Umeå University's logo

Locoregional recurrences and distant metastases are major problems for patients with squamous cell carcinoma of the head and neck (SCCHN). Because SCCHN is a heterogeneous group of tumours with varying characteristics, the present study concentrated on the subgroup of squamous cell carcinoma of the oral tongue (SCCOT) to investigate the use of machine learning approaches to predict the risk of recurrence from routine clinical data available at diagnosis. The approach also identified the most important parameters that identify and classify recurrence risk. A total of 66 patients with SCCOT were included. Clinical data available at diagnosis were analysed using statistical analysis and machine learning approaches. Tumour recurrence was associated with T stage (P=0.001), radiological neck metastasis (P=0.010) and diabetes (P=0.003). A machine learning model based on the random forest algorithm and with attendant explainability was used. Whilst patients with diabetes were overrepresented in the SCCOT cohort, diabetics had lower recur‑ rence rates (P=0.015 after adjusting for age and other clinical features) and an improved 2‑year survival (P=0.025) compared with non‑diabetics. Clinical, radiological and histological data available at diagnosis were used to establish a prognostic model for patients with SCCOT. Using machine learning to predict recurrence produced a classification model with 71.2% accuracy. Notably, one of the findings of the feature importance rankings of the model was that diabetics exhibited less recur‑ rence and improved survival compared with non‑diabetics, even after accounting for the independent prognostic variables of tumour size and patient age at diagnosis. These data imply that the therapeutic manipulation of glucose levels used to treatdiabetes may be useful for patients with SCCOT regardless of their diabetic status. Further studies are warranted to investigatethe impact of diabetes in other SCCHN subtypes.

PURPOSE: The field cancerization concept indicates the presence of pre-cancerous changes in clinically normal tissue surrounding the tumor. In squamous cell carcinoma of the oral tongue (SCCOT) which is infrequently linked to human papillomavirus infection, we have previously reported that clinically normal tongue contralateral to tumor (NTCT) is molecularly abnormal. Here, combining our transcriptomic and genomic data, we aimed to investigate the contribution of molecular changes in NTCT to cancer development.

METHODS: Microarray gene expression data of 14 healthy controls, 23 NTCT and 29 SCCOT samples were investigated to characterize transcriptional profiles in NTCT. Whole exome sequencing and RNA-sequencing data of paired NTCT and tumor samples from 15 SCCOT patients were used to study correlation between copy number variation and differential gene expression.

RESULTS: Using supervised multivariate partial least squares discriminant analysis, a total of 61 mRNAs that distinguish NTCT from healthy tongue were selected. Functional enrichment analysis of the 22 upregulated genes showed increased "positive regulation of nitrogen compound metabolic process" in NTCT. All 12 genes involved in this process have roles in apoptosis (anti- and/or pro-apoptotic). Compared to healthy controls, Zinc Finger Protein 395 (ZNF395), a pro-apoptotic tumor suppressor located on chromosome 8p, was the only gene showing increased mRNA level in NTCT whereas decreased in SCCOT. Given the frequent loss of chromosome 8p in SCCOT, the impact of ZNF395 copy number variation on gene expression was further examined, revealing a positive correlation between copy number and mRNA level (correlation coefficient = 0.572, p < 0.001).

CONCLUSION: NTCT is susceptible to malignant transformation, where tissue homeostasis is maintained at least partly through regulation of apoptosis. Loss of the pro-apoptotic gene ZNF395 could thus initiate cancer development.

Background: Interpretable machine learning (ML) for early detection of cancer has the potential to improve risk assessment and early intervention.

Methods: Data from 261 proteins related to inflammation and/or tumor processes in 123 blood samples collected from healthy persons, but of whom a sub-group later developed squamous cell carcinoma of the oral tongue (SCCOT), were analyzed. Samples from people who developed SCCOT within less than 5 years were classified as tumor-to-be and all other samples as tumor-free. The optimal ML algorithm for feature selection was identified and feature importance computed by the SHapley Additive exPlanations (SHAP) method. Five popular ML algorithms (AdaBoost, Artificial neural networks [ANNs], Decision Tree [DT], eXtreme Gradient Boosting [XGBoost], and Support Vector Machine [SVM]) were applied to establish prediction models, and decisions of the optimal models were interpreted by SHAP.

Results: Using the 22 selected features, the SVM prediction model showed the best performance (sensitivity = 0.867, specificity = 0.859, balanced accuracy = 0.863, area under the receiver operating characteristic curve [ROC-AUC] = 0.924). SHAP analysis revealed that the 22 features rendered varying person-specific impacts on model decision and the top three contributors to prediction were Interleukin 10 (IL10), TNF Receptor Associated Factor 2 (TRAF2), and Kallikrein Related Peptidase 12 (KLK12).

Conclusion: Using multidimensional plasma protein analysis and interpretable ML, we outline a systematic approach for early detection of SCCOT before the appearance of clinical signs.

Oral cancer is a paradigm of Slaughter's concept of field cancerization, where tumors are thought to originate within an area of cells containing genetic alterations that predispose to cancer development. The field size is unclear but may represent a large area of tissue, and the origin of mutations is also unclear. Here, we analyzed whole exome and transcriptome features in contralateral tumor-distal tongue (i.e. distant from the tumor, not tumor-adjacent) and corresponding tumor tissues of 15 patients with squamous cell carcinoma of the oral tongue. The number of point mutations ranged from 41 to 237 in tumors and from one to 78 in tumor-distal samples. Tumor-distal samples showed mainly clock-like (associated with aging) or tobacco smoking mutational signatures. Tumors additionally showed mutations that associate with cytidine deaminase AID/APOBEC enzyme activities or a UV-like signature. Importantly, no point mutations were shared between a tumor and the matched tumor-distal sample in any patient. TP53 was the most frequently mutated gene in tumors (67%), whereas a TP53 mutation was detected in only one tumor-distal sample, and this mutation was not shared with the matched tumor. Arm-level copy number variation (CNV) was found in 12 tumors, with loss of chromosome (Chr) 8p or gain of 8q being the most frequent events. Two tumor-distal samples showed a gain of Chr8, which was associated with increased expression of Chr8-located genes in these samples, although gene ontology did not show a role for these genes in oncogenic processes. In situ hybridization revealed a mixed pattern of Chr8 gain and neutral copy number in both tumor cells and adjacent nontumor epithelium in one patient. We conclude that distant field cancerization exists but does not present as tumor-related mutational events. The data are compatible with etiologic field effects, rather than classical monoclonal field cancerization theory. © 2022 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

The human homologues of murine double minute 2 (MDM2) and 4 (MDM4) negatively regulate p53 tumour suppressor activity and are reported to be frequently overexpressed in human malignancies, prompting clinical trials with drugs that prevent interactions between MDM2/MDM4 and p53. Bone marrow samples from 111 patients with acute myeloblastic leukaemia, myelodysplastic syndrome or chronic myelomonocytic leukaemia were examined for protein (fluorescence-activated cell sorting) and messenger RNA (mRNA) expression (quantitative polymerase chain reaction) of MDM2, MDM4 and tumour protein p53 (TP53). Low protein expression of MDM2 and MDM4 was observed in immature cells from patients with excess of marrow blasts (>5%) compared with CD34⁺/CD45^low cells from healthy donors and patients without excess of marrow blasts (<5%). The mRNA levels were indistinguishable in all samples examined regardless of disease status or blast levels. Low MDM2 and MDM4 protein expression were correlated with poor survival. These data show a poor correlation between mRNA and protein expression levels, suggesting that quantitative flow cytometry analysis of protein expression levels should be used to predict and validate the efficacy of MDM2 and MDM4 inhibitors. These findings show that advanced disease is associated with reduced MDM2 and MDM4 protein expression and indicate that the utility of MDM2 and MDM4 inhibitors may have to be reconsidered in the treatment of advanced myeloid malignancies.

Cellular stress conditions activate p53-dependent pathways to counteract the inflicted damage. To achieve the required functional diversity, p53 is subjected to numerous post-translational modifications and the expression of isoforms. Little is yet known how p53 has evolved to respond to different stress pathways. The p53 isoform p53/47 (p47 or ΔNp53) is linked to aging and neural degeneration and is expressed in human cells via an alternative cap-independent translation initiation from the 2nd in-frame AUG at codon 40 (+118) during endoplasmic reticulum (ER) stress. Despite an AUG codon in the same location, the mouse p53 mRNA does not express the corresponding isoform in either human or mouse-derived cells. High-throughput in-cell RNA structure probing shows that p47 expression is attributed to PERK kinase-dependent structural alterations in the human p53 mRNA, independently of eIF2α. These structural changes do not take place in murine p53 mRNA. Surprisingly, PERK response elements required for the p47 expression are located downstream of the 2nd AUG. The data show that the human p53 mRNA has evolved to respond to PERK-mediated regulation of mRNA structures in order to control p47 expression. The findings highlight how p53 mRNA co-evolved with the function of the encoded protein to specify p53-activities under different cellular conditions.

Background: Salivary gland tumors (SGTs) include a large group of rare neoplasms in the head and neck region, and the heterogeneous and overlapping features among the subtypes frequently make diagnostic difficulties. There is an urgent need to understand the cellular mechanisms underlying the heterogeneity and overlap among the subtypes, and explore the subtype-specific diagnostic biomarkers.

Methods: The tumor tissue and the adjacent normal tissue from the 6 most common types of SGTs were processed for organoid culture which only maintained tumor epithelial cells. Organoids were histologically evaluated based on phenotype markers, followed by transcriptional profiling using RNA-sequencing. The transcriptomic similarities and differences among the subtypes were analyzed by subtype consensus clustering and hierarchical clustering. Furthermore, by comparative transcriptional analysis for these 6 types of SGTs and the matched organoids, the potential diagnostic biomarkers from tumor epithelium were identified, in which two selected biomarkers were evaluated by qPCR and confirmed by immunohistochemistry staining using a tissue microarray.

Results: We generated a biobank of patient-derived organoids (PDOs) with 6 subtypes of SGTs, including 21 benign and 24 malignant SGTs. The PDOs recapitulated the morphological and transcriptional characteristics of the parental tumors. The overlap in the cell types and the heterogenous growth patterns were observed in the different subtypes of organoids. Comparing the bulk tissues, the cluster analysis of the PDOs remarkably revealed the epithelial characteristics, and visualized the intrinsic relationship among these subtypes. Finally, the exclusive biomarkers for the 6 most common types of SGTs were uncovered by comparative analysis, and PTP4A1 was demonstrated as a useful diagnostic biomarker for mucoepidermoid carcinoma.

Conclusions: We established the first organoid biobank with multiple subtypes of SGTs. PDOs of SGTs recapitulate the morphological and transcriptional characteristics of the original tumors, which uncovers subtype-specific biomarkers and reveals the molecular distance among the subtype of SGTs.

Transporter associated with antigen processing 1 (TAP1) and TAP2 serve pivotal roles in adaptive immunity. Tumor cells often show reduced antigen presentation on their surface as one mechanism to escape immune recognition. Whether downregulation of TAPs is a common mechanism of tumor immune evasion in squamous cell carcinoma of the oral tongue (SCCOT) is unclear. In the present study, samples from 78 patients with SCCOT and 17 patients with benign hyperplastic tongue lesions were analyzed for TAP1 and TAP2 expression by immunohistochemistry. The percentage of positive cells and staining intensity were scored. Associations with clinicopathological variables and survival outcome were also investigated. The results demonstrated that TAP1 and TAP2 levels were highly associated with each other in individual samples and were upregulated in SCCOT compared with benign lesions (P<0.001). The proportion of TAP1‐ or TAP2‐positive tumor cells was >80% in all but two of the tumors, whereas 25.6 and 23.0% of the tumors showed weak intensity of TAP1 and TAP2, respectively. There were no significant associations with clinicopathological variables or survival outcomes between TAP‐intermediate/strong and TAP‐weak tumors. However, in patients <70 years old and with early stage SCCOT, male patients had better outcomes than female patients (log‐rank P<0.05), and the best outcome was observed in male patients with intermediate/strong TAP expression. In conclusion, loss of TAP was not a frequent event in SCCOT and stronger TAP expression in male patients was associated with improved survival, providing further evidence for sex‐specific immune modulation in cancer.

Background: Patients with squamous cell carcinoma of the head and neck (SCCHN) have a high-risk of recurrence. We aimed to develop machine learning methods to identify transcriptomic and proteomic features that provide accurate classification models for predicting risk of early recurrence in SCCHN patients.

Methods: Clinical, genomic, transcriptomic and proteomic features distinguishing recurrence risk were examined in SCCHN patients from The Cancer Genome Atlas (TCGA). Recurrence within one year after treatment was classified as high-risk and no recurrence as low-risk.

Results: No significant differences in individual clinicopathological characteristics, mutation profiles or mRNA expression patterns were seen between the groups using conventional statistical analysis. Using the machine learning algorithm, extreme gradient boosting (XGBoost), ten proteins (RAD50, 4E-BP1, MYH11, MAP2K1, BECN1, NF2, RAB25, ERRFI1, KDR, SERPINE1) and five mRNAs (PLAUR, DKK1, AXIN2, ANG and VEGFA) made the greatest contribution to classification. These features were used to build improved models in XGBoost, achieving the best discrimination performance when combining transcriptomic and proteomic data, providing an accuracy of 0.939 and an Area Under the ROC Curve (AUC) of 0.951.

Conclusions: This study highlights machine learning to identify transcriptomic and proteomic factors that play important roles in predicting risk of recurrence in patients with SCCHN and to develop such models by iterative cycles to enhance their accuracy, thereby aiding the introduction of personalized treatment regimens.

Circulating lipoproteins as risk factors or prognostic indicators for various cancers have been investigated previously; however, no clear consensus has been reached. In this study, we aimed at evaluating the impact of serum lipoproteins on the prognosis of patients with squamous cell carcinoma of the head and neck (SCCHN). Levels of total cholesterol, low-density lipoproteins (LDL), high-density lipoproteins (HDL), triglycerides and lipoprotein(a) were measured in serum samples from 106 patients and 28 healthy controls. We found that HDL was the only lipoprotein exhibiting a significant difference in concentration between healthy controls and patients (p = 0.012). Kaplan–Meier survival curves indicated that patients with high levels of total cholesterol or LDL had better overall survival than patients with normal levels (p = 0.028 and p = 0.007, respectively). Looking at patients without lipid medication (n = 89) and adjusting for the effects of TNM stage and weight change, multivariate Cox regression models indicated that LDL was an independent prognostic factor for both overall (p = 0.005) and disease-free survival (p = 0.013). In summary, our study revealed that high LDL level is beneficial for survival outcome in patients with SCCHN. Use of cholesterol-lowering medicines for prevention or management of SCCHN needs to be evaluated carefully.