Prognostic value of hypoxia-responsive gene expression profile in patients diagnosed with head and neck squamous cell carcinoma

Highlights • The mRNA expression of hypoxia-responsive genes is head and neck squamous cell carcinoma (HNSCC)-specific.• KIF14 mRNA is markedly altered in the group of patients responding to radiotherapy, which was reflected in the clinical outcome of HNSCC patients.• Silencing of KIF14 reverses its radiosensitizing capability.


Introduction
Head and neck cancer is the seventh most common cancer worldwide [1].90 % of head and neck cancers are squamous cell carcinomas (HNSCC) that originate from the oral cavity, oropharynx, larynx, hypopharynx, and nasopharynx.The major risk factors behind HNSCC are use of tobacco, alcohol, and exposure to high-risk human papilloma virus.The treatment modalities of HNSCC depends on the TNM stage and primary tumor site [1,2].A single modality treatment, surgery or radiotherapy is used for early-stage tumor.Advanced stages of the disease require multimodal treatment based on radiotherapy combined with cisplatin-based chemotherapy.Despite the progress in treatment strategies, advanced HNSCC still yields a crude prognosis (5-year overall survival, <50 %), carries a high risk of distant metastasis and local recurrence [3].Therapy failure is often associated with radioresistance and its effectiveness is greatly attenuated by hypoxia within the tumoral tissue [4,5].
HNSCC has shown significantly worse outcome in hypoxic tumors compared to tumors expressing less hypoxic features [6].The importance of hypoxia-induced radioresistance has been acknowledged for decades pointing at complexity of interaction between tumor hypoxia and radiotherapy in the search for radiosensitizing agents [4].Furthermore, hypoxic tumors have shown to exhibit aggressive tumor phenotypes with increased risk of progression, metastasis, and recurrence [7,8].To combat the hypoxic state, cells have acquired several adaptive mechanisms, in which hypoxia inducible factor (HIF) pathway is central.Hypoxic tumor microenvironment is a non-cellular factor that induces epithelial-mesenchymal transition (EMT) by HIF-1-mediated activation of Twist or Snail that are important drivers of EMT [9].EMT serves as one of the key factors affecting tumor cell proliferation, invasion and metastasis, angiogenesis or even cancer-stem cell phenotype of cancer cells [10].Hypoxia is believed to have impact on cancer stem cell (CSC) phenotype and leads to dysregulation of cancer stemness transcription factors such as Sox2 and Nanog [11].
Moreover, HIF mediates the transcription of several hundreds of genes implicated in cell division, angiogenesis, and cell metabolism as well as regulates expression of CA9 [12].CA9 has been associated with hypoxia and poor prognosis in a range of tumor types [13,14].In HNSCC, hypoxia is often correlated with changes in cellular metabolism and responses to radio-and chemotherapy.Moreover, genes responsible for glucose metabolism (GLUT1 and GLUT3) and angiogenesis (VEGF) are also activated by tumor hypoxia [15,16].
In our previous study [17], hypoxia was found to induce an increased survival capacity after radiation and an increase expression of EMT-and CSC-associated genes in HNSCC cell lines.Furthermore, a microarray analysis identified a panel of hypoxia-responsive genes and among genes highly upregulated by hypoxia were CA9, CASP14, LOX, GLUT3, SER-PINE1 and highly downregulated AREG, EREG, CCNB1 and KIF14.Moreover, further network analysis has revealed the involvement of the microarray-identified genes in cancer, cell death and survival as well as cellular assembly and organization.
In the present study we further evaluated the impact of the abovementioned genes in 32 HNSCC biopsies from patients treated with radiotherapy on treatment response and overall survival.

Patient material and clinicopathological data
The study material included 32 fresh-frozen, biopsy specimens obtained before initiation of radiotherapy from patients diagnosed with HNSCC.The use of patient material has been approved by the Ethical Committee of Linköping.The biopsies were collected between the years of 2003 and 2009.All patients received radiotherapy as accordingly: 13 (40.6 %) solely received radiotherapy, 15 (46.9 %) preoperative radiotherapy, 3 (9.4%) postoperative and 1 (3.1 %) received chemotherapy in addition to radiotherapy and surgery.The patient group was retrospectively divided as described in our preceding study [18,19] into responder and non-responder according to therapeutic response.If the tumor size was reduced during radiotherapy and no recurrent disease was observed with within 1 year following radiotherapy, a patient was considered a responder.A patient was considered a non-responder if the tumor grew during ongoing radiation treatment or patients had a relapse of disease within 6 months of radiation treatment.Within the above-mentioned panel of 32 HNSCC patients included in the analysis, 16 patients were identified as non-responders and 16 as responders (Table 1).
The non-responder group consisted of 3 (18.8%) samples from the gingiva, 4 (25 %) from the larynx, 6 (37.5 %) from the tongue, 1 (6.3 %) from the trigonum retromolare and 2 (12.5 %) from the buccal mucosa.The responder group consisted of 3 (18.8%) samples from the gingiva, 3 (18.8%) from the larynx, 5 (31.3) from the tongue, 1 (6.3 %) from the trigonum retromolare, 1 (6.3 %) from the buccal mucosa, 1 (6.3 %) from the hypopharynx, 1 (6.3 %) from the mucosa of the palate and 1 (6.3 %) from the floor of the mouth.Patient characteristics, primary tumor site, degree of differentiation and TNM staging are shown in Table 1.Noncancerous oral tissues, collected from the sites as far as possible from the tumor margin were included in the study as controls.

RNA extraction
Total RNA extraction from frozen tissues was performed using the AllPrep DNA/RNA/miRNA Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions.Tissues were disrupted and homogenized in RLT buffer containing β-marcaptoethanol and one stainless steel bead using TissueLyser II instrument.The RNA fraction was separated using RNeasy Mini spin columns.RNA was dissolved in RNase-free water and stored at − 70 • C. RNA concentration was measured using NanoDrop (TermoFisher, Massachusetts, USA).

Western blot
Whole cell extracts were prepared from the HNSCC cells using RIPA buffer for 30 min at 4 • C and the protein concentration was determined using the Bio-Rad DC Protein Assay, and 30 µg of total cell extracts was subjected to Western blotting.The membranes were incubated with anti-KIF14 antibody (1:1000; Abcam, UK), followed by a goat antirabbit antibody conjugated to HRP (1:5000; Santa Cruz Biotechnology, USA).The bands were visualized with the Western Blotting Luminol Reagent (Bio-Rad, USA).Equal loading was verified by reprobing the membranes with an HRP-conjugated anti-βactin antibody (1:2000; Santa Cruz Biotechnology, USA).

RNA interference
Cells were seeded at a density of 12 000 cells/cm 2 and were transfected 24 h later with 10 nmol/l FlexiTube siRNA (Cat.No 1,027,416; Qiagen, Germany) against KIF14 (SI02781324 and SI02781163; 1:1 ratio of the two siRNA clones was used for each transfection), or a nontargeting siRNA with no homology to any known human gene (AllStars Negative Control siRNA: 1,027,280) with the HiPerFect transfection reagent (Qiagen, Germany).Knockdown was verified by RT-qPCR and Western blot 24 h post transfection.For the cell proliferation study, cells were seeded into 12-well plates (BD Falcon, USA) and irradiated (2, 4 or 6 Gy) 48 h after transfection.

Cell proliferation assay
Tumor cells were seeded in 12-well plates at densities of 300-800 cells/cm 2 , depending on the plating efficiency of each cell line.Selected cells were irradiated (2, 4 or 6 Gy) with 4 MeV photons generated by a linear accelerator (Clinac 4/100, Varian, Palo Alto, USA), delivering a dose-rate of 2.0 Gy/min.Nine days after treatment, cells were fixed in 4 % paraformaldehyde (20 min), stained with crystal violet solution (0.04 % in 1 % ethanol) for 20 min at room temperature and then washed and air-dried.After solubilization in 1 % SDS, the optical density at 550 nm was measured using a Victor plate reader (EG & G Wallac).

Statistical analysis
Statistical analysis was performed using Prism 9.0 (GraphPad Software, Inc., La Jolla, CA, USA) and SPSS IBM version 27.0 (IBM Corporation, Armonk, NY, USA).The log2 transformed mRNA expression data were used in the analysis.Shapiro Wilk test was used to assess if the dataset is normally distributed.The high and low mRNA levels were assessed based on the median mRNA expression.The multiple comparison analysis was performed using one-way ANOVA followed by posthoc Tukey's multiple correction test.The mRNA expression of the genes of interest was analyzed for correlation with overall survival of the patients using the Mantel-Cox log-rank statistics in SPSS software and presented as Kaplan-Meier survival curves depending on low or high mRNA expression.Differences between two groups (non-targeting siRNA and KIF14 siRNA) were analyzed with the unpaired Student t-test.All values obtained were represented as mean ± SD of at least three independent experiments.*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

The expression of hypoxia-responsive genes in patients with HNSCC
In the search for potential biomarkers in HNSCC, we examined mRNA expression of five highly upregulated (CA9, CASP14, LOX, GLUT3, SERPINE1) and four highly downregulated (AREG, EREG, CCNB1 and KIF14) hypoxia-responsive genes in 32 HNSCC tumors and six adjacent normal oral tissue.Depending on response to radiotherapy, the patients were divided into responders (n = 16) or non-responders (n = 16) group.Non-tumoral oral tissues were used as control (n = 6).The mRNA expression analysis of the selected genes was assessed with RT-qPCR.
Our results show a significantly higher mRNA expression of the hypoxia marker CA9 and SERPINE1 in all tumor biopsies (responders and non-responders) compared to normal tissue.Although CA9 mRNA expression was lower in the responder group in comparison to nonresponder, the difference did not reach significance in the analyzed patient material.The SERPINE1 mRNA expression was slightly higher in responder group, however this difference was not significant.The mRNA expression of other highly upregulated genes (LOX and GLUT3) in hypoxic HNSCC cell lines was higher in HNSCC patients compared to normal tissues but the differences were not statistically significant.The mRNA expression of CASP14, highly upregulated gene in hypoxic HNSCC cell lines remained at similar level when compared non-tumoral and tumoral tissue.There were no significant differences regarding mRNA expression of analyzed hypoxia-upregulated genes between responder and non-responder groups (Fig. 1).
Regarding the hypoxia-downregulated genes, the mRNA expression level of AREG, CCNB1 and KIF14 did not demonstrate the same pattern as in hypoxic HNSCC cell lines.In turn, we observed higher expression of AREG mRNA in HNSCC patients with trend to higher expression in the responder group.Like AREG, KIF14 mRNA level was higher in HNSCC samples than that in control group.Interestingly, the responder group exhibited significantly higher level of KIF14 mRNA compared to the non-responder group.
The EREG mRNA expression was lower in both responder and nonresponder group compared to normal tissue, but its expression differed significantly between the responder and control group.The CCNB1 mRNA expression in HNSCC samples did not deviate much from nontumoral tissue with slight but not significant increase in the responder group (Fig. 2).

The impact of the candidate hypoxia-responsive genes on overall survival in HNSCC
We further evaluated the association of hypoxia-responsive genes with overall survival of the HNSCC patients.The obtained relative mRNA expression level was divided into low and high mRNA expression.The median value for log2 fold change was used as a cut-off, where values > median were considered as high mRNA expression and values < median were considered as low mRNA expression.We observed no statistically significant differences between low and high mRNA expression regarding the analyzed hypoxia-upregulated genes (Fig. 3).Regarding the hypoxia-downregulated genes (Fig. 4), we demonstrated that patients with high KIF14 mRNA expression showed significantly longer survival rate than patients with low KIF14 mRNA expression (p = 0.05).Interestingly, simultaneous high mRNA expression of KIF14 and low mRNA expression of CA9 correlated with noticeably better overall survival of HNSCC patients (p = 0.01; Fig. 5).Furthermore, the overall survival of patients with simultaneous high SERPINE1 and KIF14 mRNA (p = 0.021) as well as high AREG and KIF14 mRNA (p = 0.028) was better than patients with the opposite expression pattern (Fig. 5).

The expression of KIF14 in HNSCC
Since KIF14 mRNA expression was significantly higher in the analyzed group of HNSCC patients responding to radiotherapy, we next evaluated the expression of KIF14 on the mRNA and protein level in HNSCC cell lines that were classified as radioresistant or radiosensitive.Four radioresistant HNSCC cell lines (LK0532, LK0824, LK0827, LK0927) and five radiosensitive HNSCC cell lines (LK0412, LK0850, LK0863, LK0902, LK0949) were included in the study based on their previously established intrinsic radiosensitivity (IR) [20].
Our results show that higher expression of KIF14 revealed by RT-qPCR (Fig. 6A and 6B) and Western blot (Fig. 6C) analysis is not restricted to sensitive cell lines as compared to patient material.However, we observed a trend toward higher expression of KIF14 mRNA in the radiosensitive HNSCC cell lines.

The impact of KIF14 on radiotherapy response in HNSCC
To explore the potential radiosensitizing role of KIF14, we implemented siRNA-mediated downregulation of KIF14 and evaluated its effect on radiotherapy response in the radiosensitive cell lines.We observed a significant decrease in HNSCC cell growth after downregulation of KIF14 compared to control cells (Fig. 7A).Moreover, silencing of KIF14 was associated with decreased sensitivity towards irradiation in the analysed radiosensitive HNSCC cell lines (Fig. 7B-D).

Discussion
Treatment modalities for HNSCC are heavily based on radiotherapy combined with chemotherapy or surgery.Given the everyday functions in the head and neck region, the consequences of HNSCC and its treatment have a large impact on health-related quality of life [21].The choice of initial treatment, administration, and sequencing involves consideration of toxic effects, morbidity, and preservation of function [22].Identifying biomarkers that predict the response to radiotherapy may assist the choice of most effective treatment for the patient.
The response to radiation therapy is greatly hindered by hypoxia [4,17].We have recently identified a panel of hypoxia-responsive genes in HNSCC cell lines that might have impact on prognosis and therapy response in HNSCC patients [17].One of the highly upregulated genes in hypoxic HNSCC cells as well as analyzed patient material was CA9 [23,24].CA9 has been shown to be upregulated at both mRNA level and protein level in hypoxic tumors and correlates with poor prognosis in various types of cancer [13,25].Hypoxia has also been shown to negatively affect radiotherapy and chemotherapy in HNSCC patients [17,[26][27][28].Our results show a distinct increase in CA9 mRNA expression in HNSCC patients compared to non-tumoral tissues.The responder group exhibited markedly lower expression of CA9 mRNA than non-responder group, which might be associated with the radiotherapy response.Despite this, CA9 mRNA expression alone was not shown to be a predictor of overall survival in the analyzed patient material.In another study [29] involving 39 HNSCC samples, CA9 expression did not have a significant impact on overall survival but on local relapse free survival (LRFS).
Another highly hypoxia-responsive gene in HNSCC cell lines, namely SERPINE1showed to be upregulated in HNSCC patients.High expression of SERPINE1 has been associated with poor prognosis in head and neck patients [30].Moreover, SERPINE1 expression is involved in epithelial to mesenchymal transition (EMT), gain of stem cell properties and resistance to anti-cancer therapy in head and neck cancer [31].In our study, expression of SERPINE1 mRNA did not show to be independent prognostic marker, however high expression of SERPINE1 mRNA in combination with high KIF14 mRNA expression correlated with better overall survival in HNSCC.This observation points at potential interconnection between SERPINE1 and KIF14 that warrants further investigation.The hypoxia-induced mRNA expression of CASP14, GLUT3 and LOX in HNSCC cell lines was partially reflected in the analyzed patient cohort.These genes have been shown to drive carcinogenesis in many types of cancer [32][33][34][35].Upregulation of the CASP14 as a novel hypoxia-regulated gene bound by HIF1 has previously been observed in breast cancer [36] and our previous study in HNSCC cell lines has pointed at CASP14 as highly hypoxia-upregulated.Although hypoxia-specific marker CA9 was significantly higher in the analyzed patient material, CASP14 mRNA seemed not to be hypoxia-dependent.Such discrepancy might be a result of different signaling between 2D-grown HNSCC cells and 3D tumor mass.Moreover, our patient cohort is too small to draw conclusion from current observation.Hypoxia has also been shown to upregulate LOX expression thereby promoting invasion and metastasis of breast and ovarian cancer cell lines [37,38].Moreover, hypoxia leads to increase of glucose uptake by upregulating GLUT1 and GLUT3 and GLUT3 expression correlates with poor prognosis in various cancers including oral squamous cell carcinoma.The trend towards higher mRNA expression of LOX and GLUT3 in HNSCC patients was observed, however no correlation with prognosis of the HNSCC patients regarding the mRNA expression of these genes was noted [39][40][41][42][43].
Among the genes downregulated by hypoxia in HNSCC cell lines, a decreased mRNA expression level of EREG was also detected in the analyzed HNSCC patient material.EREG serves as a ligand of the epidermal growth factor receptor (EGFR) and its deregulation in different types of cancer leads to activation of EGFR signaling pathways and tumorigenesis [44].EREG has also shown to be involved in reprogramming of cancer associated fibroblasts (CAF) and EMT induction in oral squamous cell carcinoma [45].The results from CCNB1 mRNA analysis in HNSCC patients differ from the results observed from HNSCC cell lines.Deregulation of this mitosis-linked molecule has been shown in various types of cancer [46][47][48].In our study, we did not observe significant differences between CCNB1 mRNA expression in non-tumoral and tumoral tissue, specifically in the non-responder group.Hypoxia itself is associated with resistance to radiotherapy and hypoxia-mediated downregulation of CCNB1 could support this process.In another study, overexpression of CCNB1 in HNSCC tumors has been shown as a marker of resistance to radiotherapy and the risk of locoregional recurrence and metastasis in patients treated with radiotherapy [49].Whether expression of CCNB1 and its potential prognostic significance is hypoxia-specific must be analyzed in a bigger patient cohort.
Regarding the hypoxia-downregulated genes, these genes were highly expressed in the analyzed patient cohort.Like EREG, AREG is a ligand of the EGFR and in HNSCC high expression of AREG was associated with better overall survival and progression free survival as well as greater benefit from cetuximab combined with chemotherapy [50].Noteworthily, high expression of AREG mRNA in combination with high expression of KIF14 mRNA correlated with better overall survival in the analyzed patient cohort.Interestingly, KIF14 mRNA expression was significantly higher in the responder group that might be beneficial regarding the response to radiation therapy.KIF14 belongs to the kinesin-3 family of proteins that are involved in various processes including vesicle transport, chromosome segregation, mitotic spindle formation, and cytokinesis [51,52].There are many reports showing oncogenic nature of KIF14 and its correlation with poor prognosis [53,54].Overexpression of KIF14 mRNA has been observed in several types of cancer including oral cancer [55][56][57][58][59][60].On the other hand, many reports point at tumor-suppressive role of KIF14 in some tumors [61,62].In our study, patients with high KIF14 mRNA expression had longer overall survival than those with low KIF14 mRNA expression.Like our results, low expression of KIF14 mRNA has been related to worse overall survival in lung adenocarcinoma and colorectal cancer [61,63].Interestingly, the prognostic value of KIF14 mRNA expression in HNSCC patients was significantly increased by simultaneous high mRNA expression of KIF14 and low mRNA expression of CA9.CA9 is highly related to hypoxia and high levels of CA9 might accelerate survival of cancer cells and negatively affect response to treatment.It has been reported that high CA9 expression is associated with reduced overall survival in oral cavity squamous cell carcinoma and poor pathological T-stage [64,65].Increased expression of KIF14 mRNA was also observed in the responder group suggesting a predictive role of KIF14 in radiotherapy response.Radiotherapy treatment is known to be beneficial for rapidly dividing cells and KIF14 is involved in many biological functions including proliferation.High KIF14 mRNA expression in responder group seems to be beneficial and might indicate sensitization of tumor cells to radiotherapy.Moreover, only patients with low CA9 mRNA and high KIF14 mRNA expression showed better overall survival pointing at hypoxia as regulator of KIF14.
Regarding the fact that patients responding to radiotherapy have higher mRNA expression of KIF14 than non-responders, we next determined the potential role of KIF14 in radiotherapy response.We show that downregulation of KIF14 leads to inhibition of cell proliferation and at the same time cells with lower expression of KIF14 become more resistant to radiotherapy.The possible explanation could be that ionizing radiation is based on the premise that rapidly proliferating cancer cells are more sensitive to DNA damage compared to cells with slow proliferation rate.

Conclusions
In conclusion, our results indicate that the pattern of hypoxiaresponsive genes in HNSCC cell lines is HNSCC tumor-specific.We show also that mRNA expression of KIF14 is a potential diagnostic marker and might serve as a predictor of radiation treatment response in HNSCC.However, more studies using larger patient cohort are required to study the hypoxia-associated expression signatures and their utility in HNSCC diagnosis and treatment.Additionally, a more mechanistic insight into KIF14 in HNSCC biology and the implication for radiotherapy are needed.

Fig. 3 .
Fig. 3. Prognostic significance of hypoxia-upregulated genes in HNSCC patients.Kaplan-Meier log-rank survival analysis of HNSCC patients regarding different mRNA expression of previously identified, hypoxia-upregulated genes.

Fig. 4 .
Fig. 4. Prognostic significance of hypoxia-downregulated genes in HNSCC patients.Kaplan-Meier log-rank survival analysis of HNSCC patients regarding different mRNA expression of previously identified, hypoxia-downregulated genes revealed by microarray analysis.

Fig. 6 .
Fig. 6.The expression of KIF14 in HNSCC cell lines.The relative KIF14 mRNA expression in each individual cell line (A) and combined KIF14 mRNA expression in normal oral tissue (N), radioresistant and radiosensitive cell lines (B).The mRNA levels are shown relative to the mRNA expression in normal oral tissue.Data are shown as means ± SD from 3 independent experiments (n = 3).Statistical analysis was performed using unpaired t-test.Western blot analysis of KIF14 expression in radioresistant and radiosensitive HNSCC cell lines (C); ns=not significant.

Fig. 7 .
Fig. 7.The effect of KIF14 silencing on radiotherapy response in HNSCC cell lines.Crystal violet assay was used to assess the proliferative ability of irradiated HNSCC cells (LK0412, LK0863 and LK0850) in the presence of KIF14 siRNA.Following downregulation of KIF14, the growth rate of the analyzed HNSCC cell lines was measured and presented as % of control siRNA (non-targeting siRNA) (A).Cell proliferation of the irradiated cells in the presence of KIF14 siRNA is presented as the percentage of the untreated controls, and data are presented as the mean ± SD from three independent experiments performed in triplicate.Western blot analysis of knockdown efficacy of KIF14 in the analyzed cell lines is presented next to the cell proliferation analysis (B-D).Statistical analysis was performed using unpaired Student t-test.*p ≤ 0.05, ***p ≤ 0.001, ****p ≤ 0.0001, ns=not significant.

Table 1
Clinical characteristics of patients included in the study.