Nature | www.nature.com | 7 LNMs, we compared Pla2g2d + and Pla2g2d − LNMs, and observed a strong upregulation in Pla2g2d + LNMs of numerous genes that have previously been shown to define T-cell-zone macrophages 43 (TZMs; Fig. 4n–p and Extended Data Fig. 8c,d), including Timd4 , Il22ra2 , Rxrg and Pla2g2d itself. By contrast, typical markers of other LNM subsets, such as Adgre1 (encoding F4/80), showed an opposite trend (Fig. 4n and Extended Data Fig. 8c,d), indicating that PLA2G2D-producing LNMs have a TZM phenotype. These observations fit with our identification of PLA2G2D -expressing myeloid cells in the TDLN paracortex, which is the anatomical region that corresponds to the T cell zone 43,44 . To identify factors that induce PLA2G2D, we exposed human monocyte-derived macrophages (MDMs) to activated T cell culture supernatant or inflammatory cytokines (IFNγ and TNF) in vitro, which resulted in a robust induction of PLA2G2D expression (Extended Data Fig. 8e,f). Analysis of TCGA data revealed strong correlations between the levels of PLA2G2D RNA and of IFNG and TNF RNA across tumour types (Extended Data Fig. 8g), suggesting that T-cell-derived inflam- matory cytokines have a role in driving the production of PLA2G2D by myeloid cells. In the LN6-987AL model, however, antibody-mediated neutralization of IFNγ did not significantly reduce LNM frequencies or Pla2g2d expression levels, indicating that additional factors contrib- ute to PLA2G2D induction in vivo (Extended Data Fig. 8h,i). Of note, Pla2g2d induction in LNMs was similar in the LN6-987AL and B16F0 models (Extended Data Fig. 8j). Thus, PLA2G2D is produced mainly by macrophages in the LN T cell zone, where it can be targeted therapeuti- cally to inhibit tumour progression. Dual targeting of PLA2G2D and PD-1 In melanoma tumour models, treatment with anti-PLA2G2D antibodies induced the expression of genes encoding effector molecules, such as chemokines, granzymes and cytokines, in CD8 + T cells, but also induced the expression of genes that encode co-inhibitory receptors (for exam- ple, Pdcd1 , Lag3 and Entpd1 ; Fig. 4q). This indicates that anti-tumour activity increases in the face of imminent checkpoint activation, which could offer opportunities for combination therapies with existing ICIs. Given that PLA2G2D directly affected CD8 + T cell phenotypes in vitro and in vivo, we set out to further characterize the effects of PLA2G2D blockade on anti-tumour CD8 + T cell responses. Depleting CD8 + T cells in the LN6-987AL model abrogated the treatment benefit, indicating that CD8 + T cells are indeed necessary for anti-PLA2G2D efficacy in vivo (Fig. 4r). Longitudinal monitoring of CD8 + T cell responses across the blood and tissues of melanoma-bearing mice revealed that treatment with anti-PLA2G2D antibodies transiently increased the abundance of Ki-67 + CD8 + T cells—including tumour-specific (TRP2-tetramer posi- tive) and memory T cells—in the TDLNs and blood at early time points (Fig. 5a–c). Notably, whereas Ki-67 + CD4 + T cells were also increased in the blood, inhibition of PLA2G2D did not induce the activation of T reg cells (Fig. 5d). Tumour-specific CD8 + T cells subsequently increased substantially in the primary tumour (day 13), coinciding with delayed cancer progression, although these numbers returned to levels observed in untreated mice at later time points (Fig. 5a). This inability to sustain the improved anti-tumour immunity induced by anti-PLA2G2D treatment was accompanied by high levels of PD-1 + CD8 + T cells in the blood, TDLNs and tumour (Fig. 5e and Extended Data Fig. 9a). Of note, tetramer-positive CD8 + T cells in tumours remained highly proliferative and did not upregulate TOX, suggesting that these cells have not yet adopted an exhausted phenotype (Extended Data Fig. 9b). Moreover, PLA2G2D was prominently upregulated in NSCLC tumours that acquired resistance to anti-PD-1 therapy 16 (Extended Data Fig. 9c), providing further rationale for combining PLA2G2D blockade with PD-1 ICI. Indeed, mixed lymphocyte reactions using mouse or human T cells and DCs showed that the induction of IFNγ secretion by anti-PD-1 blocking antibodies was suppressed by PLA2G2D, with genetic deletion or antibody-mediated inhibition of PLA2G2D ameliorating the suppression of cytokine release (Extended Data Fig. 10a–e). Combining anti-PD-1 inhibition with genetic or antibody-mediated inactivation of PLA2G2D in the immunogenic MC38 tumour model revealed that combined inhibition suppressed tumour growth beyond the efficacy achieved with monotherapy (Fig. 5f and Extended Data Fig. 10f). Nota- bly, treating PLA2G2D -humanized mice bearing ICI-insensitive B16F10 or metastatic LN6-987AL tumours with human-specific anti-PLA2G2D antibodies either sensitized tumours to (B16F10) or further improved the efficacy of (LN6-987AL) anti-PD-1 ICI (Fig. 5g–i and Extended Data Fig. 10g,h). This improved efficacy with combination therapy was accompanied by increased frequencies of tumour-specific and IL-2-producing CD8 + T cells in the TME (Fig. 5j). Thus, PLA2G2D and PD-1 represent non-redundant immune checkpoints that could be exploited in combination immunotherapy strategies. Discussion Here we show that the cellular composition of TDLNs differs between patients with melanoma who have a poor prognosis after surgery and those who have a favourable one. Specifically, we found that TDLNs from patients with metastatic recurrence exhibited an accumulation of immunosuppressive myeloid cells expressing PLA2G2D in T-cell-rich cellular neighbourhoods of the paracortex. In mice, we pinpointed Pla2g2d expression to a specific population of macrophages, called LNMs, that are enriched in TDLNs and absent from tumours. Inhib- iting PLA2G2D boosted tumour-specific CD8 + T cell responses and reduced tumour growth, indicating that immune checkpoints that are active mainly in TDLNs—outside the local TME—can control systemic anti-tumour immunity. PLA2G2D + myeloid cells were highly abundant in TDLNs from patients with stage III melanoma who later developed distant metastases, and Fig. 3 | PLA2G2D suppresses T cell activity and controls tumour growth in multiple models of cancer. a , Tumour volume measurements in Pla2g2d +/+ and Pla2g2d − / − mice ( n = 16 per group) after inoculation with B16F10, E.G7-OVA or MC38 cancer cells. b , Tumour volume measured over time in Pla2g2d +/+ ( n = 8) and Pla2g2d − / − C57BL/6 mice after inoculation with MC38. Pla2g2d − / − mice received PLA2G2D–Fc ( n = 8) or control Fc protein ( n = 7) by intraperitoneal (i.p.) injection. c , Proliferation (top; carboxyfluorescein succinimidyl ester (CFSE) dilution assay) and IL-2 or IFNγ secretion (bottom, Meso Scale Discovery immunoassay) were measured in activated (or non-stimulated; no stim.) cultured healthy donor CD4 + , CD8 + or total T cells after exposure to increasing doses of PLA2G2D–Fc, catalytically inactive PLA2G2D(H47Q)–Fc or control Fc proteins ( n = 3 donors). d , Intracellular levels of IL-2, IFNγ and granzyme B (GZMB) in CD8 + T cells treated for 72 h with PLA2G2D–Fc ( n = 4 donors). e , Top, surface marker levels of the indicated proteins during CD8 + T cell activation in vitro ( n = 3 donors). Bottom, representative histograms of PLA2G2D surface binding. f , DEGs ( P adj < 0.05) 24 h after exposing human CD8 + T cells ( n = 4 donors) to PLA2G2D protein and TCR (anti-CD3 and CD28) stimulation. The bar graph on the right shows top enriched pathways within upregulated genes ( n = 222). g , Representative CellTrace dilution profiles (measure of proliferation) in activated CD8 + T cells that were exposed to PLA2G2D (red) or left untreated (blue) at the indicated time points. h , Secretion of IL-2 ( n = 2 donors) or IFNγ ( n = 3 donors) in anti-CD3- and anti-CD28-stimulated peripheral blood mononuclear cell (PBMC) cultures exposed to PLA2G2D–Fc in the absence or presence of human-specific (44.H1 or 44.H2) or mouse cross-reactive (48.M1, 47.M2 or 48.M3) anti-PLA2G2D monoclonal antibodies. i , j , Tumour volume measurements in wild-type ( i ) or humanized PLA2G2D (hPLA2G2D; j ) C57BL/6 mice after inoculation with the indicated cancer cells (isotype, n = 9; all others, n = 10). Antibodies were administered by i.p. injection. Data are mean ± s.e.m. P values by two-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) ( a – c , i , j ), two-way ANOVA with Šidák correction for multiple testing ( h ) or two-sided paired t -test ( d ). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 (exact P values in Supplementary Data).