Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Oct;125(8):1122-1134.
doi: 10.1038/s41416-021-01491-x. Epub 2021 Jul 21.

The combined detection of Amphiregulin, Cyclin A1 and DDX20/Gemin3 expression predicts aggressive forms of oral squamous cell carcinoma

Ekaterina Bourova-Flin  1 , Samira Derakhshan  2 , Afsaneh Goudarzi  3 , Tao Wang  1 , Anne-Laure Vitte  1 , Florent Chuffart  1 , Saadi Khochbin  1 , Sophie Rousseaux  4 , Pouyan Aminishakib  5   6
Affiliations

The combined detection of Amphiregulin, Cyclin A1 and DDX20/Gemin3 expression predicts aggressive forms of oral squamous cell carcinoma

Ekaterina Bourova-Flin et al. Br J Cancer. 2021 Oct.

Abstract

Background: Large-scale genetic and epigenetic deregulations enable cancer cells to ectopically activate tissue-specific expression programmes. A specifically designed strategy was applied to oral squamous cell carcinomas (OSCC) in order to detect ectopic gene activations and develop a prognostic stratification test.

Methods: A dedicated original prognosis biomarker discovery approach was implemented using genome-wide transcriptomic data of OSCC, including training and validation cohorts. Abnormal expressions of silent genes were systematically detected, correlated with survival probabilities and evaluated as predictive biomarkers. The resulting stratification test was confirmed in an independent cohort using immunohistochemistry.

Results: A specific gene expression signature, including a combination of three genes, AREG, CCNA1 and DDX20, was found associated with high-risk OSCC in univariate and multivariate analyses. It was translated into an immunohistochemistry-based test, which successfully stratified patients of our own independent cohort.

Discussion: The exploration of the whole gene expression profile characterising aggressive OSCC tumours highlights their enhanced proliferative and poorly differentiated intrinsic nature. Experimental targeting of CCNA1 in OSCC cells is associated with a shift of transcriptomic signature towards the less aggressive form of OSCC, suggesting that CCNA1 could be a good target for therapeutic approaches.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1
Fig. 1. Identification of candidate prognostic biomarkers in OSCC.
a Overview of the strategy to identify prognostic biomarkers in OSCC (oral squamous cells carcinoma). From the analysis of RNAseq data publicly available from normal human tissues (Arrayexpress: https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-1733/), 3430 candidate genes were identified with a predominant expression in one specific tissue-type, and either silent or with a very low expression level in all normal non-germline adult tissues. Step 1: exploration: By analysing the expression of these genes in transcriptomic data of the first series of 97 oral squamous cell carcinoma (OSCC) tumours with survival data (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE41613), 15 genes were selected as prognostic marker candidates according to three criteria: (1) significant association between overexpression and a shorter survival probability (Cox model P value < 0.05), (2) existence of expression thresholds stratifying patients with significantly different survival probabilities (log-rank test P value < 0.05), (3) interval of significant thresholds >50% (in percentiles). Step 2: selection: Of the 15 genes satisfying these three criteria, three were chosen, on the basis of the availability of a reliable antibody able to specifically detect the corresponding proteins. Step 3: 1st validation of the 3-genes combination: The efficiency of their combination to stratify patients with different prognoses was validated in an independent cohort constituted of the 467 HNSC (head and neck squamous carcinoma) patients from the TCGA (from https://portal.gdc.cancer.gov/projects/TCGA-HNSC) in univariate as well as multivariate analyses taking into account clinical and biological parameters associated with prognosis in OSCC. Step 4: 2nd validation of the 3-genes combination by immunohistochemistry (IHC): This 3-genes-based classifier was used to define an IHC test based on the detection of the corresponding proteins, which successfully predicted survival in our series of 66 patients. b, c Stratification of the OSCC patients using the 3-genes signature in the training dataset (GSE41613, n = 97, Affymetrix technology). b Individual ectopic expression of each gene: Kaplan–Meier survival curves illustrating the association between the abnormal expression of each of the three selected genes (AREG, CCNA1, and DDX20) and overall survival probability in OSCC patients. For each gene, the survival probabilities are compared between OSCC patients whose tumour had activated the gene (black line) and those whose tumour had not activated the gene (blue line). The P values of the Cox and log-rank models are shown. c Combination of three genes by the number of ectopically expressed genes: Kaplan–Meier survival curves comparing the overall survival (OS) probability between OSCC patients grouped according to the number of activations of the combination of the three selected genes (3-genes encompassing AREG, CCNA1 and DDX20). In the left panel, the patients are grouped according to the total number of gene activations (as indicated), whereas in the right panel the survival is compared between two groups of patients whose tumour activates none or only one of the three genes (blue curve) or two or three of the three genes (black curve). The number of activated genes is used as the explanatory variable in the cox model, whereas the log-rank tests the significance of a different survival probability between groups.
Fig. 2
Fig. 2. Stratification of the TCGA-HNSC patients using the 3-genes signature in the validation dataset (n = 467, RNAseq).
Kaplan–Meier survival curves illustrating TCGA-HNSC patients’ overall survival (OS) according to the number of activations of the three genes (AREG, CCNA1 and DDX20), considering all patients (n = 467) (a) or subsets of patients defined according to clinical and biological parameters (b, c or d) including TNM early or late stages (b), human papillomavirus (HPV) status (c) or tumour localisation (d). a All patients: in the left panel, the patients are grouped according to the total number of gene activations, and in the right panel survival probabilities are compared between patients with tumours activating none or one gene (0–1) or 2 or 3 of the three genes (2–3). b–d Kaplan–Meier survival curves illustrating TCGA-HNSC patients’ overall survival (OS) according to our 3-gene classifier when applied to the following subsets of patients. b Survival probabilities of subsets of patients at an early TNM stage ((T1 or T2) and (N0 or N1), left panel) or TNM late-stage ((T3 or more) or (N2 or more), right panel). c Survival probabilities of subsets of HPV-negative patients (left panel) or HPV-positive patients (right panel, note that only one HPV-positive patient is negative for our 3-gene classifier and P values and hazard ratio are not statistically relevant in this case). d Survival probabilities of subsets of patients with tonsil or base of tongue localisation (TBOT: left panel) or other localisation (other than TBOT: right panel).
Fig. 3
Fig. 3. Stratification of the OSCC patients according to the protein/antigenic signature using IHC on tumour sections in our validation cohort of OSCC patients.
a Left panel: examples of immunohistochemical detection of AREG, CCNA1 and DDX20, as indicated, showing samples with “high” and “low” labelling at an x400 magnification. Right panel: scoring criteria are shown (see the methods section for details). b Prognostic of patients grouped according to the IHC detection of each of the three proteins individually: high labelling = 1 and low labelling = 0. c Prognostic of patients stratified with the three proteins in combination (the score corresponds to the total of “high” scorings for the three antibodies).
Fig. 4
Fig. 4. Differential analysis between 3-genes positive and negative OSCC tumours identifies a genome-wide expression signature in dataset TCGA-HNSC.
a Volcano plot showing log ratio (x axis) and −log10 (P value) (y axis) of differential expression between 3-gene-positive OSCC samples (>= 2 activated genes) and the low-risk 3-genes negative samples (no or one gene activation). b Box plots showing the distribution of expression values of the three genes between the two groups of tumours defined above. 0 = low-risk 3-gene-negative samples; 1 = high-risk 3-genes positive samples. c Heatmap illustrating the expression of 339 genes up- or downregulated (t-test P value < 0.01, absolute fold change value > 2) in the high-risk 3-gene-positive OSCC samples compared with the low-risk 3-gene-negative samples. d Correlation plot between the 3-genes positive versus negative signature (x axis) and HPV-negative versus positive signature (y axis) (correlation coefficient = 0.72). e The gene expression profile of HPV-negative versus -positive OSCC tumours shares similarities and differences compared to the signature of 3-genes positive versus negative OSCC tumours in TCGA-HNSC dataset. GSEA plots showing the following genesets ranked according to the log ratio between 3-genes positive versus negative OSCC tumours in the TCGA-HNSC dataset. Left panel: genes depleted in HPV-negative versus positive OSCC tumours (fold change < −2 and P value < 0.01) (n = 357 genes). A large proportion of these genes are also downregulated in 3-genes positive tumours compared to 3-genes negative. Right panel: genes enriched in HPV-negative versus positive OSCC tumours (fold change > 2 and P value < 0.01) (n = 393 genes). Many of these genes are also upregulated in 3-genes positive tumours compared to 3-genes negative.
Fig. 5
Fig. 5. Downregulation of CCNA1 decreases cell proliferation and induces a shift of the gene expression profile toward that of 3-genes negative good prognosis OSCC tumours.
a CCNA1 knockdown in FaDu OSCC cell line. Western blots showing the decreased expression of the CCNA1 protein in FaDu cells after knockdown of the gene. In all, 20 µg of urea extracts from cells transfected either with siCCNA1 or with SiControl were analysed by immunoblotting using anti-CCNA1 (upper panel) and anti-tubulin (lower panel) antibodies. Three replicates of each condition were done. b FACS analysis of cell cycle in FaDu cells after siCCNA1. The cell cycle of FaDu cells transfected either with siCCNA1 or with SiControl was analysed by Flow Cytometry using the Accuri C6 Flow Cytometer. Three replicates of each condition were done. c Volcano plot showing the log ratio (x axis) and −log10 (P value) (y axis) of differential expression between control and siCCNA1 FaDu cells. d The gene expression profile of control versus siCCNA1 treated FaDu cells is similar to that of 3-genes positive versus negative OSCC tumours in both datasets. GSEA plots showing the following genesets ranked according to the log ratio between control (CCNA1 high) versus siCCNA1 (CCNA1 low) FaDu cells in our RNAseq experiment: d1, d4. Genes enriched or depleted in CCNA1 high versus low FaDu cells with absolute fold change > 2 and P value < 0.01 in control versus siCCNA1 FaDu cells) (respectively, n = 314 genes or n = 482 genes) (to illustrate the position of the most enriched genes on this ranked genes GSEA analysis). d2, d5: genes enriched or depleted in 3-genes positive versus negative OSCC samples from dataset GSE41613 with absolute fold change > 2 and P value < 0.01 (respectively, n = 48 genes or n = 112 genes). d3, d6: genes enriched or depleted in 3-genes positive versus negative OSCC samples from dataset TCGA-HNSC with absolute fold change > 2 and P value < 0.01) (respectively, n = 43 genes or n = 49 genes). e GSEA plots showing the enrichment/depletion of genesets indicative of sensitivity or resistance to therapeutics of aggressive 3-genes positive versus negative OSCC tumours from the GEO GSE41613 (1st column) and TCGA-HNSC (2nd column) datasets, FaDu cell line expressing high versus low levels of CCNA1 after experimental downregulation of CCNA1 (3rd column), and HPV-negative versus positive tumours from the TCGA-HNSC dataset (4th column). The following selection of three genesets are shown whose enrichment suggest (i) resistance to doxorubicin (5-fluorouracil): genes upregulated in doxorubicin-resistant vs -sensitive gastric cancer cell lines [28]), (ii) a potential sensitivity to Aplidin, a marine-derived compound with potential anticancer properties: genes downregulated in the MM1S cells (multiple myeloma) after treatment with aplidin [29], (iii) a sensitivity to Dasatinib, a multitargeted kinase inhibitor: genes whose expression positively correlated with sensitivity of breast cancer cell lines to Dasatinib: [30]). The details about the genesets are available on the MsigDB website (http://www.gsea-msigdb.org/gsea/msigdb/search.jsp). For detailed enrichment score values and composition of the genesets, please refer to Supplementary Table S6. Supplementary Fig. S9 shows more GSEA plots characterising the molecular profile of aggressive forms of OSCCs.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Blatt S, Krüger M, Ziebart T, Sagheb K, Schiegnitz E, Goetze E, et al. Biomarkers in diagnosis and therapy of oral squamous cell carcinoma: a review of the literature. J Craniomaxillofac Surg. 2017;45:722–30. doi: 10.1016/j.jcms.2017.01.033. - DOI - PubMed
    1. Gupta S, Kong W, Peng Y, Miao Q, Mackillop WJ. Temporal trends in the incidence and survival of cancers of the upper aerodigestive tract in Ontario and the United States. Int J Cancer. 2009;125:2159–65. doi: 10.1002/ijc.24533. - DOI - PubMed
    1. Safi A-F, Grochau K, Drebber U, Schick V, Thiele O, Backhaus T, et al. A novel histopathological scoring system for patients with oral squamous cell carcinoma. Clin Oral Investig. 2019. 10.1007/s00784-019-02804-4. - PubMed
    1. Rivera C, Oliveira AK, Costa RAP, De Rossi T, Paes Leme AF. Prognostic biomarkers in oral squamous cell carcinoma: a systematic review. Oral Oncol. 2017;72:38–47. doi: 10.1016/j.oraloncology.2017.07.003. - DOI - PubMed
    1. Almangush A, Heikkinen I, Mäkitie AA, Coletta RD, Läärä E, Leivo I, et al. Prognostic biomarkers for oral tongue squamous cell carcinoma: a systematic review and meta-analysis. Br J Cancer. 2017;117:856–66. doi: 10.1038/bjc.2017.244. - DOI - PMC - PubMed

Publication types

MeSH terms

-