to non-reconstituted mouse thymus (g; left panel; scale bar: 500 μm) and spleen (h; left panel; scale bar: 200 μm). Dense repopulation of human lymphocytes in mouse kidney (renal capsule) grafted with human thymus (I; scale bar: 200 μm). A representative example of staining is shown in this figure. All observations (g–i) were consistent, and the experiment was repeated more than 2×. All mice were euthanized by CO₂ inhalation/cervical dislocation and organs harvested 24 weeks after CD34⁺ cell injections. Data are presented as mean values ± SD for b–f. A one-sided paired t-test was used for statistical analysis when p-values are specified. Source data are provided as a Source Data file.

Fig. 2. Functional characterization of human immune cells in humanized mice.

a–c T- and B-cell response to hTERT vaccine. a Schema for hTERT DNA vaccination. Hu-mice, as described in Fig. 1, received a total of three injections of hTERT vaccine (hTERT DNA [50 μg; i.m] followed by electroporation) every 2 weeks and the mice were euthanized by CO₂ inhalation/cervical dislocation and organs harvested 1 week after the last injection to determine T- and B-cell responses. b Anti-TERT T-cell responses. Spleen cells from Hu-mice (n = 3) were stimulated overnight (18 h) with pools of overlapping hTERT peptides (15 mer; 2 μg/ml/peptide) spanning the entire hTERT protein. Human IFNγ was detected in ELISPOT assay using a kit. Data are represented as SFU (spot forming units; mean ± SE) per 10⁶ splenocytes. hTERT-specific T-cell (IFNγ) response from vaccinated mice was compared to age-matched Hu-mice that received pMV101, a modified pVAX1 vector, as control or untreated age, sex-matched NSG mice controls. c Anti-TERT antibody (IgG) responses. Endpoint binding titer was determined in sera of hTERT vaccinated mice (n = 3) after 3 immunizations and compared to sera from NSG mice as controls. Data are presented as mean values ± SD. d–h Functional ability of immune T cells to restrict tumor growth. d Schema for Hu-mice tumor challenge experiment. e Hu-mice with T-cell reconstitution can restrict tumor growth of HLA-A2 matched A375 melanoma cells. Hu-mice, as described in Fig. 1 that have 90 to 120 cells/μl circulating CD45+/CD8+ cells (closed circle, blue line) when challenged with melanoma cells (10⁵; s.c.), can restrict tumor growth significantly (*p = 0.0281) when compared to non-reconstituted NSG mice (n = 3; closed circles, brown line) and Hu-mice with high circulating B-cells (n = 4; >65% [550 to 600 cells/μl] CD20⁺; open circles, magenta line) have unrestricted tumor growth. Tumor growth measurements are recorded using a digital caliper by an independent researcher. f–h Treatment with anti-PD-1 antibody can restrict tumor growth of melanoma cells. f Schema for anti-PD-1 therapy. Hu-mice with CD45⁺/CD8⁺ (90 to 120 cells/μl) and cells were randomized, and they receive melanoma cells (10⁵; s.c.). When tumors are palpable, mice receive anti-PD-1 (10 mg/kg; i.p. injections) every week for four injections and tumor growth measurements are recorded. g, h Anti-PD-1 therapy restrict melanoma growth. Hu-mice (n = 10) treated with anti-PD-1 antibody can restrict tumor growth of two different melanoma cells [(WM3629 [HLA-A3]; g] and [A375 [HLA-A2]; h]) significantly (blue line, closed circles; *p = 0.0437 for WM3629 and p = 0.0547 for A375) when compared to Hu-mice treated with control IgG (n = 5; magenta; open circles) or non-reconstituted age and sex-matched NSG mice (n = 10; brown; closed circles) treated with the ant-PD-1 antibody. Two different batches of Hu-mice were used for this experiment and they had comparable CD45⁺/CD8⁺ T-cell counts. Unrestricted tumor growth in presence of anti-PD-1 antibody was observed when Hu-mice were challenged with an aggressive phenotype of the tumor (Supplementary Fig. 8). Data are presented as mean values ± SEM for e–h. A one-sided paired t-test was used for analysis when p-values are provided. Source data are provided as a Source Data file.

Fig. 3. Immune and tumor heterogeneity as a possible cause of therapy resistance to anti-PD-1.

a–d Heterogeneous distribution of leukocytes and immune cells in tumors after PD-1 treatment. a. Tumor (A375) bearing Hu-mice that received anti-PD-1 as in Fig. 2h, showed dense leukocyte infiltration of leukocytes (bottom panel; scale bar: 500 μm) when compared to mice that received control mouse IgG (top panel; scale bar: 100 μm) as determined by H&E staining. b Tumor (A375) bearing Hu-mice that received anti-PD-1 showed a heterogeneous distribution of CD4⁺ and CD8⁺ T cells (stitched image right panel; scale bar: 500 μm) when compared control Ig treated Hu-mice that had a sparse distribution of T cells (right panel). Please see Supplementary Fig. 9 for digital quantification of CD4⁺ and CD8⁺ T cells. c Tumor (A375) bearing Hu-mice that received anti-PD-1 showed either low to moderate (left panel; scale bar: 50 μm) or robust (right panel; scale bar: 50 μm) tumor-infiltration of CD4⁺ (brown) and CD8⁺ (blue) T cells within the same tumor. d MassCyTOF staining shows heterogeneous and higher distribution of CD8⁺ T cells (magenta) within the nestin⁺ tumor (A375) cells (dark blue) in anti-PD-1 treated tumor-bearing mice (lower panels) as compared to low infiltration of CD8⁺ T cells (upper panels) in untreated Hu-mice (see f for quantification). Distribution of GrB⁺ T cells (yellow arrows; 2nd to right bottom panel) was heterogeneous as they were higher on the bottom half (digital quantification:131 counts) of the tumor section when compared to the remainder of other nestin⁺ tumor cell areas (digital quantification: 37 counts). e CD8⁺ T cells are of memory phenotype as they stain for CD45RO (light blue; top panel) and areas not infiltrated by CD8⁺ T cells reveal the presence of CD4⁺/FOXP3⁺ cells (magenta arrows; bottom panel). f Quantification of MassCyTOF images by ImageJ software indicates a significant increase (p = 0.0049) in CD8⁺ T cells and Granzyme B⁺ T cells; CD45RO⁺ memory T cells compared to untreated mice and increase in FOXP3⁺ Treg cells in the indicated top or bottom half panels. All histology and MassCyTOF staining were consistent and confirmed in replicates of 2. g, h Downmodulation of HLA class I (white arrows) was observed in tumor areas that were associated with higher FOXP3⁺ cells (p = 0.0351; see e bottom panel; a representative image). Confirmation by quantitative analysis of mean intensity (MI) of HLA-class I expression (h). A one-sided paired t-test was used for analysis when p-values are provided. Representative scale bars in d, e, g represent 50 μm. All observations (a–e, g) were consistent, were observed in technical replicates (CyTOF), confirmed in histology staining, and in replicate experiments.

Fig. 4. Increase in mast cells after anti-PD-1 therapy.

a CIBERSORT analysis of the RNA-seq data set (GSE161353) showed a higher abundance of mast related genes in tumors (WM3629 and A375) obtained from Hu-mice after anti-PD-1 treatment and increased expression of CXCL10, a chemoattractant for mast cells (b). The presence of mast cells was further confirmed by mast cell tryptase IHC staining. c, d Stitched image (top left panels) and three different fields from an untreated tumor (WM3629) had negligible staining for mast cells (top right panel and bottom two panels; scale bar: 100 μm), whereas stitched image (top right middle panel) and three different fields from anti-PD-1 treated tumor had robust staining for mast cells, which was highly significant compared to control-treated mice (p = 0.005407; top right panel and bottom 2 panels; scale bar: 100 μm). e Co-localization of FOXP3⁺ Treg and mast cells after anti-PD-1 therapy. Co-localization of these cells as determined by IHC staining was significant (p = 0.000047; scale bar: 200 μm) suggesting crosstalk. f Downmodulation of HLA class I. HLA class I molecules as determined by staining with anti-HLA class I antibody (red) were downmodulated in tumor (WM3629) areas (black arrows) that were infiltrated by mast cells (blue). Scale bar represents 50 μm. g Mast cells co-express CXCR3. Mast cells were co-stained with anti-MCT (blue) and anti-human CXCR3 (red; white arrows) antibodies (see Supplementary Fig. 11 for digital quantification). Scale bar represents 50 μm. h Melanoma (WM3629) cells co-express CXCL10. Tumor cells were co-stained with anti-melanoma (HMB45 [blue]) and anti-human CXCL10 (red; white arrows) antibodies. Scale bar represents 100 μm. Histology staining (c–h) was confirmed in repeat experiments (2×). Source data are provided as a Source Data file (data are included as part of Supplementary Fig. 10).

Fig. 5. Anti-PD-1 therapy response is modulated by the presence of tumor-infiltrating mast cells.

a Increase in mast cells in melanoma patients’ tumors after anti-PD-1 therapy. Immunostaining of human melanoma patients’ tumor showed an increased presence of mast cells after anti-PD-1 therapy (right panel) when compared to untreated individuals (left panel). Scale bars in both panels represent 200 μm. A representative staining is shown. b, c CIBERSORT analysis of three independent data sets (GSE123728 [16 pre and 23 on-therapy patients] and GSE91061 [43 pre-/on-therapy patients]) obtained from melanoma patients undergoing immune-checkpoint therapy showed a higher abundance of mast cell-related genes when compared to pre-therapy tumors and this was significant in b (n = 38; p = 0.0007) and c (n = 73; p = 0.049) due to higher number of matched pair biopsies (before and on-therapy samples). Box plots for pre-therapy tumors are represented as median (0.01153732; quartile 0.01153734), minimum (0; quartile 0.00561458), and maximum (0.07344924; quartile 0.02074330), and for on-therapy tumors are represented as median (0.01577157; quartile 0.01577157), minimum (0; quartile 0.00651041) and maximum (0.05185949; quartile 0.02311874). d Increased levels of mast cells in immune-checkpoint therapy non-responders. CIBERSORT analysis of RNASeq data set from MD Anderson trial (pre- and on-therapy patients (n = 23; Helmink et al.)) revealed higher mast cell levels in anti-PD-1 therapy non-responder population. However, this increase was not significant due to smaller patient numbers. Data are presented as mean values ± SEM. A one-sided paired t-test was used for analysis when p-values are provided. Source data are provided as a Source Data file (data are included as part of Supplementary Fig. 10).

Fig. 6. Complete regression of tumors after a combination of sunitinib and anti-PD therapy.

a Using an independent batch of Hu-mice, established tumors (A375) (details as in Fig. 2g, h; n = 5/group) were treated with sunitinib (20 mg/kg) daily by oral gavage and after 72 h, anti-PD-1 therapy (10 mg/kg) was given weekly for a total of 6 injections. Complete tumor regression was observed in presence of combination therapy (black inverted triangle; p = 0.0001; 2nd panel), while sunitinib alone (gray circles), anti-PD-1 alone (blue line, closed circles) or control IgG (magenta; open circles) did not have any effect of tumor growth. We observed similar results when drug imatinib (50 mg/kg; daily by oral gavage) was used in combination with anti-PD-1 (brown line, closed circles; p = 0.0282; n = 7; 3rd panel from the bottom). Imatinib alone had no effect on tumor growth. Cediranib (6 mg/kg; daily by oral gavage) either alone or in combination with anti-PD-1 antibody was unable to shrink the tumors (n = 3 per group; bottom panel). Data are presented as mean values ± SD. A one-sided paired t-test was used for analysis when p-values are provided. b Survival curve from the above-treated mice indicates significant (p = 0.0013 as determined by the log-ranked test) survival advantage of tumor-bearing Hu-mice that received combination therapy of sunitinib and anti-PD-1 group when compared to sunitinib alone (gray), or anti-PD-1 alone (dark blue) or control IgG (magenta) groups. c, d Depletion of mast cells in spleen in sunitinib treated mice (bottom [1000] μm left and right panels [100] μm) when compared to control mice (top left [1000] μm and right panels [100] μm panel). Histology staining (c, d) was confirmed in repeat experiments (2×). e Schema of mast cell-induced resistance mechanism to anti-PD-1. Source data are provided as a Source Data file.