The Hippo Pathway Component TAZ Promotes Immune Evasion in Human Cancer through PD-L1

Almost all neoplasms show some degree of immune cell presence (1). In the tumor microenvironment, cancer cells interact with immune cells often restraining their antineoplastic roles and enhancing tumor-promoting functions. Cancer cell–intrinsic processes (i.e., aberrant expression of oncogenes and tumor suppressors) are critical for this polarization of the immune response. Indeed, dysregulated signaling pathways within malignant cells can upregulate genes that lead to immune evasion such as PD-L1 (CD274).

PD-L1 is a key protein that governs the interaction between tumor-infiltrating T lymphocytes and cancer cells. As an immune checkpoint molecule, PD-L1 on cancer cells binds to its receptor, PD-1, on T cells to suppress their function. Specifically, PD-1/PD-L1 binding inhibits T-cell activation/proliferation/IL2 production, promotes anergy/exhaustion, and initiates T-cell apoptosis (2). Blockade of PD-1 or PD-L1 with mAbs reverses many of these phenomena and can restore T-cell function. Thus, the PD-1/PD-L1 axis has gained recognition as an exciting therapeutic target for treating multiple cancer types. While PD-1 and PD-L1 blocking agents have shown tremendous success in treating melanoma, bladder cancer and non–small cell lung cancer, response rates to these therapies are low in breast cancer clinical trials (3–6). Therefore, there is an urgent need to develop a better understanding of the immunobiology of breast cancer to establish methods for stratifying patients who will respond well to immunotherapy. Furthermore, given that there is great interest in exploiting PD-L1 as a biomarker, it is vital that we fully comprehend the cellular networks affecting its expression across varying cancer types.

Over the past decade, the Hippo signaling pathway has emerged as a central player in regulating many aspects of tumor biology (7–9). When the Hippo pathway is activated by upstream signals, MST1/2 kinases (homologs of Drosophila “Hippo”) phosphorylate and activate LATS1/2 kinases (10–12). LATS1/2 subsequently phosphorylate transcriptional coactivators TAZ and YAP at key serine residues (S89 and S127, respectively) to inhibit their translocation into the nucleus, interaction with the TEAD family of transcription factors and activation of downstream genes (e.g., CTGF, CYR61; refs. 13–15). TAZ is an oncogene in breast and lung cancer where alterations in its activity have been associated with chemotherapy resistance and metastasis (16–18). There is early evidence that the Hippo pathway can influence immune cell recruitment and activation as well as the anticancer immune response (19–22). However, there has been no systematic search for TAZ and YAP transcriptional targets that act on immune cells during development and disease. Indeed, whether dysregulated Hippo signaling affects the immune system in the context of breast cancer is unknown. Finally, it remains unclear whether the relationship between Hippo signaling and the immune system is conserved between human cancers and the model organisms that are commonly used to study these diseases.

In our study, we have performed the first comprehensive screen for immune-related transcriptional targets of TAZ and YAP using NanoString gene expression profiling. We identified 71 genes that are transcriptionally regulated by TAZ or YAP overexpression that are relevant to immunology. We have further validated PD-L1 as a bona fide transcriptional target of the Hippo signaling pathway in human cells. Surprisingly, we observed that the relationship between TAZ and PD-L1 is not conserved in multiple mouse cell lines due to differences between the human and mouse PD-L1 promoter sequences. Therefore, we provide evidence that the Hippo pathway plays critical roles in modulating the immune system and directing human cancer immune evasion.

To induce TAZ, YAP, or LATS overexpression in our inducible cell lines, cells were treated with 1 μg/mL doxycycline (Dox) for 48 hours. For some experiments, cells were treated as follows: 5 nmol/L 12-O-tetradecanoylphorbol-13-acetate (TPA) 1 hour, 200 ng/mL EGF 30 minutes, 4 μg/mL insulin 30 minutes, 1 μmol/L sphingosine-1-phosphate (S1P) 30 minutes, 10 μmol/L forskolin + 100 μmol/L 3-isobutyl-1-methylxanthine (IBMX) 1 hour, 2 μmol/L glucagon 1 hour, 100 nmol/L wortmannin 1 hour, 10 μmol/L GSK2334470 4 hours, serum starvation 4 hours, 10 μmol/L LY3009120 4 hours, 100 nmol/L dasatanib 24 hours, 1 μmol/L fluvastatin 24 hours, 10 μmol/L pazopanib 24 hours, 10 μmol/L rottlerin 18 hours. Culture media for cell lines are listed in the Supplementary Data. Cell lines were purchased from ATCC. The identity of our cell lines has not been recently authenticated or Mycoplasma tested. All experiments were conducted using cells with passage number less than 40.

Site-directed mutagenesis was performed using overlapping PCR. Mouse TAZ and TEAD4 cDNAs were synthesized by reverse transcription from E10 cells. For transient gene expression, cDNAs were cloned into pCDNA3.1. For inducible overexpression, cDNAs were cloned into a puromycin-resistant modified pTRIPZ. Stable overexpression constructs were created in a HA-tagged, hygromycin-resistant modified WPI. Methods for lentivirus production/infection are as described previously (15). To create the human PD-L1 promoter reporter, nucleotides −221 to +21 were PCR-amplified from HeLa cell gDNA and cloned into pGL3-basic. The mouse Pd-l1 promoter (nucleotides −1723 to +220) was constructed using gDNA extracted from C57BL/6 mice. Deletions in the promoter reporters were made by overlapping PCR. See Supplementary Data for primers used in cloning.

RNA was collected from each condition in biological replicates using RNAzol RT reagent (Sigma) and was cleared of residual gDNA using the RNeasy Mini Kit (Qiagen). Three-hundred nanograms of RNA was used per NanoString reaction using the nCounter Human Immunology v2 Gene Expression Panel or the nCounter Mouse Immunology v1 Gene Expression Panel. nSolver 3.0 software was used for background subtraction, normalization, and data analysis.

Total RNA collection and quantitative (q)RT-PCR protocols using SYBR Green reagents are as previously described (15). Primers are in Supplementary Data.

Methods for Western blot analysis are as described in ref. 15. Antibodies for Western blot analysis were as follows: PD-L1 (E1L3N), PD-L2 (D7U8C), YAP/TAZ (D24E4), MST1 (D8B9Q), MST2, LATS1 (C66B5), and phospho-YAP (S127) from Cell Signaling Technology; YAP (H-125) from Santa Cruz Biotechnology; TAZ (M2-616) from BD Biosciences; TEAD1-4 (EPR15629) from Abcam; LATS2 (BL2213) from Bethyl Laboratories; β-actin (AC-15) from Sigma. Flow cytometry was performed according to a standard protocol from Abcam using a PE-conjugated antibody for human PD-L1 (MIH1) from eBioscience (isotype control mouse IgG1 (B11/6) from Abcam) or a PE-conjugated antibody for mouse PD-L1 (MIH5) from eBioscience (isotype control rat IgG2a from eBioscience).

siRNAs were transfected into cells using Lipofectamine RNAiMAX reagent (Invitrogen) at a final concentration of 50 nmol/L and according to the manufacturer's instructions. MST1/2 were knocked down using siRNA from the TriFECTa RNAi kit from IDT. See Supplementary Data for other siRNA sequences. Knockdown efficiency was determined 48 hours after transfection by Western blot analysis.

CRISPR–Cas9 constructs for TAZ knockout were generated using the lentiCRISPRv1 vector (23, 24). Clonal cell lines were established by limiting dilution plating. CRISPR-Cas9–resistant addback TAZ constructs were created by mutating the PAM sequence in the TAZ cDNA by overlapping PCR. sgRNA sequences and primers for the addback mutations are listed in the Supplementary Data. Addback constructs were cloned into WPI.

One-hundred nanograms of promoter reporter was cotransfected into SK-BR-3 using PolyJet transfection reagent (SignaGen) alongside 200 ng of TAZ/YAP construct, 200 ng of TEAD construct, and 10 ng of Renilla luciferase pRL-TK plasmid as an internal control. Total DNA was adjusted to 510 ng/well using pCDNA3.1. After 48 hours, dual luciferase assay was performed using a kit from Promega. Relative promoter activity was calculated as the ratio of Firefly luciferase signal to Renilla luciferase signal. All measurements were normalized to promoter reporter alone.

Chromatin immunoprecipitation (ChIP) was performed using the SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads) from Cell Signaling Technology. MCF10A-TAZ-S89A cells were induced with doxycycline for 72 hours. Ten micrograms of chromatin was incubated overnight with 2 μg antibody for TAZ (M2-616 from BD Biosciences) or with 2 μg normal mouse IgG (Santa Cruz Biotechnology). The PD-L1, CTGF, and CYR61 promoters were PCR-amplified using primers in the Supplementary Data. Total chromatin extract (“input”) was used as a positive control for PCR.

Cancer cell TAZ-S89A overexpression was induced with doxycycline for 48 hours. Jurkat T cells were activated with 1 μg/mL phytohaemagglutinin (PHA) and 50 ng/mL phorbol 12-myristate 13-acetate (PMA) for 48 hours to induce PD-1 expression. Following doxycycline treatment, 5 × 10⁵ cancer cells were plated in Jurkat cell media into each well of a 12-well plate. The following day, 5 × 10⁴ activated T cells were added to each well of the 12-well plate. After 24 hours (MCF10A coculture) or 5 hours of coculture (A549, H1299 coculture), T cells were collected. T-cell apoptosis was measured using the Caspase-Glo 3/7 assay from Promega. For PD-L1 inhibition, H1299 cells were plated in Jurkat cell media containing 20 μg/mL purified anti-PD-L1 antibody (29E.2A3 from BioLegend). T-cell coculture was performed the following day in the presence of 20 μg/mL anti-PD-L1 antibody.

A549 TAZ-S89A overexpression was induced with doxycycline for 48 hours while Jurkat T cells were stimulated with 1 μg/mL PHA and 50 ng/mL PMA for 24 hours to initiate IL2 and PD-1 expression. A total of 1 × 10⁵ A549 cells were plated with 1 × 10⁴ activated T cells in 100 μL T-cell media with PMA/PHA in a 96-well plate. The following day, coculture supernatants were collected and IL2 levels were measured by ELISA using the IL-2 Human ELISA kit from Invitrogen. For PD-L1 inhibition, cells were cocultured in the presence of 20 μg/mL purified anti-PD-L1 antibody (29E.2A3 from BioLegend).

Student t test (two-tailed) and ANOVA (with post hoc analysis) were used for all statistical analysis.

To identify immune-related transcriptional targets of TAZ and YAP, we subjected RNA from MCF10A breast epithelial cell lines overexpressing doxycycline-inducible, constitutively active TAZ (TAZ-S89A), or YAP (YAP-S127A) to NanoString profiling using the nCounter Human Immunology v2 panel of 579 immunology-related genes (Fig. 1A). Genes regulated by TAZ or YAP were defined as having at least a 2-fold, statistically significant change in expression after TAZ-S89A or YAP-S127A induction (P < 0.1). By this criterion, 25 genes were upregulated by TAZ-S89A, while 34 genes were downregulated. A further 17 genes were upregulated by YAP-S127A, whereas 19 genes were downregulated. Top candidate TAZ-S89A- and YAP-S127A–regulated genes are heatmapped in Fig. 1B. Many genes were commonly regulated by TAZ-S89A and YAP-S127A (e.g., S1PR1, NLRP3, PD-L1/CD274; Fig. 1C). The fold-change regulations are depicted in Fig. 1D and E. The top TAZ-S89A- and YAP-S127A–upregulated genes were validated by qRT-PCR and were further confirmed to be upregulated by TAZ-S89A/YAP-S127A overexpression in a mesenchymal-like breast cancer cell line, MDA-MB-231 (Fig. 1F; Supplementary Fig. S1A and S1B). In addition, as a control, we verified that none of the top candidates identified by our screen were regulated by doxycycline treatment in wild-type MCF10A (Supplementary Fig. S1C). Therefore, TAZ and YAP modulate the expression of many genes that are relevant to immunology and the tumor immune microenvironment.

Given the well-established role of PD-L1 in modulating the interaction between cancer and immune cells, we chose this gene for further study. Consistent with our NanoString data, TAZ-S89A and YAP-S127A induced PD-L1 expression at the protein level in MCF10A (Fig. 2A). PD-L2 (PDCD1LG2), a paralog of PD-L1 and a candidate YAP-regulated gene identified by our screen, was also induced at the protein level by YAP-S127A overexpression. TAZ-S89A–induced PD-L1 expression was apparent by flow cytometry in nonpermeabilized cells, indicating that TAZ-S89A enhances PD-L1 expression at the plasma membrane (Fig. 2B). The relationship between TAZ-S89A overexpression and PD-L1 upregulation was also observable in human immortalized lung/bronchus epithelial cells (HBE-135; Fig. 2C).

We next evaluated whether upstream Hippo pathway components regulate PD-L1 through TAZ and YAP. Transient knockdown of MST1/2 or LATS1/2 in wild-type MCF10A suppressed inhibitory YAP-S127 phosphorylation, enhanced CYR61 mRNA expression, and increased PD-L1 expression at the protein and mRNA level (Fig. 2D–G; Supplementary Fig. S2A and S2B). Furthermore, LATS2 overexpression in MDA-MB-231 (high endogenous PD-L1) increased YAP-S127 phosphorylation and reduced PD-L1 expression (Fig. 2H). Therefore, the Hippo pathway regulates PD-L1.

Many upstream signaling pathways converge on Hippo signaling (Fig. 2I). We next explored whether our finding that the Hippo pathway affects PD-L1 expression might provide insights into other stimuli that can regulate PD-L1. We exposed MCF10A (low TAZ) and MDA-MB-231 (high TAZ) to culture conditions that inhibit and activate Hippo signaling, respectively (25–31). We identified four treatments [PKC activator (TPA), EGF, insulin and GPCR inhibitor (S1P); Fig. 2J; Supplementary Fig. S2C] that induced both PD-L1 and CYR61 mRNA expression in MCF10A and six treatments [PKA activator (Forskolin/IBMX), glucagon, PI3K inhibitor (Wortmannin), PDK inhibitor (GSK2334470), serum starvation, RAF inhibitor (LY3009120); Fig. 2K; Supplementary Fig. S2D] that repressed PD-L1 mRNA expression in a similar pattern to CYR61 in MDA-MB-231. Thus, the Hippo pathway may link PD-L1 expression to various other signaling networks and stimuli.

We next set out to determine whether TAZ and PD-L1 are coexpressed in cancer cell lines. Indeed, the expression levels of TAZ are correlated with those of PD-L1 in multiple breast cell lines and a similar (albeit weaker) relationship was observed in lung cell lines (Fig. 3A and B). This was consistent with data from The Cancer Genome Atlas (TCGA) database indicating that TAZ and PD-L1 mRNA expression cooccurs in multiple cancer types including breast invasive carcinomas and lung adenocarcinomas (Supplementary Fig. S3A). Interestingly, relatively fewer datasets from TCGA demonstrated an association between YAP and PD-L1 mRNA expression (Supplementary Fig. S3B).

To test whether TAZ is required for high PD-L1 expression in certain cancer cells, we knocked out TAZ in Hs578T and MDA-MB-231 using CRISPR-Cas9. TAZ knockout decreased PD-L1 protein in both cell lines (Fig. 3C and D). Clonal cell lines derived from the heterogeneous TAZ knockout MDA-MB-231 cell lines also showed reduced PD-L1 expression compared with control cells (Fig. 3E). PD-L1 expression in TAZ-knockout MDA-MB-231 cell lines was rescued by addback of CRISPR-Cas9–resistant TAZ constructs (Fig. 3F and G). We further confirmed that PD-L1 expression in wild-type MDA-MB-231 depends on TAZ activity by pharmacologically inhibiting TAZ/YAP function (32, 33). Inhibition of TAZ and YAP reduced PD-L1 mRNA expression to a greater extent than CTGF (Fig. 3H). Therefore, PD-L1 expression in cancer cell lines is determined by the activity of TAZ.

TAZ and YAP regulate gene expression by binding to transcription factors such as the TEAD family (i.e., TEAD1–4; refs. 34, 35). To establish whether TEADs play a role in the regulation of PD-L1 by TAZ, we overexpressed a constitutively active, TEAD-binding mutant form of TAZ (TAZ-S89A-F52/53A) in MCF10A and found that this construct had a diminished ability to induce PD-L1 (Fig. 4A and B). Similarly, knockdown of TEAD1/3/4 in MCF10A-TAZ-S89A reduced TAZ-induced PD-L1 expression (Fig. 4C). Finally, addback of TEAD-binding mutant TAZ (TAZ-F52/53A) into TAZ-knockout MDA-MB-231 cells could not restore PD-L1 expression (Fig. 4D). Therefore, TAZ regulates PD-L1 expression primarily through TEAD transcription factors.

To further characterize the transcriptional regulation of PD-L1 by TAZ, we constructed a luciferase-based reporter for the minimal PD-L1 promoter (nucleotides −221 to +21). We cotransfected this reporter into SK-BR-3 breast cancer cells alongside TAZ-S89A and TEAD1-4. TAZ-S89A and TEAD1-4 coexpression dramatically increased PD-L1 promoter activity (Fig. 4E). TEAD-binding mutant TAZ (TAZ-F52/53A), TAZ without its C-terminal transcriptional coactivation domain (TAZ-Δ227), and TAZ-binding mutant TEAD4 (TEAD4-Y429H) all could not activate the PD-L1 promoter (Fig. 4F). Like TAZ, a similar increase in PD-L1 promoter activity was induced by YAP-S127A and TEAD4 (Fig. 4G).

We performed a deletion scan to determine the PD-L1 promoter region regulated by TAZ-S89A and TEAD4. Deletion of nucleotides −100 to −40 abolished activation by TAZ-S89A and TEAD4 (Fig. 4H). Within this region, we identified a putative TEAD-response element spanning positions −74 to −62 (CAGGAAAGTCCAA) (Fig. 4I). Deletion of this region dramatically reduced the activation of the PD-L1 promoter by TAZ-S89A and TEAD4 (Fig. 4J). We confirmed that TAZ binds at the PD-L1 promoter by ChIP. Specifically, we found that fragments from the proximal PD-L1 promoter (i.e., positions −221 to +21 and −183 to +58), the CTGF promoter and the CYR61 promoter precipitated with TAZ whereas a more distal fragment from the PD-L1 promoter (positions −1039 to −808) did not (Fig. 4K). Therefore, TAZ regulates PD-L1 expression by binding to the PD-L1 promoter through the TEAD family of transcription factors thereby enhancing promoter activity.

To explore the functional significance of TAZ in modulating the anticancer immune response, we cocultured TAZ-overexpressing cancer cell lines with Jurkat T cells. Jurkat T cells were activated with PMA and PHA prior to coculture to induce PD-1 expression (Fig. 5A). TAZ-S89A overexpression in MCF10A was sufficient to increase apoptosis in Jurkat T cells cocultured with these cells (Fig. 5B and C). A similar increase in apoptosis was observed in T cells cocultured with TAZ-S89A–overexpressing A549 or H1299 lung cancer cells (Fig. 5D–G). To evaluate whether TAZ enhances T-cell apoptosis through PD-L1, we cocultured TAZ-S89A–overexpressing H1299 with Jurkat T cells in the presence of a PD-L1 blocking antibody. Indeed, inhibition of PD-L1/PD-1 binding completely suppressed TAZ-S89A–induced T-cell apoptosis (Fig. 5H). Thus, cancer cell TAZ expression regulates T-cell viability through PD-L1.

As a second measure of T-cell function, we measured IL2 production by PMA/PHA–activated Jurkat T cells in coculture with A549. In this system, TAZ-S89A overexpression in A549 was sufficient to suppress IL2 production by T cells and this was reversed when cells were cocultured in the presence of a PD-L1 blocking antibody (Fig. 5I). Collectively, these data are consistent with a model in which TAZ/YAP/TEAD and the upstream Hippo pathway direct cancer immune evasion through the transcriptional regulation of PD-L1 (Fig. 5J).

Many targets of TAZ and YAP are similarly regulated in human and mouse cell lines (e.g., CTGF). However, it has been suggested that there is substantial divergence in the transcriptional programs that act on the human and mouse immune systems (36, 37). To determine whether TAZ upregulates PD-L1 in murine cell lines, we established TAZ-S89A–overexpressing mouse cell lines [three mammary cell lines (HC11, NMuMG, E0771), one lung cell line (E10) and one melanoma cell line (B16-OVA); Fig. 6A)]. While Ctgf was upregulated by TAZ in each of these cell lines, TAZ-S89A overexpression caused no change in Pd-l1 mRNA expression in any of the cell lines examined (Fig. 6B). As a control, Pd-l1 mRNA was easily detectable in 293T cells overexpressing mouse Pd-l1 cDNA using the same primers for qRT-PCR (Fig. 6C). Likewise, human TAZ-S89A, mouse TAZ-S89A, or human YAP-S127A overexpression all had no effect on PD-L1 protein levels (Fig. 6D–F; Supplementary Fig. S4A and S4B). Thus, PD-L1 appears to be differentially regulated by TAZ in human and mouse cells.

To investigate the mechanisms underlying this distinction we constructed a luciferase reporter for the mouse Pd-l1 promoter (nucleotides −1723 to +220). While the human PD-L1 promoter was activated equally by both human and mouse TAZ-S89A/TEAD4, the mouse Pd-l1 promoter had a dramatically diminished response to both human and mouse constructs (Fig. 6G). Therefore, the relationship between TAZ and PD-L1 is not conserved in mouse cells and this is likely due to regulatory differences between the human and mouse PD-L1 promoters. It is notable that other top candidate TAZ-regulated genes from our profiling in human cells were also not upregulated by human or mouse TAZ-S89A in HC11 or NMuMG (e.g., S1pr1, Nlrp3; Fig. 6H and I; Supplementary Fig. S4C and S4D).

Our observation that multiple candidate TAZ targets identified by our human screen were not upregulated in mouse cell lines suggests that there may be broader species-specific differences in the TAZ transcriptional program that have been unrealized in previous work. Given the importance of murine models for studies of immunology, we set out to explore the extent of divergence between human and mouse TAZ immune-related gene targets. We subjected RNA from HC11-TAZ-S89A and NMuMG-TAZ-S89A to NanoString gene expression profiling using the nCounter Mouse Immunology panel (Fig. 7A). Twenty-eight genes were upregulated by human TAZ-S89A in these cell lines while 56 genes were downregulated by TAZ-S89A (Fig. 7B–D). We observed substantial differences in the gene targets regulated by TAZ in human and mouse cells. Only a minority (14/83) of the candidate TAZ targets identified by our screen in mouse cells were also candidates from our human screen (e.g., Pdgfb, Il12a, Cd14; Fig. 7E). We used qRT-PCR to validate the top upregulated candidate TAZ targets from our screen and confirmed that none of these genes were significantly affected by doxycycline treatment in wild-type cells (Fig. 7F; Supplementary Fig. S5A and S5B). We also explored whether these genes were affected by mouse TAZ-S89A overexpression (in HC11 and NMuMG) or human TAZ-S89A overexpression (in MCF10A). Indeed, both human and mouse TAZ-S89A upregulated the expression of multiple genes identified by our screen (Fig. 7F and G). Furthermore, several of the top candidates from our screen in mouse cells including Tigit, Ptpn22, Masp1, and Il7 were uniquely upregulated by TAZ in mouse cell lines but not in human cells (Fig. 7F–H). Therefore, we have uncovered multiple genes that may be differentially regulated by TAZ between these two species.

The Hippo signaling pathway plays critical roles in cancer development and progression. In cancer cells, dysregulation of the Hippo transducers TAZ and YAP leads to aberrant expression of their downstream gene targets and endows cells with numerous “hallmarks of cancer” (9). In our study, we performed the first comprehensive screen for immune-related transcriptional targets of TAZ and YAP using NanoString profiling. In doing so, we uncovered a novel function for TAZ in promoting cancer immune evasion through the transcriptional regulation of PD-L1.

PD-L1 is a critical mediator in the interaction between effector T lymphocytes in the tumor microenvironment and tumor cells. Indeed, the PD-1/PD-L1 axis is a key target for immunotherapies. In our study, we showed that both TAZ and YAP transcriptionally regulate PD-L1 in human cancer cells. We characterized the molecular mechanisms by which TAZ enhances PD-L1 expression by binding to the PD-L1 promoter through the TEAD family of transcription factors. Most significantly, we demonstrated that these observations have functional importance in coculture experiments where TAZ overexpression in cancer cells was sufficient to disrupt T-cell function through PD-L1.

These findings provide new insights into how PD-L1 is regulated in breast cancer. PD-L1 is expressed by a significant portion of triple-negative breast cancers (38). However, whether high tumor cell PD-L1 levels predict a favorable or unfavorable prognosis for breast cancer patients is unclear and likely depends on the circumstances surrounding PD-L1 upregulation (39, 40). PD-L1 may be expressed due to tumor cell–intrinsic processes (e.g., oncogenic signaling through PI3K, STAT3, HIF-1α, or TAZ) but may also occur secondary to a robust anticancer immune response (e.g., through IFNγ signaling and NF-κB; ref. 41). In each of these scenarios, PD-L1 expression likely reflects different disease pathology. Thus, knowledge of tumor PD-L1 status on its own may be insufficient as a biomarker for predicting patient prognosis and response to immunotherapy. A better understanding of the mechanisms regulating PD-L1 expression may allow us to more accurately interpret what PD-L1 positivity means and may reveal superior biomarkers and therapeutic targets for cancer treatment. Furthermore, given that ongoing clinical trials applying anti-PD-1 and anti-PD-L1 therapies to breast cancer patients have shown only limited success, there is an urgent need to develop tests that can stratify patients for immunotherapy (3–5). Our findings suggest that future work exploring tumor TAZ status as a prognostic and predictive factor for cancer immunotherapy may be warranted. In addition, as we have shown that pharmacologic inactivation of TAZ/YAP significantly inhibited PD-L1 expression, our data also highlight the therapeutic potential of targeting the Hippo pathway for cancer treatment either as a monotherapy or in combination with PD-L1–targeted immunotherapies.

It should also be noted that PD-L1 plays multiple roles in immune evasion and cancer biology outside of its effects on T-cell function. For example, PD-1/PD-L1 binding regulates tumor cell phagocytosis by tumor-associated macrophages and several groups have proposed that PD-L1 has tumor cell–intrinsic functions (42–44). Therefore, it is possible that TAZ regulates the immune system more broadly through PD-L1 or that signaling downstream of PD-L1 contributes to the oncogenic potential of TAZ. These areas represent interesting directions for future work.

While characterizing PD-L1 as a transcriptional target of TAZ, we were surprised to find that the relationship between PD-L1 and TAZ was not apparent in mouse cell lines. This was unexpected, as many TAZ and YAP transcriptional targets are conserved between these two species including CTGF (35). Moreover, both human and mouse PD-L1 genes are regulated by other stimuli such as hypoxia (45). However, immune-related transcriptional programs, and regulatory sequences in general, can differ dramatically between human and mice and an alignment of the human and mouse PD-L1 promoters shows that there are notable differences between these sequences. Thus, our data reveal important species-specific differences in PD-L1 regulation that were previously unrealized in the literature. To the best of our knowledge, this is the first (and only) description of a mechanism of PD-L1 regulation that occurs in human cells but not in mice. This finding could have tremendous significance for future studies and underscores the need to choose appropriate model systems when studying tumor immunology. This conclusion may also help reconcile notable discrepancies between earlier publications that argue that the Hippo pathway both promotes and antagonizes cancer immune evasion. Our study, and another recent study in lung cancer, supports the conclusion that human TAZ suppresses antineoplastic immune responses (46). However, Moroishi and colleagues have observed that loss of LATS1/2 in fact inhibits immune evasion through activation of TAZ and YAP in three different mouse syngeneic tumor models (20). It is possible that species-specific transcriptional regulation by TAZ underlies this distinction.

A number of recent publications have tied the Hippo pathway to other immune-related phenomena including cerebral ischemia reperfusion injury, neurodegeneration, and post-myocardial infarction cardiac remodeling (47–49). The targets of TAZ and YAP revealed by our screen may also contribute to our understanding of how the Hippo pathway influences each of these processes. TAZ and YAP induced dramatic changes in expression for many immune-related genes regulating diverse phenomena ranging from inflammasome formation to complement. Interestingly, many of the top candidate TAZ- and YAP-regulated genes we identified are involved in normal immune cell maturation and differentiation (e.g., IL7R; ref. 50). Future work will be necessary to validate these candidates as bona fide transcriptional targets of TAZ and YAP, to determine whether these genes are regulated in a species-specific manner and to appreciate the functional significance of these relationships in development and disease.

In summary, in this study we have implicated the Hippo pathway and TAZ as key players in directing cancer immune evasion and have discovered differences in the transcriptional regulation of human and mouse PD-L1. These findings offer compelling evidence that the Hippo pathway and its effectors TAZ and YAP are important regulators of the immune response and present exciting opportunities for future studies.

No potential conflicts of interest were disclosed.

Conception and design: H.J. Janse van Rensburg, T. Azad, C.H. Graham, X. Yang

Development of methodology: H.J. Janse van Rensburg, T. Azad, Y. Hao, B. Snetsinger, P. Khanal, C.H. Graham, M.J. Rauh, X. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.J. Janse van Rensburg, T. Azad, M. Ling, Y. Hao, B. Snetsinger, P. Khanal, L.M. Minassian

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.J. Janse van Rensburg, T. Azad, B. Snetsinger

Writing, review, and/or revision of the manuscript: H.J. Janse van Rensburg, T. Azad, P. Khanal, C.H. Graham, M.J. Rauh, X. Yang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.J. Janse van Rensburg, Y. Hao, B. Snetsinger

Study supervision: X. Yang

This work was supported by grants from the Canadian Institute of Health Research (#119325, 148629 to X. Yang), the Canadian Breast Cancer Foundation/Canadian Cancer Society (to X. Yang), and the Canada Foundation for Innovation (to M.J. Rauh). H.J. Janse van Rensburg is supported by a Canada Graduate Scholarship and Queen Elizabeth II Graduate Scholarship in Science and Technology. We would like to thank Canadian Institute of Health Research, the Canadian Breast Cancer Foundation/Canadian Cancer Society, and the Canada Foundation for Innovation for their financial support. We would like to thank Shannyn MacDonald-Goodfellow for her technical expertise and Scott Gerber for providing us with the B16-OVA cell line.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.