Targeting Plk1 Sensitizes Pancreatic Cancer to Immune Checkpoint Therapy

Pancreatic ductal adenocarcinoma (PDAC) has the highest mortality rate of all cancers (1). PDAC formation starts with acinar-to-ductal metaplasia (ADM), the hallmark of acute pancreatitis (AP), followed by pancreatic intraepithelial neoplasia (PanIN)—the precursor lesion of most PDACs (2). In contrast, only a minority of PDACs arise from cystic neoplasms, such as an intraductal papillary mucinous neoplasm (3). The standard-of-care therapy, in addition to surgery in eligible patients, is gemcitabine; however, this treatment strategy has shown limited success (4). Interestingly, involvement of polo-like kinase 1 (Plk1) in cell-cycle progression and in the response to cellular stresses has shown that Plk1 plays an important role in aberrant cell proliferation and survival associated with tumorigenesis (5). Inhibition and knockdown of Plk1 induces apoptosis in many cancers, and it is currently a promising antitumor target (6). Importantly, a recent study suggests that Plk1 might be a mediator for gemcitabine resistance in pancreatic cancer. Among approximately 40 potential target genes, Plk1 was the only gene that could distinguish gemcitabine-sensitive versus gemcitabine-resistant tumors (7). However, a Plk1 inhibitor as a single reagent to treat PDAC showed little efficacy (8). As a result, the further investigation of additional pathways is required.

Programmed death ligand 1 (PD-L1) is a ligand for the programmed death 1 (PD-1), a costimulatory molecule that negatively regulates T-cell responses (9). Binding of PD-L1 to PD-1 suppresses T-cell activation and proliferation and induces apoptosis. In patients with PDAC, the intratumoral expression of PD-L1 correlates with a poor prognosis (10). This observation makes sense because increased PD-L1 expression by cancer cells and in the tumor microenvironment (TME) is a fundamental escape mechanism from host immunity. Therefore, blockage of T-cell checkpoint receptors with neutralizing Abs has emerged as a promising immunotherapeutic approach. Unfortunately, to date, attempts at immunotherapy in PDAC have achieved limited clinical benefits as single agents (11). This finding may be a result in part of a uniquely and more complex immunosuppressive TME, which is dominant in most human PDACs. The immunosuppressive TME is an important regulator of disease progression and is responsible for poor responses to conventional therapy. Recent studies have shown that the TME of PDAC—including cancer-associated fibroblasts (12, 13), extracellular matrix (14), immune cells (15), and cytokines (16)—is involved in the progression of PDAC through complex interactions with cancer cells.

Understanding these interactions requires a fundamental understanding of the altered pathways commonly present in PDAC. For example, the retinoblastoma protein, RB, is a tumor suppressor known to regulate the cell cycle, DNA damage response, and cellular differentiation. In addition, cyclin-dependent kinase inhibitor 2A (CDKN2A) is either mutated or its promoter methylated in 95% of PDACs (17). The CDKN2A gene products, p16/Ink4a and p14/Arf, are inhibitors of cyclin-dependent kinase 4/6 (CDK4/6; refs. 18, 19). CDK4/6 hyperactivation leads to inactivation of the RB tumor suppressor, resulting in increased proliferation (18, 20). Jin and colleagues reported the interaction between RB and NF-κB protein p65, which is primarily dependent on CDK4/6-mediated serine-249/threonine-252 (S249/T252) phosphorylation of RB. RNA sequencing (RNA-seq) analysis shows a subset of NF-κB pathway genes, including PD-L1, are selectively upregulated by RB knockdown or CDK4/6 inhibition (21).

Although many studies have shown that inhibition of Plk1 benefits tumor therapy, its role in tumor immunosuppression has not been described. Particularly, the function of Plk1 has not yet been examined in the context of mutant Kras^G12D, a common driver mutation in PDAC. In this study, we examined the effects of concomitant conditional Plk1 overexpression and Kras^G12D activation. By unbiased bioinformatics analyses, we observed dynamic changes of Plk1-associated inflammation during PDAC development. To our surprise, we observed a delayed phenotype of ADM, leading to repression of inflammation and a significantly slower progression of PanIN in mice upon Plk1 induction. Of interest, we discovered a negative association between Plk1 and PD-L1 in PDAC cells and human specimens, indicating that Plk1 regulates expression of PD-L1. Mechanistically, inhibition of Plk1 activates the NF-κB pathway in an RB-dependent manner, leading to upregulation of PD-L1. Furthermore, combination therapy with a Plk1 inhibitor significantly potentiated the efficacy of anti–PD-L1 in treating PDAC cell line–derived allograft tumors by reactivating antitumor responses. Finally, the levels of Plk1 and PD-L1 are associated with outcomes of patients with PDAC, which can be distinguished by differences in T-cell infiltration. Our data have demonstrated a tumor-suppressive role of Plk1 during the early stage of pancreatic cancer and a negative regulation of PD-L1 by Plk1 in PDAC, suggesting a new possibility to improve immunotherapy treatment of PDAC by combination with inhibition of Plk1.

In brief, to prevent chromosomal positional effects associated with the random nature of transgenic integration, we inserted the cassette into the well-characterized euchromatic locus Rosa26. Standard transfection procedures using this exchange vector were followed by recombinase-mediated integration and resulted in more than 90% positive embryonic stem (ES) cell clones. Induction of CreER activity was accomplished by administration of tamoxifen (200 μL of 20 mg/mL) via oral gavage to adult mice (6–8 weeks). Mice, housed in the animal facility with free access to standard rodent chow and water, were under pathogen-free conditions and maintained in a 12-hour light/12-hour dark cycle.

The Plk1 knock-in mouse (Plk1^LSL) was designed by the Xiaoqi Liu lab at Purdue University, and constructed by Taconic, Germany. Plk1^LSL mice and were subsequently obtained from these recombinant ES cells. Mist1^Cre/+/Kras^G12D/+ mice were generated by the Konieczny lab at Purdue University, whereas Rosa26^CreERt2 mice were purchased from the Jackson Laboratory. To generate Mist1^CreER/+/Plk1^LSL mice, we crossed the Plk1^LSL line to the Mist1^CreER/+ line. To generate Mist1^CreER/+/Kras^G12D/+/Plk1^LSL mice, we crossed the Mist1^CreER/+/Kras^G12D/+ line to the Plk1^LSL line. To generate Rosa26^CreERt2/Plk1^LSL mice, we crossed the Plk1^LSL line to the Rosa26^CreERt2 line. All procedures involving mice were approved by Purdue University Animal Care and Use Committee (Protocol no.: 1111000133E001) and University of Kentucky Animal Care and Use Committee (Protocol no.: 2020–3685).

AP was induced by caerulein via intraperitoneal injections. Adult mice (6–8 weeks) were given 8-hourly intraperitoneal injections of caerulein (Sigma-Aldrich) for 2 consecutive days (50 μg/kg body weight), with control mice receiving PBS. Mice were sacrificed and pancreata samples were harvested for paraffin blocks, following the final caerulein injection. For all analyses, 2 to 5 mice per time point/genotype/experimental condition were analyzed.

While Panc1 and HeLa cells were purchased from ATCC, HEK293T was purchased from Sigma. Mouse KPC tumor cell lines were generated from individual primary tumors of KPC (Kras^G12D/+/p53^R172H/+/Elas^CreER/+) mice. Panc1, HeLa, and HEK293T cells were cultured in DMDM (Sigma) supplemented with 10% FCS (Atlantic), 100 U/mL penicillin–streptomycin (Sigma), whereas KPC cells were maintained in RMPI1640 (Gibco) supplemented with 10% FCS (Atlantic) and 100 U/mL penicillin–streptomycin (Sigma) and L-glutamine (Sigma). IFNγ treatment was performed over a period of 24 hours at a concentration of 10 ng/mL, if not indicated otherwise. Short tandem repeat (STR) profiling, provided by ATCC was used for testing and authentication of all cell lines. All cells were within 50 passages and Mycoplasma were detected every 3 months using MycoAlert PLUS Mycoplasma Detection Kit (Lonza, LT07–705).

All RNA-seq raw counts were normalized to transcripts per kilobase million (TPM) by R17. Because small sample sizes would result in unstable gene lists and a poor prediction power, we developed a rank method for differential expression gene analysis to fit small sample size scenarios. Each gene expression profile was ranked and a one-way ANOVA analysis was used for the differential expression analysis among groups for AP, PanIN, and PDAC. With a P value of <0.001, we selected 4,097 of 18,130 significantly differentially expressed genes.

To observe the Plk1-associated functions in disease progression, differential expression analyses were performed in baseline (no caerulein injection), AP (2-day post caerulein injection), and PanIN (25-day post caerulein injection) samples of mice with or without Plk1 knock-in and PDAC. Differential expression was calculated by the Wilcoxon signed-rank test after the TPM scale normalization. The top genes of the 4,097 significantly modulated genes were used for pathway function enrichment and network analyses by the QIAGEN Ingenuity Pathway Analysis (QIAGEN IPA) software. For pathway enrichment analysis, we used the DAVID Bioinformatics Resources 6.8 tool to annotate these significant genes’ molecular functions referring to the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways annotations. Heatmaps of these genes were generated by the GENE-E (https://software.broadinstitute.org/GENE-E) online tool.

Mouse KPC tumor cell lines were generated from individual primary tumors derived from KPC (Kras^G12D/+/p53^R172H/+/Elas^CreER/+) mice. For the subcutaneous tumor studies, cells were injected into the interscapular region. For the TME study, we injected KPC cells orthotopically into the pancreas of Rosa26^CreERT2/Plk1^f/+mice, in which Plk1 was depleted upon tamoxifen injection. Tumors were measured over time, and the tumor volume was calculated as, Volume = (Length × Width²)/2.

Kinase assays were performed with TBMD buffer [50 mmol/L Tris (pH 7.5), 10 mmol/L MgCl₂, 5 mmol/L dithiothreitol, 2 mmol/L EGTA, 0.5 mmol/L sodium vanadate, 20 mmol/L p-nitrophenyl phosphate] supplemented with 125 μmol/L ATP and 10 μCi of [γ-³²P] ATP at 30°C for 30 minutes. After the mixtures were resolved by SDS-PAGE, the gels were stained with Coomassie brilliant blue, followed by autoradiography.

The simulation box size is determined to ensure that the protein does not interact with its mirror. The 3.5-nm shell of NaCl solution is used to ensure the protein is surrounded by enough aqueous solution. Table 1 lists the numbers of molecules in the three simulation boxes. This work deploys the AMBER14 force field with OL15 modifications (22) to describe bonded and nonbonded interactions in the systems. The nonbonded interactions are a sum of short-range Lennard-Jones 12–6 potential and long-range coulombic potential, as shown in Eq. A. The bonded interactions are a sum of the bond, angle, and dihedral potentials, as described in the force field.

where E_ij is the potential energy due to the nonbonded interactions between atoms i and j, r_ij is the distance between atoms i and j, ε_ij is the energetic parameter, σ_ij is the geometric parameter, and e_i is the partial charge of atom i.

A three-step simulation process is conducted for every simulation system. First, energy minimization was conducted to remove any too-close contacts between atoms. Second, a 200-ns canonical (NVT, T = 310 K) ensemble molecular dynamics (MD) simulation (integral step = 2 fs) was conducted to let the system reach thermodynamic equilibrium. Third, a 300-ns canonical (NVT, T = 300 K) ensemble MD simulation (integral step = 2 fs) was conducted to collect the trajectory at a frequency of 200 ps. The Berendsen method (23) is used to control the temperature and pressure of the system in the second step because it allows the system to reach the desired pressure and temperature at a fast pace. The velocity-rescaling method (24) is used to control the temperature of the system in the third step. The short-range van der Waal's interactions use a 1.2-nm cutoff, and the long-range electrostatic interactions were calculated using the particle mesh Ewald sum (25). All bonds involving H atoms were constrained during the simulations. The energy minimization and MD simulations for all the systems were conducted using Gromacs-2021 (26).

All-atom models were used to describe the wild-type (WT) RB protein and the mutant (RB-S758D), water molecules and ions. The initial structures of WT RB and RB-S758D were predicted by AlphaFold2 (27) using the sequences published by UniProt (P06400). The simulation systems were created by placing an RB protein or Rb-mutant protein in the center of a cubic box measuring 24 × 24 × 24 nm³ and the protein was associated with a 3.5-nm thick shell of 0.15 mol/L NaCl solution. Counter-ions were added to neutralize the system.

Inflammation scoring, immunofluorescence (IF) staining, chemokine and cytokine quantification, qRT-PCR, IHC staining, tumor mass, and flow cytometry data were analyzed by an unpaired Student t test, with Welch correction being applied when unequal variances were present between groups. The statistic of survival was calculated by the Kaplan–Meier log rank analysis. All data were presented as the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; n.s., not significant.

The data generated in this study are available upon request from the corresponding author.

More details can be found in Supplementary Materials and Methods.

In our previous study, we reported that the inhibition of Plk1 sensitized PDAC to gemcitabine, a standard chemotherapy for advanced pancreatic cancer (7). This finding renders a plausibility to investigate the efficacy of Plk1 inhibition in the treatment of PDAC. To this end, we firstly evaluated the levels of Plk1 in human PDAC tumors and found that the levels of Plk1 were gradually elevated in tumors with higher grades. The P value for Plk1 with respect to G1 tumors versus G2 tumors was 0.0017, which is statistically significant. The P value for G2 tumors versus G3 tumors was 0.0656, which implies an increasing trend with respect to Plk1 expression. The P value for G3 tumors versus G4 tumors was 0.843, indicating no significant difference, likely due to a small sample size of G4 group (n = 2) tumors. In general, the levels of Plk1 were clearly elevated in high-grade pancreatic tumors. Moreover, tumors with higher grades gained more TP53 mutations than those of tumors with lower grades (Fig. 1A). Intriguingly, a significantly higher level of Plk1 was observed in TP53-mutant patients than that in TP53 nonmutant patients, implying that Plk1 may facilitate PDAC progression in a p53-dependent manner. Interestingly, TP53-mutant tumors were most likely to be higher grade than TP53 WT tumors (Fig. 1B). To investigate the clinical significance of Plk1 in PDAC, we grouped tumors into those that were well differentiated, moderately differentiated, and poorly differentiated. As shown in Fig. 1C, a higher level of Plk1 expression was observed in patients with more poorly differentiated tumors, indicating the role of Plk1 in the aggressiveness of PDAC. To investigate the roles of Plk1 on the clinical outcomes of patients with PDAC, we used a pancreas tissue microarray with matched normal, chronic pancreatitis (CP), PanIN, and PDAC to analyze the survival outcomes in patients. In agreement with the observed properties of Plk1 in PDAC progression, patients with a higher level of Plk1 had a significantly shorter overall survival compared with those with a lower level of Plk1 (Fig. 1D). These collective data suggest the possibility of Plk1 as a therapeutic target for PDAC treatment. Indeed, a number of Plk1 inhibitors have been explored in the laboratory or in clinical studies (28). One phase II study with the single-agent Plk1-inhibitor BI2536 was completed in patients with advanced, unresectable pancreatic cancer (NCT00710710). Unfortunately, such inhibition as a single agent failed to induce tumor regression in patients with advanced cancer (8). Herein, we must investigate the reasons resulting in the failure of Plk1 inhibition so that we can create alternative strategies to improve the efficacy of treatment for PDAC. Toward this aim, we sought to explore the role that Plk1 plays in PDAC development. Although the potential that Plk1 plays in PDAC prognosis was mentioned above, the role that Plk1plays in the initiation and progression of PDAC is not fully understood.

While CP predisposes to the development of PDAC, both CP and PDAC share a common etiology. A central role has been proposed for the development of ADM in the inflammatory environment that is associated with CP. This metaplastic change is now recognized as a precursor to PDAC (29). In the current study, we explored the roles of Plk1 in the initiation and progression of pancreatitis, PanIN, and PDAC. To this end, we developed an inducible transgenic mouse line (Fig. 1E) with the specific expression of Plk1 in the pancreas by crossing Plk1^LSL mice to Mist1^CreER/+ mice to generate Mist1^CreER/+/Plk1^LSL offspring (Supplementary Fig. S1A). Plk1 expression was observed as early as 1 week after tamoxifen treatment (Supplementary Fig. S1B and S1C). As indicated by hematoxylin and eosin staining, no overt phenotype was observed in the acinar, ductal, or islet cell compartments of Mist1^CreER/+/Plk1^LSL mice (Supplementary Fig. S1D and S1E). To investigate the role of Plk1 in PanIN development, we generated Mist1^CreER/+/Kras^G12D/+ and Mist1^CreER/+/Kras^G12D/+/Plk1^LSL mouse lines.

Caerulein injection was administrated to induce AP in Mist1^CreER/+ and Mist1^CreER/+/Plk1^LSL mice 2 days posttreatment (Fig. 1F) and in PanIN with Mist1^CreER/+/Kras^G12D/+ and Mist1^CreER/+/Kras^G12D/+/Plk1^LSL mice 25 days posttreatment (Fig. 1G), respectively. To understand the mechanism of PDAC progression, an unbiased bioinformatics analysis was performed with mouse AP, PanIN, and PDAC samples (Fig. 1H). A total of 4,097 genes were identified to be differentiated among various disease statuses (Supplementary Table S1). All genes were classified into eight patterns, which were enriched in different pathways. As showed in Fig. 1I, a dynamic change of pathways was observed during the progression of PDAC in mice. Upon annotation, we found that pathways enriched in patterns 3 and 4 were involved in the process of disease development. Of note, the majority of the pathways were associated with immune response, including chemokine, cytokine, T cell, B cell, and TGFβ pathways. These collective data suggest that inflammation may be involved in the progression of PDAC. Herein, we hypothesize that Plk1 may be associated with inflammation pathways during disease development. To test this concept, we constructed an interaction network to investigate the interactions between inflammation pathways and Plk1. As shown in Fig. 1J, we performed IPA and identified the potential connections among inflammation pathways, including cytokine, chemokine, collagen, NF-κB, and TGFβ (Supplementary Fig. S2A and S2B). These connections might interact with Plk1, implying that Plk1 possibly engages in immune response during PDAC progression via the NF-κB pathway. To further study the possible interaction between Plk1 and inflammation during PDAC progression, we identified the top genes associated with Plk1 by a differential expression analysis (Supplementary Table S2), and we overlapped the genes of multiple inflammation pathways identified in patterns 3 and 4. Upon annotation, Plk1-associated pathways were classified into five new patterns in which different pathways were enriched. Of note, major pathways were clustered with the immune response and were involved in AP and PanIN (pattern 1) or PanIN and PDAC (pattern 3; Fig. 1K). This observation suggests that Plk1-associated inflammation may be involved in PDAC progression. To validate our findings, we then mined public databases of mouse pancreatitis and human PDAC samples. Upon investigation, differences in the pathways identified in Fig. 1I can be observed in the reported mouse AP (Supplementary Fig. S3A) and human PDAC samples (Supplementary Fig. S3B). In summary, Plk1 is involved in PDAC progression possibly by regulating inflammatory pathways.

ADM of the pancreas is a process by which pancreatic acinar cells differentiate into ductal-like cells. To evaluate the importance of Plk1 during ADM, we characterized Plk1 expression during the damage and subsequent recovery phases of AP (30). The major phenotype associated with caerulein-induced AP is a loss of acinar cell integrity (31, 32). To validate the findings in Fig. 1, Mist1^CreER/+ and Mist1^CreER/+/Plk1^LSL mice were used to investigate the role of Plk1 in AP. Caerulein treatment in Mist1^CreER/+/Plk1^LSL mice led to a significant and rapid damage to the exocrine acinar cells. By 2 days post-AP, significant increases in inflammatory cell infiltration were observed, accompanied by the extensive formation of ADM lesions (Supplementary Figs. S1E and S2A–S2D). Likewise, we observed significant decreases of IL1b, IL3, IL10, IL12(p70), IL17, IFNγ, TNFα, KC, IL6, IL9, MCP-1a, and G-CSF upon Plk1 induction in mice with caerulein-induced AP, suggesting that Plk1 attenuates AP via decreasing levels of inflammation (Fig. 2B). Intriguingly, expression of Plk1 in the pancreas reduced edema and the inflammatory infiltrate caused by caerulein. Despite the significant development of ADM lesions, AP metaplasia was transient, as the lesions resolved 7 days post-AP (Fig. 2C). In addition, the expression of the acinar cell marker amylase was decreased and was accompanied with an increased expression of the ductal cell marker K19 at 2 days post-caerulein treatment of Mist1^CreER/+ mice. In striking contrast, such a process was clearly antagonized in Mist1^CreER/+/Plk1^LSL mice. In all cases, K19 protein levels returned to normal states at 7 days post-caerulein treatment (Fig. 2D).

Because the induction of Plk1 prevents caerulein-induced pancreatitis that may lead to PanIN, we next examined whether PanIN development was also delayed. For this purpose, we generated Mist1^CreER/+/Kras^G12D/+ and Mist1^CreER/+/Kras^G12D/+/Plk1^LSL mouse lines with caerulein injection to induce pancreatitis. Histologic examination showed evidence of accelerated PanIN development in Mist1^CreER+/Kras^G12D/+ mice 25 days post-AP. In contrast, Mist1^CreER+/Kras^G12D/+/Plk1^LSL mice exhibited limited early-stage PanIN lesions and were well protected from PanIN formation (Fig. 2E and F). We also observed a reduction in the inflammatory infiltrate and the necrosis induced by caerulein treatment upon Plk1-overexpression in the pancreas (Fig. 2G). We further characterized the impact of Plk1-induction on Kras-induced lesion formation by alcian-blue staining, which confirms areas of authentic pancreatic lesion formation. As we show, overexpression of Plk1 indeed reduced alcian-blue staining in the pancreas, indicating a reduction in lesion formation (Fig. 2H and J). Next, the conversion of acinar cells to ductal cells was analyzed by IF staining against amylase (red) and K19 (green). In agreement, pancreata of Mist1^CreER+/ Kras^G12D/+/Plk1^LSL mice showed little formation of K19, and they exhibited limited early-stage PanIN lesions in comparison with that in Mist1^CreER+/Kras^G12D/+ mice (Fig. 2I and J), indicating that Plk1 induction significantly delayed the development of PanIN. Taken together, these data suggest that the overexpression of Plk1 delays Kras^G12D/+-initiated PanIN.

As shown in Fig. 2, the induction of Plk1 inhibited caerulein-induced inflammation in the pancreas. To further understand the underlying mechanisms, we performed RNA-seq with mouse pancreas lines described in Fig. 1F and G. We observed that Plk1-overexpression was associated with a reduction of chemokines and interleukins and served to suppress immune cells in mice with AP. Further, while reductions of IL2, IL4, IL7, IL9, and IL21 could suppress T-cell proliferation/survival, the decrease of IL12 and IL27 could inhibit INFγ production, thus inhibiting an inflammatory response. The reduction of chemokines, including CXCL9, CCL5, CCL19, CCL21, CCL22, CXCR5, CCR7, and CCR8, could affect functions of immune cells by affecting cell trafficking, cell–cell interaction, migration, and homing (Fig. 3A). This reduction of chemokines and interleukins could be partially recovered when it progressed to PanIN (Fig. 3B). However, how Plk1 inhibits inflammation in pancreas is still not clear. As suggested in Fig. 1J, Plk1 may regulate inflammation in the pancreas through the NF-κB pathway. In addition, caerulein induces AP by activating NF-κB and cytokine expression (33). Thus, we performed IHC staining against phospho–NF-κB in AP and observed a decrease in the nuclear localization of phospho–NF-κB in mice with Plk1 overexpression (Fig. 3C and D). To dissect the alteration of the NF-κB pathway in PanIN, we also examined phospho–NF-κB in tissues from Mist1^CreER+/Kras^G12D/+ and Mist1^CreER+/Kras^G12D/+/Plk1^LSL mouse pancreata. Similarly, we observed a reduction in the nuclear localization of phospho–NF-κB in mouse pancreas upon Plk1-overexpression (Fig. 3E and F). Furthermore, we performed immunoblotting with lysates of mouse pancreas. In agreement, we observed a reduction in phospho–NF-κB at S536 and S468 upon Plk1-overexpression, suggesting that there is inactivation of the NF-κB pathway in Plk1-overexpressing samples (Fig. 3G). These data indicate that overexpression of Plk1 in the mouse pancreas inactivates the NF-κB pathway by blocking the translocation of NF-κB to the nucleus.

Given the suppressive effects of Plk1 in AP and PanIN, we next sought to investigate the dynamic changes of Plk1 during disease progression in both mouse and human tissues. As showed in Fig. 3H, we observed an increase of the level of endogenous Plk1 in mouse PDAC in comparison with AP and PanIN. However, no significant changes of Plk1 were observed between pancreatitis and PanIN. Furthermore, we then investigated the levels of Plk1 in human patients by IHC and found a higher level of Plk1 in human tumors than in human pancreatitis and PanIN. Similarly, no obvious changes of Plk1 were observed between pancreatitis and PanIN in human patients, which is in accordance with the observations in mice (Fig. 3I). Moreover, increased Plk1 was observed in samples derived from the same patients, indicating a dynamic role of Plk1 in the human disease (Fig. 3J). In summary, the collective data implies that Plk1 could prevent pancreatic inflammation and PanIN via inactivating NF-κB pathway.

Although Plk1 shows important roles in PDAC progression, the mechanisms that attribute to the failure of Plk1 inhibitors as single agents are still elusive. Dissection of this puzzle could largely benefit the clinical practice. As shown above, Plk1-induction reduced the level of IFNγ (Fig. 2B). Considering the established role of IFNγ in inducing PD-L1 expression and the role of PD-L1 in inhibiting antitumor immunity, we hypothesize that Plk1 inhibition might induce an unexpected increase in PD-L1, thus antagonizing an antitumor immune response and the eventual failure of Plk1 inhibition. To test this hypothesis, we harvested pancreas from Mist1^CreER/+ and Mist1^CreER/+/Plk1^LSL mice and determined the levels of PD-L1 by immunoblotting. As shown in Supplementary Fig. S4A, the level of PD-L1 was surprisingly reduced upon Plk1 induction.

Because an elevation of Plk1 was observed in PDAC, it is reasonable to further investigate the possibility of Plk1-targeted therapy in PDAC. However, the limited efficacy for Plk1 inhibition alone was observed in a clinical trial (NCT00710710), implying that unexpected mechanisms might exist to contribute to the poor response rates (8). To validate the association between Plk1 and PD-L1, we used murine cancer cells, derived from individual primary tumors of KPC (Elas^CreER/+/Kras^G12D/+/p53^R172H/+) mice. As indicated, depletion of Plk1 in KPC cells increased the level of PD-L1 (Supplementary Fig. S4B), whereas Plk1-overexpression dramatically suppressed PD-L1 in KPC cells (Supplementary Fig. S4C). To further validate the association between Plk1 and PD-L1 in human patients, IHC staining against Plk1 and PD-L1 was performed. As indicated, a higher level of Plk1 corresponded to a lower level of PD-L1 expression in both tumor cells and tumor-infiltrating lymphocytes and monocytes (Supplementary Fig. S4D and S4E). Taken together, Plk1 expression was negatively associated with the level of PD-L1, implying that Plk1 might be a negative regulator of PD-L1. These findings could partially explain the failure of Plk1 inhibitors as monotherapy in PDAC.

Because Plk1 negatively associates with the level of PD-L1, we then sought to explore whether and how Plk1 regulates the expression of PD-L1. First, we showed that treatments with the Plk1 inhibitors, BI6727 and GSK461364A, significantly elevated the expression level of PD-L1 with or without IFNγ administration (Fig. 4A). To ask whether the elevation of PD-L1 was due to Plk1 inhibition-induced mitotic arrest, we treated cells with nocodazole and GSK461364A. More elevation of the levels of PD-L1 was observed in the group treated with GSK461364A than that in the group treated with nocodazole, suggesting that the Plk1 inhibition-induced elevation of PD-L1 is not only due to mitotic arrest (Supplementary Fig. S5A). In agreement, the depletion of Plk1 also elevated the expression of PD-L (Supplementary Fig. S5B), and the elevation of PD-L1, was likely due to increased transcription upon Plk1-depletion (Fig. 4B). Further, Plk1 expression–induced inhibition of PD-L1 is kinase activity dependent, as overexpression of constitutively active Plk1-T210D mutant showed a more dramatic effect on the level of PD-L1 (Fig. 4C). Because TNFα can induce PD-L1 expression via the NF-κB pathway and because Plk1 was associated with inactivation of the TNF pathway and the NF-κB pathway in mice and human pancreatic cancers (Supplementary Table S3), we hypothesized that Plk1 regulates PD-L1 via the NF-κB pathway. To this end, we investigated the possible activation of the NF-κB pathway by inhibiting the activity of Plk1. As indicated, inhibition of Plk1 dramatically increased the NF-κB phosphorylation on Ser468 but not Ser536 (Fig. 4D; Supplementary Fig. S5C). It has been documented that the translocation of NF-κB from the cytoplasm to the nucleus activates its transcription factor activity. Accordingly, we performed IF to examine the translocation of NF-κB upon Plk1 overexpression. As indicated in Fig. 4E, expression of the kinase-defective Plk1-K82M mutant also resulted in more nuclear localization of NF-κB than that in cells expressing WT and the constitutively active Plk1-T210D mutant form.

It has been reported that phosphorylation of RB inhibits NF-κB activation by inferring translocation of NF-κB from the cytoplasm to the nucleus (21). To examine whether the translocation of NF-κB upon Plk1 depletion is RB-dependent, we performed immunoblotting and found that RB was elevated upon Plk1 inhibition (Fig. 4F). To rule out whether the elevation of RB is due to mitotic arrest, we treated cells with nocodazole ± GSK461364A. In agreement, inhibition of Plk1 resulted in the elevation of RB beyond a level observed in cells treated with nocodazole, which only slightly resulted in an increase in RB (Fig. 4G). Moreover, Plk1 and RB can be coimmunoprecipitated with each other (Fig. 4H), suggesting that RB is an interacting partner of Plk1. We then tested whether Plk1 directly phosphorylates RB with in vitro kinase assays and showed that a recombinant GST-fusion RB fragment containing amino acids (aa) 379 to 792 was a robust Plk1 substrate (Fig. 4I). To map the phosphorylation site(s), we mutated every single serine/threonine within the region to alanine and eventually identified S758 as a Plk1 phosphorylation site (Fig. 4J). We also observed an increase in the phosphoserine level of RB upon transfection of constitutively activate Plk1-T210D (Fig. 4K). Furthermore, nocodazole treatment significantly increased ubiquitination of RB, which can be attenuated by BI2536 cotreatment, suggesting that Plk1 phosphorylation of RB induces degradation of RB (Fig. 4L; Supplementary Fig. S5D).

To further investigate the function of the Plk1 phosphorylation of RB, we predicted the initial structures of RB-WT and the RB-S758D mutant based upon the sequences published on UniProt (P06400) using AlphaFold2⁽²⁷⁾. We also investigated the structure and flexibility of the RB-WT and the RB-S758D mutant in explicit aqueous solution using MD simulations (Fig. 4M). The final configurations of the MD simulations showed that the RB-S758D (yellow) mutant presents a conformation different from the RB-WT (blue; Fig. 4N). In addition, the positive value of the root mean square deviation for C_α atoms also showed that the conformation of RB-S758D should differ from that of the RB-WT protein. The difference in conformation implies a distinct function between RB-WT and RB-S758D mutant forms (Supplementary Fig. S6A). Next, we characterized the flexibility of RB-WT and RB-S758D proteins using the root mean square fluctuation (RMSF) and ΔRMSF of their C_α atoms. As illustrated in Supplementary Figs. S6B and S4O, the flexibility of RB-S758D presents the largest change around RES874-879 (ΔRMSF < –1.0 nm) on the RB-C domain, and RES822-826 presents a positive ΔRMSF value (>0.2 nm) on the same domain. Such variation in this observed flexibility may impact the functions of the RB-S758D protein compared with the RB-WT protein. Furthermore, residues on the RB-N domain also present a variation in flexibility upon the mutation: RES28-43 presents ΔRMSF < −0.4 nm and RES20-24 shows ΔRMSF > 0.2 nm. It is reported that the RB-N domain interacts with NF-κB, which can be enhanced by its phosphorylation (21). Thus, the changes of residue flexibility on the RB-N and RB-C domain may impact the binding affinity of the RB protein.

We then investigated the mutation-induced changes of secondary and tertiary structures of the RB protein. The secondary structure versus time plots do not show significant differences (Supplementary Fig. S6C) between RB-WT and RB-S758D. However, the mutation induces a change in the tertiary structure. We calculated the residue–residue contact map for RB-N, RB-P, and RB-C domains. The contact map of RB-C domain shows the loss of some close residue–residue contact within the domain. (Fig. 4P). RB-N domain also presents some changes in the residue–residue contact map but not as significant as the RB-C domain (Supplementary Fig. S6D and S6E). These collective data render the plausibility that the S758D mutation induces protein structure changes, leading to distinct functions and downstream consequences.

Considering that the phosphorylation of RB inhibits the translocation of NF-κB from the cytoplasm to the nucleus (21), we hypothesized that Plk1 phosphorylation induces a structure change in RB that might cause dysregulation of NF-κB translocation. To ask how the Plk1 phosphorylation of RB regulates the NF-κB pathway, we showed that the RB-S758D mutant repressed translocation of NF-κB to the nucleus in comparison with WT RB (Fig. 4Q and R). To elucidate the molecular mechanisms by which RB regulates NF-κB function, we examined whether RB interacts with NF-κB. Coimmunoprecipitation demonstrated that endogenous NF-κB interacted with exogenous RB-GFP, and the interaction was weakened by RB phosphorylation at S758 (Fig. 4S). These findings validate the prediction for the distinct function of a S758D mutation induced-protein structure change using ΔRMSF of their C_α atoms. Consequently, we observed that cells expressing the RB-S758D mutant showed less PD-L1 expression than those expressing WT RB (Fig. 4T). To further determine whether Plk1-mediated regulation of PD-L1 is NF-κB–dependent, we treated NF-κB–depleted cells with GSK461364A and observed that a Plk1 inhibition–induced elevation of PD-L1 was rescued by depletion of NF-κB (Fig. 4U). Taken together, we concluded that depletion/inhibition of Plk1 increases the level of PD-L1 via promoting nuclear translocation of NF-κB in an RB-dependent manner (Fig. 4V).

Given that PDAC is an immunological “cold” cancer, it is not surprising to observe the limited efficacy of immune checkpoint inhibitors in this disease. Although elevation of Plk1 in PDAC implies administration of Plk1 inhibitors to treat PDAC, there has been limited efficacy for Plk1 inhibition alone, probably due to the undesirable elevation of PD-L1. However, the rather unexpected elevation of PD-L1 upon Plk1 inhibition provides the rationale to treat PDAC through combination of Plk1 inhibition and anti-PD-L1. To explore this opportunity, we tested the efficacy of anti-PD-L1 blockade plus inhibition of Plk1 in mouse models. To recapitulate the intrapancreatic immune response to PDAC, we constructed a pancreatic orthotopic allograft model by injecting KPC cells into WT mouse pancreas with intact immune functions, followed by treatments with anti–PD-L1 and BI2536. As indicated, monotherapies of anti–PD-L1 and BI2536 showed minimal effects on tumor suppression, but the effects of combination treatment were significant. Combination treatment synergistically reduced tumor volume and tumor weight (Fig. 5A and B). IF staining indicated a reduction in cell proliferation and a proclivity for apoptosis (Fig. 5C; Supplementary Fig. S7A). A significant prolongation in overall survival was observed in mice treated with the combination of BI2536 and anti–PD-L1 as compared with mice receiving isotype control or single treatments (Fig. 5D). Long-term exposure to this treatment regimen did not alter the body weight of these animals, indicating its tolerability during a sustained period of administration (Supplementary Fig. S7B).

Next, we sought to dissect how Plk1 inhibition resensitizes PDAC to anti–PD-L1 treatment. One of the roles of PD-L1 in cancer is to prevent cancer cells from immune attack. Because the immunosuppressive TME in PDAC contributes to the limitation of immune checkpoint blockade, one approach to treat PDAC is to reverse this protumorigenic microenvironment. We firstly evaluated various immune cell markers in patient samples by IF to examine the properties of the TME. In tumors, we observed fewer toxic T-cells but more immunosuppressive cells, such as Foxp3⁺ and CD206⁺ cells, infiltrated in the tumors, indicating immunosuppression in these tumors (Fig. 5E; Supplementary Figs. S8A).

Because Plk1 is associated with an immunosuppressive TME in human PDAC, the reestablishment of the efficacy of anti–PD-L1 by administration of Plk1i in PDAC supports the possibility that Plk1-inhibition may alter the immunosuppressive TME in PDAC. To test this concept, we injected KPC cells into WT mice orthotopically, followed by treatment with BI2536 and anti–PD-L1 (Fig. 5F). As observed, monotherapy with anti–PD-L1 inhibited tumor growth. However, the tumor-suppressive effect was more significant upon combination treatment, whereas a single treatment with BI2536 showed no dramatic effect on tumorigenicity (Fig. 5G). To further dissect the role of Plk1 on the TME, we analyzed the immune infiltrates in PDAC. As shown in Fig. 5H and Supplementary Fig. S8B, we analyzed the infiltrations of immune cells in tumors, including dendritic cells (DC), B cells, MDSC, macrophages, CD4⁺ T cells, CD8⁺ T cells, and NK cells. As demonstrated, monotherapy with anti–PD-L1 increased CD4⁺ cells but decreased macrophages in orthotopic tumors and no changes of infiltrations of any immune cells were observed in tumors treated with BI2536 alone. However, combination treatment increased the populations of CD4⁺ cells, CD8⁺ cells, and DCs, and decreased NK cells and macrophages but not B cells and MDSCs in the tumors. These collective data indicate combinational treatment of anti–PD-L1 and BI2536 suppresses tumorigenicity through enhancing an antitumor immune response by increasing immune infiltrates.

To further validate the effects of Plk1 inhibition on the TME, we created Rosa26^CreERt2/Plk1^f/+ mice, in which Plk1 was depleted upon tamoxifen injection (Supplementary Fig. S8C). These Plk1-depleted mice were injected with KPC cells, followed by anti–PD-L1 treatment. As indicated in Fig. 5I–K and Supplementary Fig. S8D, a PD-L1 Ab alone was not enough to inhibit tumor growth in WT mice but did suppress KPC tumor formation in Plk1-depleted mice. Of note, tumors implanted in Rosa26^CreERt2/Plk1^f/+ mice appeared smaller than those in WT mice, indicating that the level of Plk1 in the TME affects tumor growth as well. Next, we sought to examine the effects of Plk1 on immune cell infiltration. As illustrated in Fig. 5K and Supplementary Fig. S8E, anti-PD-L1 alone in WT mice did not affect toxic T cells or MDSCs in tumors. However, upon Plk1 depletion, anti–PD-L1 increased CD8⁺ cells but decreased MDSCs in tumors, eventually reactivating the immune response in the tumor. Moreover, anti–PD-L1 alone showed no effect on CD8⁺ and CD4⁺ cell counts in lymph nodes and B cells in spleen of WT mice, but it induced a reduction in CD8⁺ cell counts in peripheral lymph nodes of Plk1-depleted mice (Fig. 5K).

Plk1 is known to control mitotic progression and chromosomal stability (34–37). An alternative explanation for the enhanced immunotherapy response is that Plk1 antagonism might expose cryptic antigens via induced chromosomal instability that leads to enhanced checkpoint blockade by targeting PD-L1. To exclude this possible explanation for the observed antineoplastic actions, we investigated chromosomal instability upon Plk1-inhibition in pancreatic cancer cells. No obvious changes for γH2AX upon Plk1-inhibition by administrating BI6727 and GSK461364A were observed, indicating there is no DNA damage upon Plk1 inhibition in the pancreatic cancer cells (Supplementary Fig. S9). In summary, the depletion of Plk1 reestablishes the antitumor responses by altering the immunosuppressive TME.

Based on the finding that Plk1 regulates expression of PD-L1 through the NF-κB pathway, we hypothesized that Plk1 and PD-L1 may play a role in the clinical outcome of patients with PDAC. To test this possibility, we explored the correlations between the levels of Plk1 and PD-L1 and the survivability in patients with PDAC. As demonstrated, we observed a higher level of Plk1 associated with a shorter overall survival in The Cancer Genome Atlas data (Fig. 6A, top). In addition, patients with a higher level of PD-L1 expression were associated with a dramatically shorter overall survival as well (Fig. 6A, bottom). To further investigate the effects of the identified negative correlation between Plk1 and PD-L1 on the clinical outcomes of patients with PDAC, we grouped tumors into those with low and high PD-L1 expression, and then investigated the relationship between the level of Plk1 and patient survival. As illustrated in Fig. 6B, a higher level of Plk1 was associated with a shorter overall survival than those with a lower level of Plk1 in patients, regardless of the level of PD-L1. Similarly, we also grouped tumors into those with low and high Plk1 expression and then investigated the relationship between the levels of PD-L1 and patient survival. Interestingly, a high level of PD-L1 was associated with a shorter overall survival than those with a low level of PD-L1 only in patients with a higher level of Plk1. No significant changes in overall survival were observed in those with lower Plk1 expression (Fig. 6C). These collective data suggest that high levels of Plk1 and PD-L1 are associated with a poorer prognosis in PDAC, and the level of PD-L1 does not affect the association between the level of Plk1 and overall survival. However, even though the level of PD-L1 is associated with the overall survival, only patients with a higher level of Plk1 can be distinguished by PD-L1 expression.

As shown in Fig. 5, a combination of a Plk1-inhibitor and anti–PD-L1 reactivated an antitumor immune response by regulating T cells. To explore the correlations of Plk1 and PD-L1 with clinical outcomes in the context of different immune infiltrates, we grouped patients based on enrichments of CD4⁺, CD8⁺, and regulatory T cells, respectively. As illustrated in Fig. 6D, in patients with decreased CD4⁺ memory T cells, a higher level of Plk1 was associated with a shorter overall survival than those with a lower level of Plk1. However, no significant changes were observed in those enriched with CD4⁺ memory T cells. In addition, a higher level of PD-L1 was associated with a shorter overall survival than those with a lower level of PD-L1 in patients with enriched CD8⁺ T cells but not in those with decreased CD8⁺ T cells (Fig. 6E). Similarly, a higher level of PD-L1 was associated with a shorter overall survival than those with a lower level of PD-L1 in patients with enriched regulatory T-cells (Tregs) but not in those with decreased Tregs (Fig. 6F).

To validate our findings in Fig. 6, we analyzed the data cohorts from ICGC database. In agreement with our findings in Fig. 6A–C, we found that high levels of Plk1 and PD-L1 were associated with a shorter overall survival (Supplementary Fig. S10A–S10C). To further investigate the roles of Plk1 and PD-L1 in the clinical outcomes of patients with different immune cell infiltrations, we grouped patients based on the infiltration status of different immune cell subtypes. In accordance with our findings in Fig. 6D–F, we observed that a higher level of Plk1 was associated with a shorter overall survival than those with a lower level of Plk1 in patients with decreased CD4⁺ memory T cells. However, no significant changes were observed in those with enriched CD4⁺ memory T cells (Supplementary Fig. S10D). No significant changes were observed in patients with either enriched or decreased CD8⁺ T cells (Supplementary Fig. S10E). Similarly, a higher level of PD-L1 was associated with a shorter overall survival than those with a lower level of PD-L1 in patients with enriched regulatory T cells (Tregs) but not in those with decreased Tregs (Supplementary Fig. S10F). These collective data suggest that the levels of Plk1 and PD-L1, associated with outcomes of patients with PDAC, can be used to distinguish patients under differential T-cell infiltration.

As described in a phase II study, Plk1 inhibition via a single agent has failed to induce tumor regression in advanced pancreatic cancer patients (NCT00710710) (8). Thus, it is of high clinical relevance to dissect the underlying mechanisms and to develop novel approaches to reverse the limited efficacy of Plk1 inhibitors in PDAC. Our data in murine PDAC models support the clinical observations, as we found disease stabilization rather than PDAC regression in KPC mice (Fig. 5). Due to the fact that anti–PD-L1 Abs seem to be effective with only certain cancer types, combination therapy trials are being actively pursued in a variety of clinical settings. The data presented here suggest that combination therapies with inhibition of Plk1 and anti-PD-L1 may achieve durable tumor regression in PDAC. Herein, we show that the pharmacologic or genetic targeting of Plk1 augmented tumor immunity, and thereby potentiated the efficacy of immunotherapy in PDAC.

In this study, we focus on the role of Plk1 signaling in malignant cells and its role as a driver of the inflammatory TME that alters immune surveillance and limits immunotherapy. Plk1 represents a new negative regulator of the pathway that induces IFN (38), which is well known for its role in T-cell proliferation, antigen sensitivity, cytokine production, and migration in these contexts (39). The effect of pharmacologic inhibition of Plk1 on the function of T cells in PDAC is likely important to the clinical application of inhibition of Plk1 and immunotherapeutic combinations (Fig. 5).

The next frontier of immune checkpoint therapy is improvement of the efficacy of checkpoint inhibitors, such as Abs against PD-1 and PD-L1, through combination approaches, especially in cancers that have shown little benefit, such as PDAC. As Plk1 plays an important role in cell-cycle progression in normal proliferating tissues, it is not surprising that Plk1 is overexpressed in numerous cancer types (40). However, the role of Plk1 in driving the suppressive TME is not well understood. We found that inhibition of Plk1 altered the production of proinflammatory cytokines and blunted to the avoidance of immune surveillance. Together, these findings indicate that Plk1 is a driver of immune escape in PDAC; and, thus, it is a target for combination with immunotherapy.

The response to immune checkpoint therapy in patients is expected to be highly correlated with PD-L1 expression in tumor cells (41). Our study showed a negative association between Plk1 and PD-L1 in mouse and human PDAC (Supplementary Fig. S4) and identifies a previously uncharacterized role of S758-phosphorylated RB in suppressing NF-κB activity and PD-L1 transcription (Fig. 4). In agreement with the essential role of RB in repressing PD-L1 expression, we demonstrated that inhibition of S758 phosphorylation by Plk1-inhibition and depletion can enhance PD-L1 expression. Thus, our findings support a model in which Plk1 inhibition may trigger an undesirable upregulation of PD-L1, thereby causing immune evasion of cancer cells (Fig. 7). This finding could partially explain the failure of Plk1 inhibitors as single reagents to treat PDAC. In addition to the role of Plk1 in the downregulation of PD-L1 in pancreatic cancer cells, we also demonstrated that Plk1 inhibited pancreatitis and delayed PanIN formation in mouse models (Fig. 2). Of interest, the novel dual roles of Plk1 at the early stage of PDAC (pancreatitis and PanIN) and in PDAC are both through dysregulating NF-κB pathway. For the first time, we demonstrate that the Plk1-associated inactivation of NF-κB plays significantly distinct roles in different stages of PDAC (Fig. 7, left). More interestingly, the observations that the negative association of Plk1 and PD-L1 and Plk1-associated inactivation of the NF-κB pathway in precursor of PDAC strongly support the notion that Plk1 regulates PD-L1 through NF-κB pathway. This hypothesis has been verified by the experiments in Fig. 4. The rather unwanted role of Plk1 in upregulating PD-L1 raises the promise to treat PDAC through combining Plk1-inhibition plus anti–PD-L1 (Fig. 5). In other words, Plk1-inhibition reestablishes the antitumor response by sensitizing immune checkpoint blockade (anti–PD-L1) in PDAC. Upon Plk1 inhibition or depletion, absence of phosphorylation of RB prevents the translocation of NF-κB to the nucleus, resulting in the undesirable upregulation of PD-L1. The combination of a Plk1i and anti–PD-L1 is expected to change an immunologically “cold” tumor to an immunological “hot” tumor, resulting in an antitumor immune response to PDAC (Fig. 7, right). This novel combination strategy kills two birds with one stone. In other words, Plk1 inhibition-induced an undesirable upregulation of PD-L1, resulting in a failure of Plk1 inhibition in PDAC. However, upregulation of PD-L1 also reestablishes the responses to anti–PD-L1 and finally induces tumor inhibition, which in turn overcomes the failure of Plk1 inhibition in PDAC. To support this conclusion, we showed that the combination of Plk1 inhibition and anti–PD-L1 increased CD8⁺ T-cell infiltration into tumors (Fig. 5). A positive response of a patient to immune checkpoint therapy depends mainly on T-cell infiltration into the tumors (41, 42). T-cell infiltration into tumors will render a subject more likely to respond to immunotherapy. Combination of Plk1 inhibition and anti–PD-L1 could improve cell trafficking and change the profile of T cells in the tumor. In view of the multiple facets of Plk1 in regulating the progression of PDAC and immune suppression, this study presents a reasonable approach for a clinical application. To be noted, the orthotopic tumors we generated are characterized by high DC infiltrates, which may explain the finding of response to immune checkpoint therapy combination. In contrast, other groups have shown that DCs comprise only about 1% to 3% of CD45⁺ cells in human PDAC and KPC tumors (43, 44). Our study could benefit from the same treatment studies in additional KPC cell lines that are reflective of the immune phenotype shared by the majority of human PDAC and demonstrating the immune cell composition of the TME at baseline and after treatment.

In human pancreatic cancer, the TP53 gene is frequently mutated (50%–75%), predominantly through missense mutations (45). Interestingly, we observed that the level of Plk1 was higher in TP53 mutant patients than those of TP53 nonmutant patients (Fig. 1B). Both Plk1 and PD-L1 showed positive associations with a poor prognosis in patients with PDAC. However, the roles of Plk1 and PD-L1 in clinical outcomes are distinct. In contrast to the expression levels of PD-L1 in patients, there are no differences for the roles of Plk1 in prognosis. Upon grouping patients into those with low and higher levels of Plk1, we observed that PD-L1 could distinguish the patients with a higher level of Plk1 but no significant differences in those with a lower level of Plk1 were observed (Fig. 6A–C). These results imply that Plk1 may be involved in an immune evasion process through PD-L1 and could be used to classify patients for targeted therapy. As we have shown, the status of T-cell infiltration exhibited differences of patient survival between the levels of Plk1 and PD-L1 in PDAC (Fig. 6D–F). Of note, the level of PD-L1 can significantly distinguish survival for the patients who have tumors enriched with CD8⁺ T cells, suggesting the potential efficacy of anti–PD-L1 in this patient population.

No disclosures were reported.

Z. Zhang: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. L. Cheng: Software, formal analysis. J. Li: Formal analysis, investigation, methodology. Q. Qiao: Software, formal analysis. A. Karki: Methodology. D.B. Allison: Formal analysis, writing–review and editing. N. Shaker: Formal analysis. K. Li: Validation, investigation, visualization. S.M. Utturkar: Software, methodology. N.M.A. Atallah Lanman: Software, methodology. X. Rao: Methodology. P. Rychahou: Methodology. D. He: Validation. S.F. Konieczny: Resources, methodology. C. Wang: Validation. Q. Shao: Software, formal analysis. B.M. Evers: Methodology. X. Liu: Supervision, funding acquisition, writing–review and editing.

This research was supported by NIH R01 CA157429 (to X. Liu), NIH R01 CA264652 (to X. Liu), NIH R01 CA256893 (to X. Liu), and NIH R01 CA196634 (to X. Liu). This research was also supported by the Biospecimen Procurement & Translational Pathology, Biostatistics and Bioinformatics, Flow Cytometry and Immune Monitoring Shared Resources of the University of Kentucky Markey Cancer Center (P30CA177558). The authors deeply appreciate the editing service provided by Eleanor Erikson.

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.