NEK2 inhibition triggers anti-pancreatic cancer immunity by targeting PD-L1

Despite the substantial impact of post-translational modifications on programmed cell death 1 ligand 1 (PD-L1), its importance in therapeutic resistance in pancreatic cancer remains poorly defined. Here, we demonstrate that never in mitosis gene A-related kinase 2 (NEK2) phosphorylates PD-L1 to maintain its stability, causing PD-L1-targeted pancreatic cancer immunotherapy to have poor efficacy. We identify NEK2 as a prognostic factor in immunologically “hot” pancreatic cancer, involved in the onset and development of pancreatic tumors in an immune-dependent manner. NEK2 deficiency results in the suppression of PD-L1 expression and enhancement of lymphocyte infiltration. A NEK binding motif (F/LXXS/T) is identified in the glycosylation-rich region of PD-L1. NEK2 interacts with PD-L1, phosphorylating the T194/T210 residues and preventing ubiquitin-proteasome pathway-mediated degradation of PD-L1 in ER lumen. NEK2 inhibition thereby sensitizes PD-L1 blockade, synergically enhancing the anti-pancreatic cancer immune response. Together, the present study proposes a promising strategy for improving the effectiveness of pancreatic cancer immunotherapy.

Pancreatic ductal adenocarcinoma (PDAC) is the most common exocrine cell tumor of the pancreas and the 7th leading cause of death in cancer patients worldwide1,2. Despite advances in diagnosis, perioperative management, radiotherapy techniques, and systematic treatments that enhance the survival rate of pancreatic cancer, the numbers of such cases and related deaths have steadily increased1,3. Immunotherapy is considered the most promising strategy to reverse this trend, especially immune checkpoint blockade targeting programmed cell death 1 (PD-1) and its ligand 1 (PD-L1, also known as B7-H1)4,5. Multiple clinical trials of PD-1/PD-L1 blockade have been launched as treatments for melanoma, lung cancer, and kidney cancer, with promising clinical results5,6,7. However, PD-1/PD-L1 blockade has so far demonstrated limited effectiveness in PDAC. For example, Royal et al. reported that no in-group patients with advanced pancreatic cancer responded, preventing an evaluation of the effectiveness of anti-PD-L1 antibodies8. Furthermore, a phase II randomized clinical trial of 65 patients with metastatic PDAC suggested that the objective response rate of patients receiving durvalumab monotherapy was 0%, although that of patients receiving combination therapy was 3.1%9. Thus, an in-depth understanding of the detailed mechanism of PD-L1-based combinatorial therapy may contribute to the development of novel therapeutic strategies10.

Accumulating evidence has demonstrated that the therapeutic efficacy of drugs targeting PD-L1 can be seriously affected by post-translational modifications (PTMs), including ubiquitination, phosphorylation, and glycosylation. Such PTMs play critical roles in regulating the stability, internalization, and localization of PD-L1, in addition to its physiological and pathological functions11. For instance, N-glycosylation of PD-L1 facilitates its folding, intracellular transport, and function12,13,14. Epidermal growth factor (EGF) signaling upregulates b-1,3-N-acetylglucosaminyltransferase (B3GNT3) expression that promotes N-glycosylation at positions N192, N200, and N219 of PD-L1 in triple-negative breast cancer cells. This stabilizes PD-L1 through the prevention of glycogen synthase kinase 3 beta (GSK3β) binding and phosphorylation of PD-L1 at the T180 and S184 residues, subsequently recognized by beta-transducin repeat containing E3 ubiquitin-protein ligase (BTRCP)-mediated PD-L1 poly-ubiquitination and degradation15,16. More importantly, the combination of EGFR inhibition and PD-L1 blockade significantly enhances mono-drug therapeutic efficacy. In addition, metformin-activated AMP-activated protein kinase directly phosphorylates PD-L1 at the S195 residue, inducing abnormal PD-L1 glycosylation that leads to PD-L1 degradation through an endoplasmic reticulum-associated protein degradation pathway16. Therefore, identification of critical regulators of PD-L1 PTM in pancreatic cancer is urgently required to improve the efficacy of PD-L1 blockade in this specific cancer type.

Never in mitosis gene A (NIMA)-related kinase 2 (NEK2) is a member of NEKs, a serine-threonine kinase family including proteins structurally related to the essential mitotic regulator17. NEK2 is a multifunctional protein physiologically involved in cell cycle regulation, such as centrosome duplication and separation17,18, microtubule stabilization19, kinetochore attachment20, in addition to spindle assembly checkpoint21,22. NEK2 is also considered a key factor in the maintenance of normal development and function of B cells23. Intriguingly, aberrant expression of NEK2 and NEK2-mediated phosphorylation of downstream proteins (e.g. p53 at Ser31524 and GAS2L1 at Ser35225) are frequently observed to be associated with tumor initiation, progression, metastasis, and consequently poor prognosis in multiple cancers, including but not limited to, breast cancer26,27, colorectal cancer28, and PDAC29. However, the potential role of NEK2 in cancer immune resistance so far remains undefined.

In this work, we demonstrate that PD-L1 is a substrate of NEK2, and combinatorial inhibition of NEK2 and PD-L1 largely improves the therapeutic efficacy of pancreatic cancer in preclinical models.

NEK2 is a prognostic factor in pancreatic cancer

To assess whether members of the NEK family (from NEK1 to NEK11) are associated with pancreatic cancer prognosis, we first compared their expression levels in tumors and normal tissue using pancreatic adenocarcinoma (PAAD) datasets from The Cancer Genome Atlas (TCGA) database. Bioinformatic analyses across multiple cancers demonstrated that NEK2 is the most significantly upregulated member of the NEK family in PAAD compared with expression in normal tissues (Fig. 1a and Supplementary Fig. 1a). The elevated expression of NEK2 was further validated in paired PDAC tissue samples by immunohistochemistry (Fig. 1b, c). More importantly, the significantly negative prognostic relevance of NEK2 was observed in a tumor tissue microarray (Fig. 1d). The analysis of TCGA data further confirmed that the higher expression of NEK2 was correlated with inferior overall survival (OS) (Fig. 1e and Supplementary Fig. 1b). Furthermore, high expression of NEK2 was also found to be significantly associated with vascular invasion and tumor grade as opposed to TNM stage and other clinical indices in multiple patient cohorts of pancreatic cancer (Table 1 and Supplementary Fig. 2a–c). Together, these results suggest that the elevated expression of NEK2 is unfavorable for patients with PDAC.

a–c Expression profile of NEK2 in pancreatic cancer. Relative NEK2 expression in pancreatic cancer was analyzed using large-scale RNA-Seq datasets of PDAC from the TCGA database (n (T) = 179, n (N) = 171; n: number of patients) (a). NEK2 expression was measured in paired tumor and normal pancreatic tissues by IHC staining, with representative images (b) and statistical results (c) shown (n = 30; n: number of patients) (N: Normal pancreatic tissue; T: Pancreatic tumor tissue). Scale bars: 100 μm. d Correlation of NEK2 with prognostic factors in pancreatic cancer. Tissue microarray analysis of the prognostic role of NEK2 in pancreatic cancer (n = 64 weakly positive; n = 50 strongly positive) (p < 0.0001). e Overall survival (OS) of patients with pancreatic cancer with high or low expression of NEK2 (n = 177). f Overall survival (OS) of pancreatic cancer patients with high numbers of CD8+ T cells and with high or low expression of NEK2 (n = 76). g Overall survival (OS) of pancreatic cancer patients with decreased CD8+ T cells and with high or low expression of NEK2 (n = 101). h Bioinformatics analysis of the correlation between NEK2 and immune effector cells using TCGA datasets. In a, data are represented as boxplots where the middle line is the median; the lower and upper hinges correspond to the first and third quartiles; the upper whisker extends from the hinge to the largest value no further than 1.5 × IQR from the hinge (where IQR is the inter-quartile range); the lower whisker extends from the hinge to the smallest value at most 1.5 × IQR of the hinge, while data beyond the end of the whiskers are outlying points that are plotted individually. *P < 0.05, **P < 0.01, ***P < 0.001 using a two-tailed t-test; ns: not significant. Kaplan–Meier method and a Gehan–Breslow–Wilcoxon test are indicated in d. The Hazard Ratios (HR) and p-values by the log-rank (Mantel–Cox) test are indicated in e–g.

Full size image

NEK2 expression indicates poor prognosis in immunologically “hot” pancreatic cancer

Since tumors are classified as immunologically “cold” (CD8+ T cell-decreased) or “hot” (CD8+ T cell-enriched), and their immune status is closely associated with tumor prognosis30, we further investigated the prognostic role of NEK2 in increased or decreased CD8+ T cell pancreatic cancer. Intriguingly, NEK2 expression was only significantly associated with OS in immunologically “hot” cells, rather than those that were “cold” (Fig. 1f, g). Furthermore, the opposite relationship was observed between the expression of NEK2 and the number of immune effector cells (Fig. 1h). Interestingly, NEK2 was also expressed in multiple immunosuppressive populations, such as myeloid-derived suppressor cells (MDSCs), dendritic cells (DCs), and macrophages, although the expression-positive rate was lower than that in tumor cells (Supplementary Fig. 3a, b). Furthermore, GO-enrichment analysis revealed that NEK2 overexpression led to a dramatic increase in the expression of proteins related to the positive regulation of leukocyte proliferation, migration, and activation, and lymphocyte activation (Supplementary Fig. 3c, d). These results indicate that NEK2 may influence the efficacy of immunotherapy in multiple ways. Additionally, given the critical role of tumor mutation burden (TMB) in the immunotherapy response, we next investigated the prognostic relevance of NEK2 in immunologically “hot” pancreatic cancer with high or low TMB, respectively. Surprisingly, NEK2 displayed a significant correlation with poor prognosis in low mutational tumors as opposed to high mutational tumors, further indicating the precise prognostic value of NEK2 in specific patients (Supplementary Fig. 4a, b). Taken together, NEK2 was found to be a prognostic factor in immunologically “hot” pancreatic cancer.

NEK2 deficiency improves pancreatic cancer immunogenicity

A KPC-NEK2 knockdown (KD) pancreatic cancer cell line, stably transfected with Nek2double nickase plasmids, was generated to determine whether NEK2 is associated with an anti-tumor immune response. In comparison with wild-type (WT) control cells, NEK2 KD-cells displayed significantly weakened resistance to T cell-mediated tumor cell-destruction in vitro (Supplementary Fig. 5a, b). The growth of NEK2 KD cell-derived tumors was consistently significantly repressed in immunocompetent C57BL/6 mice but not in immunodeficient nude mice (Fig. 2a–h). Flow cytometry further revealed a significant increase in the number and function of CD8+ T cells in NEK2 KD cell-derived tumors (Fig. 2i, j). Furthermore, immunocompetent mice with such tumors exhibited markedly extended survival that was not observed in immunodeficient nude mice (Fig. 2k–m). Taken together, the results suggest that NEK2 is involved in the development and progression of PDAC in an immune-dependent manner, with deletion of NEK2 improving pancreatic cancer immunogenicity.

a, c Schematic protocols displaying WT and NEK2-depleted pancreatic cancer cells separately, and s.c. injection into immunocompetent and immunodeficient mice (n = 7). b, e Representative images of tumors and growth curves of immunocompetent mice. Tumors were measured at specified time points then dissected at the endpoint (n = 7). d, f Representative images of tumors and growth curves of tumors from immunodeficient mice. Tumors were measured at specified time points and dissected at the endpoint (n = 7). g, h The weight of tumors from immunocompetent and immunodeficient mice was reported at the endpoint (n = 7). i, j Representative images and statistical results of tumor-infiltrating lymphocytes (n = 7). k Schematic protocols showing WT and NEK2-depleted pancreatic cancer cells separately injected into immunocompetent and immunodeficient mice (n = 13). l, m Survival of immunocompetent and immunodeficient mice bearing NEK2-depleted pancreatic cancer cells (n = 13). Kaplan–Meier survival curves with log-rank test assessing the significance between WT and NEK2 KD immunocompetent (l) (p < 0.0001) and immunodeficient (m) (p = 0.33) mice. Results represent means ± SD of one representative experiment in e–j. All data are representative of three independently performed experiments. *P < 0.05, **P < 0.01, ***P < 0.001 using a two-tailed t-test; ns: not significant. Kaplan–Meier method and a Gehan–Breslow–Wilcoxon test are indicated in m and l.

Full size image

NEK2 inhibition activates an anti-pancreatic cancer immune response

To investigate the influence of NEK2 inhibition on pancreatic tumorigenesis and growth, pancreatic cancer cells both with and without pretreatment using a NEK2 inhibitor were injected s.c. into C57BL/6 and nude mice, respectively (Fig. 3a). Pretreatment with NEK2 inhibitor resulted in a lower incidence of tumors and postponed tumor occurrence (Fig. 3b), whereas such a trend was not observed in nude mice (Fig. 3c, d). Furthermore, tumor-bearing C57BL/6 and nude mice were individually administrated with NEK2 inhibitor and tumor growth was significantly suppressed in C57BL/6 mice (Fig. 3e–g) but not in nude mice (Fig. 3h–j). Moreover, the NEK2 inhibitor imposed no significant effect on the weight of mice compared with the control group (Fig. 3k, l). Additional immunohistochemical assays indicated a significant decrease in PD-L1 expression and increased CD8+ T cell infiltration in NEK2 inhibitor-treated tumors, suggesting that NEK2 inhibition may result in downregulation of PD-L1 and an enhancement of T cell function (Fig. 3m–o). In support of this, the specific NEK2 inhibitor significantly enhanced T cell-mediated tumor cell-destruction in vitro (Supplementary Fig. 5c, d). Collectively, the results demonstrate that NEK2 inhibition suppressed the onset and development of pancreatic cancer, possibly associated with the decreased expression of PD-L1.

a, c Schematic protocols of pancreatic cancer cells with or without pretreatment with a kinase-specific inhibitor of NEK2 (10 μM, 24 h) separately and s.c. injected into immunocompetent and immunodeficient mice (n = 7). b, d Tumor incidence in immunocompetent and immunodeficient mice was recorded at the times indicated. Tumors were further treated using an NEK2 inhibitor (100 μg/mouse, 2 weeks). e, f Schematic protocol of the treatment schedule and tumor growth curve in immunocompetent mice (n = 5). Tumors were measured at the time points indicated and removed at the endpoint. (h, i) Schematic protocol of the treatment schedule and tumor growth curve in immunodeficient mice (n = 5). Tumors were measured at specified time points and dissected at the endpoint. Tumor and mouse weight of immunocompetent (g, k) (n = 5) and immunodeficient mice (j, l) (n = 5) as reported at the endpoint. m–o Representative images and further quantification of tumor-infiltrating lymphocytes. Scale bars: 100×: 50 μm; 200×: 100 μm. Kaplan–Meier method and a Gehan–Breslow–Wilcoxon test are indicated in b and d. Results represent means ± SD of one representative experiment in f, g, i, j, k, l, m, and n. All data are representative of three independently performed experiments. *P < 0.05, **P < 0.01, ***P < 0.001 using a two-tailed t-test; ns: not significant.

Full size image

NEK2 positively correlates and interacts with PD-L1 in pancreatic cancer

Since PD-L1 is a well-recognized and quite important factor in the mediation of tumor immune resistance, we further investigated the regulatory effects of NEK2 on PD-L1 expression. The relationship between NEK2 and PD-L1 was first identified as co-occurrence at a genomic alteration level by mutual exclusivity analysis (Supplementary Fig. 6a). It was further confirmed to have a positive association in multiple assays, including, but not limited to, at the transcriptional level in the TCGA database (Supplementary Fig. 6b) and at the protein level in paired clinical tissue samples (Fig. 4a), in addition to tumor tissue microarrays (Fig. 4b–d). Moreover, a significant correlation was observed in KPC (KrasLSL-G12D; Trp53LSL-R172H; Pdx1-Cre) and KTC (KrasLSL-G12D/+; Tgfbr2flox/wt; Ptf1a-Cre) autochthonous pancreatic tumors in a genetically engineered mouse model (GEMM) (Supplementary Fig. 6c–e). Expression levels of PD-L1 were significantly reduced in NEK2 KD tumors compared with WT control tumors in a xenograft mouse model (Supplementary Fig. 7a, b). Furthermore, an endogenous interaction between NEK2 and PD-L1 in multiple pancreatic cell lines was observed (Fig. 4f, g). Moreover, glutathione S-transferase (GST)-pull down assay showed that NEK2 bound to PD-L1 directly (Fig. 4h). Consistently, Duo-link assay demonstrated the real binding between NEK2 and PD-L1 in the cell (Fig. 4i). In further investigation of the localization of NEK2–PD-L1 interaction, no signals were detected with antibodies against both cytosolic and luminal proteins after trypsinization in the permeable fraction; instead, signals of the cytosolic domain of IRE1α were rapidly reduced after trypsinization, whereas signals of NEK2 and HSP90B1 were maintained in the non-permeable fraction (Supplementary Fig. 8a). Taken together, the results indicate a positive correlation and direct interaction of NEK2 with PD-L1 inside the ER lumen in pancreatic cancer.

a Western blot analysis of NEK2 and PD-L1 in clinical pancreatic tissue samples from patients (n = 20) (N: Normal pancreatic tissue; T: Pancreatic tumor tissue). b–d Representative images and statistical results of IHC staining of NEK2 and PD-L1 in a tissue microarray (n = 156). e–g Cell lysates from SW1990, KPC, and CFPAC-1 separately analyzed by IP and Western blotting using the antibodies indicated. Representative image is shown n = 3 independent experiments. h GST-pull down assay of NEK2-His and GST-PD-L1 protein. Representative image is shown n = 3 independent experiments. i Representative images of individual immunofluorescence staining of NEK2 and PD-L1 interaction in KPC cells by Duolink assay. The red dots (NEK2/PD-L1 interaction) indicate their interaction. Representative image is shown n = 3 independent experiments. The Spearman correlations and p-values by Spearman’s test are indicated in d.

Full size image

NEK2 inhibits ubiquitination-mediated proteasomal degradation of PD-L1

To further investigate the manner in which NEK2 regulates PD-L1, we measured the expression levels of PD-L1 in NEK2 inhibitor-treated, NEK2 KD, or overexpressed cells. NEK2 inhibition or depletion decreased PD-L1 expression in pancreatic cancer cells (Fig. 5a, b), while the overexpression of NEK2 induced the upregulation of PD-L1 (Fig. 5c). Interestingly, downregulated PD-L1 expression caused by NEK2 depletion or inhibition could be restored by treatment with the proteasome inhibitor MG132, suggesting that NEK2 prevents proteasome-mediated degradation of PD-L1 (Fig. 5d, e). Indeed, PD-L1 in both NEK2 KD and inhibition groups exhibited a shorter half-life than control groups, a reduction that was also rescued by MG132 (Fig. 5f–i). Furthermore, increased levels of ubiquitination were observed in NEK2-deficient conditions (Fig. 5j, k). Collectively, these results suggest that NEK2 controls the stability of PD-L1 in a kinase-dependent manner. As demonstrated by Hung’s group, glycogen synthase kinase 3β (GSK3β) is an additional key protein that interacts with PD-L1 and induces phosphorylation-dependent proteasome degradation of PD-L115. Thus, the relationship between the role of NEK2 and GSK3β in regulating the expression level of PD-L1 was further explored. Interestingly, our results revealed that NEK2 inhibitor reduced the phosphorylation of GSK3β, but reversed the up-regulation of PD-L1 caused by the down-regulation of p-GSK3β, and affected the PD-L1–GSK3β interaction (Supplementary Fig. 9a–d). In contrast, the effect of the GSK3β inhibitor on NEK2 and PD-L1 expression was not very stable and significant in pancreatic cancer cells (Supplementary Fig. 9e, f). These results suggest that NEK2 plays a more important role than GSK3β in the regulation of PD-L1, at least in pancreatic cancer.

a, b Western blot analysis and flow cytometry of PD-L1 expression in pancreatic cancer cell lines after treatment with NEK2 inhibitor (10 μM, 24 h) and NEK2 knockdown. Representative image is shown n = 3 independent experiments. c LC–MS proteomics quantitative analysis of pancreatic cancer cells overexpressing NEK2. d Western blot analysis of PD-L1 expression in KPC cells treated with NEK2 inhibitor (10 μM, 24 h) after treatment with MG132 (100 μM, 24 h). Representative image is shown n = 3 independent experiments. e Western blot analysis of PD-L1 expression in WT and NEK2 KD KPC cells treated with MG132 (100 μM, 24 h). Representative image is shown n = 3 independent experiments. f Stability analysis of PD-L1 in KPC cells treated with NEK2 inhibitor (10 μM, 24 h) after treatment with cycloheximide (CHX) (20 μg/mL). Representative image is shown n = 3 independent experiments. g Stability analysis of PD-L1 in WT and NEK2 KD KPC cells treated with CHX (20 μg/mL). Representative image is shown n = 3 independent experiments. h, i Statistical analysis of three independent experiments is displayed. The intensity of PD-L1 protein expression was quantified using a densitometer. j Ubiquitination assay of PD-L1 in KPC cells treated with NEK2 inhibitor (10 μM, 24 h), subjected to anti-PD-L1 IP and anti-ubiquitin Western blot analysis after treatment with MG132 (50 μM, 24 h). k Ubiquitination assay of PD-L1 in WT and NEK2 KD KPC cells treated with MG132. Results represent means ± SD of one representative experiment in h and i. All data are representative of three independently performed experiments. *P < 0.05, **P < 0.01, ***P < 0.001 using a two-tailed t-test; ns: not significant.

Full size image

NEK2 maintains the stability of PD-L1 by phosphorylation at T194/T210

In addition to upregulation of PD-L1 expression, its glycosylation at the N192/N200/N219 residues was also stimulated by NEK2 overexpression (Supplementary Fig. 10a–c). The influence of glycosylation on PD-L1 stability can be regulated by kinase-mediated phosphorylation, suggesting NEK2 controls the protein level of PD-L1 in such a manner. Two evolutionarily conserved NEK phosphorylation-specific motifs (F/LXXS/T) were identified across multiple species, centering at the T194 and T210 residues (Mus musculus: T193 and T209) in the glycosylation-rich region (N192/N200/N219) (Fig. 6a). Specific antibodies designed to recognize PD-L1 phosphorylated at T194 and T210 (T193/T209 in the mouse) were specifically generated to detect potential modification of NEK2 on PD-L1 (Fig. 6b). A significant decrease in T193/T209 phosphorylation levels of PD-L1 was observed in the presence of the NEK2 inhibitor and in NEK2 KD cells (Fig. 6c). In vitro kinase assay demonstrated that NEK2 directly phosphorylated PD-L1 (Fig. 6d). To further investigate the influence of NEK2-mediated phosphorylation on PD-L1 expression levels, constitutively non-phosphorylated (T193/209A) and phosphorylated (T193/209D) mutants of PD-L1, in addition to a kinase-dead (K37R) mutant of NEK2, were individually constructed. The results demonstrated that the T193/209A mutation in PD-L1 reduced its expression but the T193/209D mutation increased its expression in pancreatic cancer cells, while the K37R mutation in NEK2 caused its loss of regulatory effect on the expression and phosphorylation of PD-L1 (Fig. 6e–g). Moreover, a shortened or extended half-life of PD-L1 was observed with T193/209A and T193/209D mutants, respectively (Supplementary Fig. 11a–d), while no effect is observed with the NEK2 K37R mutant on PD-L1 half-life (Supplementary Fig. 11e, f). Furthermore, increased and decreased ubiquitination of PD-L1 were observed in PD-L1 T193/209A and T193/209D mutants, respectively (Fig. 6h), while no influence was consistently observed with the NEK2 K37R mutant on PD-L1 ubiquitination (Fig. 6i). Taken together, the results demonstrate that NEK2 maintains PD-L1 stability through phosphorylation at the T193/T209 residues.

a Schematic diagram of the NEK-binding motif (F/LXXS/T) in the glycosylation-rich region and amino acid sequences around the potential binding sites of PD-L1 were aligned in evolutionarily divergent species. The F/LXXS/T motifs are highlighted in blue. b Generation of site-specific antibodies against T194 and T210 (Mus musculus: T193 and T209)-phosphorylated PD-L1. c Western blot analysis of T193 and T209-phosphorylated PD-L1 in KPC cells with NEK2 inhibitor (10 μM, 24 h) and NEK2 KD KPC. Representative image is shown n = 3 independent experiments. d In vitro kinase assay and western blot analysis of pT193-PD-L1 and pT209-PD-L1 expression of recombinant PD-L1 WT and NEK2 (active) protein. Representative image is shown, n = 3 independent experiments. e, fWestern blot analysis of PD-L1 expression in flag-PD-L1 WT and T193/209A or T193/209D-transfected KPC cells with or without MG132 treatment. Representative image is shown n = 3 independent experiments. g Western blot analysis of pT193-PD-L1, pT209-PD-L1, and PD-L1 expression in WT and K37R transfected NEK2 KD KPC cells. Representative image is shown n = 3 independent experiments. h Ubiquitination assay of PD-L1 in Flag-PD-L1 WT and T193/209A or T193/209D-transfected KPC cells, subjected to anti-PD-L1 IP and anti-ubiquitin Western blot analysis after treatment with MG132 (50 μM, 24 h). Representative image is shown n = 3 independent experiments. i Ubiquitination assay of PD-L1 in WT and K37R transfected NEK2 KD-KPC cells subjected to PD-L1 IP and Western blot analysis with anti-ubiquitin after treatment with MG132 (50 μM, 24 h). Representative image is shown n = 3 independent experiments.

Full size image

NEK2 inhibition sensitizes PD-L1-targeted pancreatic cancer immunotherapy

Since NEK2 regulates the stability of PD-L1, we hypothesized that inhibition of NEK2 may enhance the therapeutic efficacy of PD-L1-targeted drugs. Indeed, NEK2 inhibitor plus anti-PD-L1 antibody significantly improved the T cell-mediated cytotoxic effect in vitro compared with mono-NEK2 inhibitor or mono-anti-PD-L1 antibody (Supplementary Fig. 12a, b). In light of this, we additionally administrated NEK2 inhibitor and/or anti-PD-L1 antibody to KPC-bearing mice (Fig. 7a), after which both tumor volumes (Fig. 7b, c) and tumor weight (Fig. 7d) decreased significantly in mice that received the combined treatment. Additionally, the combination of NEK2 inhibitor and anti-PD-L1 antibody did not influence the weight of the mice (Fig. 7e). Additional flow cytometric analysis revealed a significant elevation in the quantity and activation of T cells infiltrated into the tumors in which both therapies were administered (Fig. 7f, g). Moreover, the combinatorial strategy was also evaluated in an orthotopic model (Supplementary Fig. 13a), where a similar outcome was observed (Supplementary Fig. 13b–d). Additionally, it was confirmed that the NEK2 inhibitor functionally inhibited NEK2 activity in mice (Supplementary Fig. 13e). Collectively, the results suggest that the NEK2 inhibitor enhanced the efficacy of PD-L1 blockade both in vitro and in vivo, alleviating immuno-resistance in pancreatic cancer.

a Schematic protocol of the combination of anti-PD-L1 antibody and NEK2 inhibitor therapy. bTumor growth curve of mice treated with anti-PD-L1 antibody (200 μg/mouse), NEK2 inhibitor (100 μg/mouse), or their combination (n = 5). c Representative images displaying tumors harvested from mice bearing KPC cells treated with anti-PD-L1 antibody, NEK2 inhibitor, or their combination (n = 5). d, e Tumor weight and mouse body weight (n = 5). f, g Flow cytometric analysis and statistical results of lymphocytes that have infiltrated the tumors (n = 5). Results are presented as means ± SD from one representative experiment in b, d, e, and g. All data are representative of three independently performed experiments. *P < 0.05, **P < 0.01, ***P < 0.001 using a two-tailed t-test; ns: not significant.

Full size image

Increasing evidence suggests that conventional mediators of the cell cycle are also involved in cancer immune regulation. For instance, cyclin-dependent kinase 4, 5, and 6 (CDK4/5/6), three serine/threonine kinases that are conserved throughout eukaryotes, act as the fundamental driving forces for cell cycle progression and thus play an important role in the occurrence and development of multiple malignant tumors31,32. Intriguingly, it has been reported that CDK4/6 inhibitors not only suppress tumor cell growth by inducing cell cycle arrest, they also trigger tumor immune evasion through inhibition of cullin 3–SPOP E3 ligase-mediated proteasomal degradation of PD-L1 that increases the abundance of the protein33. Furthermore, combination therapy of CDK4/6 inhibitors and PD-1/PD-L1 blockade antibodies improve the efficacy of immunotherapy in multiple cancer types32,33,34,35. CDK5 regulates the expression of PD-L1 via the interferon regulatory factors IRF2 and IRF2BP2 in medulloblastoma36. As mentioned above, NEK2 is a multifunctional protein involved in non-immune and immune function including cell cycle regulation, microtubule stabilization, kinetochore attachment, spindle assembly checkpoint, phosphorylation of downstream proteins, and the maintenance of normal development and function of B cells, and has been associated with tumor progression and clinical prognosis in multiple cancers. In the present study, NEK2 has been identified as another critical regulatory factor in cancer immune resistance. Overall, these findings provide a theoretical basis for the development of an anti-cancer treatment strategy combining cell cycle-related kinase inhibitors with PD-1/PD-L1-targeted immunotherapy.

PD-L1 is expressed in a variety of cells, especially cancer cells and macrophages, and is critical for tumors to escape immune surveillance37. To date, PD-1/PD-L1 blockade has become the most successful strategy in cancer immunotherapy against solid tumors38. However, the overall response rate for PD-1 or PD-L1 blockade rarely exceeds 40%39. One reason that accounts for this phenomenon is the PTM of PD-L1, which potentially impacts the therapeutic efficacy of PD-L1 blockade. The mechanism of PD-L1 PTM has been extensively investigated and multiple regulators affecting PD-L1 stability via PTMs have been identified, providing methods to improve the efficacy of PD-L1 blockade15,40,41,42. For instance, EGF signaling induces PD-L1 glycosylation, while Gefitinib (an EGFR inhibitor) sensitizes PD-L1 blockade therapy in a syngeneic mouse model15. Based on these findings, the present study further demonstrated that NEK2 is a critical regulator of PD-L1 phosphorylation, and is involved in tumorigenesis and the progression of pancreatic cancer. NEK2 directly interacts with and phosphorylates PD-L1 to maintain its stability. The combination of a NEK2 inhibitor and anti-PD-L1 antibody synergistically suppresses tumor growth both in vivo and in vitro, thus representing the potential for sensitizing PD-L1-targeted therapy in pancreatic cancer.

Given the considerable influence of the tumor immune microenvironment on immune checkpoint blockade (ICB) therapy, we investigated the prognostic roles of NEK2 in pancreatic cancers with different immune statuses. NEK2 was significantly correlated with the prognosis of patients of immunologically “hot” status, while no significant correlation was observed in immunologically “cold” PDAC. The presence of tumor-killing T cells demonstrates the capability of the immune system to destroy tumors and its sensitivity to ICB43. PDAC is characterized by substantial immunological heterogeneity and differential levels of T cell infiltration44,45,46. “Hot” tumors typically respond well to checkpoint treatments47,48,49. Hence, NEK2 may be considered a therapeutic biomarker for ICB, and intervention with NEK2 may enhance the efficacy of PD-L1 blockade in “hot” pancreatic tumors. In addition to T cells, a great number of immunosuppressive populations are present in PDAC, also greatly influencing the efficacy of checkpoint therapy. Interestingly, the expression of both NEK2 and PD-L1 was also observed in macrophages, DCs, and MDSCs in addition to tumor cells37. In light of this, we speculated that NEK2 may also regulate the expression of PD-L1 in these suppressive populations. TMB is a measure of the number of somatic mutations in tumors and can be treated as a biomarker that reflects the potential therapeutic efficacy of immunotherapy, allowing the identification of patients who may benefit50. A high mutation load correlates with increased numbers of neoantigens, leading to T cell immunoreactivity and sensitivity to ICB treatments51,52. Interestingly, the expression levels of NEK2 can be used to predict the prognosis of patients with low TMB, suggesting that inhibition of NEK may be a promising strategy for increasing the effectiveness of immunotherapy in tumors with low TMB.

The effect of phosphorylation on the regulation of PD-L1 stability is complex, indirect, and sometimes even contradictory, but ultimately closely associated with modification by glycosylation. Recently, Hung’s group revealed that IL-6-activated JAK1 phosphorylates PD-L1 at Tyr112, catalyzing PD-L1 glycosylation through recruitment of endoplasmic reticulum-associated N-glycosyltransferase STT3A and so maintaining its stability53. Glycosylation of PD-L1 at N192, N200, and N219 contributes to PD-L1 stability and is important for its immunosuppressive function. Therefore, it is possible to generate a therapeutic antibody (STM108-ADC) that specifically recognizes the N192 and N200 glycosylation sites of PD-L115,42. However, it was revealed in another study that phosphorylation of PD-L1 S195 can induce abnormal glycosylation and prevent the translocation of PD-L1 to the cell membrane and subsequently, PD-1 binding, increasing the activity of CTL16. In the present study, we identified the T194/T210 residues of PD-L1 as targeting sites for NEK2-mediated phosphorylation. Since the newly revealed T194 site is close to S195, its phosphorylation may interrupt S195 modification, potentially promoting normal PD-L1 glycosylation, thus increasing its stability. Nonetheless, specific drugs that target the T194/T210 sites of PD-L1 and analysis of their specificity and clinical safety remain to be defined in the future.

In conclusion, NEK2 is a prognostic factor of immunologically “hot” and low mutational pancreatic cancer. NEK2 interacts with and phosphorylates PD-L1 at the T194/T210 residues, maintaining its stability, and thus inhibition of NEK2 leads to expedited degradation of PD-L1, activating cytotoxic T cells to obliterate pancreatic cancer cells (Fig. 8). Consequently, while further preclinical studies and clinical trials are warranted for demonstration, NEK2 is nevertheless a promising target for pancreatic cancer therapy, and the combinatorial inhibition of NEK2 and PD-L1 will synergistically benefit pancreatic cancer patients.

A schematic model is proposed to illustrate how PD-L1 protein stability is regulated by NEK2 in pancreatic cancer. NEK2 positively regulates and interacts with PD-L1 largely through PD-L1 phosphorylation at the T194/T210 residue in ER of pancreatic cancer. Therefore, treatment with NEK2 inhibitor unexpectedly suppressed PD-L1 protein expression, largely by inhibition of PD-L1 phosphorylation that promotes its degradation.

Full size image