PD-L1 signaling selectively regulates T cell lymphatic transendothelial migration

Programmed death-1 (PD-1) and its ligand PD-L1 are checkpoint molecules which regulate immune responses. Little is known about their functions in T cell migration and there are contradictory data about their roles in regulatory T cell (Treg) function. Here we show activated Tregs and CD4 effector T cells (Teffs) use PD-1/PD-L1 and CD80/PD-L1, respectively, to regulate transendothelial migration across lymphatic endothelial cells (LECs). Antibody blockade of Treg PD-1, Teff CD80 (the alternative ligand for PD-L1), or LEC PD-L1 impairs Treg or Teff migration in vitro and in vivo. PD-1/PD-L1 signals through PI3K/Akt and ERK to regulate zipper junctional VE-cadherin, and through NFκB-p65 to up-regulate VCAM-1 expression on LECs. CD80/PD-L1 signaling up-regulates VCAM-1 through ERK and NFκB-p65. PD-1 and CD80 blockade reduces tumor egress of PD-1high fragile Tregs and Teffs into draining lymph nodes, respectively, and promotes tumor regression. These data provide roles for PD-L1 in cell migration and immune regulation.

PD-L1 is constitutively expressed on LECs, and is increased in inflamed tissues or the tumor microenvironment (TME) 22,23 . In lymph nodes (LNs), PD-L1 is most highly expressed by LECs compared to other stromal cells such as blood endothelial cells or fibroblastic reticular cells 24 . Since LECs are important for TEM of lymphocytes, these observations raise the possibility that LEC PD-L1 not only directly regulates lymphocyte activation at sites of inflammation but may also regulate other functions such as migration.
In the present study we found activated Tregs expressed the highest level of PD-1 among different T cell subsets. This prompted us to investigate if Treg PD-1 interacted with LEC PD-L1 to regulate other aspects of Treg function. Our investigations uncovered another role for the PD-1-PD-L1 interaction in Treg migration across LECs and into the draining lymphatics.
PD-L1 was highly expressed on both mouse and human LECs, while minimal PD-1 was detected on the surface of the LECs (Fig. 1c). Immunohistochemistry and flow cytometry analysis of permeabilized LECs showed intracellular PD-1 expression while PD-L1 was predominantly on the cell surface (Fig. 1d). To determine PD-L1 expression on LEC in physiological conditions, fresh primary dermal LECs (Lyve-1 + CD31 + ) were isolated and found to have comparable levels of PD-L1 expression as the cultured primary LECs (Fig. 1c). Minimal levels of PD-L1 or PD-1 were detected on Lyve-1 − CD31 − non-endothelial cells.
Tregs and Teffs differentially engage PD-L1 on LECs for TEM. We previously validated our model of lymphatic TEM, using primary mouse LEC monolayers grown on the membrane of a Boyden chamber and measuring the migration of various leukocyte subsets from the basal (abluminal) to the apical (luminal) side of the endothelial cells in response to a chemokine gradient 25,27 . Using this assay, blocking PD-1 with anti-PD-1 mAb (Rmp1-14) on mouse iTregs or tTregs inhibited TEM across LECs toward CCL19, a potent T cell chemokine, in a dosedependent fashion ( Supplementary Fig. 3a, b). In contrast, blocking PD-L1 on Tregs with anti-PD-L1 mAb (10 F.9G2) did not inhibit TEM (Fig. 2a). Blocking PD-1 or PD-L1 on Teffs or naïve CD4 did not inhibit TEM (Fig. 2b, Supplementary Fig. 3c, d). No inhibition of Treg TEM was observed when they migrated across the plastic membrane of the Boyden chamber without LECs (Fig. 2a, b), suggesting the interaction of Treg PD-1 with PD-L1 on LECs was essential for regulating migration. To test the possible effect of antibody-FcR interactions or antibody crosslinking on the blockade, we generated (Fab')2 and Fab fragments. Both F(ab')2 and Fab anti-PD-1 inhibited Treg TEM (Fig. 2c), showing that blockade was not FcR dependent or due to crosslinking of PD-1. PD-1 −/− iTregs had reduced migration and motility compared to wild-type (Fig. 2d, e), while PD-L1 −/− iTregs had enhanced motility (Fig. 2e), suggesting Treg PD-L1 may constitutively engage with PD-1 in cis to prevent PD-1 from interacting with LEC PD-L1. These data show that both genetic and pharmacologic inhibition resulted in similar effects. Importantly, PD-1 blockade of activated human naïve Tregs but not Teffs with anti-human PD-1 mAb (EH12.2H7) also inhibited TEM across human LECs (Fig. 2f), demonstrating similarities between murine and human migration mechanisms. Anti-PD-1 mAbs also inhibited migration to another T cell chemoattractant, sphingosine 1-phosphate (S1P) ( Supplementary Fig. 3e), showing that TEM blockade was not limited to one chemokine.
CD80, which was more elevated on Teffs, is another ligand for PD-L1. To test whether these cells used the CD80/PD-L1 for TEM, we pretreated murine Teffs or iTregs with anti-CD80 (1G10) to block CD80/PD-L1. Masking CD80 inhibited Teff but not iTreg TEM (Fig. 2g) in a dose-dependent manner (Fig. 2h). Stimulation of naive CD8 T cells and B cells also increased PD-1 expression ( Supplementary Fig. 2a, b), however, blocking PD-1, CD80, or PD-L1 on activated CD8 T cells or B cells did not affect cell migration ( Supplementary Fig. 3f, g). Similarly, these mAbs did not block TEM of matured BMDCs ( Supplementary Fig. 3h), suggesting PD-1/PD-L1 signaling did not favor CD8 T cell, B cell, or BMDC TEM, but was used uniquely by Tregs for TEM.
Differential signaling in iTregs and Teffs by PD-L1 Ig engagement. We investigated signaling in T cells by PD-L1 engagement during murine iTreg or Teff migration. Mouse PD-L1 Ig (Immunoglobulin) (extracellular domain fused with human IgG1) was immobilized on the Boyden chamber membrane. Migration of iTregs but not Teffs across the PD-L1 Ig-coated (LEC-free) membrane toward CCL19 or S1P ( Supplementary  Fig. 3d) was significantly enhanced (Fig. 2i) compared to membranes coated with human IgG1, CD80 Ig or LTβR Ig as controls. Engagement by PD-L1 Ig induced rapid phosphorylation of Akt (also known as protein kinase B) on threonine 308 (Thr308) in iTregs after 10 min. PD-L1 Ig suppressed classical NFκB-p65 phosphorylation, while Akt phosphorylation was maintained (Fig. 2j). Only low and transient phosphorylation of extracellular signal-regulated kinase (ERK) was observed. In contrast to iTregs, Teffs showed strong constitutive phosphorylation of ERK (Fig. 2k) which was suppressed by PD-L1 Ig. PD-L1 Ig engagement on Teffs had no specific effect on NFκB or Akt phosphorylation. These data thus indicated different responses of iTregs versus Teffs after PD-L1 engagement. PD-1 blocking mAb (Rmp1-14) pretreated iTregs and CD80 blocking mAb (1G10) pretreated Teffs did not demonstrate specific NFκB or ERK activation, while phosphorylation of Akt (Thr308) was observed ( Supplementary Fig. 3i, j). These data suggested that the steric effects of the PD-1 or CD80 blocking mAbs on the Tregs or Teffs, respectively, were most important in blocking TEM rather than directly inducing signals that impaired TEM of either T cell subset.

Discussion
In the present study, we demonstrated that Tregs and especially activated iTregs, preferentially expressed PD-1 to ligate PD-L1 for lymphatic transendothelial migration (TEM). The direct role in migration and the preferential use by iTreg but not non-Treg also suggests that PD-1/PD-L1 signaling may be important to modulate the ratio of suppressive Tregs and reactive Teffs at inflammatory sites. PD-1 signaling in Teffs recruits SHP2 to terminate Zap70/ERK and PI3K/PKCθ and counteracts T cell receptor signal transduction and CD28 co-stimulation 16 . In contrast in activated iTregs we found that PD-1 signaling suppressed classical NFκB-p65 and induced upregulation of ERK and Akt phosphorylation, indicating alternative PD-1 signaling compared to Teffs.
iTregs expressed higher levels of PD-1 than Teffs, while Teffs expressed higher levels of PD-L1 and CD80 than Tregs. These differences were more pronounced for human Tregs and Teffs. In earlier studies 30,40 , CD80 reportedly trans-binds PD-L1 and inhibits immune responses. More recent studies suggest CD80 binds exclusively in cis to PD-L1 on APC 35,36 , promoting immunity. These differences might be due to variable spatiotemporal dynamics of surface expression and different cell types with unique functions. Differing from studies that relied on engineered tumor cell lines with gene overexpression in B cells or APCs, our model exclusively used primary LECs, which do not express CD80 or PD-1, to show a trans interaction of CD80/PD-L1. In particular, we demonstrated that: (i) CD80 Ig induced strong PD-L1-ERK signaling in wild-type but not PD-L1 −/− LECs which was non-overlapping with PD-1 Ig induced PD-L1-PI3K/AKT signaling; (ii) masking CD80 on Teffs or blocking PD-L1 on LEC with specific PD-L1 antibody (10 F.2H11), which solely blocks the PD-L1 and CD80 interaction, inhibited exclusively Teff TEM; (iii) protein binding assays indicated CD80 Ig and PD-1 Ig but not CTLA4 Ig bound to LEC PD-L1; and LEC PD-L1 co-immunoprecipitated with CD80 Ig; (iv) Teffs bound to immobilized PD-L1 Ig which was blocked by masking PD-L1 on PD-L1 Ig or CD80 on Teff. Further, the PD-L1 Ig coated on the Boyden chamber increased Teff TEM which was blocked by anti-CD80. Masking PD-L1 Ig with anti-PD-L1 also blocked Teff TEM. (v) incubating Teffs with increasing doses of CD80 Ig did not alter PD-L1 expression; and (vi) anti-PD-L1 treatment of Teffs did not affect TEM. Taken together, we propose the coexistence of both cis and trans binding of CD80 to PD-L1. It is possible that high levels Teff-PD-L1 occupy the lower levels of PD-1 in cis, freeing CD80 to engage LEC PD-L1 in trans for TEM ( Supplementary Figs. 7, 8b). In contrast, the higher expression of PD-1 enables Tregs to engage PD-L1 on LECs and overshadows the lower expression of CD80 that precludes its participation in TEM. Other activated immune cells such as CD8 T cells, B cells, or imBMDC did not use these molecules for TEM, although they also expressed various levels of them. This highlights some of the unique attributes of Treg migration, including their ability to facilitate the migration of other cells across LEC 25 (Rmp1-14), anti-CD80 (1G10), or isotype rat IgG (2A3). Scheme of tumor treatment (d). Tumor growth curve (e). 8 mice/group; *p < 0.05, Two-way ANOVA, Holms-Šidák correction for multiple comparisons. Representative dot plots (f) and frequencies (g) of Foxp3 + CD25 + CD4Tregs, Foxp3 + CD25 − CD4 Tregs, and Foxp3 − CD25 + CD4 Teffs in total CD4 T cells of tumor-infiltrating lymphocytes (TILs) and of draining LNs (dLNs) or non-draining control LNs (non-dLNs) analyzed by flow cytometry. f, g Three to four mice/group. Data representative of 3 (a-c, f, g) and 2 (e) independent experiments. Mean ± SEM. *P < 0.05 by one-way ANOVA. Source data are provided as a Source Data file.
ZO-1 and the zipper junctional VE-cadherin remained. PD-L1 deficiency also upregulated Lyve-1 expression, the receptor of hyaluronan. Changes in each of these junctional and cell surface proteins could contribute to impaired T cell TEM. PD-L1 −/− LECs and lymphatic vessels also showed altered morphology compared to WT. These observations suggested endothelial PD-L1 is required for cytoskeletal integrity during homeostasis and morphologic changes during TEM.
PD-1 blockade with anti-PD-1 mAbs is now accepted clinical immunotherapy for melanoma 43 . The efficacy has been attributed to its reinvigoration of Teff functions. Immune suppressive Tregs in TILs are considered a barrier to effective antitumor immunity and their depletion by anti-CD25 mAbs improves checkpoint blockade 44 . The effects of PD-1 blockade on intratumoral Tregs have been inconsistent in various clinical observations or murine models. PD-1 blockade reportedly decreased CD4 Tregs: Teffs ratios in TILs of a murine osteosarcoma model 45 . In contrast, an enhanced ratio was observed in squamous cell carcinomas 46 . In line with several prior reports 14,47 , we observed that PD-1 blockade decreased CD25 + Foxp3 + CD4 Tregs, but increased CD25 − Foxp3 + CD4 Tregs in TILs. It is plausible that the PD-1 blockade induced conversion of CD25 + Foxp3 + CD4 Tregs into CD25 − Foxp3 + CD4 Tregs 38 , since we observed the intratumoral transferred CD25 + Tregs with PD-1 blockade had decreased tumor egress, along with increased conversion to IFNγ-producing CD25 -Tregs. Despite the significant increase in the CD25 − Treg population in the TILs, this subset was significantly decreased in the dLNs by PD-1 blockade, consistent with inhibition of tumor egress. The CD25 − but not CD25 + Tregs in TILs expressed high levels of PD-1, which is coincident with the in vitro migration assays showing that migrated Tregs retained high PD-1 expression, while PD-1 low Tregs remained non-migrated. These results imply that the PD-1 high CD25 − Foxp3 + CD4 Tregs were targeted for migration inhibition by PD-1 blockade. Importantly, T-betassociated IFNγ production was elevated in the PD-1 high CD25 − Treg subset, which drives attenuated suppressive function and promotes antitumor immunity 39 . Human TIL PD-1 high Tregs are also reportedly converted to a dysfunctional signature and exhibit enhanced secretion of IFNγ after anti-PD-1 treatment 48 . It is important to note that non-mutually exclusive mechanisms for the accumulation of IFNγ-producing CD25 − Foxp3 + CD4 Tregs in TILs due to PD-1 blockade may not only involve Treg conversion and migration inhibition but also a proliferation of the CD25subset. However, the PD-1 high CD25 -Foxp3 + CD4 Tregs had minimal proliferation as assessed by Ki67 expression, suggesting Treg conversion and inhibition of Treg tumor egress rather than proliferation as the causes of accumulation in TILs. Whether the Treg conversion is a direct effect of anti-PD-1 on CD25 + Tregs or the PD-1 high CD25 − Tregs are first targeted for migration inhibition which is then followed by conversion remains to be determined.
The preferential migration of TILs from primary tumors to dLNs via afferent lymphatic vessels was demonstrated by using photoconvertible Kaede transgenic mice, and the majority of these migrated T cells have an effector rather than regulatory phenotype 49 , consistent with our observations here. We observed that CD80 blockade caused CD4 Teff accumulation in melanoma TILs and a concomitant decrease in dLNs, suggesting lymphatic migration was inhibited by blockade of CD80/PD-L1. CD80 blockade also promoted accumulation of T-bet + IFNγ-producing CD25 + CD4 Teffs, and reinvigorated exhausted CD8 in TILs. Reinvigorating the exhausted Teffs by blockade of the PD-1 pathway has proven efficacy in cancer therapy. In our study, PD-1 blockade not only reversed T cell exhaustion but also unleashed Teff immunity since IFNγ-producing PD-1 high CD25 − Tregs have less suppressive function 39,50 . Thus, the increased frequencies of IFNγ-producing Tregs or CD4 Teffs, and granzyme B high CD8 Teffs in TILs by anti-PD-1 and anti-CD80 treatment may all have contributed to the melanoma regression, suggesting a potential combination therapy for melanoma with PD-1 and CD80 Abs. Anti-PD-1 combined with anti-cytotoxic Tlymphocyte-associated protein 4 (CTLA4) exhibits superior antitumor efficacy compared with single-agent therapy 51 , suggesting that CTLA4 contributes to mechanisms of effector T cell exhaustion or anergy. Anti-CD80 (1G10) also blocks CD80 binding to CTLA4 30 , which is constitutively expressed on Tregs and reportedly depletes CD80/CD86 from APCs via transendocytosis 36,52 . Whether CTLA4 also depletes Teff CD80, and whether Teff CD80 and Treg CTLA4 regulate Teff reinvigoration or migration, and hence tumor regression, need further investigation. Of note, the increased IFNγ in TILs could also promote PD-L1 expression and the apoptosis of LEC, which might affect Treg or Teff tumor egress. Further investigation will be required to assess the regulation of Tregs and Teffs in TILs by PD-L1blockade. However, interpretation of results may be difficult since PD-L1 is widely expressed by these T cell subsets, by other lymphoid cells, and by non-hematopoietic endothelial cells. Overall, our study provides previously undescribed functions for PD-1 or CD80-driven PD-L1 signaling in endothelial cells for Treg or Teff migration and function. This information may improve the efficacy and strategies for therapies for autoimmune diseases and cancer.

Methods
Mice. C57BL/6 J (CD45.2 and CD45.1) (female, 7-10 weeks old) were purchased from The Jackson Laboratory (Bar Harbor, ME). C57BL/6.Foxp3GFP mice were kindly provided by Dr. A. Rudensky (Memorial Sloan Kettering Cancer Center) 53 . PD-1 −/− and PD-L1 −/− mice have been described 2 . All animal care and experiments were carried out using protocols approved and overseen by the University of Maryland IACUC in compliance with state and federal guidelines.
Primary LECs and tumor cells. Primary dermal LECs of C57BL/6 mouse (C57-6064L) or human (H-6064L) were from Cell Biologics, Inc. (Chicago, IL), and were cultured according to the manufacturer's instructions in manufacturer-provided mouse endothelial cell medium supplemented with 5% FBS, 2 mM L-glutamine, 100 IU /mL penicillin, vascular endothelial growth factor, endothelial cell growth supplement, heparin, epidermal growth factor, and hydrocortisone. Primary skin LECs were freshly isolated from the ears of wild-type C57BL/6 mice as previously described 25 . Briefly, ears were digested in 4 mg/ml collagenase D (Roche, Indianapolis, IN) at 37°C for 1 h, washed, resuspended in mouse endothelial cell medium (Cell Biologics, Inc), and plated in six-well tissue culture plates overnight. The adherent cells were harvested for flow cytometry analysis. B16F10-Fluc/eGFP was purchased from Imanis Life Sciences (Rochester, MN), and maintained in DMEM supplemented with 10% FBS, 1× Penicillin/Streptomycin, 0.8 mg/mL G418, and 1 µg/mL puromycin.
Cell viability and apoptosis assays. For viability, LECs were treated as indicated for 72 h, washed, and incubated for 3 h with 0.5 mg/mL MTT (3-(4, 5-Dimethylthiazol-2-yl)-2, 5 diphenyl tetrazolium bromide) (Sigma-Aldrich). Fifty microliters of DMSO were added to cells before reading OD at 550 nm and 690 nm. For apoptosis, LECs were treated as indicated for 36 h, washed, and stained with PE Annexin V Apoptosis Detection kit (BD Biosciences) following the instructions.
Flow cytometry. Cells were incubated with antibodies for flow cytometry for 30 min at 4°C, washed with PBS, and fixed with 4% paraformaldehyde, and run on an LSR Fortessa flow cytometer (BD Biosciences). For intracellular staining, cells were permeabilized with BD perm/fix buffer prior to incubation with antibodies. Results were analyzed with FlowJo 10.7.1 (Treestar).
Protein binding assay and co-immunoprecipitation. Plated LECs were incubated with CD80 Ig, PD-1 Ig, or CTLA4 Ig (all fused to C-terminal human IgG1) or control human IgG1 for 1 hour at 37°C. After washing, the LECs were either stained with mouse anti-human IgG1 (clone HP6069, Thermo Fisher) and rat antimouse PD-L1 (10F9G2) for immunohistochemistry, or stained with PE antihuman IgG1 for flow cytometry analysis of bound Igs. Alternatively, after washing, the LECs collected and lysed in the protein lysis buffer (see Immunoblotting method). The PD-L1-CD80 Ig immune complex was immunoprecipitated by sequential incubations and washes at 4°C with 1 μg/mL anti-PD-L1 Ab (10F9G2) overnight and 30 μL of 50% protein A/G agarose slurry (Thermo Fischer) for 4 h, and then immunoblotted with mouse anti-human IgG1 (Clone HP6069, Thermo Fisher). The same whole-cell lysates were also immune blotted with anti-PD-L1. For Teff-PD-L1 Ig binding assay, 3 × 10 5 Teffs were incubated in a 96-well plate coated with 1 μg/mL PD-L1 Ig or human IgG1 for 3 h. MTT (0.5 mg/mL) was added 2 h before the plate reading.
Immunohistochemistry. Cell monolayers or tissues were fixed for 20 min at 4°C with 4% (w/v) paraformaldehyde (Affymetrix, Santa Clara, CA), then permeabilized with PBS 0.2% (v/v) Triton X-100 (Sigma-Aldrich), and treated with 4% donkey serum for 30 min then incubated with primary antibodies for overnight at 4°C. The bound antibodies were detected with Alexa Fluor 448, 647 (Cy5), or 546 (Cy3)-conjugated secondary antibodies (Jackson ImmunoResearch, West Grove, PA) for 1 h at 4°C. The mounted slides were visualized by fluorescent microscopy (Zeiss LSM 510 Meta and LSM5 Duo). The mean fluorescence intensity (MFI) of images and the T cell PD-1 or CD80 colocalization with PD-L1 on LEC were analyzed with Volocity version 6.3 software. Quantification of the junctional VEcadherin in ×60 magnified images of LVs of whole mounted LVs or adherent LECs was performed with ImageJ. Length of zipper junctions and button junctions were measured. The percentage of zipper junction was calculated as: length of zipper junction × 100 / (length of zipper junction + length of button junction) 31 .
Statistical analysis. Numerical data are presented as mean ± SEM. Asterisks mark data statistically different from the controls, with p-values noted in the figure legends. A p-value of <0.05 was considered significant for one-way ANOVA and unpaired, two-tailed t-tests using Prizm 8 software. The number of replicates is noted in the figure legends.
Reporting summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.