Dendritic cell PD-L2 restrains intratumoral CD8⁺ T cell immunity.

Dendritic cell PD-L2 restrains intratumoral CD8⁺ T cell immunity.

Keywords

• PD-1 ligands

• Immune checkpoint blockade

• Dendritic cells

Main Findings

PD-1 blockade is central to cancer immunotherapy, yet the context-dependent function and cellular source of its ligands, PD-L1 and PD-L2, remain incompletely understood. While previous work focused on PD-L1 on tumor cells, there has been a recent appreciation of PD-L1 on conventional dendritic cells (cDCs), which sustain the stem-like CD8⁺ T cells in tumor-draining lymph nodes (tdLNs) that fuel treatment responses. PD-1 also engages a second, higher-affinity ligand, PD-L2, whose in vivo function has remained poorly defined. In this study, the authors ask whether cDC-derived PD-L1 and PD-L2 are redundant, and delineate the spatial bounds of their function.

Using single-cell RNA sequencing of dendritic cells from B16-OVA melanomas, tdLNs, and non-tdLNs, the authors showed that Pdcd1lg2(PD-L2) was among the most strongly induced genes in tumor-associated CCR7⁺ cDC1s and cDC2s, whereas Cd274 (PD-L1) was expressed in both tumor and lymph node. The same restriction of PDCD1LG2 to CCR7⁺ cDCs held in human melanoma. Using a new conditional Pdcd1lg2allele crossed to Clec9a-cre, the authors then deleted either ligand from cDCs and followed adoptively transferred OVA-specific OT-I cells. Loss of PD-L1 expanded stem-like and progenitor exhausted (Tpex) OT-I cells in tdLNs and raised their IL-7R expression, but did not increase intratumoral effector responses, cytokine production or tumor control. Loss of PD-L2 did the reverse, where the tdLN progenitor pool was unchanged, while intratumoral OT-I cells shifted from Tpex towards granzyme B⁺, IFNγ⁺ and TNF⁺ effector/exhausted states, and B16-OVA and ID8 tumors were controlled. The effect was abolished with PD-1-deficient OT-I cells, and RGMb (PD-L2’s alternative receptor) was undetectable on CD8⁺ T cells, placing the signal on PD-1. Only tumor models rich in CCR7⁺ cDC1s (B16-OVA, ID8, but not MC38-OVA or EO771-OVA) responded, and diphtheria-toxin depletion of intratumoral PD-L2⁺CCR7⁺ cDC1s alone restrained growth.

To understand how PD-L2 could be regulating intratumoral CD8⁺ T cell differentiation, the authors conducted spatial transcriptomics (10X Xenium, Visium HD) of non-tdLNs, tdLNs, and tumors from B16-OVA bearing mice with and without OT-I transfer. Spatial transcriptomics showed that CCR7⁺ cDC1 neighbourhoods were enriched for Tpex but not effector/exhausted cells, with both enriched in Ccl19 hotspots where CCL19-high cancer-associated fibroblasts were the dominant chemokine source. These findings pointed towards a potential retention mechanism, given that CCR7 upregulation licenses cDC egress to the tdLN. These findings are recapitulated in human melanoma, which showed LAMP3⁺cDCs adjacent to TCF1⁺CD8⁺ T cells. Finally, GM-CSF drove PD-L2 induction on CCR7⁺ cDCs, suggesting a local feedback loop.

Limitations

· While the evidence for CCR7⁺ cDC1s as the principal mediators is very persuasive, the study does not fully isolate that subset. Clec9a-cre deletes completely in cDC1s but also in 50–70% of cDC2s. A comparison using Xcr1-cre, or sorting CCR7⁺ cDC1s and cDC2s and testing them separately would be useful. Furthermore, while the authors do control for Clec9a deficiency itself using Clec9acre/crelittermates, it would be useful to exclude a Clec9a-cre-specific effect by investigating whether the phenotype holds a pan-cDC driver, such as with Zbtb46-cre.

· All tumors here were implanted subcutaneously. Since the model depends on CCL19⁺cancer-associated fibroblasts creating a spatial niche, it would be informative to know whether that architecture forms the same way in tissue-appropriate stroma. An orthotopic ID8 model, or an autochthonous melanoma, would test whether the CAF, CCR7⁺ cDC1, and Tpex axis is a general feature or partly a property of the subcutaneous site.

· The study’s claim that PD-L1 and PD-L2 act in distinct compartments rests on the contrasting steady-state phenotypes and the spatial data, but the site of action is inferred rather than demonstrated. Do the additional effector cells in PD-L2ΔDC tumors differentiate in situ, or arrive from the draining lymph node? Blocking egress after the Tpex population is established would distinguish these possibilities and would considerably strengthen the spatial argument. Since the therapeutic prediction (using CCR7⁺cDC1 abundance to determine whether PD-L1- or PD-L2-directed blockade is effective) is the most clinically consequential claim, the authors could perform antibody blockade in models beyond B16-OVA, with a range of CCR7+ cDC expression.

· Additional human data would further strengthen this study. While the immunofluorescence shows the localization of LAMP3⁺ DCs near TCF1⁺CD8⁺ T cells, PD-L2 is not stained. Furthermore, due to the limitations of TCGA deconvolution (bulk RNA sequencing), it is not possible to resolve cDC1 from cDC2, or annotate patients who received immunotherapy.

· The authors link CCR7+ cDC abundance and worse survival in TCGA, but CCR7+ is used here as a proxy for PD-L2+. The authors’ own data show that only a fraction of CCR7+ cDC1s are PD-L2+, ranging from ~20% to ~80% across lymph nodes. Relating survival to PDCD1LG2 expression, or to a PD-L2+CCR7+ cDC signature, would be more accurate.

· Lastly, the study does not confirm the source of GM-CSF. Conditional Csf2deletion in T cells or NK cells would be needed to show whether Tpex/NK-derived GM-CSF is what induces PD-L2.

Significance/Novelty

In this study, the authors reframe the two PD-1 ligands as spatially divided checkpoints. While PD-L1 governs the progenitor reservoir in the tdLN, PD-L2 mediates the Tpex-to-effector transition inside the tumor. The study also extends CCR7⁺/mregDC biology beyond priming to the active control of T cell state transitions, and supplies a spatial circuit involving CCL19⁺CAF-mediated retention and GM-CSF-driven PD-L2 induction to explain why these cells become suppressive in some tumors and not in others.

For therapy, the work offers a rationale for a choice between anti-PD-1 versus anti-PD-L1, and nominates CCR7⁺ cDC1 (and PD-L2) abundance as a candidate biomarker, with the potential that sparing PD-L2 blockade where these cells are absent might reduce toxicity

Credit

Reviewed by Bhavya Singh as part of a cross-institutional journal club between the Icahn School of Medicine at Mount Sinai, the University of Oxford, MD Anderson, and the University of Toronto.

The author declares no conflict of interests in relation to their involvement in the review.