The CD98 Heavy Chain Is a Marker and Regulator of Head and Neck Squamous Cell Carcinoma Radiosensitivity

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide with about 550,000 new cases diagnosed annually (1). Despite advances in treatment management, patients with locally advanced disease often experience locoregional and distant recurrence and have a 5-year survival rate of only about 50% (1–3). Primary radiochemotherapy (RCTx) or postoperative radiochemotherapy (PORT-C) are standard treatments for patients with locally advanced HNSCC (4–8). Because of the biological heterogeneity of HNSCC, the patients' response to treatment is highly diverse and reliable stratification of patients with HNSCC for prediction of outcomes is of utmost importance (9).

In addition to clinical stage, other clinical, pathologic, and biological indicators that may have a prognostic value in HNSCC have been reported, such as expression of hypoxia-associated gene signatures, tumor metabolic volume measured by 2-[18F]fluoro-2-deoxy-D-glucose (FDG) uptake, cancer stem cell (CSC) marker expression, CD8⁺ tumor-infiltrating lymphocytes, and human papillomavirus (HPV) infection (10–16). A number of clinical studies demonstrated that infection with HPV is a strong prognostic factor in HNSCC after PORT or primary RCTx (10–12). To further stratify patients with HPV-negative HNSCC for the risk of recurrence, additional markers are warranted.

Retrospective biomarker analyses in patients with locally advanced HNSCC treated with curatively intended cisplatin-based PORT (PORT-C) or primary RCTx revealed that the putative CSC marker SLC3A2 is a prognostic biomarker of locoregional tumor control (LRC) in patients with HPV-negative tumors, with significantly higher LRC rates in tumors with low SLC3A2 mRNA expression (11, 17).

SLC3A2 encodes for the heavy chain of a cell surface, transmembrane protein that is also known as CD98 (CD98hc). Together with the large neutral amino acid transporters LAT1 (SLC7A5), CD98hc constitutes a heterodimeric transmembrane amino acid transporter LAT1/CD98hc that preferentially transports large neutral amino acids including isoleucine, leucine, methionine, valine, histidine, cysteine, tryptophan, and tyrosine (18, 19). A recent study also showed an interaction of CD98hc with other proteins: LAT2 (SLC7A8), which is an amino acid transporter specific for isoleucine and leucine and a mediator of glutamine efflux (20); and xCT (SLC7A11), which is a cystine/glutamate exchange transporter essential for synthesis of the antioxidant glutathione (GSH; ref. 21). Owing to its function, a high expression of CD98hc is associated with the development and aggressiveness of a number of cancers including HNSCC (19, 22, 23). Although the role of CD98hc as a potential regulator of CSC maintenance in HNSCC has been reported previously (22), its contribution to the molecular mechanisms governing tumor response to radiotherapy is not yet understood.

The main objective of this study is the biological validation of SLC3A2 as a potential biomarker of HNSCC radioresistance and identification of potential molecular mechanisms driving survival of SLC3A2-overexpressing cells during radiotherapy.

Additional methods not described here are included in the Supplementary Data.

In this study, a total of 197 patients from two different patient cohorts were included. All patients were diagnosed with locally advanced HNSCC arising in the oral cavity, oropharynx, hypopharynx, or larynx and received curative-intended, state-of-the-art primary radiotherapy (10/197), or cisplatinum-based radiochemotherapy (187/197) between 1999 and 2015, covering the tumor region and regional neck nodes as well as a boost to the tumor region and involved regional lymph nodes. Inclusion criteria have been described in detail before (11). For staging, the 7th edition of the UICC classification was used. The first patient cohort included 134 patients selected according to available tumor material from an original population of 158 patients of the retrospective, multicenter study of the German Cancer Consortium Radiation Oncology Group (DKTK-ROG), all of whom received primary radiochemistry as described previously (11). From the remaining 24 patients, insufficient tumor material was available. Thus, this patient cohort was further supplemented by a second, monocentric cohort including 63 patients, who were treated with primary radio(chemo)therapy between 1999 and 2015 at the Dresden University Hospital (Dresden, Germany). The latter cohort was not part of the DKTK cohort but fulfilled the same inclusion criteria. Patient characteristics are shown in Supplementary Table S1. All patients were followed-up for a minimum of 24 months.

Written informed consent was obtained from all patients. The study was conducted in accordance with the Declaration of Helsinki. Ethical approval for retrospective analyses of clinical and biological data was granted by the local ethics committee (Dresden) as well as by the ethics committees of all DKTK partner sites for the respective patients of the multicenter study.

In this study, we used established human squamous cell carcinoma (hSCC) cell lines including Cal33 derived from squamous cell carcinomas of the tongue (Deutsche Sammlung von Mikroorganismen und Zellkulturen DSMZ GmbH), FaDu derived from pharyngeal squamous cell carcinoma (ATCC), UTSCC5 originated from squamous cell carcinomas of the tongue (established at the University Turku, Finland) and SAS derived from squamous cell carcinomas of the tongue (Health Science Research Resources Bank, Osaka, Japan). All cell lines including CD98hc wild-type (WT) and monoallelic knockout (maKO) clones of Cal33 RR cells were maintained in DMEM (Sigma-Aldrich, GE) containing 10% FBS (PAA Laboratories, GE) and supplemented with 1 mmol/L l-glutamine (Sigma-Aldrich, GE), 1% HEPES (1 mol/L, PAA Laboratories, GE), 1% sodium pyruvate (100 mmol/L, Sigma-Aldrich, GE), 1% MEM nonessential amino acids (100×, Sigma-Aldrich, GE). Radioresistant (RR) cell lines of Cal33 and FaDu cells were established as described previously (24). All cell lines were cultured in a humidified 37°C incubator supplemented with 5% CO₂. The cell lines (i.e., FaDu, FaDu RR, SAS, UTSCC5, Cal33, Cal33 RR, Cal33 RR WT1, Cal33 RR WT2, Cal33 RR maKO1, Cal33 RR maKO2) were genotyped using microsatellite polymorphism analyses by Eurofins Medigenomix Forensik GmbH and tested for Mycoplasma directly prior to experimentation.

The results of colony formation assays, γH2AX foci assay, Seahorse metabolism analysis, metabolic mass spectrometry analysis, flow cytometry, and Western blotting were analyzed by paired t tests. The differences between cell survival curves were analyzed using the statistical package for the social sciences (SPSS) v23 software as described by Franken and colleagues (25) by fitting the data into the linear-quadratic formula S(D)/S(0) = exp(αD+βD²) using stratified linear regression. Multiple comparison analysis for the data depicted on Figs. 3A, 4A, and 5B was performed using one-way ANOVA analysis by GraphPad Prism software. A P < 0.05 was regarded as statistically significant. Correlation was evaluated by SUMO software using the Pearson correlation coefficient. For the presented patient cohort, the primary endpoint was LRC, which was calculated from the first day of radiotherapy to the occurrence of a local or regional recurrence or censoring. Corresponding survival curves were estimated by the Kaplan–Meier method and compared by log-rank tests. In addition, the impact of CD98 and LAT1 on LRC was evaluated using univariable and multivariable Cox regression. In multivariable regression, CD98 or LAT1 were included together with nodal stage (0,1 vs. 2,3), p16 status and the natural logarithm (ln) of tumor volume, as performed in ref. 11. Patients were classified according to the fraction of CD98- or LAT1-positive tumor cells using the cutoff ≥10%. This cutoff was adapted from the literature (26) to give the highest power for a significant patient stratification regarding LRC as described in ref. 17.

Previous retrospective multicenter studies demonstrated that SLC3A2 mRNA levels are significantly associated with LRC in patients with locally advanced HNSCC, who were treated with PORT-C or primary RCTx (11, 17). As the correlation between mRNA and protein levels has been reported to be notoriously low (27), we attempted to assess the levels of the SLC3A2-encoding protein CD98hc by IHC analysis of tumor tissues from patients with locally advanced HNSCC treated with primary RCTx (n = 197, DKTK cohort and monocentric Dresden cohort; Fig. 1). First, the percentage of CD98hc-positive tumor cells within the specimens was evaluated. Patients with at least 10% of CD98hc-positive tumor cells showed significantly lower LRC in univariable and multivariable analyses (P = 0.002 and P = 0.005, Fig. 1A and B; Supplementary Table S2) compared with those with less positive or negative tumors. Tumors with more intensive staining for CD98hc protein showed significantly lower LRC rates compared with those with less intensive staining in univariable analyses (P = 0.002, Supplementary Table S2; Fig. 1C) and in multivariable analyses (P = 0.006, Fig. 1A). Taken together, these data suggest that both overall expression of CD98hc in tumor bulk and the size of cell population, which is highly positive for CD98hc expression, are prognostic biomarkers of LRC in patients with locally advanced HNSCC treated with primary RCTx.

Next, we analyzed whether radiation response of the established HNSCC cell lines xenografted into mice correlates with CD98hc expression. We assessed the protein expression levels of CD98hc in seven different HNSCC cell lines (SAS, UTSCC5, Cal33, FaDu, UTSCC8, XF354, and SAT) by Western blot analysis (Fig. 2A) and found a positive correlation between previously determined tumor control dose 50 (TCD₅₀) values after fractionated irradiation of HNSCC xenograft tumors (28, 29) and CD98hc protein levels in corresponding cell lines (R = 0.83; P < 0.05; Fig. 2B).

Plasma membrane proteins from the parental Cal33 and FaDu cell lines and their previously described radioresistant (RR) derivative sublines (24, 30) were enriched by differential centrifugation and analyzed by LC/MS-MS as described in refs. 31 and 32 (Supplementary Tables S3 and S4). CD98hc was identified as one of the top scoring proteins upregulated in the RR sublines as shown in Fig. 2C. Western blot analysis revealed that expression of CD98hc protein is dynamically regulated after X-ray irradiation, which can indicate its role in the tumor response to radiotherapy (Supplementary Fig. S1A).

To validate the role of CD98hc in regulation of cell radiosensitivity, siRNA-mediated knockdown of SLC3A2 expression in several HNSCC cell lines including Cal33, Cal33 RR, FaDu, UTSCC5, and SAS was used. The results of 2D and 3D clonogenic assays demonstrated that knockdown of CD98hc led to a significant increase of radiosensitivity in all analyzed cell lines (Fig. 2D; Supplementary Fig. S1B and S1C). Taking into account that the effect of siRNA-mediated gene silencing is transient, CRISPR/Cas9 gene editing technology was employed to achieve a stable knockout of CD98hc expression in the Cal33 RR cell line (Supplementary Fig. S2A). Analyses of the cell clones demonstrated that homozygous knockout was not achievable. This may be explained by a nearly-tetraploid genotype of the Cal33 RR cell line and, potentially, also by the special importance of CD98hc expression for survival and proliferation. Two wild-type clones (WT1, WT2) and two clones with monoallelic knockout of SLC3A2 expression (maKO1, maKO2) were further used. Metaphase chromosome analysis of each of these selected clones showed that both WT2 and maKO2 have a hyperdiploid karyotype, and the WT1 and maKO1 are near-tetraploid (Supplementary Fig. S2B). Therefore, WT and maKO cells were compared according to their chromosome numbers (WT1 with maKO1, and WT2 with maKO2, respectively).

Radiobiological clonogenic survival assays showed that both maKO1 and maKO2 clones are more radiosensitive than the corresponding WT1 and WT2 clones (Fig. 2E and F; Supplementary Fig. S2C). Residual DNA double-strand breaks (DSB) were analyzed by γH2A.X foci count 24 hours after cell irradiation with 4 Gy of X-rays (Fig. 2G). SLC3A2 maKO cell lines showed less efficient DNA DSB repair as compared with their wild-type counterparts. Gene expression analyses showed that the SLC3A2 loss-of-function phenotype is associated with downregulation of key DNA repair genes (Fig. 2H).

Gene expression analyses showed that the SLC3A2 maKO1 phenotype is associated with inhibition of cell proliferation and deregulation of amino acid metabolism (Fig. 3A–C; Supplementary Fig. S3A). Cell energy metabolic analyses using seahorse technology revealed that maKO cells have a low basal oxygen consumption rate (OCR) and a more quiescent phenotype, which is in line with a reduced proliferation rate of maKO1 compared with WT1 cells (Fig. 3C and D; Supplementary Fig. S3B).

The intracellular levels of Krebs cycle metabolites were measured in SLC3A2 maKO1 and WT1 cells by LC/MS-MS (33). Most Krebs cycle metabolites such as α-ketoglutarate, fumarate, citrate, cis-aconitate, isocitrate, and pyruvate were found to be significantly downregulated in SLC3A2 maKO1 cells, whereas the level of glutamine was increased. These data demonstrate that maKO1 cells have a decreased CD98hc/LAT1-mediated uptake of amino acids, which serve as a source for Krebs cycle intermediates and are transported in exchange with internal glutamine (Fig. 3E; Supplementary Fig. S3C). Inhibition of amino acid transport resulted in the activation of p53 signaling mechanisms and in the increase of expression of the common stress-responsive transcriptional factor ATF3 (Supplementary Fig. S4A–S4D). Of importance, ATF3 expression was found to be downregulated in HNSCC-radioresistant cell lines (Supplementary Fig. S4E).

A recent study demonstrated that CD98hc-associated amino acid transport is crucial to control reactive oxygen species (ROS; Supplementary Fig. S3C; ref. 34). Gene expression analysis of maKO1 and WT1 cells showed that SLC3A2 loss-of-function results in deregulation of a number of genes involved in oxidative stress response (Fig. 3F). To determine whether CD98hc silencing disrupts redox balance in HNSCC cells, the levels of ROS, as well as reduced and oxidized glutathione (GSH and GSSG, respectively) were measured in SLC3A2 maKO and WT cell lines. Reduced CD98hc expression was associated with decreased basal GSH/GSSG ratio and significant upregulation of ROS after irradiation (Fig. 3G and H).

CD98hc (SLC3A2) and LAT1 (SLC7A5) constitute a heterodimeric transmembrane complex that mediates amino acid transport (18, 34, 35). Previous studies showed that CD98hc plays a crucial role for transport activity by regulating LAT1 trafficking to the plasma membrane (36–38). Indeed, analyses of the LAT1 and CD98hc protein expression in SLC3A2 maKO and WT cell lines revealed a more pronounced cytoplasmic localization of LAT1 in maKO cells compared with WT clones (Fig. 4A). Analyses of tumor tissues from patients with locally advanced HNSCC treated with primary R(C)Tx (n = 63; Dresden monocentric cohort) showed a significant correlation of membrane localization of CD98hc and LAT1 proteins (Fig. 4B). In addition to the membrane colocalization, protein and mRNA levels of CD98hc also significantly correlate with LAT1 expression in SLC3A2 maKO and WT clones and in HNSCC cell lines (R = 0.78; P < 0.05; Fig. 4C and D; Supplementary Fig. S5A and S5B). Similar to CD98hc, TCD₅₀ values of the xenograft models were found to significantly correlate with expression levels of LAT1 in the corresponding HNSCC cell lines (R = 0.73; P < 0,05; Fig. 4E).

Recent studies showed that the tumorigenic potential of CD98hc KO cells can be attributed to residual LAT1 transport activity (34). To validate the role of LAT1-mediated amino acid transport in regulation of radiation response in SLC3A2-deficient cells, SLC3A2 maKO1 and WT1 cell lines were transfected with SLC3A2 siRNA or SLC7A5 siRNA alone or in combination, and analyzed by clonogenic survival assays after different doses of X-rays (Supplementary Fig. S6A and S6B). Knockdown of both proteins resulted in a significant increase in cell radiosensitivity. Strikingly, the effect of SLC7A5 knockdown on cell radiosensitivity depends on the levels of CD98hc protein and is more pronounced in WT cells. Knockdown of expression of both SLC3A2 and SLC7A5 results in inhibition of the mTOR signaling pathway in WT cells (Fig. 4F and G). In contrast, SLC3A2 maKO cells showed low basal level of PI3K/mTOR pathway activation, which was not inhibited but rather increased upon SLC3A2 or SLC7A5 knockdown (Fig. 4F and G). To test whether low basal activation levels of the prosurvival PI3K/mTOR signaling pathways may lead to a higher sensitivity of SLC3A2 maKO1 cells to inhibition, the effect of the dual PI3K/mTOR inhibitor BEZ235 was investigated. It was found to be more potent for the inhibition of viability of SLC3A2 maKO1 cells as compared with their WT1 counterparts (Supplementary Fig. S6C). Consistent with these results, targeting of the PI3K/mTOR pathways with BEZ235 yielded a significant increase in cell radiosensitivity only in maKO1 cells (Fig. 4H; Supplementary Fig. S6D). Of importance, treatment of maKO1 cells with BEZ235 results in activation of autophagy, a prosurvival mechanism which was also described for other types of cancer after inhibition of PI3K/AKT/mTOR pathway (Fig. 4I; refs. 39–42). Taken together, these data suggest that expression of CD98hc in HNSCC is crucial for the LAT1 membrane localization, activation of the prosurvival mTOR/PI3K signaling pathway and regulation of cell radiosensitivity. Similar to knockout of CD98hc, inhibition of mTOR/PI3K signaling results in increase of cell radiosensitivity and induces autophagy as a mechanism of cellular stress response.

Gene expression analysis revealed that autophagy-associated genes are activated in SLC3A2 maKO1 cells (Supplementary Fig. S7A). A high basal level of autophagy in maKO cells was confirmed by Western blotting and by cytometry-based analysis of Autophagy Green as an autophagosome marker (Fig. 5A and B; ref. 43). In contrast to WT1 cells, which showed radiation-induced autophagy, maKO1 cells exhibit a high basal level of autophagy, which did not further increase after irradiation (Fig. 5B).

Knockdown of the key autophagy-regulating protein ATG5 has been previously shown to suppress autophagy in HNSCC cells (44). Furthermore, ATG5 was found to be significantly overexpressed in HNSCC-radioresistant cells as compared with their parental counterparts (Supplementary Fig. S7B). Analysis of the TCGA gene expression dataset for 517 patients with HNSCC revealed that low expression of ATG5 significantly correlates with better overall survival (Supplementary Fig. S7C). Inhibition of autophagy by knockdown of ATG5 and by Bafilomycin A1 (45) in WT1 and maKO1 cells resulted in significant radiosensitization in both cell lines (Fig. 5C–F; Supplementary Fig. S7D and S7E).

Autophagy inhibitors Bafilomycin A1 and Chloroquine, which prevent autolysosome maturation more potently reduce the viability of SLC3A2 maKO1 cells compared with their WT counterparts (Supplementary Fig. S7F). Consistent with these results, treatment with Bafilomycin A1 resulted in activation of apoptosis with more pronounced effect in maKO1 than in WT cells (Supplementary Fig. S7G). Taken together, these data suggest that activation of autophagy in CD98hc-deficient cells acts as a prosurvival mechanism, and its inhibition could be an effective approach for tumor cell radiosensitization.

Expression levels of LAT1 and CD98hc were significantly correlated in the HNSCC cell lines studied here (Fig. 4C and D). Consistent with these data, analysis of the TCGA HNSCC dataset (n = 519) revealed the highest correlation between the expression values of SLC3A2 and SLC7A5 compared with the correlation of these genes' expression to the rest of the tumor transcriptome (Fig. 6A). This analysis also revealed a significant correlation of the SLC3A2 gene and other CD98hc-associated amino acid transporters such as LAT2 (SLC7A8) and xCT (SLC7A11; Fig. 6A). IHC analyses of tumor tissues from patients with locally advanced HNSCC treated with primary R(C)Tx confirmed that expression levels of CD98hc and LAT1 proteins are highly correlated (P < 0.001; monocentric Dresden cohort; Fig. 6B; Supplementary Table S5). In univariate analyses, LAT1 overexpression was also found to be significantly associated with poor LRC (all patients: P = 0.047; patients with HPV-negative HNSCC only: no events, Cox model did not converge; Supplementary Table S2; Fig. 6C). A statistical trend was also revealed for an association of the percentage of LAT1-positive tumor cells with LRC. Patients with at least 10% of LAT1-positive tumor cells were found to have a poor prognosis compared with those with low or negative LAT1 staining (all patients: P = 0.077; Supplementary Table S2; Fig. 6D). However, in multivariate analyses corrected for N stage, tumor volume and p16 status, no significant impact on LRC was revealed for LAT, which may, in part, be explained by the small patient cohort (Supplementary Table S6).

Our previous studies demonstrated that mRNA levels of SLC3A2 gene are significantly associated with LRC in patients with locally advanced HNSCC who received PORT-C or primary RCTx (11). In this study, we found that the protein levels of CD98hc correlate with the TCD₅₀ of HNSCC xenograft tumors, which is a functional endpoint of CSC inactivation in vivo (46). Consistent with these data, IHC analysis of pretherapeutic tumor specimens from patients with locally advanced HNSCC, who underwent primary R(C)Tx revealed a correlation between LRC and CD98hc protein levels. Furthermore, high expression levels of CD98hc protein were found to be associated with a poor prognosis after primary RCTx compared with those patients with low CD98hc-expressing tumors.

A recent study by Rietbergen and colleagues has been shown that CD98 overexpression in p16/HPV16 DNA-positive oropharyngeal carcinoma is significantly associated with decreased overall survival and shorter progression-free survival (26). We could not show this effect, which is likely due to the low number of p16-positive/HPV16 DNA-positive oropharyngeal tumors in the cohorts analyzed in our study. Our cohorts are mainly composed of patients with HPV-negative HNSCC that is characterized by a less favorable outcome after radio(chemo)therapy. Additional biomarkers and treatment modifications for patients with HPV-negative HNSCC are urgently needed to improve outcome of RCTx in these patients (13, 16, 17).

The curative potential of radiotherapy depends on its ability to induce irreparable DNA damage in tumor cells by direct ionization of DNA or by production of ROS (47, 48). For the defense against ROS-induced oxidative stress, cancer cells produce a high level of an antioxidant GSH (49). CD98-related amino acid transporter complexes such as CD98hc/xCT are essential for the transport of cystine and thus for the synthesis of GSH and control of ROS level (50). Our study showed that downregulated CD98hc expression is associated with a significant decrease in the level of reduced GSH and upregulation of ROS levels after irradiation suggesting that CD98hc is important for tumor cell protection against oxidative stress.

Restriction of amino acid availability in CD98hc-deficient cells, which is associated with inhibition of mTOR signaling and decreased levels of Krebs cycle intermediates as a source of energy and biosynthesis, also contributes to the activation of autophagy. When autophagy is activated upon amino acid restriction, cells start to utilize its own components, which provide the building blocks and energy to survive (51). Our data suggest that restriction of amino acid availability in CD98hc-deficient cells is associated with inhibition of mTOR/PI3K signaling and significant increase in basal levels of autophagy, which can be increased further by mTOR/PI3K inhibition. In agreement with these findings, recent studies revealed that targeting of mTOR in HNSCC results in feedback autophagy activation, which serves as a prosurvival mechanism (40–42). Conversely, autophagy inhibition can sensitize HNSCC cells to anticancer therapies targeting mTOR/PI3K signaling (40, 41). Consistently, we found that expression levels of one of the key autophagy regulators, the ATG5 gene correlates with overall survival of patients with HNSCC, and inhibition of autophagy by ATG5 knockdown or by Bafilomycin A1 resulted in tumor cell radiosensitization and induction of apoptosis. Interestingly, global gene expression analyses revealed that cells with monoallelic knockout of SLC3A2 show significantly decreased expression levels of key DNA repair genes and upregulation of the wild-type p53-dependent signaling. This might suggest additional mechanisms for the CD98-dependent regulation of tumor radiosensitivity that warrant further studies. For further discussion on the possible role of CD98hc for the p53-dependent signaling, see Supplementary Data. Of note, we did not reveal a correlation between overexpression of CD98hc and 11q13 chromosomal locus amplification, which is common in HNSCC and is associated with a deregulation of DNA damage response in HNSCC (Supplementary Fig. S8A and S8B; ref. 52).

We found a significant coregulation and membrane colocalization of CD98hc and LAT1 in HNSCC cell lines and tumor tissues. This is in line with previous observations showing that expression of CD98hc and LAT1 are coregulated, and that CD98hc is required for the membrane localization of LAT1 (53). Inhibition of LAT1 expression resulted in a significant downregulation of cell radioresistance only in wild-type but not in SLC3A2 maKO cells, suggesting that the role of LAT1 in regulating tumor radiosensitivity may be mediated by the CD98hc protein. IHC analysis of LAT1 protein expression in tumor tissues from patients with locally advanced HNSCC treated with primary RCTx showed poor LRC rates in patients with high LAT1 expression levels. Analysis of the TCGA dataset for patients with HNSCC revealed a significant correlation of the SLC3A2 gene and other CD98hc-associated amino acid transporters such as LAT2 and xCT. The potential role of these genes as biomarkers for radiotherapy warrants further investigation.

Previously published data demonstrated the role of CD98 as a marker of the HNSCC CSC population (22). Consistent with this finding, our data suggest that not only the overall expression of CD98hc in the tumors, but also the size of CD98hc-positive cell populations are prognostic biomarkers of LRC in patients with locally advanced HNSCC treated with primary R(C)Tx. Overall, our findings indicate that CD98hc is not only a potential biomarker of the CSC number, but also a promising target for tumor radiosensitization. Current clinical trials are testing CD98-targeted treatment in a variety of malignancies (Supplementary Table S7). Integration of such approaches into radiotherapy might open up a promising avenue for clinical translational studies.

Future research is needed to validate our findings by using additional SLC3A2 knockout cell cultures, by combination of CD98hc-targeted therapy with radiotherapy in patient-derived HNSCC xenograft mouse models and by prospective validation of CD98hc and LAT1 expression levels as prognosticators of LRC after radio(chemo)therapy.

M. Baumann attended an advisory board meeting of Merck KGaA (Darmstadt), for which the University of Dresden received a travel grant; received funding for his research projects and for educational grants to the University of Dresden by Teutopharma GmbH, IBA, Bayer AG, Merck KGaA, and Medipan GmbH; as former chair of OncoRay (Dresden) and present CEO and Scientific Chair of the German Cancer Research Center (DKFZ, Heidelberg), signed/s contracts for his institute(s) and for the staff for research funding and collaborations with a multitude of companies worldwide; and for the German Cancer Research Center (DKFZ, Heidelberg), is on the supervisory boards of HI-STEM gGmbH (Heidelberg). None of these funding sources were involved in the study design or materials used, nor in the collection, analysis and interpretation of data nor in the writing of the paper.

M. Krause received funding for her research projects by Merck KGaA and Medipan GmbH, and as chair of OncoRay - National Center for Radiation Research in Oncology she signs contracts for her institutes and for the staff for research funding and collaborations with a multitude of companies worldwide. None of these funding sources were involved in the present study. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Digomann, I. Kurth, L.A. Kunz-Schughart, M. Baumann, A. Linge, A. Dubrovska

Development of methodology: D. Digomann, I. Kurth, A. Tyutyunnykova, C. Peitzsch, A. Linge, A. Dubrovska

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Digomann, I. Kurth, A. Tyutyunnykova, O. Chen, I. Gorodetska, C. Peitzsch, I.-I. Skvortsova, B. Aschenbrenner, G. Eisenhofer, S. Richter, S. Heiden, J. Porrmann, A.A. Dowle, L. Hein, F. Lohaus, A. Linge, A. Dubrovska

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Digomann, I. Kurth, A. Tyutyunnykova, S. Löck, C. Peitzsch, I.-I. Skvortsova, G. Negro, B. Aschenbrenner, S. Richter, B. Klink, A.A. Dowle, A. Abdollahi, A. Linge, A. Dubrovska

Writing, review, and/or revision of the manuscript: D. Digomann, A. Tyutyunnykova, O. Chen, S. Löck, I.-I. Skvortsova, G. Eisenhofer, S. Richter, B. Klink, A. Abdollahi, F. Lohaus, M. Krause, M. Baumann, A. Linge, A. Dubrovska

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Digomann, I.-I. Skvortsova, G. Negro, C. Schwager, L.A. Kunz-Schughart, F. Lohaus, M. Krause, M. Baumann, A. Linge, A. Dubrovska

Study supervision: I. Kurth, M. Krause, M. Baumann, A. Linge, A. Dubrovska

The authors thank Vasyl Lukiyanchuk for analysis of gene expression data, TCGA datasets and Celigo imaging, co-localization analysis, PCR analysis of DNA repair genes as well as for design and cloning of sgRNAs, screening for clones with SLC3A2 deletion, and their genotyping. They thank Liane Stolz-Kieslich for her technical assistance with IHC staining. They thank Lydia Koi (DKTK) for providing the HNSCC cell samples and Aliona Bogdanova (MPI-CBG, Dresden) for providing the Cas9-HF1 plasmid construct. They also thank the DKTK-ROG for providing access to the tumor material and to the clinical data of the subset of patients from previously published work (11). Work in the Dubrovska laboratory was partially supported by grants from the Deutsche Forschungsgemeinschaft (DFG; 273676790, 401326337, and 416001651), from Wilhelm Sander-Stiftung (2017.106.1), the DLR Project Management Agency (01DK17047), and the German Federal Ministry of Education and Science (BMBF; 03Z1NN11). D. Digomann was supported by the Preiss Daimler foundation Medical Equipment and Research and by the Else Kröner-Promotionskolleg. The York Centre of Excellence in Mass Spectrometry was created thanks to a major capital investment through Science City York, supported by Yorkshire Forward with funds from the Northern Way Initiative, and subsequent support from EPSRC (EP/K039660/1; EP/M028127/1). Work in Baumann and Krause laboratories was supported by German Federal Ministry of Education and Science (BMBF, Oncoray), the German Cancer Consortium DKTK, and by the Deutsche Forschungsgemeinschaft (DFG) grants BA 1433/5-2 and BA 1433/6.

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.