CDX2 enhances natural killer cell–mediated immunotherapy against head and neck squamous cell carcinoma through up‐regulating CXCL14

Abstract (NK) cells are at the first line of defence against tumours, but their anti‐tumour mechanisms are not fully understood. We aimed to investigate the mechanism by which NK cells can mediate immunotherapy against head and neck squamous cell carcinoma (HNSCC). We collected fifty‐two pairs of HNSCC tissues and corresponding adjacent normal tissues; analysis by RT‐qPCR showed underexpression of CXCL14 in HNSCC tissues. Primary NK cells were then isolated from the peripheral blood of HNSCC patients and healthy donors. CXCL14 was found to be consistently under‐expressed in the primary NK cells from the HNSCC patients. However, CXCL14 expression was increased in IL2‐activated primary NK cells and NK‐92 cells. We next evaluated NK cell migration, IFN‐γ and TNF‐α expression, cytotoxicity and infiltration in response to CXCL14 overexpression or knockdown using gain‐ and loss‐of‐function approach. The results exhibited that CXCL14 overexpression promoted NK cell migration, cytotoxicity and infiltration. Subsequent in vivo experiments revealed that CXCL14 suppressed the growth of HNSCC cells via activation of NK cells. ChIP was applied to study the enrichment of H3K27ac, p300, H3K4me1 and CDX2 in the enhancer region of CXCL14, which showed that CDX2/p300 activated the enhancer of CXCL14 to up‐regulate its expression. Rescue experiments demonstrated that CDX2 stimulated NK cell migration, cytotoxicity and infiltration through up‐regulating CXCL14. In vivo data further revealed that CDX2 suppressed tumorigenicity of HNSCC cells through enhancement of CXCL14. To conclude, CDX2 promotes CXCL14 expression by activating its enhancer, which promotes NK cell–mediated immunotherapy against HNSCC.


| INTRODUC TI ON
Head and neck squamous cell carcinoma (HNSCC), which usually arises from the mouth, nose and throat, with metastasis to neighbouring lymph nodes, is the 6th most common cancer in the world. 1,2 Besides, smoking and alcohol consumption are the major risk factors to cause HNSCC, and additionally, human papillomavirus (HPV) infection and TP53 mutation are also associated with HNSCC. 3 Every year, around 500 000 cases are newly diagnosed with HNSCC globally, with approximately 40 000 new cases and consequent 7890 deaths occurring annually in the United States. 4 Surgery, radiation and chemotherapy, along or in combination with targeted therapy, are used for HNSCC therapy. Given the cytotoxicity of conventional treatment and their failure in the treatment of recurrent and/or metastatic stages of HNSCC, cell-based immunotherapies are being explored as promising strategies for the treatment of HNSCC. 5 Cell-based immunotherapy refers to a treatment strategy that uses the immune system of patients to fight against infections and other diseases such as cancer through a targeted activation or repression of their white blood cells and lymphatic system. 6 In this scenario, immune mediators including natural killer (NK) cells, dendritic cells (DC), macrophages and cytotoxic T lymphocytes (CTLs) protect the host and eliminate tumours by targeting surface antigens of tumour cells. 7 Natural killer (NK) cells are so-named due to their innate and cytotoxic characteristics, which were initially identified as a population of spleen-derived cytotoxic lymphocytes that were neither B nor T cells. 8 The NK cells function as the first line of defence against pathogens and tumours through their effects on death-receptor pathways and granule exocytosis, which are similar to those of CTLs. 9 The chemokine (C-X-C motif) ligand 14 (CXCL14), also known as BRAK, is expressed in a wide range of normal cells, but is especially abundant in epithelial cells. CXCL14 is also expressed in immune cells, where it is involved in immune surveillance by recruiting NK, dendritic cells and T cells. 10 In this paper, we investigated the underlying molecular mechanism whereby CXCL14 mediates NK cells to target HNSCC.
Our research findings reveal that caudal type homeobox 2 (CDX2), upstream regulator of CXCL14, activates CXCL14 enhancer to up-regulate its expression and enhance the therapeutic efficacy of immunotherapy against HNSCC by NK cells. Furthermore, the CDX2 pathway also provides a potential immunotherapeutic target to improve the treatment of advanced stages of HNSCC. The protocols involved in this paper were approved by the Ethics Committee of Jilin University Second Hospital. All procedures in this study involving humans were in accordance with the Declaration of Helsinki. Before sample collection, all participants provided written informed consents for entry in the research project.

| Natural killer cell isolation and culture
Peripheral blood mononuclear cells (PBMCs) from 10 healthy volunteers and 10 HNSCC patients were isolated according to the manufacturer's protocol of the Ficoll-Paque Premium kit (GE healthcare; VWR), and NK cells were isolated using a Miltenyi NK cell isolation kit (LS) (Miltenyi Biotech) following manufacturers' procedures.
Purity of the NK cells exceeded 85% and contained less than 1% CD3 + T cells. Isolated NK cells were cultured in a human NK cell culture medium with R10 + 5% human AB serum (Sigma).
The NK92 cell line was purchased from cell bank of China Academy of Sciences and was grown in a Minimum Essential Medium (MEM) α medium (Thermo Fisher Scientific) supplemented with 10% foetal bovine serum (FBS) (Thermo Fisher Scientific), 1.5 g/L NaHCO 3 , 2 mM L-glutamine, 0.2 mM inositol, 0.02 mM folic acid, 0.1 mM 2-mercaptoethanol and 100 U/ml recombinant human interleukin-2 (IL-2). Cells were cultured in a 37℃ incubator under standard conditions with 5% CO 2 .

| RT-qPCR
Total RNA from cells or tumour tissues was extracted using TRIzol reagent (Invitrogen). Agarose gel electrophoresis was used to examine the integrity and purity of isolated RNA. Reverse transcription was done using a Primescript™ RT Reagent Kit (RRO37A; TaKaRa). qPCR was performed with an ABI 7500 instrument (Applied Biosystems) to amplify targeted genes and internal control sequences. The PCR system was set up for 25 μL volumes, and GAPDH served as internal control. The 2 −ΔΔCt method was used to quantify the relative expression of mRNA. 13 Primers used for qPCR are listed below (Table 1).
Next, the membrane was washed three times with Tris-buffered saline Tween-20 (TBST) and incubated with horseradish peroxideconjugated secondary antibody (ab6721, 1:2000; Abcam) for 1 hour at room temperature. After being washed three times with TBST, the membrane was developed using enhanced chemiluminescence (ECL) (BB-3501; Ameshame) and photographed with a Bio-Rad imaging system (Bio-Rad Laboratories). Images were analysed by Quantity One v4.6.2 software by quantifying the grey value of the band with GAPDH as loading control. Experiments were repeated in triplicate.

| Immunoprecipitation analysis
Nuclear protein extraction was performed as previously described. 14 Protein G Dynabeads (Invitrogen) used for Immunoprecipitation

| Enzyme-linked immunosorbent assay
The supernatant of NK cell cultures was collected and transferred to an Eppendorf (EP) tube, followed by centrifugation at 8000 g for 8 minutes. The amount of IFNγ and TNFα in the supernatant was measured according to the protocol of the Enzyme-linked immunosorbent assay (ELISA) kit (Abcam).

| Tumour spheroid assays
In order to study the infiltration of NK cells, SCC-25 cell line was used for tumour spheroid assays. SCC-25 and IL-2-stimulated primary NK cells were seeded with a 1:1 ratio into agarose coated 96-well plates.

| Immunofluorescence
Mouse tumours were fixed in 4% (w/v) PFA for 3 hours at 4°C, followed by 30% (w/v) sucrose incubation overnight. The next day, samples were embedded in OCT (Tissue-Tek), frozen on dry ice, cut into 10μm-thick sections with a cryomicrotome and stored at −20°C until staining. Slides were air-dried for 20 minutes at room temperature and incubated with PBS to remove the OCT. Samples were then incubated in PBS/ 0.2% Triton X-100 for permeabilization at room temperature for 10 minutes. Next, the samples were blocked with 10% goat serum and 0.1% Triton X-100 for 1 hour at room temperature, followed by addition of diluted anti-CD45 primary antibody (ab10558, 1:1000; Abcam) for incubation in staining buffer at 4°C overnight. The next day, samples were washed with staining buffer and then incubated with diluted secondary antibody (ab175475, 1:200; Abcam) incubation in staining buffer for 1 hour in the dark, followed by washing with staining buffer. DAPI was used for counterstaining, and samples were washed with PBS and deionized water.
Immunofluorescence (IF) images were captured under an inverted fluorescent microscope (Eclipse Ti), which were quantified by ImageJ (NIH) in at least four randomly selected individual fields of view per sample.

| Statistical analysis
All data in this paper were processed using SPSS 21.0 (IBM Corp., Armonk, NY, USA). Measurement data were presented as mean ± standard deviation (mean ± SD). Data from tumour and adjacent normal tissues that were normally distributed were analysed by paired t test, while data of other two groups were compared using unpaired t test. Data among multiple groups were processed via one-way analysis of variance (ANOVA) and Tukey's post hoc test, and tumour volume data from different time-points were analysed by repeated measures ANOVA. Statistical difference was presented as P < .05.

| CXCL14 expression is down-regulated in HNSCC and the derived NK cells
It has been reported that chemokine CXCL14 expression correlates with the prognosis of colorectal cancer after resection. 16 Figure 1B), which suggested that CXCL14 was poorly expressed in the inactivated NK cells. NK cell-mediated anti-tumour effect has been shown to depend on IL-2, which is known as a survival factor of NK cells and an enhancer of the cytotoxic properties. 18 Next, we applied IL-2 to stimulate the primary NK cells isolated from healthy donors and NK-92 cells to study the function of CXCL14 on NK cells. RT-qPCR analysis revealed that CXCL14 expression was considerably increased in IL-2-activated primary NK cells and in NK-92 cells compared with untreated control cells ( Figure 1C). Altogether, these data indicated a down-regulation of CXCL14 in HNSCC and the derived NK cells.

| CXCL14 enhances NK cell migration and cytolytic activity
To study whether CXCL14 participates in cancer immunotherapy by NK cells, we first knocked down or overexpressed CXCL14 using TNFα were elevated by CXCL14 overexpression; however, the levels of these two factors in the medium were remarkably suppressed by CXCL14 silencing ( Figure 2D).
NKG2D is an important protein associated with the cytotoxic activity of NK cells. 19 Therefore, we performed Western blot assay to detect the expression of NKG2D in NK cells. The data revealed that NKG2D expression was considerably increased by oe-CXCL14 and decreased by sh-CXCL14-1 ( Figure 2E). Next, we evaluated the cytotoxicity of IL-2-stimulated NK cells against HNSCC, CAL27 and SCC-25 cells, and found that the cytotoxicity of IL-2-stimulated NK cells against HNSCC cells was significantly enhanced by oe-CXCL14, but was reduced by sh-CXCL14 ( Figure 2F,G). Meanwhile, to study the infiltration of NK cells, we conducted tumour spheroid assay using the SCC-25 cell line. The immunohistochemistry results demonstrated that the number of infiltrated NK cells (CD45 + cells) was considerably increased by oe-CXCL14, whereas their number was remarkably decreased by sh-CXCL14-1 ( Figure 2H). These data revealed that CXCL4 increased NK cell migration and cytolytic activity.

| CXCL14 inhibits tumorigenesis
To study the role of CXCL14 in HNSCC and whether this role was associated with effects of NK cells in vivo, we established a tumourbearing mouse model using SCC7 cells and then injected CXCL14 overexpression (oe-mCXCL14) or knockdown (sh-mCXCL14) lentivirus or corresponding control vectors into the mice. At three weeks after injection, the expression of CXCL14 mRNA in tumours was determined by RT-qPCR, which showed that CXCL14 expression was significantly up-regulated in the tumour-bearing mice injected with oe-mCXCL14, but notably down-regulated in those injected with sh-mCXCL14 ( Figure 3A). In addition, it was also found that the tumour Note: The data were analysed by chi-square test.
Abbreviations: HNSCC, head and neck squamous cell carcinoma; N0, no regional lymph node metastasis; N1, unilateral cervical and/or unilateral or bilateral retropharyngeal node (s), ≤6 cm in greatest dimension, above supraclavicular fossa; N2, bilateral cervical node (s), ≤6 cm in greatest dimension, above supraclavicular fossa; T1, nasopharynx, oropharynx or nasal cavity; T2, parapharyngeal extension; T3, bony structures and/or paranasal sinuses; TNM, tumour node metastasis. We then used IF to detect the NK cell infiltration in tumours, which demonstrated that the number of infiltrating CD45 + cells in the tumour-bearing mice was remarkably elevated in response to CXCL14 overexpression, but was decreased in response to CXCL14 knockdown. In the meantime, the number of CD45 + cells elevated by CXCL14 was notably decreased after treatment with anti-asialo GM1 ( Figure 3D). Additionally, the NKG2D protein expression was measured by Western blot assay. The results showed that CXCL14 overexpression led to increased NKG2D protein expression, while CXCL14 knockdown caused a reduction in its expression. However, the CXCL14-induced elevation of NKG2D protein expression was repressed by anti-asialo GM1 treatment ( Figure 3D,E). These results revealed that CXCL14 considerably inhibited the tumour growth in mice which was possibly mediated by NK cells. RT-qPCR showed that CDX2 and CXCL14 expression was notably increased by oe-CDX2, while the elevated CXCL14 expression by oe-CXCL14 was suppressed by sh-CDX2-1 ( Figure 4E). Furthermore, Western blot assay revealed no difference in CDK2/CXCL14 ratio among NK cells of different densities ( Figure S1C). In summary, these results revealed that CDX2 positively regulated CXCL14 expression.

| CDX2 increases CXCL14 expression via activating its enhancer in primary NK cells
It has been reported that PAX-6 and CDX2 interact with coactivator p300 to synergistically activate glucagon gene expression. 20 Thus, we speculated that CDX2 might regulate CXCL14 expression by interacting with the co-activator p300. To test this prediction, we used an speculation, IP assay to determine the interaction between CDX2 and p300 in NK cells, which showed that CDX2 indeed interacted with p300 ( Figure 4F). Next, to investigate the role of p300 in the regulation of CXCL14 by CDX2, we used lentivirus to knock down p300.
Meanwhile, we silenced p300 in the presence of CDX2 and conducted RT-qPCR analysis, which revealed that CDX2 expression was not affected by sh-p300, but that CXCL14 expression was significantly down-regulated by sh-p300 in the presence of CDX2 ( Figure 4H).
Previous investigations demonstrated that CDX2 could bind to YAP1 enhancer to regulate its expression. 21  , and p300 in NK cells. *P < .05 compared with oe-NC group, and # P < .05 compared with sh-CDX2-1 + oe-NC group. (F) IP assay to analyse the interaction between p300 and CDX2 in primary NK cells. (G) Western blot assay to analyse the expression of p300 after knocking p300 down in primary NK cells. *P < .05 compared with sh-NC group. (H) RT-qPCR to detect the expression of CDX2, CXCL14 and p300. *P < .05 compared with oe-CDX2 + sh-NC group. (I) ChIP to analyse the effect of CDX2 on the enrichment of H3K27ac, p300 and H3K4me1 in the enhancer of CLCX14. (J) Western blot assay to analyse the expression of H3K27ac and H3K4me1. *P < .05 compared with oe-NC group. Measurement data were presented as mean ± standard deviation. Data from tumour and adjacent normal tissues that were normal distributed were analysed by paired t test, while data of two groups were compared using unpaired t test. Data among multiple groups were compared via one-way ANOVA and Tukey's post hoc test. Experiments were repeated 3 times hypothesized that CDX2-p300 complex might activate the enhancer through binding to CXCL14 enhancer. Since H3K27ac and H3K4me1 are enhancer markers, we applied ChIP assay to detect the enrichment of CDX2, H3K27ac, p300 and H3K4me1 in the enhancer region of CXCL14. Results demonstrated that the enrichment of CDX2, H3K27ac, p300 and H3K4me1 in the enhancer region of CXCL14 was notably increased upon overexpression of CDX2 ( Figure 4I), whereas the expression of H3K27ac and H3K4me1 was not affected ( Figure 4J). In the meantime, no significant difference was detected in the expression of H3K27ac and H3K4me1 in NK cells of peripheral blood from healthy donors and HNSCC patients ( Figure S1B).

| CDX2 enhances NK cell migration and cytolytic activity via up-regulating CXCL14 expression
We next investigated the effects of CXCL14-mediated CDX2 on the migration of NK cells and cytolytic activity. Our Transwell assay results revealed that NK cell migration was significantly increased by oe-CDX2, but notably decreased by sh-CDX2-1. However, the NK cell migration suppressed by sh-CDX2-1 was substantially rescued by oe-CXCL14 ( Figure 5A). Then, ELISA results revealed that the expression of IFNγ and TNFα in the medium was significantly increased upon CDX2 overexpression, yet notably decreased by CDX2 knockdown, while the reduction in expression of IFNγ and TNFα caused by sh-CDX2-1 was reversed by oe-CXCL14 ( Figure 5B).
Western blot results demonstrated that NKG2D expression and PD-1 expression were remarkably increased after CDX2 overexpression, but considerably decreased after CDX2 silencing. The reduced NKG2D expression and PD-1 expression induced by sh-CDX2-1 were rescued by oe-CXCL14 ( Figure 5C).
Next, we investigated the cytotoxic effect of IL-2-stimulated NK cells on HNSCC cell lines, CAL27 and SCC-25. The cytotoxicity of IL-2-stimulated NK cells against HNSCC cells was significantly increased by oe-CDX2, but notably decreased by sh-CDX2-1.
Finally, we performed tumour spheroid assays using the SCC-25 cell line in conjunction with immunohistochemistry. Results indicated that the number of infiltrated NK cells in SCC-25 tumour spheroids was significantly increased after CDX2 overexpression, but was notably decreased after CDX2-1 knockdown. The reduced number of infiltrated NK cells in SCC-25 tumour spheroid caused by sh-CDX2-1 was rescued by oe-CXCL14 treatment ( Figure 5F).
These data revealed that CDX2 accelerated NK cell migration and enhanced cytolytic activity via up-regulating CXCL14 expression.

| CDX2/CXCL14 inhibits in vivo tumorigenesis
To investigate the role of CDX2/CXCL14 in HNSCC in vivo, we estab- was considerably down-regulated in the sh-mCDX2 group compared to that in the sh-NC group. However, compared with sh-mCDX2 + oe-NC, CDX2 expression was unchanged, and CXCL14 expression was notably up-regulated in the sh-mCDX2 + oe-mCXCL14 group ( Figure 6A).
As expected, the tumour volume and weight were significantly increased when CDX2 was knocked down, both of which effects were remarkably decreased after restoration of CXCL14 ( Figure 6B,C).
Meanwhile, the data showed that the number of infiltrating CD45 + cells was remarkably decreased by CDX2 knockdown in vivo, which effect was counteracted after restoration of CXCL14 ( Figure 6D).
Consistent with the in vitro results, NKG2D protein expression was reduced by CDX2 knockdown, while CXCL14 overexpression reversed this reduction ( Figure 6E). Taken together, our data revealed that CDX2/CXCL14 inhibited the growth of HNSCC in mice, which was associated with the function of NK cells.

| D ISCUSS I ON
Head and neck squamous cell carcinoma is a relative common cancer and one of leading causes of cancer death worldwide. NK cells serve as key players of innate immune system and exert important roles in the defence against tumours. 24 In this paper, our primary goal was to elucidate the molecular mechanism by which NK cells exert an We have now shown that CXCL14 expression was down-regulated in HNSCC and patient NK cells, which suggests its role in the devel-  our results also showed the up-regulation of natural killer group 2D (NKG2D) by CXCL14 overexpression. NKG2D is a receptor that is expressed on the surface of NK cells as well as T cell subsets, which facilitates crucial immunosurveillance roles by recognizing ligands that are usually overexpressed by cancerous cells. 33 It has been reported that NKG2D expression is correlated with NK cell infiltration, 34,35 which also supports our present result. The in vivo studies in the tumour-bearing mouse model also showed that overexpression of CXCL14 can inhibit the tumour growth and increase NK cell infiltration.
Next, we ask whether there is any upstream molecule that regulates the expression of CLCX14. To answer this question, we conducted bioinformatics analysis, which predicted CDX2 as an upstream regulator of CLCX14. CDX2 belongs to homeobox transcription factor family, which was initially identified as an important transcription factor for the development of intestinal epithelium. 36 In colorectal cancer, CDX2 expression was found to be down-regulated. 37 Present results show that CDX2 expression is also down-regulated in HNSCC and NK cells, while overexpression of CDX2 could increase the expression of CXCL14. Molecular biological analysis further revealed that CDX2 interacts with histone acetyltransferase p300 and that CDX2/p300 activates the enhancer of CXCL14 to promote its expression. Our data thus uncover the potential molecular mechanism by which CXCL14 is regulated. Finally, we investigated whether CDX2 contributes to the anti-tumour potential of NK cells. Our in vitro and in vivo data consistently indicate that CDX2 induces the migration and infiltration of NK cells, increases the secretion of IFNγ and TNFα by NK cells, enhances NK cell toxicity against HNSCC, and suppresses the tumour growth in mice.
Taken together, our research elucidates the molecular mechanism how the CDX2/CXCL14 axis enhances NK cell-mediated immunotherapy against HNSCC, in which CDX2 stimulates the cytotoxicity of NK cells through up-regulation of CXCL14 (Figure 7).
Our study may also inform a novel immunotherapeutic strategy to be translated into the treatment of HNSCC. However, the possibility of detrimental immunotherapy side effects such as attack of healthy cells or tissues calls for careful consideration in future investigations.

ACK N OWLED G EM ENTS
We would like to acknowledge the reviewers for their helpful comments on this paper.

CO N FLI C T O F I NTE R E S T
The authors declare that there is no conflict of interest. Resources (equal).

DATA AVA I L A B I L I T Y S TAT E M E N T
The data sets generated and/or analysed during the current study are available from the corresponding author on reasonable request.