C57BL/6 mice were from Charles River/NCI. Nur77-GFP reporter (C57BL/6-Tg(Nr4a1-EGFP/cre)820^Khog/J) (37), OT-I transgenic (C57BL/6-Tg(TcraTcrb)1100^Mjb/J) (38), Thy1.1 congenic mice (B6.PL-Thy1^a/CyJ) (39), Rag1^ko mice (B6.Rag1^em10Lutzy) and Perforin^ko (C57BL/6-Prf1tm1Sdz/J) (40) mice, all from Jackson Laboratories, were bred and maintained in a pathogen-free facility at the University of Virginia. Six to 12-week-old Nur77-GFP x (OT-I x Thy1.1) F1 mice were the source of mice expressing OT-I⁺Thy1.1⁺Nur77-GFP⁺ cells used for adoptive transfer. All procedures were approved by the University of Virginia Animal Care and Use Committee in accordance with the NIH Guide for Care and Use of Laboratory Animals.

B16-F1 cells transfected to express cytoplasmic ovalbumin (ova) have been described (41). Ova-transfected B16-F1 (B16-ova) were cultured in RPMI-1640 (Corning) supplemented with 5% FBS (Sigma), 15 mM HEPES and 2 mM L-glutamine (both from Gibco). Blasticidin (10μg/ml) (Gibco) was added to maintain ova expression in B16-ova. Cells were authenticated by visual confirmation of melanin pigment production in vitro and in vivo, and OVA expression confirmed by staining with the H-2K^b+ova peptide specific antibody 25-D1.16. Where indicated, B16-ova cells were transduced to express the fluorescent protein IRFP720 (42). All cultured cells injected into mice were within 2–8 passages after thaw and mycoplasma free.

B16-ova cells (4×10⁵) in 200 μL phosphate-buffered saline (PBS) were injected subcutaneously (SC) in the neck scruff. Where indicated B16-ova cells were transduced with IRFP720 fluorescent protein to enable tumor cell identification prior to implantation. Mice were monitored for weight loss, signs of distress and tumor size every 2–3 days. Where indicated, mice were injected intraperitoneally (IP) daily with 5 μg/mL FTY720 (Novartis) or saline control for indicated periods of time during tumor growth. Where indicated mice were treated IP with 250 μg Brefeldin A (Sigma) 4–6 hours prior to harvest. At the time of harvest mice were euthanized and tumor, TDLN, NDLN, and/or spleen were collected and processed as outlined below.

B16-ova tumors from C57BL/6 or Perforin^ko mice were harvested in RPMI-1640 (Corning) supplemented with 2% FBS, 0.05 mM β-mercaptoethanol, 40 μg/mL DNase (all from Sigma), 15 mM HEPES, 2 mM L-glutamine, 10 mM sodium pyruvate, 1X essential and non-essential amino acids, gentamicin (1 μg/mL) (all from Gibco), and liberase™ (76 μg/mL) (Roche). Tumors were digested for 15 min at 37°C, manually homogenized, and filtered through 70 μm mesh (Miltenyi) to prepare single cell suspensions. The CD45⁺ fraction was enriched with CD45 MicroBeads mouse (Miltenyi) using an AutoMACS instrument and analyzed by flow cytometry. Fractions with CD45⁺ cells removed were stained for CD31⁺ cells in selected experiments or a second enrichment was performed using CD31 MicroBeads (Miltenyi).

B16-ova tumors from C57BL/6 mice were harvested and CD8⁺ cells isolated using MACS beads (Miltenyi) as described above. Cells were stimulated with CD3/CD28 T activator beads (10 μL/mL, Gibco) for 12 hours in the presence of anti-CD107a antibody (AF488, Biolegend) to mark degranulating cells. Brefeldin A (10 μg/mL) was added during the last 4-6 hours of culture to block secretion of intracellular cytokines.

LN were manually shredded using needles, and spleens were cut up manually with scissors. Tissue suspensions were digested using liberase™ (76 μg/mL) for 30 min at 37°C and then manually homogenized and filtered through 70 μm mesh (Miltenyi) prior to staining and analysis by flow cytometry.

Single-cell suspensions of splenocytes from naïve Nur77-GFP⁺OT-I⁺Thy1.1⁺ mice were prepared smashing spleens between frosted glass slides and filtering cells through a 100 μm filter. Red blood cells were lysed using RBC lysis buffer (Invitrogen). Cells were washed twice, resuspended in PBS, and immediately injected IV (5x10⁴) into Thy1.2⁺ C57BL/6 mice. The next day, recipient mice were implanted SC with B16-ova cells. Seven days post tumor implantation spleen, TDLN, and tumor were isolated and single cell suspensions prepared as described above.

Starting 2 days prior to tumor implantation NK cells were depleted by injection of 100 μg anti-NK1.1 (BioXcell) or isotype control (BioXcell) antibody IP into mice. Injections were repeated every 3 days until tumors were harvested. Depletion was confirmed in all mice used for analysis by flow cytometry (example tumor depletion shown in Supplementary Figures 2A–E ).

Starting on D3 post tumor implantation 200 μg of anti-PD-1 blocking antibody (BioXcell) or isotype control (BioXcell) was injected IP into tumor bearing mice with or without NK cell depletion. Injections were repeated every 3 days until tumors were harvested.

Mice were injected IP on D3 and D5 post tumor injection with 250 μg anti-CD40L blocking antibody (BioXcell) or an isotype control antibody (BioXcell) in 200 μl of PBS.

Cells were Fc blocked with 1:1000 anti-CD16/CD32 (2.4G2, BioXcell) and stained with Live/Dead Fixable Aqua (Life Technologies) in PBS for 20 min at 4°C. Subsequently cells were stained with fluorescently labeled antibodies in PBS supplemented with 2% FBS and 0.1% sodium azide for 30 min at 4°C. For extracellular stains only, cells were fixed in 2% paraformaldehyde (Thermo Scientific) for 10 min at 4°C. For intracellular and intranuclear stains, BD Cytofix/Cytoperm and BD Transcription Factor Staining kits were used according to manufacturer’s protocol. Data was acquired on Cytoflex (Beckman Coulter) or Attune (BD Biosciences) flow cytometers and analyzed using FlowJo software. In experiments that analyzed Nur77 expression, cells were directly analyzed without fixation. OT-I CD3⁺ CD8⁺ T cells were subsequently gated on Thy1.1⁺ before assessing Nur77 experiments. Normalized MFI values were generated on positive cells by dividing all values generated that day by the average of the isotype control group.

Following extracellular staining, cells were stained with Annexin V-APC in Annexin buffer (Biolegend). Cells were immediately run on the cytometer while still in the buffer.

Single cell suspensions from tumors or LN were incubated with SIINFEKL-Dextramer or irrelevant dextramer control (Immudex) for 1 h at 37°C. Cells were subsequently stained for other surface or intracellular markers as described above.

Data is displayed as mean with error bars representing SEM. Groups were compared using a one-way ANOVA with Tukey’s multiple comparisons test or a Students T test with Welch’s correction for comparisons only involving two groups. All analysis and graphs were performed using PRISM software (Graphpad).

To examine differences in early-stage and late-stage TME, we used flow cytometry of single cell suspensions to compare subcutaneous B16-ova tumors on day 7 (D7), the earliest time at which we could consistently locate them, to day 14 (D14) tumors. While the total number of CD31⁺ tumor vascular endothelial cells was higher in D14 tumors ( Figure 1A ), the numbers per g of tumor were similar on both days ( Figure 1B ), indicating that early-stage tumors are well-vascularized. The number and percentage of CD31⁺ cells expressing ICAM-1 was unchanged between D7 and D14, but VCAM-1 and CXCL9 (gating shown in Supplementary Figure 1A ) were expressed by a higher number ( Figures 1B, C ) and percentage ( Figures 1D, E ) of these cells on D7, albeit at a similar mean fluorescence intensity (MFI) ( Figure 1F ). This suggests that effector T cells should be readily able to enter D7 tumors. Indeed, T cell infiltrates were evident, but the numbers per g of tumor were 46% lower for CD4⁺ cells and 65% lower for CD8⁺ cells on D7 than D14 ( Figure 1G ). This is consistent with the possibility that cells other than T cells might upregulate the expression of HRL, particularly VCAM-1 and CXCL9, on early-stage tumor endothelial cells.

In contrast to T cells, we found that NK cells, and CD11c⁺MHC-II⁺, CD11c^negCD11b⁺MHCII⁺, and CD11c^negCD11b⁺MHCII^neg myeloid cells (gating shown in Supplementary Figure 1B ), were present at similar levels per g of tumor at all stages of tumor growth ( Figures 1H, I ). Innate immune cells can express inflammatory cytokines that upregulate HRL expression, and our previous work implicated IFNγ in controlling expression of both VCAM-1 and CXCL9 (9). To determine if the intratumoral innate cells produced IFNγ, mice were treated with Brefelden A (BFA) prior to tumor harvest. While no myeloid subpopulations in D7 tumors expressed IFNγ directly ex vivo ( Figure 1J ), NK cells in D7 tumors did do so, although this was lost by D14 ( Figure 1K , gating shown in Supplementary Figure 1C ). We found that B16-ova cells did not express any NK cell ligands ( Figure 1L ). However, small fractions of B16-ova cells in D7 tumors expressed the NK activating NKG2D ligands Rae1 and H60 and some also upregulated MHCI, an inhibitory ligand for NK cells ( Figure 1L ). Together these data suggest that the production of IFNγ by NK cells in early-stage tumors is driven by B16-ova tumor cells. Overall, this suggests that IFNγ produced by NK cells increases HRL expression on CD31⁺ tumor vasculature and promotes T cell infiltration in early-stage tumors.

To test this hypothesis, we depleted NK cells by injection of anti-NK1.1 prior to tumor implantation and maintained depletion with additional injections for the duration of the experiment ( Supplementary Figures 2A–E ). The fractions of CD31⁺ cells that expressed ICAM-1 and VCAM-1 were unchanged by NK cell depletion ( Figure 2A ), but the expression levels on positive cells were modestly but significantly lower ( Figure 2B ). Thus, NK cells augment HRL expression on CD31⁺ early-stage tumor vasculature but are not the only cells responsible. However, despite this reduced HRL expression, the numbers of CD8⁺ T cells per g of tumor in NK-depleted D7 tumors were significantly increased, and those of CD4⁺ T cells were trending toward an increase ( Figure 2C ), while APC and myeloid populations were unchanged ( Figure 2D ). This was not due to changes in tumor weight as NK cell depletion did not significantly alter tumor weights ( Figure 2E ). Approximately 80% of CD8⁺ T cells in D7 tumors were antigen experienced (CD44⁺) and consisted of CD62L^neg effector cells and CD62L⁺ cells that had either not yet downregulated CD62L after activation or were central memory ( Figure 2F , gating shown in Supplementary Figure 2F ). CD44^neg CD62L⁺ naïve cells and CD44^neg CD62L^neg early activated cells were present at lower levels. The distribution of these intratumoral CD8⁺ T cell subpopulations was unchanged by NK depletion ( Figure 2F ). This demonstrates that NK cells, despite increasing HRL expression on D7 tumor vasculature, mediate an unexpected reduction in intratumoral CD8⁺ T cells.

To test the functionality of the antigen-experienced CD8⁺ T cells in NK-depleted D7 tumors, we examined cytokine production after in vivo BFA blockade. In tumors containing NK cells, ~10% of CD8⁺ T cells produced IFNγ, while only ~5% in tumors from NK-depleted mice did so ( Figure 3A gating shown in Supplementary Figure 3A ). However, the numbers of IFNγ producing cells per g were the same in both tumors, and there were more than twice as many cells per g not producing IFNγ in NK-depleted tumors ( Figure 3B ). The majority of IFNγ⁺ cells were CD44⁺CD62L^neg and were similar per g tumor in isotype and NK depleted mice ( Figure 3C ). A small portion of IFNγ⁺ cells were CD44⁺CD62L⁺ and these were increased in NK depleted tumors ( Figure 3C ). The increase in IFNγ^neg CD8⁺ T cells per g in NK-depleted tumors was statistically significant in all subpopulations, but the majority of the increase was in CD44⁺ antigen-experienced CD8⁺ T cells ( Figure 3D ). To determine if this effector deficiency was due to the TME or intrinsic to the T cells, we isolated CD8⁺ T cells from D7 tumors and re-stimulated them ex vivo with anti-CD3 and anti-CD28. All T cells became CD44⁺ following this restimulation. However, a smaller percentage of restimulated CD8⁺ T cells from tumors of NK-depleted mice produced IFNγ compared to restimulated CD8⁺ T cells from tumors of normal mice, and this was evident only in the CD62L^neg population ( Figure 3E ). The percentage of cells showing degranulation marked by CD107a was also decreased in restimulated CD62L^neg T cells from tumors of NK-depleted mice ( Figure 3E , gating shown in Supplementary Figure 3B ), demonstrating that the effector function deficiency is generalized. Thus, depletion of NK cells selectively increases the accumulation in early-stage tumors of antigen-experienced CD8⁺ T cells that intrinsically lack at least one effector function.

One common mechanism leading to intrinsic T cell dysfunction in tumors is exhaustion, characterized by increased expression of inhibitory molecules, including PD-1. However, the percentages and numbers per g of PD-1⁺ CD8⁺ T cells in early-stage tumors of mice containing or lacking NK cells were not significantly different ( Figure 3F ). To determine whether PD-1 was blocking T cells that might have otherwise expressed IFNγ, we treated mice containing or lacking NK cells with anti-PD-1 starting on D3 after tumor implantation and harvested on D7. There was no significant increase in total CD8⁺ T cells or IFNγ⁺ CD8⁺ T cells per g of tumor as a consequence of anti-PD-1 treatment in mice containing or lacking NK cells ( Figures 3G, H ). These results suggest that the increase in non-functional CD8⁺ T cells in tumors from NK depleted mice is not due to PD-1 mediated exhaustion.

Recent studies have identified an exhausted stem-cell-like T cell progenitor population with a SLAMF6^HiTCF7⁺ phenotype (43). Cells with this phenotype comprised ~13% of the total CD8⁺ TIL ( Figure 3I , gating shown in Supplementary Figure 3C ). However, there was no difference in the percentage or numbers of SLAMF6^HiTCF7⁺ CD8⁺ T cells in D7 tumors from mice containing or lacking NK cells ( Figure 3I ). This population was also not increased following anti-PD-1 treatment. These results suggest that NK depletion does not influence the development of SLAMF6^HiTCF7⁺ T cell progenitors, and that the population that is increased in D7 tumors of NK-depleted mice is therefore non-progenitor.

Anergic CD4⁺ and CD8⁺ T cells show reduced responses to TCR engagement and lack resulting effector functions (44–46). While T cell effector cytokine production and proliferation are often linked, anergic T cells continue to proliferate (47–49). To evaluate whether the non-functional CD8⁺ T cells in NK depleted tumors were anergic, we first examined their ability to proliferate within the tumor. Using Ki67, there were no changes in the proliferating fractions of CD8⁺ T cells overall, or of CD8⁺ T cell subpopulations, in early-stage tumors from mice containing or lacking NK cells ( Figure 4A , gating shown in Supplementary Figure 4A ). To determine whether sensitivity to TCR signals was reduced in CD8⁺ T cells in tumors from NK-depleted mice, we transferred congenically marked ovalbumin specific OT-I CD8⁺ T cells that expressed Nur77-GFP prior to tumor implantation. Nur77 expression is directly tied to TCR signaling (37). The percentage of OT-I T cells that expressed Nur77-GFP was decreased in tumors from NK-depleted mice, and this was most evident in the CD44⁺CD62L^neg effector population ( Figure 4B , gating shown in Supplementary Figure 4B ). The normalized Nur77-GFP MFI of Nur77⁺ OT-I cells in tumors from NK-depleted mice was also reduced ( Figure 4C ). These results indicate that intratumoral CD8⁺ T cells activated in mice lacking NK cells had reduced sensitivity to TCR signals, but similar levels of proliferation, consistent with an anergic phenotype.

Since CD8⁺ T cells in tumors from mice containing or lacking NK cells proliferated similarly, this cannot explain their increased numbers in tumors from NK-depleted mice. Anergic T cells have also been shown to have increased resistance to apoptosis (46). Using a viability marker (Live/Dead Aqua), we observed a significant decrease in the percentage of dead CD8⁺ T cells overall, and of antigen experienced CD44⁺CD62L⁺ and CD44⁺CD62L^neg subpopulations, in NK-depleted tumors ( Figure 4D ). We also observed a significant decrease in the percentage of Annexin V^Hi (late apoptosis) CD44⁺CD62L^neg CD8⁺ T cells ( Figure 4E , gating shown in Supplementary Figure 4C ). We considered that reduced apoptosis of intratumoral T cells in tumors lacking NK cells might be reflective of direct NK cell killing, which requires perforin (35). To test this hypothesis, WT and Perforin^ko mice, with and without NK depletion, were implanted with B16-ova, and tumors analyzed on D7. CD8⁺ T cell numbers in tumors from WT and Perforin^ko mice containing NK cells were comparable, as were the numbers in tumors from NK-depleted mice of both genotypes ( Figure 4F ). Thus, the increase in CD8⁺ T cells in tumors from NK-depleted mice is not due to a lack of direct NK cell killing. These results suggest that the increased level of CD8⁺ T cells in D7 tumors from NK-depleted mice is due to a reduction in apoptotic death in the antigen experienced population, consistent with an anergic phenotype.

To determine if T cell dysfunction continued in late-stage tumors, we harvested D14 tumors from mice that were NK cell depleted either prior to tumor implantation (D-2) in the same manner as our early-stage tumors, or after initial T cell priming (D7). NK depletion on D7 did not result in any changes in the number of CD8⁺ T cell in D14 tumors ( Figure 4G ). However, in D14 tumors from mice depleted of NK cells prior to tumor implantation we observed a significant reduction in CD8⁺ T cell numbers overall compared to mice with intact NK cells ( Figure 4G ). This contrasts with the elevated numbers of CD8⁺ T cells observed in D7 tumors from NK depleted mice. We did not find a significant reduction in the numbers per g of tumor of either IFNγ^neg or IFNγ⁺ cells in NK depleted mice, although there was a trend toward a reduction for IFNγ⁺ cells although this data showed significant variation ( Supplementary Figures 4D, E ). However, there was a significant reduction in the percentage of IFNγ⁺ CD8⁺ T cells in late-stage tumors lacking NK cells ( Figure 4H ) consistent with what was observed in D7 tumors. It is notable that these changes were accompanied by an overall significant increase in CD8⁺ T cell numbers per g of tumor from D7 to D14 in both NK depleted ( Supplementary Figure 4F ) and non-depleted ( Figure 1G ) mice. While depletion of NK cells prior to tumor implantation had no effect on the size of D7 tumors ( Figure 2E ), it did result in a significant increase in the size of D14 tumors ( Figure 4I ). However, there was no change in D14 tumor size when NK cells were depleted starting on D7 ( Figure 4I ). We utilized RAG mice lacking T cells to determine if this effect was directly mediated by NK cells. There was no increase in late-stage tumor size in RAG mice depleted of NK cells prior to tumor implantation ( Figure 4J ), suggesting that the impact of NK cell depletion on tumor control in WT B6 mice is mediated by its effect on adaptive immunity. This indicates that NK cell depletion reduces the functionality of intra-tumoral CD8⁺ T cells in both early and late stage tumors, and that this results in reduced long-term tumor control.

T cell anergy normally develops during initial activation from an imbalance among signals via the TCR, costimulatory molecules, and cytokines (46). To determine whether NK depletion altered activation of CD8⁺ T cells in the TDLN, we treated mice starting on day 3 (D3) post implantation with FTY720 to prevent T cell egress. The overall number of CD8⁺ T cells in the TDLN or non-draining LN (NDLN) was unchanged by NK depletion ( Supplementary Figures 5A, B ). However, the percentage ( Figures 5A, B ) and number ( Supplementary Figures 5C, D ) of CD8⁺CD44⁺CD62L^neg effector T cells were reduced specifically in the TDLN and not the NDLN. The percentage of Ki67⁺ proliferating CD8⁺ T cells was reduced in the TDLN in all populations except the naïve CD44^negCD62L⁺ cells ( Figure 5C ). The percentage of endogenous ovalbumin-specific CD8⁺ T cells, representing tumor antigen specific T cells, among all CD8⁺ T cells, was also reduced in TDLN but not NDLN of NK-depleted mice ( Figure 5D , Supplementary Figure 5E ). This was associated with decreased proliferation of ovalbumin specific CD8⁺ T cells overall ( Figure 5E ), a reduced number of the CD44⁺CD62L⁺ subpopulation ( Figure 5F ), and an increased percentage of naïve (CD44^negCD62L⁺) ovalbumin specific CD8⁺ T cells ( Supplementary Figure 5F ). We did not observe a change in the subpopulation distribution of ovalbumin specific T cells in the NDLN, where naïve cells were the primary population ( Supplementary Figure 5G ). These results demonstrate that NK depletion leads to a reduction in T cell activation, proliferation, and tumor specific effector generation in the TDLN.

The development of anergic T cells has been shown to be promoted by elevated TCR stimulation and/or a lack of co-stimulation (46, 50–53). We therefore tested the hypothesis that the reduced number of proliferating and differentiated effector T cells in the TDLN of NK-depleted mice reflected increased sensitivity to TCR stimulation. We tested this by examining the activation of OT-I T cells expressing the Nur77-GFP reporter that were transferred prior to tumor implantation. OT-I T cells did not show a significant difference in subpopulation distribution in TDLN of isotype and NK-depleted mice ( Figure 5G ). Importantly however, the fraction of Nur77⁺ OT-I T cells ( Figure 5H ) and their Nur77 MFI ( Figure 5I ), were significantly higher, indicating that they had received enhanced TCR signals. Nur77 signaling was specific to the TDLN, as the Nur77 positive signal was nearly absent in the spleens of both isotype and NK depleted mice ( Supplementary Figure 5H ). Taken together, these results demonstrate that NK depletion leads to enhanced TCR stimulation of CD8⁺ T cells in TDLN, decreased proliferation, and differentiation of those cells. This is consistent with what others have shown to promote the development of anergic T cells.

Based on the changes in CD8⁺ T cell phenotype that accompanied NK depletion, we examined the phenotypes of APC in both tumor and TDLN. The numbers of intratumoral APC ( Figure 2D ) and their subsets ( Figure 6A ) were not altered by NK depletion. While the numbers of CD11b⁺CD11c^negMHCII⁺ APC in TDLN of NK-depleted mice were not reduced, the numbers of CD11c⁺MHCII⁺ APC were, and this was evident in both CD11b⁺ and CD103⁺ subsets ( Figure 6B ). This was specific for the TDLN, as NDLN APC numbers were unchanged ( Supplementary Figure 6A ). The lower numbers of these APC subsets in TDLN could be due to reduced entry of migratory APC, or reduced survival or proliferation of resident APC. Cells with an CD11c^HiMHCII^Int phenotype, characteristic of the majority of LN resident APC (54), were selectively reduced in the TDLN of NK-depleted mice, while cells with a CD11c^LoMHCII^Hi phenotype, characteristic of migratory DC, were not altered ( Figure 6C , Supplementary Figure 6B ). This is consistent with the unchanged number of intratumoral APC and suggests that APC migration from the tumor remains unchanged in NK depleted mice. This suggests that one consequence of NK depletion is a reduction in the numbers of CD11b⁺ and CD103⁺ subsets of resident DC in TDLN.

We also examined the maturation status of APC in tumor and TDLN. All subsets of intratumoral APC from NK depleted mice expressed lower levels of MHCII ( Figure 6D ), suggesting that they are less mature than those from non-depleted mice. As immature APCs in tissues are more phagocytic and have reduced proteolytic capacity (55, 56) we used tumors expressing the fluorescent protein iRFP720 to identify APC that retained unprocessed tumor derived antigen. All subsets of intratumoral APC from NK depleted mice showed significantly greater retention of tumor derived antigen ( Figure 6E ), again consistent with a less mature phenotype. The percentage of intratumoral APC expressing CD80 ( Figure 6F ) and their level of CD80 expression ( Supplementary Figure 6C ), which typically increase with APC maturation (55, 56) was unchanged by NK depletion. Overall, this suggests that intratumoral APC from NK depleted mice are less mature.

In APC from TDLN of either NK-depleted or non-depleted mice retention of tumor derived antigen could not be detected. However, compared to APC in NDLN, CD80 was expressed on approximately 25-100% more cells in most CD11c⁺ APC subsets in TDLN but was not different on CD11b⁺CD11c^negMHCII⁺ cells or resident APC ( Figure 6G ). Thus, APC in TDLN appear more mature compared to those in NDLN. Neither the percentage of APC expressing CD80 ( Figure 6H ), nor their CD80 expression level ( Supplementary Figure 6D ) was changed in any subsets in TDLN of mice lacking NK cells. However, MHCII expression levels were increased in all populations of CD11c⁺ APC (CD11b⁺, CD103⁺, resident, and migratory; gating shown in Supplementary Figure 6E ) and in CD11c^negCD11b⁺MHCII⁺ cells ( Figure 6I ). In TDLN of intact mice, CD11c⁺MHCII⁺ APC that expressed CD80 (gating shown in Supplementary Figure 6F ) also expressed higher levels of MHCII ( Figure 6J ). However, in TDLN from NK depleted mice, both CD80⁺ and CD80^neg APC showed an increase in MHCII MFI, and they were no longer different from one another ( Figure 6J ). Thus, APC in early stage TDLN of NK depleted mice display increased MHCII expression regardless of CD80 expression. Increased MHCII expression in mature APC usually corresponds with an increased level of costimulatory molecule expression (57). Increased MHCII expression without increased CD80 suggests an incomplete APC maturation, which may have resulted in the aberrant activation and anergic T cell phenotypes we observed in tumor bearing mice lacking NK cells.

NK cells in tumor bearing mice express elevated levels of CD40L (58), and interactions with CD40 cause increased APC maturation (59). CD40L blockade also results in reduced T cell numbers and function in tumor bearing mice (60). We found that some NK cells, as well as some CD4⁺ T cells, in early-stage tumors and TDLN expressed CD40L ( Supplementary Figure 7 ). We utilized CD40L blockade to determine if this resulted in an increase in dysfunctional T cells in early-stage tumors similar to NK depletion and resulted in similar effects on APC and T cell phenotypes in the TDLN. In contrast to NK depletion, CD40L blockade beginning on D3 led to almost complete absence of T cells in early-stage tumors ( Figure 7A ), making it impossible to evaluate changes in T cell function or phenotype. However, in the TDLN, CD40L blockade resulted in a reduction in the numbers of CD8⁺ T cells overall, naïve (CD44^negCD62L⁺), and effector (CD44⁺CD62L^neg) T cells, but no change in CD44⁺CD62L⁺ CD8⁺ T cells ( Figure 7B ), and similar results were seen with either NK depletion alone or combination NK depletion and CD40L blockade. While the observed reduction in overall and naïve CD8⁺ T cell numbers was not seen in the TDLN of NK depleted mice in Figure 5A , this is likely because the mice in Figure 7 were not treated with FTY720. We augmented these studies by comparing the effects of CD40L blockade and NK depletion on tumor antigen specific cells. As above, the total number of ova-dextramer⁺ CD8⁺ T cells in the TDLN was reduced comparably by NK cell depletion, CD40L blockade, and the combination ( Figure 7C ). This was primarily due to a reduction in the number of antigen-experienced (CD44⁺) ova-dextramer⁺ CD8⁺ T cells ( Figure 7C ). As we observed with bulk T cells there was no significant difference in the reduced number of effector cells in the TDLN of CD40L blocked, NK depleted, or dual treated mice. These results suggest that NK cell depletion and CD40L blockade target a common pathway that results in diminished T cell differentiation and that the effect of NK cells on T cell differentiation in the TDLN during early-stage tumor growth is via CD40L. However, CD40L blockade has effects beyond those of NK cell depletion as there is complete blockage of T cell infiltration into the early-stage tumors. This suggests that NK cells provide an important source of CD40L during the initial stages of CD8⁺ T cell activation but are not the only source of CD40L in the microenvironment.

We next determined if CD40L blockade also resulted in incomplete APC maturation in the TDLN similar to that induced by NK depletion. In keeping with Figures 6A, B , CD40L blockade resulted in reduced numbers of CD11b⁺, CD103⁺, and resident APC, but unchanged migratory APC, in the early-stage TDLN, and was not significantly different from NK cell depletion alone or combination treatment ( Figures 7D, E ). In keeping with Figure 6D and Supplementary Figure 6C , neither the percentage ( Figure 7F ) or expression level ( Figure 7G ) of CD80 were changed on CD11c⁺MHCII⁺ APC as a consequence of any treatment. Finally, in keeping with Figures 6E, F , and Supplementary Figures 6D, E , MHCII expression on all APC subsets Figures 7H, I , regardless of CD80 expression was increased similarly by CD40L blockade, NK cell depletion, or the combination ( Figures 7J, K ). This suggests that CD40L blockade during early-stage tumor growth promotes incomplete maturation of APC in TDLN, similar to NK depletion.