Introduction

In cancer cells, mitochondrial structure and function are altered, resulting in mitochondrial dysfunction1. Mitochondrial dysfunction impairs cellular energy production, increases glycolysis and acidosis, and leads to deregulation of cellular metabolism, a hallmark of cancer2,3,4,5. This process is associated with cancer aggressiveness, invasion, metastasis, and drug resistance6,7. These traits are also characteristic of epithelial to mesenchymal transition (EMT)8,9,10, a complex, variable, and context-specific cellular program that enables carcinoma cells to suppress epithelial characteristics in favour of mesenchymal differentiation11. In cancer, EMT is usually incomplete and reversible, resulting in coexpression of epithelial and mesenchymal markers12. EMT-associated reprogramming is regulated by various transcription factors, including SNAI1, SNAI2, ZEB, and TWIST13,14. In head and neck cancer, SNAI2 plays a dominant role13,15. Among others, the transforming growth factor beta (TGF-β1) regulatory pathway is a canonical trigger of EMT16,17.

Mitochondrial dysfunction and EMT regulatory pathways are interrelated. Mitochondrial dysfunction can activate glycolytic pathways, which interact with EMT-related transcription factors such as Snail. Snail represses enzymes of the glycolytic pathway as well as enzymes of the gluconeogenesis pathway, such as fructose-1,6-bisphosphatase, which leads to enhanced glycolytic flux18. Downregulation of the expression of mitochondrial proteins involved in oxidative phosphorylation (OXPHOS) was associated with EMT gene signatures and enhancement of EMT7,19. Induction of mitochondrial dysfunction via oligomycin A and antimycin A resulted in downregulation of E-cadherin expression, upregulation of vimentin expression, and increased migration and invasion, suggesting activation of the EMT cellular program20. Although these studies were not performed in head and neck cancer, they suggest a relationship between mitochondrial dysfunction and EMT.

Head and neck squamous cell carcinoma (HNSCC) is a common malignancy derived from mucosal surfaces of the oral cavity, sinonasal cavity, pharynx, and larynx. HPV-induced HNSCC differs substantially from HPV-negative HNSCC; however, both entities show high inter- and intratumoral tumour cell heterogeneity21,22. In this study, we chose 4 different HPV-negative HNSCC cell lines for the experiments. Our aim was to examine whether induced mitochondrial dysfunction leads to EMT in different HPV-negative HNSCC cell lines. Mitochondrial dysfunction was induced by blocking mitochondrial OXPHOS and measured using extracellular flux analysis and expression of the cystine/glutamate transporter xCT (SLC7A11)23. In extracellular flux analysis, the oxygen consumption rate (OCR) serves as a measure of OXPHOS, and the extracellular acidification rate (ECAR) serves as a measure of glycolysis. A high OCR/ECAR ratio indicates a metabolic state driven by OXPHOS. Solute carrier family 7 member 11 (SLC7A11) is the coding gene for xCT, which is known to be overexpressed in cells with mitochondrial damage with low OXPHOS24. The OCR/ECAR ratio and xCT expression are inversely correlated. EMT was assessed by gene expression analysis of SNAI2 and vimentin, flow cytometric analysis of vimentin/cytokeratin double-positive cells, and holotomographic microscopy12.

Methods

Cell lines

The 4 HNSCC cell lines SCC-2523, UPCI-SCC-00325, HN26 and CAL 2727 were purchased from the German Collection of Microorganisms DSMZ, Braunschweig, Germany. Cells were thawed at 37 °C in 8 mL of medium in 15 mL Falcon tubes, centrifuged for 5 min at 290g/at 4 °C and cultivated in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham’s F12 with additional 1% penicillin/streptomycin, 1% MEM with nonessential amino acids (NEAAs), 1 mM sodium pyruvate 100 mM (all from PAN-Biotech, Aidenbach, Germany), 10% foetal bovine serum and 1% l-glutamine 200 mM (both from Gibco, Grand Island, NY, USA). The cells were stabilized by two passages and cultured for 2 weeks prior to the experiments at 5% CO2 and 37 °C.

Induction of mitochondrial dysfunction

For induction of mitochondrial dysfunction, a mixture of carbonyl cyanide-4(trifluoromethoxy) phenylhydrazone (FCCP) and oligomycin was used (#103275-100, Agilent Bioscience, Sta. Clara, USA)28,29 or prepared analogously as described in Supplement 3. Cells were plated in XFp Cell Culture Miniplates (#103022-100, Agilent Bioscience, Sta. Clara, USA) in DMEM/F12 (0.3%) at a density of 2 × 104 cells per mL, which created a confluence of approximately 70%. For TGF-β1 controls, 1 μL/mL TGF-β1 (#P01137, R&D Systems, Minneapolis, USA) was added to the cell culture medium, and the cells were incubated overnight (5% CO2/37 °C) prior to the addition of FCCP/oligomycin.

Assessment of mitochondrial dysfunction

Extracellular flux analysis29 with the Seahorse XFp Analyzer (Agilent Bioscience, Sta. Clara, USA) and cystine/glutamate transporter (xCT) expression served to confirm mitochondrial dysfunction. The FCCP/oligomycin mixture (XFp Cell Energy Phenotype Stress Kit; #103275-100, Agilent Bioscience, Sta. Clara, USA) was assembled as described in the manufacturer’s protocol and added immediately before the measurement. The Seahorse XFp Analyzer was used for measurement 4× during baseline and 4× during FCCP/oligomycin conditions in a 5 min interval (Supplement 1). Parameters were normalized to 104 cells/well (Supplement 2).

For cystine/glutamate transporter (xCT) expression, 2 × 105 cells/mL were plated in four petri dishes in 10 mL of DMEM/F12 (0.3% FBS) each, two of them with an additional 1 ng/mL TGF-β1 (positive control for EMT). After 24 h, two dishes (one with TGF-β1 and one without) were exposed to 1 µM FCCP/oligomycin (Supplement 3) for 50 min in a CO2-free incubator at 37 °C to mimic the XFp Seahorse measurement. The other two dishes were incubated the same way without FCCP/oligomycin.

After FCCP/oligomycin treatments, cells were cultivated for one additional day without any further treatments. On this day, cells were allowed to adapt and react prior to RNA and protein isolation. For RNA and protein measurement, cells were lysed as described in the “RNA isolation, reverse transcription and RT-PCR” section below. For qPCR, the primer sequences for xCT (SLC7A11) were downloaded from the PrimerBank of the Massachusetts General Hospital, Boston, MA, USA28. Primer sequences are listed in Suppl. Table 1. Primers were synthesized by Invitrogen (Darmstadt, Germany) and were used together with the Sensifast Sybr Fluorescein Kit of Bioline (Labconsulting, Vienna, Austria) in a Bio-Rad MyiQ™ (Bio-Rad, Laboratories, Inc., Hercules, CA, USA) cycler according to the manufacturer's protocol. GAPDH was used as a housekeeping gene, and the relative quantities of SLC7A11 were calculated as described above.

Assessment of epithelial mesenchymal transition (EMT)

EMT was assessed by gene expression of vimentin and SNAI2 using quantitative PCR and by vimentin/cytokeratin coexpression with flow cytometry in the FCCP/oligomycin-treated cells (see FCCP/oligomycin treatment above). In addition, all cell lines were examined under a holotomographic microscope to detect EMT-typical morphological changes. Primer sequences for vimentin and SNAI2 were downloaded from PrimerBank at Massachusetts General Hospital, Boston, MA, USA (Suppl. Table 1). After FCCP/oligomycin treatment and one additional cultivation day, the cells were lysed, and RT-PCR was performed as described for xCT. GAPDH was used as a housekeeping gene, and the relative amounts of vimentin and SNAI2 transcripts were calculated as described above.

The percent of vimentin-cytokeratin double-positive cells was determined in a CytoFLEX flow cytometer (Beckman Coulter). Therefore, 2 × 105 cells/ml were plated in Petri dishes with 10 ml of DMEM/F12 (0.3% FBS) and treated as described for xCT analyses. After FCCP/oligomycin treatment and an extra cultivation day to allow the cells to respond, the cells were collected by trypsinization, counted and suspended in FBS (#26140087, Gibco, MA, USA). Then, the cell suspensions were incubated with antibodies against cytokeratin and vimentin (Suppl. Table 2) for 20 min at room temperature using the Perfix NC kit from Beckman Coulter (Marseille, France) according to the manufacturer's instructions. One reaction was set for 2 × 106 cells. Negative event gates were set by the isotype controls containing 99% of the isotype events. Positive events were defined as events with higher signals than the negative control gates. At least 2 replicates were analysed, and all samples were measured in duplicate.

For holotomographic microscopy, the same conditions and media were used as for extracellular flux analysis. The only difference was that 2 × 105 cells/mL were plated in IbiDi dishes (IbiDi, Ltd., Planegg, Germany) instead of Agilent XFp cell culture miniplates, and FCCP/oligomycin was added manually. After the same treatment time as for the flux analyses in a CO2-free incubator, the XFp measurement medium was replaced with DMEM/F12 (0.3% FBS), and the cells were immediately analysed under a 3D Cell Explorer Holotomography microscope (Nanolive SA, Switzerland) with an air objective at 60× magnification. Three typical histomorphologic features of EMT were assessed and semiquantitatively scored from 0 to 3: fibroblast-like morphology (FM), cell individualization, and cell detachment. For these 3 features, a score of 0 indicates 0–10% of cells have this feature; 1 indicates 11–50%; 2 indicates 51–80%; and 3 indicates ≥ 80% of cells have this feature (Fig. 1).

Figure 1
figure 1

Holotomographic microscopy scoring example of EMT characteristics (SCC25). (A) Score 0 for fibroblast-like-morphology (FM). (B) Score 1 for FM. (C) Score 2 for FM and (D) score 3 for FM. (E) Score 0 for cell individualization. (F) Score 1 for cell individualization. (G) Score 2 for cell individualization and (H) score 3 for cell individualization. (I) Score 0 for surface adhesion, (J) Score 1 for surface adhesion. (K) Score 2 for surface adhesion and (L) Score 3 for surface adhesion. Live cell imager (Nanolive, Switzerland); 20 μm.

Full size image

RNA isolation, reverse transcription and RT-PCR

For gene expression analyses, cells were lysed in 1 mL of TRIzol® Reagent (Ambion®, Life TechnologiesTM, Carlsbad, CA, USA), and RNA was isolated following the instructions of the manufacturer. RNA concentrations were determined by absorption at 260 nm and fluorometric measurements (Qubit, Invitrogen, Darmstadt, Germany), and RNA quality and integrity were evaluated by a Qubit RNA IQ kit (Invitrogen). The proportion of intact RNA in the total RNA isolates was at least 70%. Two micrograms of total RNA was reverse transcribed by M-MuLV Reverse Transcriptase with 2 µg of oligo dT15 (GeneON, Ludwigshafen am Rhein, Germany) in a ThermoQ heating and cooling block (Biozym, Hessisch Oldendorf, Germany). cDNA samples representing 10 ng of original total RNA were subjected to real-time qPCR. GAPDH was used as a housekeeping gene, and relative quantities of the transcripts were calculated by pairwise differences of threshold cycles (δCT) of the gene of interest and the loading control housekeeping gene30. For the final analysis, we used the relative quantification and related the PCR signal in the cells to a control reference gene expression level12. The identity of the PCR products of genes was confirmed by Sanger sequencing by Microsynth Austria (Vienna, Austria)12.

Data analysis

The full factorial experimental design with 2 (with/without FCCP/oligomycin) × 2 (with/without TGF-ß1) × 4 (cell lines) × 4 replicates = 64 experiments was analysed with a generalized linear model. The results of PCR, flow cytometry, and extracellular flux analysis revealed a right-skewed distribution. Therefore, a gamma distribution with a log linkage function was modelled. The estimated marginal means (EMM) and their standard errors (SEM) are reported. Holotomography scores for EMT-typical cellular responses, ranging from 0 to 9, were compared with and without FCCP/oligomycin using the Fisher-Freeman-Halton exact test. Calculations were performed with SPSS Ver. 27 (IBM, Armonk, NY).

Results

Effect of FCCP/oligomycin on EMT

The effects of FCCP/oligomycin on EMT depended on the experimental cell line. Considering all four cell lines together, the addition of FCCP/oligomycin increased the relative expression of SNAI2 from 8.5 ± 0.8 to 12.0 ± 1.1 (p = 0.007). When we separately examined the effect of FCCP/oligomycin on each cell line, only CAL 27 cells showed increased SNAI2 expression (p < 0.001; Fig. 2A), while the effect was not significant in the other 3 cell lines (all p > 0.1). The addition of FCCP/oligomycin had no effect on the overall relative expression of vimentin (p = 0.137). For the effect of FCCP/oligomycin on vimentin expression in individual cell lines, an increase was observed for CAL 27 cells from 2.2 ± 0.5 × 10–3 to 33.2 ± 10.2 × 10–3 (p = 0.002; Fig. 2B). Finally, the percentage of vimentin-cytokeratin double-positive cells in flow cytometry did not change in the overall analysis of all cell lines; however, at the cell line level, it decreased in SCC03 cells (p = 0.009) and increased in CAL 27 cells (p = 0.003) following FCCP/oligomycin administration (Fig. 2C). Possible EMT scores in holotomographic microscopy ranged from 0 (no EMT) to 9. Holotomographic EMT scores in 32 experimental sets were mostly below five, indicating generally few cells with EMT-typical morphological changes (Table 1). In particular, no significant effects of FCCP/oligomycin on EMT-typical morphological cell characteristics were observed (Fischer-Freeman-Halton p = 0.9). Occasionally, FCCP/oligomycin induced EMT-typical changes in cell morphology (Fig. 3A).

Figure 2
figure 2

Relative SNAI2 (A), vimentin (VIM; B) expression, and percentage of vimentin/cytokeratin double positive cells (C) in response to FCCP/oligomycin (EMM: Estimated marginal mean; Bars: SEM; *p < 0.05, **p < 0.01, ***p < 0.001).

Full size image
Table 1 Count of cell lines with EMT-scores in holotomographic microscopy of 4 HNSCC cell lines (range of possible scores from 0 to 9) with and without FCCP/oligomycin (p = 0.9).
Full size table
Figure 3
figure 3

Holotomographic microscopy of HN cells (A) in response to FCCP/oligomycin addition (A1 + A2) and SCC25 cells (B) in response to TGF-ß1 (B1 + B2). Bars: 20 μm.

Full size image

Cell line heterogeneity

Consistent with the heterogeneity of head and neck cancer, the four HNSCC cell lines used differed markedly in all parameters investigated (Fig. 2A–C, control). The baseline expression of SNAI2 differed by up to fivefold among the four cell lines (p < 0.001; Fig. 2A) SCC25 cells had a baseline relative SNAI2 expression of 17.7 ± 2.2, and CAL 27 cells had a baseline relative expression of 3.5 ± 0.4. Marked differences between cell lines were also observed for baseline vimentin expression (p < 0.001). Again, SCC25 cells had the highest expression at 4.7 ± 1.1, and CAL 27 cells had the lowest expression at 2.2 ± 0.5 × 10–3 (Fig. 2B). Flow cytometry results also yielded different baseline percentages of vimentin/cytokeratin double positivity in the four cell lines (p < 0.001; Fig. 2C). Here, SCC25 and CAL 27 cells had the highest values. SCC03 and HN had similarly low values (Fig. 2C).

Evidence of successful induction of mitochondrial dysfunction

A high OCR/ECAR ratio indicates a metabolic phenotype driven by OXPHOS, whereas lower OCR/ECAR ratios indicate glycolysis due to mitochondrial dysfunction. The observed decrease in the OCR/ECAR ratio confirmed FCCP/oligomycin-induced mitochondrial dysfunction in all cell lines except SCC25 cells (Fig. 4A; Suppl. Table 3). Moreover, FCCP/oligomycin addition increased xCT values in all cell lines (Fig. 4B; Suppl. Table 4). As with the markers for EMT, the baseline OCR/ECAR ratio (p < 0.001) and baseline xCT expression (p < 0.001) revealed significant differences between the cell lines studied (Suppl. Fig. 1).

Figure 4
figure 4

OCR/ECAR ratio (A) and xCT expression (B) of 4 HNSCC cell lines in response to FCCP/oligomycin (EMM: Estimated marginal mean; bars: SEM; *p < 0.05, **p < 0.01, ***p < 0.001).

Full size image

Evidence that cell lines were EMT-competent

TGF-ß1 is a key mediator of EMT in cancer cells; however, not all HNSCC cell lines are capable of EMT. The responses of SNAI2 and vimentin to TGF-ß1 stimulation indicate that the four cell lines used can respond to an appropriate stimulus with EMT (Suppl. Fig. 2; Suppl. Table 5). Double positivity of vimentin and cytokeratin in flow cytometry did not significantly increase after TGF-ß1 stimulation (Suppl. Fig. 2, Suppl. Table 5), but a trend towards EMT-like histomorphological features was observed by holotomographic microscopy (p = 0.1; Fig. 3; Suppl. Table 6).

Discussion

EMT contributes to the aggressiveness and therapeutic resistance of head and neck carcinomas8. In addition to genetic reprogramming and external stimuli from the tumour microenvironment31, there is ample evidence that different kinds of cellular stress promote EMT in carcinomas32,33. Here, we investigated cellular stress induced by mitochondrial dysfunction, a particular form of cellular stress that is common in HNC2. We questioned whether cellular stress due to mitochondrial dysfunction is a significant EMT trigger. In 4 HNC cell lines, mitochondrial dysfunction was induced by FCCP/oligomycin, an established method to inhibit mitochondrial OXPHOS. This change leads to a metabolic state dominated by glycolysis, as is the case with HNC in vivo.

Clinically, HNC exhibits considerable cellular heterogeneity. Hence, we used several HNC cell lines for this study. To reduce variability, we used only HPV-negative cell lines. Consequently, the results are not applicable to HPV-positive cell lines. Different HNC cell lines might prefer different regulatory pathways and/or have a heterogeneous bioenergetic organization34,35. Indeed, some cell lines investigated were constitutively more glycolytic, and others depended more on OXPHOS. We also observed high variability among EMT parameters in the 4 cell lines (Fig. 4). This finding may in part explain why the EMT response of the HNC cell lines to mitochondrial dysfunction in this study differed.

To determine whether mitochondrial dysfunction had actually been induced, we used extracellular flux analysis, a relatively precise method to measure metabolic activity and state in cells29. We also examined the expression of xCT, a reliable surrogate marker of mitochondrial dysfunction24. Both methods indicated the successful induction of mitochondrial dysfunction with the addition of FCCP/oligomycin. The OCR/ECAR ratio decreased significantly when FCCP/oligomycin was added (p < 0.001). This result is in line with a previous report showing that glycolytic rates increased after 1 h of oligomycin treatment in different cancer cell lines34. Only SCC25 cells showed no decrease in the OCR/ECAR ratio; however, xCT expression increased substantially (Suppl. Fig. 2). Wang and coauthors also reported increased xCT expression when cells were treated with oligomycin and antimycin A, which was linked to proliferation and malignant progression24.

To confirm that the cells were all capable of epithelial mesenchymal transition, we examined the EMT response to TGF-β1, a canonical trigger of EMT12,17. EMT was determined by the gene expression of SNAI2, a master regulator of EMT in HNSCC, and the expression of vimentin, a mesenchymal marker not normally expressed in epithelial cells. In addition, the double positivity of vimentin and cytokeratin, a hallmark of EMT, was measured by flow cytometry and EMT-typical morphologic features were examined by holotomographic microscopy. TGF-β1 caused EMT in all cell lines and with all detection methods. Although not significant, holotomographic microscopy also revealed EMT induction after TGF-β1 stimulation. This finding indicates that the cell lines used were capable of undergoing EMT and our results support observations of other studies that suggested the stabilization of a hybrid epithelial/mesenchymal phenotype through SNAI236. Furthermore, the equivocal role of SNAI2 in maintenance of stem cell pluripotency and DNA stability in healthy and malignant cells is underlined37 as simultaneous increase in SNAI2 and vimentin expression was only seen in one cell line.

Mitochondrial dysfunction and EMT regulatory pathways influence each other. EMT transcription factors can affect various metabolic pathways18. Downregulation of mitochondrial protein expression involved in OXPHOS was associated with enhanced EMT7 as well as oligomycin A administration, leading to downregulation of epithelial factors, increased vimentin expression and subsequent invasion and migration20. This is in line with our results, where we could see an increased vimentin expression when cells were treated with oligomycin and FCCP.

Besides stress, several other inhibitors of EMT and its metastatic potential were recently reviewed elsewhere. Reduced proliferative potential after complete EMT and loss of epithelial characteristics has been observed in breast cancer cells38. In addition, feedback mechanisms, responsible for mesenchymal–epithelial transformation, were found to inhibit EMT in cells with epithelial characteristics39. Finally, stimulation with EMT promoting factors such as TGF-ß might require several days before actually inducing EMT40. Correlating these aspects with stress induced mitochondrial dysfunction might thus be an interesting subject of further research.

Thus, it can be assumed that mitochondrial dysfunction did indeed occur and that the cells were capable of EMT. However, the EMT response to mitochondrial dysfunction varied from cell line to cell line. An EMT response was consistently detectable in CAL 27 cells. This may be due to several reasons. One possible explanation could be the different cellular background of each cell line. SCC25 cells exhibit increased baseline expression of EMT proteins such as vimentin41. UPCI-SCC-03 cells contain wild-type p53, which is known to suppress EMT42. Therefore, it would be explanatory that it was not possible to induce EMT via mitochondrial dysfunction in these two cell lines. HN and CAL-27 cells contain a p53 mutation43. However, they cannot be directly compared because HN cells were isolated from a metastatic tumor, which is different from cell lines isolated from primary tumors, such as CAL-27 cells. Therefore, HN cells might have been exposed to other types of cellular stress before and therefore, we were unable to induce EMT with FCCP and oligomycin.

In summary, the study suggests that given the marked heterogeneity of head and neck carcinomas, EMT induced by mitochondrial dysfunction may occur in different clones of HNC in vivo. However, EMT does not appear to be a major or common mechanism of the cancer-promoting effects of mitochondrial dysfunction in HNC. Notwithstanding, other forms of cellular stress, such as hypoxic, nutritional, and heat stress, can trigger EMT in head and neck cancer44,45,46.