Breaking NGF-TrkA immunosuppression in melanoma sensitizes immunotherapy for durable memory T cell protection

Abstract

Melanoma cells, deriving from neuroectodermal melanocytes, may exploit the nervous system's immune privilege for growth. Here we show that nerve growth factor (NGF) has both melanoma cell intrinsic and extrinsic immunosuppressive functions. Autocrine NGF engages tropomyosin receptor kinase A (TrkA) on melanoma cells to desensitize interferon γ signaling, leading to T and natural killer cell exclusion. In effector T cells that upregulate surface TrkA expression upon T cell receptor activation, paracrine NGF dampens T cell receptor signaling and effector function. Inhibiting NGF, either through genetic modification or with the tropomyosin receptor kinase inhibitor larotrectinib, renders melanomas susceptible to immune checkpoint blockade therapy and fosters long-term immunity by activating memory T cells with low affinity. These results identify the NGF-TrkA axis as an important suppressor of anti-tumor immunity and suggest larotrectinib might be repurposed for immune sensitization. Moreover, by enlisting low-affinity T cells, anti-NGF reduces acquired resistance to immune checkpoint blockade and prevents melanoma recurrence.

Extended Data Fig. 1. NGF expression in melanoma and its effect on melanoma cell proliferation.

a-f, 1,000,000 B16, Braf^V600EPten^−/−, or YUMM1.7 melanoma cells were subcutaneously injected into C57BL/6J mice. Melanoma tissues were harvested when tumors grew to 100 mm³. Expression of Ngf, Bdnf, Ntf3, Ntf4, Cntf, and Lif was assessed by quantitative RT-PCR. Spleen, lung, and skin tissues from normal C57BL/6J mice were as controls. Data are pooled from four independent experiments. g, Bioinformatic assessment of NGF expression in primary and metastatic melanoma. Box plots: the horizontal lines indicate the first, second (median) and third quartiles; the whiskers extend to ±1.5× the interquartile range. h, 100,000 B16 or YUMM1.7 melanoma cells were subcutaneously injected into female C57BL/6J mice. NGF concentration in serum of tumor-bearing mice was assessed by ELISA at indicated time. n=6 (day 0), n=3 (day 10), n=3 (day 14), n=3 (day 21 for B16), n=6 (day 21 for YUMM1.7). i-l, Knockout efficiency of NGF sgRNA in B16 (i) and YUMM1.7 (k) cells. In vitro proliferation of B16 (j) and YUMM1.7 (l) cells after NGF knockout. Hereafter we termed sgRNA-2 as NGF sgRNA. n=3 for each time point. m, 100,000 B16 cells were subcutaneously injected into NSG mice. Data are presented as means ± s.d. n=6 (WT) and 8 (KO). Western blot, one of two independent experiments is shown. *P < 0.05, ****P < 0.0001. P values were determined using two-tailed Student’s t-test (a), one-way ANOVA (j, l), two-sided Wilcoxon test (g), and two-sided Mann-Whitney test (m, h).

Extended Data Fig. 2. NGF deficiency remodels the melanoma microenvironment.

a-b, 1,000,000 B16 cells were subcutaneously injected into male C57BL/6J mice. Total RNA was extracted from melanoma tissues on day 15 and analyzed by RNA-seq. t-distributed stochastic neighbor embedding (t-SNE) plot cluster visualization (a) showing segregation of tumors. GO pathway analysis (b) of NGF sgRNA versus Ctrl sgRNA tumors. Red arrows indicated immune-related pathways. P value was calculated by one-tailed fisher exact test, and adjusted using Benjamini-Hochberg false discovey rate (FDR). c-d, 1,000,000 YUMM1.7 (c) and B16 (d) cells were subcutaneously injected into C57BL/6J mice. Melanoma tissues were harvested on day 10, and proportions of myeloid populations were analyzed by flow cytometry in YUMM1.7 (c) and B16 (d) tumors. n=6 (WT) and 7 (KO) in (c). n=7 (WT) and 7 (KO) for PMN and Mϕ in (d), n=6 (WT) and 6 (KO) for DC in (d). e-h, 1,000,000 YUMM1.7 Ctrl sgRNA and 2,000,000 YUMM1.7 NGF sgRNA cells were subcutaneously injected into female C57BL/6J mice. Tumors with similar tumor size (e) were collected on day 6 (Ctrl sgRNA) and day 15 (NGF sgRNA) after inoculation. The proportion of CD4+ T (f), CD8+ T (g) and NK (h) cells were analyzed by flow cytometry. n=8 (WT) and n=7 (KO). i-l, 1,000,000 Ctrl sgRNA and NGF sgRNA B16 cells were subcutaneously injected into male C57BL/6J mice. Tumors with similar tumor size (i) were collected on day 8 (Ctrl sgRNA) and day 15 (NGF sgRNA) after inoculation. The absolute number of CD4+ T (j), CD8+ T (k) and NK (l) cells were analyzed by flow cytometry. n=4 (WT) and 6 (KO). Data are presented as means ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001. Two-sided Mann-Whitney test (c, e, f, g, h, i, j, k, l), one-tailed fisher exact test (b) and two-tailed Student’s t-test (d).

Extended Data Fig. 3. NGF decreases IFNγ sensitivity of melanoma cells.

a, 1,000,000 B16 cells were subcutaneously injected into male C57BL/6J mice. Melanoma tissues were harvested on day 10 and submitted for scRNA-seq. Uniform manifold approximation and projection (UMAP) plots. Phenotypic clusters are represented in distinct colors. b, Fraction of immune cells. c, Significantly enriched hallmark gene sets in the melanoma population from scRNA-seq. d, IFNγ response gene signature in melanoma cells. e, The apoptosis score was compared in melanoma cell population after NGF inactivation. Box plots in e: the horizontal lines indicate the first, second (median) and third quartiles; the whiskers extend to ±1.5× the interquartile range. f, Ctrl or NGF sgRNA B16 cells were treated with IFNγ for 30 min. Western blotting was used to assess the level of p-Stat1 (Y701) and total Stat1. g, B16 tumor cells were treated with 10 μM p75NTR inhibitor LM11A-31 for 24 h, followed by 20 ng/ml IFNγ stimulation for 30 min. Western blots show levels of p-Stat1 (Y701) and total Stat1. h, shTrkA knockdown efficiency in B16 cells as determined by flow cytometry. i, Rescued expression of Socs1 in NGF sgRNA B16 cells, as assessed by RT-PCR. Cell line samples were from one well of 6-well plate for each group, with 3 technical replicates. Data are presented as means ± s.d. Western blot, one of two independent experiments is shown. **P < 0.01, ****P < 0.0001. Two-sided Wilcoxon test (e) and two-tailed Student’s t-test (i). OE, overexpression.

Extended Data Fig. 4. NGF decreases IFNγ sensitivity through the MEK/MAPK/SOCS1 pathway.

NGF sgRNA B16 tumor cells were treated with NGF at the indicated concentrations for 30 min. Western blots show Jak-Stat (a) and MAPK signaling (c) pathway proteins. NGF sgRNA B16 cells were serum-starved for 5 hours, and then treated with NGF for another 2 h. NGF sgRNA B16 cells were also treated by Jak1, Stat3 (b), MEK, ERK, p38 MAPK and JNK inhibitors (d) 1 h before NGF treatment, as indicated. Socs1 expression was evaluated by RT-PCR (b, d). Samples were pooled from 3 biological replicates with 3 technical replicates. Data are presented as means ± s.d. Western blot data pooled from two independent experiments. ****P < 0.0001. P values were determined using one-way ANOVA.

Extended Data Fig. 5. NGF suppresses IFNγ signaling in the melanoma microenvironment.

a-l, 1,000,000 B16 cells were subcutaneously injected into male C57BL/6J mice. Melanoma tissues were harvested on day 10 for scRNA-seq. Figure shows GO pathway analysis of each cluster in NGF sgRNA versus Ctrl sgRNA tumors. m, 1,000,000 B16 cells were subcutaneously injected into male and female C57BL/6J mice. Melanoma tissues were harvested on day 15. The level of IFNγ in B16 tumor lysates was analyzed by western blot using antibody against IFNγ. n=3.

Extended Data Fig. 6. Lymphocyte depletion experiments in melanoma models.

1,000,000 control or NGF sgRNA B16 cells were subcutaneously injected into male C57BL/6J mice. Lymphocyte depletion efficiencies by anti-CD4, anti-CD8, and anti-NK1.1 antibodies in B16 (a, b, d, e) and YUMM1.7 (c, e)-bearing mice were assessed by flow cytometry. Representative plots (a, d) and statistics (b, c, e) are shown. n=10 (IgG) and n=9 (anti-CD4 and anti-CD8) in (b), n=6 (IgG), 6 (anti-CD4) and 7 (anti-CD8) in (c), n=10 (IgG) and 9 (anti-NK1.1) for B16, n=6 for YUMM1.7 in (e). Data are presented as means ± s.d. ***P < 0.001, ****P < 0.0001. Two-tailed Student’s t-test.

Extended Data Fig. 7. NGF-TrkA suppresses T cell activation and function.

a, CD25 expression on CD8+ T cells as measured by flow cytometry after NGF treatment for 72 h. Data are presented as means ± s.d. b, Cytokine production by effector CD8+ T cells after NGF treatment for 72 h. Data are presented as means ± s.d. c, OT-I cells were activated with 1 μg/ml OVA N4 peptide for 48 h. Expression of cell surface p75NTR was measured by flow cytometry. d, OT-I cells were stimulated with the indicated concentrations of OVA N4 and OVA Q4 peptides for 16 h. CD69 expression was assessed by flow cytometry. Data are presented as means ± s.d. e, Schematic representation of competitive T cell transfer experiments. f, Knockout efficiency of TrkA on CD8+ T cells by CRISPR/Cas 9 editing technology. g-i, 36 hr after intravenous transfer, the percentage of ZsGreen+ sgNT and BFP+ sgTrkA Pmel-1:Cas 9 T cells in B16 melanoma was analyzed by flow cytometry (g). The ratio between BFP+ sgTrkA and ZsGreen+ sgNT Pmel-1:Cas 9 T cells was calculated (h). The percentage of IFNγ-producing CD8+ T cells in ZsGreen+ sgNT and BFP+ sgTrkA Pmel-1:Cas 9 T cells was analyzed (i). Data were pooled from two independent experiments. Data are presented as means ± s.d. n=7. j, A375 melanoma cells were treated with 10 μM larotrectinib for 16 hr, followed by treatment with 20 ng/ml IFNγ for 4 h. CXCL10 expression was assessed by RT-PCR. k, SK-MEL-2 melanoma cells were treated with 10 μM larotrectinib for 16 hr, followed by treatment with 100 ng/ml IFNγ for 5 h. CXCL10 expression was assessed by RT-PCR. Samples were pooled from 3 biological replicates in 3 independent experiments, with 2–3 technical replicates for each experiment. Data are presented as means ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Two-tailed Student’s t-test (a, b, d, i),one-way ANOVA (j, k). FMO, Fluorescence Minus One.

Extended Data Fig. 8. NGF does not suppress non-antigen-specific T cells and strategies for sorting high- and low-affinity T cells.

a, Tumor growth in indicated groups of male C57BL/6 mice that rejected B16 tumors and were rechallenged with LLC lung cancer cells, as compared with naive male C57BL/6J mice. Data are combined from two independent experiments (n=3–5). Data are presented as means ± s.d. b, Tumor growth in female C57BL/6 mice that rejected YUMM1.7 NGF sgRNA tumors and were rechallenged with LLC lung cancer cells, as compared with naive female C57BL/6J mice. n=4. Data are presented as means ± s.d. c, Tumor growth of female C57BL/6 mice that rejected B16 NGF sgRNA tumors and were rechallenged with LLC lung cancer cells, as compared with naive female C57BL/6J mice. n=4. Data are presented as means ± s.d. d, Shannon entropy of TCRs in YUMM1.7 tumors. n=4. Box plots: the horizontal lines indicate the first, second (median) and third quartiles; the whiskers extend to ±1.5× the interquartile range. e, Female C57BL/6 mice were vaccinated i.v. with 0.1 LD50 of attenuated recombinant Listeria-OVA. OVA-tetramer-high and –low CD8+ T cells were FACS-sorted from spleen on day 7 after a single vaccination. Representative dot blots gated on OVA-tetramer+ CD8+ T cells are shown. f, Female C57BL/6J mice were subcutaneously injected with 50,000 Ctrl or NGF sgRNA B16-OVA cells. All mice in Ctrl sgRNA group were sacrificed due to excessive tumor growth. Tumor-free mice in the NGF sgRNA group were rechallenged with B16-OVA cells on day 78. 9 days later, OVA-tetramer+ CD8+ T cells were sorted from tumor draining-lymph nodes and submitted for TCRβ-seq. Representative dot blots gated on OVA-tetramer+ CD8+ T cells are shown.

Extended Data Fig. 9

Schematic of tumor cell-derived NGF restricting anti-tumor immunity in melanoma.

Fig. 1. NGF predicts a dismal prognosis and a poor response to immunotherapy in melanoma patients.

a-g, Tumor growth curves of C57BL/6J mice subcutaneously injected with 100,000 YUMM1.7 (a, c, d, female), 1,000,000 YUMM1.7 (b, female) or 100,000 B16 cells (e, f, g, male) with control (Ctrl) or NGF sgRNA, as indicated. Survival curves of mice bearing 100,000 YUMM1.7 (d) or B16 cells (g). Data are representative of two to three independent experiments for each cell line. n=6 in (a), n=7 in (b), n=10 in (c and d), n=6 in (e), n=9 (WT) and 10 (KO) in (f), n=8 (WT) and 9 (KO) in (g). Representative photos and tumor weight on day 21 for YUMM1.7 (a, b) and day 15 for B16 (e). h-i, SKCM gene expression and associated clinical data were retrieved from TCGA database (h) and Hugo et al cohort (i), respectively. Red and blue curves represent low and high expression of NGF, respectively. j, Schematic of generating NGF-deficiency score. k-n, NGF-deficiency score was positively correlated with ICB response in indicated patient cohorts. Box plots: the horizontal lines indicate the first, second (median) and third quartiles; the whiskers extend to ±1.5× the interquartile range. n=18 for k, n=50 for l, n=51 for m, n=82 for n. o, Schematic of generating NGF deficiency-associated immune signature by machine learning. p-q, NGF deficiency-associated immune signature was used to predict survival in indicated patient cohorts. Data are presented as means ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. P values were determined using a two-tailed Student’s t-test (a, b, c, e, f, k, l, m, n) or Mantel–Cox log-rank test (d, g, h, i, p, q).

Fig. 2. NGF establishes an immune-excluded tumor microenvironment in melanoma.

a, Heat map of immune deconvolution for Ctrl or NGF sgRNA B16 tumor tissues by RNA-seq. n=4. b-d, CD45 (b, d) and CD4, CD8, NK1.1 (c, d) immunofluorescence of Ctrl or NGF sgRNA B16 tumor tissues, as indicated. Nuclei were counterstained with DAPI. Scale bars, 50 μm. n=7 for b, c and d. e-f, Tumor-infiltrating lymphocytes were analyzed by flow cytometry of Ctrl or NGF sgRNA B16 (e) or YUMM1.7 (f) tumor tissues. Data are representative of two independent experiments. n=7 for B16, n=6 (WT), 7 (KO) for YUMM1.7. g, Volcano plot showing increased (red) and decreased (blue) gene expression (FDR-adjusted P < 0.05) in NGF sgRNA versus Ctrl sgRNA B16 tumors. A cluster of chemokines was labelled. One-sided P value from Ward test. P values were adjusted using Benjamini-Hochberg false discovey rate (FDR). n=4 mice per group, P<0.05. h, Male C57BL/6J mice were subcutaneously injected with 1,000,000 Ctrl or NGF sgRNA B16 cells and treated with Cxcr3 neutralizing antibodies. Percentages of intratumoral CD4+ T, CD8+ T, and NK cells were analyzed at day 15 by flow cytometry. Data combined from two independent experiments. n=15 (WT), 15 (IgG) and 13 (anti-CXCR3). i, Tumor growth curves of C57BL/6J mice as in (h). Data combined from two independent experiments. n=17 (WT), 19 (IgG) and 15 (anti-CXCR3). j, CD45 and CXCL10 immunofluorescence in Ctrl and NGF sgRNA B16 tumor tissues. Nuclei were counterstained with DAPI. Scale bars, 50 μm. Representative data from two independent experiments. k, Male C57BL/6J mice were subcutaneously injected with 1,000,000 Ctrl or NGF sgRNA B16 cells. CD45− and CD45+ cells were sorted out by FACS on day 10. Chemokine expression by quantitative RT–PCR analysis of CD45− (left) and CD45+ (right) cells from Ctrl or NGF sgRNA B16 tumor tissues. Samples were from 3 individual mice in each group with one technical test for each sample. Data are presented as means ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. P values were determined using a two-tailed Student’s t-test (d, e, k), a two-sided Mann-Whitney test (f), Ward test (g), or one-way ANOVA (h, i).

Fig. 3. Autocrine NGF signaling suppresses melanoma cell chemokine expression.

a, IFNγ-related genes in B16 tumor cell population from scRNA-seq. b, The correlation between IFNγ score and apoptosis score in the B16 melanoma cell population from scRNA-seq. c, Melanoma cells were divided into two subpopulations with low and high IFNγ score, respectively, as in (b). The fraction of melanoma cells in each subpopulations was shown. d, The IFNγ score was compared in each population after NGF inactivation as in (b). e, p-Stat1 (Y701) in B16 cells was assessed by flow cytometry after IFNγ treatment for 30 min. f, Chemokine expression by RT–PCR in B16 cells treated with IFNγ for 24h. n=3. g, Tumor-infiltrating lymphocytes were analyzed by flow cytometry after IFNγ neutralization. Data combined from two independent experiments. n=12 (WT), 14 (IgG), 12 (anti-IFNγ). h, Tumor growth curves as in (g). Data combined from two independent experiments. n=14 (WT), 15 (IgG), 16 (anti-IFNγ). i, p-Stat1 (Y701) levels were assessed by flow cytometry in shNT or shTrkA B16 cells. j, Chemokine expression by RT–PCR analyses of B16 cells in the presence of IFNγ. n=3. k, Tumor-infiltrating lymphocytes were analyzed by flow cytometry. n=13. l, Tumor growth curves as in (k). n=13. m, SOCS1 and SOCS3 expression were analyzed by western blotting. n, Western blots show levels of p-Stat1 (Y701) and total Stat1 in B16 cells treated with IFNγ. o, Chemokine expression by RT-PCR analyses of B16 cells treated with IFNγ. n=3. p, Tumor-infiltrating lymphocytes were analyzed by flow cytometry. n=7 (WT), 9 (KO), 11 (KO+SOCS1). q, Tumor growth curves of mice as in (p). n=9 (WT), 10 (KO), 12 (KO+ SOCS1). r, Schematic of NGF cross-regulating IFNγ signaling. Data are presented as means ± s.d. For Box plots in a and d: the horizontal lines indicate the first, second (median) and third quartiles; the whiskers extend to ±1.5× the interquartile range. Western blot, one of two independent experiments is shown. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Two-tailed Student’s t-test (e, i, k, l), one-way ANOVA (f, g, h, j, o, q), two-sided Mann-Whitney test (p) and two-sided Wilcoxon test (d).

Fig. 4. The NGF-TrkA axis restricts CD8+ T cell activation and function.

a, IFNγ−producing lymphocytes were analyzed in YUMM1.7 and B16 tumors by flow cytometry. n=6 (WT), 7 (KO) for YUMM1. n=6 (CD4, CD8 in WT), 5 (NK in WT), 7 (KO) for B16. b, Tumor growth curves of melanoma-bearing mice treated with lymphocyte depletion antibodies, as indicated. n=6 (WT), 6 (IgG), 7 (CD8), 6 (CD4), 6 (NK) for YUMM1.7. n=7 (WT), 10 (IgG), 9 (CD8), 9 (CD4), 9 (NK) for B16. c, Heat map for RNA-seq of anti-CD3ε/28-activated CD8+T cells treated with NGF for 24 h. d, TrkA expression on the surface of CD8+ OT-I cells. e, TrkA expression kinetics on CD8+ OT-I cell surface in the presence of OVA N4 or OVA Q4 peptide. f-h, Calcium flux of OT-I cells pre-treated with NGF and co-cultured with splenic cells loaded with OVA N4 or Q4, as calculated using microscopy (f). Calcium dynamics (g) and concentration integration within the first 10 min (h) were analyzed. i, Study design. OT-I cells were primed by OVA N4 or OVA Q4 peptides. B16-OVA N4 tumor-bearing mice were treated with OVA N4-primed high-affinity OT-I cells. B16-OVA Q4 tumor-bearing mice were treated with OVA Q4-primed low-affinity OT-I cells. j-k. Tumor growth curves are plotted as in (i). Combined data from two independent experiments. n=7 (WT), 7 (KO), 8 (WT+N4), 8 (KO+N4), 14 (WT+Q4), 16 (KO+Q4). l-m, Naïve CD8+ T cells from C57BL/6J mice were stimulated with anti-CD3ε antibody-coated plate for 48 h. Activated CD8+ T cells were then treated with NGF for another 48 h. Ctrl and NGF-treated CD8+ T cells were then restimulated with 2 μg/ml soluble anti-CD3ε antibodies for 10 min. Phosphorylated ZAP70 was assessed by flow cytometry. Unstimulated CD8+ T cells served as a control for basal level of pZAP70. n=3. n, Co-immunoprecipitation assay showing the interaction of SHP-1 with CD3ε. Quantification data were combined from three independent experiments. o, Schematic of CD8+ T cell suppression by NGF. Data are presented as means ± s.d. *P < 0.05, ***P < 0.001, ****P < 0.0001. Two-tailed Student’s t-test (a, m, n), one-way ANOVA (b, h) or two-sided Mann-Whitney test (k).

Fig. 5. Targeting NGF enhances anti-melanoma immunotherapy.

a, B16 cells were treated with 10 μM larotrectinib for 24 h, followed by 20 ng/ml IFNγ stimulation for 30 min. Western blot shows p-Stat1 (Y701) and total Stat1 levels. b, Cxcl10 expression by RT-PCR assay of B16 cells in the presence of 10 μM larotrectinib and 20 ng/ml IFNγ for 24 h. Samples were pooled from 3 biological replicates with 3 technical replicates. c-d, Immunohistochemical staining of NGF in human melanoma tissues from Moffitt Cancer Center. Scale bars, 50 μm. Representative images from 104 samples in one experiment. e, Experimental design. f, Growth curves of individual tumors in male C57BL/6J mice treated as indicated. Representative data from four independent experiments. n=8–9 as indicated. g, Survival curves of C57BL/6 mice treated as indicated. Combined data from four independent experiments. n=16 (WT), 16 (laro), 32 (WT+T), 32 (WT+laro+T), 16 (KO), 32 (KO+T). h, Tumor growth in male C57BL/6 mice that rejected B16 tumors in the indicated groups and rechallenged with WT B16 cells on day 140, as compared with naive male C57BL/6J mice. Data combined from two independent experiments. n=9 (Ctrl), 4 (WT+T), 9 (WT+laro+T), 8 (KO+T). i-l, Tumor growth and survival of female C57BL/6 mice that rejected NGF sgRNA B16 (i, j) or YUMM1.7 (k, l) tumors in indicated groups and rechallenged with WT B16 or YUMM1.7 cells on days 140 and 80, respectively, as compared with naïve female C57BL/6J mice. Representative data from two independent experiments. n=4 (Ctrl), 4 (KO) for B16. n=6 (Ctrl), 6 (KO) for YUMM1.7. m, Representative photos of depigmentation of C57BL/6 mice that rejected NGF sgRNA YUMM1.7 cells during rechallenge. n, Frequency of depigmentation in (m). n=6. Data are presented as means ± s.d. Western blot, one of two independent experiments is shown. *P < 0.05, ***P < 0.001, ****P < 0.0001. One-way ANOVA (b), Mantel–Cox log-rank test (g, j, l) and two-sided Chi-square test (n).

Fig. 6. NGF inhibits the expansion and function of low-affinity T cells.

a, Analysis flow chart of intratumoral TCR repertoire. b, Study design of TCRβ-seq in YUMM1.7 tumors. c, TCR similarity in YUMM1.7 tumors. d, Gini coefficiency of TCRs in YUMM1.7 tumors. Box plots in c and d: the horizontal lines indicate the first, second (median) and third quartiles; the whiskers extend to ±1.5× the interquartile range. e, Expanded TCR clones which make up >10%, >5%, >2%, >1% and >0.5% are shown as gradient segments. Segments shaded in darker colors represent TCR clones found above the cutoff percentage and those shaded in lighter colors represent TCR clones that were rarer in the whole TCR repertoire of YUMM1.7 tumors. f, Study design to discover high- and low-affinity TCRs. g-h, Expanded TCR clones in OVA-specific high-affinity (g) and low-affinity (h) TCR repertoires of B16-OVA tumors. i, Representative treatment schedule. All control mice were sacrificed when tumor growth limits were exceeded. Tumor-free mice in the NGF sgRNA group were rechallenged with WT B16-OVA cells, and OVA-tetramer+ CD8+ T cells were sorted from tumor draining lymph nodes for TCRβ-seq. j-k, Clonotype (j) and clone (k) frequencies of high- and low-affinity TCRs were analyzed as in (i). Data were derived from the pooled 6 tumor-draining lymph nodes. l, Experimental design. Tumor-free mice in both groups were rechallenged with WT B16-OVA cells, and OVA-tetramer+ CD8+ T cells were sorted for TCRβ-seq. m-n, Clonotype (m) and clone (n) frequencies of high- and low- affinity TCRs was analyzed as in (l). Data were sourced from the pooled 2–6 tumor-draining lymph nodes. n=2 (WT+anti-PD-1), n=6 (KO+anti-PD-1). o, Schematic representation showing identification of low-affinity TCRs with confirmed memory response. q, Frequency of low-affinity TCRs with confirmed memory response was analyzed in draining lymph nodes of Ctrl or NGF sgRNA B16-OVA tumor-bearing C57BL/6J mice receiving anti-PD-1 therapy. Data combined inputs from the pooled 2–6 tumor-draining lymph nodes and from 8 individual B16-OVA primary melanoma tissues, each from a separate mouse. n=2 (WT+anti-PD-1), n=6 (KO+anti-PD-1), n=8 (B16-OVA primary tumor). *P < 0.05, **P < 0.01, ****P < 0.0001. Two-tailed Student’s t-test (c, d).