The omics analysis with reference of single cells, spheroids and organoids from HNC patient samples is a major challenge. One of the primary objectives of any omics analysis is to find reliable targets for therapeutic intervention. Such a task becomes possible with the identification of cell-specific genes which need to regulated specifically. Identification of biomarkers from a transcriptomics data typically start with the computational analysis of highly differentially expressed genes. This computational analysis becomes possible through well-established and benchmarked bioinformatics strategies, which face specific challenges in the case of single cell data emerging from HNSCC. Three issues require special attention viz. (i) much subtler changes in expression levels in single cell populations when compared to bulk expression data, (ii) sparsely collected data with lots of missing values, and (iii) absence of largescale relationships between single cell changes of expression and gene sets such as pathways or ontology terms, frequently used in interpreting bulk expression data outcomes.

To address the first of these issues (weak biomarker signal in differential expression), many computational tools, dedicated to the scRNA-seq data analysis have been developed (49, 50), which have allowed for significant advances in investigating heterogeneity and single-cell specific markers. A database of tools employed for scRNA-seq analysis has been reported and can be accessed via URL http://www.scRNA-tools.org (51). Tools like SCANPY (52) and Scatter (53) are some of the powerful robust pipelines that are well-integrated for comprehensive analysis (pre-processing and post-processing analysis) of scRNA-seq data. Many of these tools and resources provide expression data analysis of single cells, which takes into account the subtle gene expression level changes.

The second bioinformatics analysis issue is that of sparseness in the data sets. Poor coverage of expression values from each sample has two implications, (a) the very absence of the expression values may lead to missing the biomarkers altogether as only 10—20% values are reliably captured, (b) these dropouts adversely impact a confident grouping of cellular profiles into their subclasses as each transcript is described by a different set of genes. Few genes are present in one face while others are available in another. One of the solutions that has been proposed by bioinformaticians to address the sparseness of expression data in scRNA-seq is to reconstruct or predict the missing gene expression values, a process well-known as “imputation” in computer science. Traditional computational methods of imputation in general have dealt with a few or a small proportion of missing values in a data set. This problem is, however far more acute in the single-cell data due to much less information available to impute the missing ones. Imputing a missing value often relies on adopting a derived value from carefully selected similar samples. In single cell analysis, groups of samples are not known a priori. Hence, the question of identifying subclasses and imputing the missing values becomes a cyclic problem. Early computational techniques, developed for imputing gene expression values have included ZIFA (Zero-Inflated Factor Analyst) (55) and CIDR (clustering through imputation and dimensionality reduction) (56). Recently, SAVER, MAGIC and scImpute dedicated specifically to reconstructing a large number of missing values or imputations, were developed (57–59) and were successful in recovering the true expression of spike-ins transcripts improving and data quality.

Beyond the algorithms and tools for scRNA-seq analysis by addressing its sparseness, a number of data resources comprising transcriptomic and genomic information are also available in the public domain. The Table 2 has listed the Gene Expression Omnibus (GEO) Dataset collection of transcriptome and associated data from HNC. Also, TCGA HNC dataset, which includes 527 cases, is a vast resource containing comprehensive integrative datasets of SNV, CNV, methylation, and slide images as well, complementing the transcriptome data. These datasets can be accessed from the GDC Data Portal (https://portal.gdc.cancer.gov/).

The discussion above is based on the review of works on the issues of scRNA-seq analysis in general, which must anyway be addressed in HNC samples as well. However, so far there is only one published study that has specifically addressed the issue of single cell transcriptomics in HNC (60), while another study from the same group has comprehensively reviewed the bioinformatic approaches and key findings derived from for the single-cell technology-based study of cancer (61). These twin papers suggest that the HNC computational analysis and results can be broadly classified into three groups viz. (a) study of cellular heterogeneity and gene expression analysis (b) study of micro-environment of cancer cells and (c) process of invasion and metastasis of cancer cells.

Among the insights gained from bioinformatics analysis of HNC data sets at single cell levels, the foremost finding arguably is reported by Qi et al. (61) in which it was shown that patient outcomes under all treatment regimens are highly dependent on the intrinsic cellular heterogeneity. Intra-tumoral heterogeneity and tumor nest architecture was largely recapitulated within lymph node metastases. Specifically, it was observed that high heterogeneity measured by mutant-allele tumor heterogeneity (MATH) scores leads to poor patient outcomes, thereby highlighting the need to understand the cell population composition of HNSC cells to improve the patient survival rates. Authors also found that malignant cells’ expression patterns could not be distinguished from those of basal tumors, suggesting that most tumors could be defined as a single ‘malignant-basal’ cohort in OSCC. This contrasts with the glioblastoma multiforme (GBM) tumor, in which the malignant cells map to multiple different subtypes. These findings suggest that HNC tumor consists of lower diversity in malignant subtypes or because the subtypes have not been confidently resolved at this stage. Another study was performed to examine the change in tumor properties by simulating single-cell events leading to macroscopic tumor development. The model was able to successfully observe adhesion-driven cell movements and nutrition dependent heterogeneous tumor growth. Different treatment plans strongly influenced the final tumor cell type composition. The growth rate was observed to be significantly decreased when metabolism in tumor cells was upregulated. The mutation rates were adjusted, and low mutation rates cell types with higher division rate and delayed cell death started dominating the tumor. The models were also used to probe treatment regimens. Shorter pulses of chemotherapy were observed to have a better effect than a uniform application. The tumor size was significantly reduced by a single strong radiotherapy pulse as compared to multiple weaker pulses. The presence of tumor stem cells was confirmed to impact treatment outcome by increasing tumor size as well as heterogeneity. In view of the above results, the single-cell simulations can be a source of information to determine the heterogeneity and also predict treatment strategy and outcomes. These proved to be highly useful in improving the understanding of tumor development on a single-cell level but also the differences/similarities from bulk tumor analysis (62).

Since, cellular heterogeneity is so critical to HNC characterization and personalized treatment, researchers have tried to establish general patterns of cellular heterogeneity so prior therapeutics for each group can be developed. Bioinformatics work has concluded that the non-malignant cells from HNC patients could be grouped into eight main clusters by cell type viz. (a) T cells, (b) B/plasma cells, (c) macrophages, (d) dendritic cells, (e) mast cells, (f) endothelial cells, (g) fibroblasts, and (h) myocytes. However, the computational analysis so far has found that these non-malignant cells did not cluster as per their origin, when their expression profiles are used for automatic grouping, suggesting that the cell types and their expression states are consistent across tumor when their expression profiles are used for automatic grouping. On the other hand, malignant cells clustered well by the patient, suggesting expression changes across patients are more diverse than across cells of the same patient. In summary, malignant cells carry patient identity. The origin of cell from the 08 groups was not well encoded into the gene expression program. In the same context, another study by Yost et al. (63) on basal cell carcinoma using a combination of scRNA-seq and TCR sequencing, have implicated cancer-associated fibroblasts (CAFs) in tumorigenesis, tumor survival, ECM remodeling, immune system suppression, and tumor invasion system suppression, and tumor invasion. Another study by Leung et al. (64) focused on single-cell DNA sequencing, exome sequencing, and targeted deep-sequencing has investigated clonal evolution during metastatic dissemination in two colorectal patients. This study has highlighted that understanding the clonality at a single-cell level in a tumor is essential to simultaneously capturing and maintaining spatial information. Another study by Casasent et al. (65) has reported a method called Topographic Single Cell Sequencing (TSCS), which utilizes a combination of LCM (66) and single-cell DNA-sequencing to measure genomic copy number profiles of single tumor cells in breast cancer patients. This approach preserves single cells’ spatial context, which is critical to the location-specific therapeutic targeting strategies. Although the studies mentioned here are performed on cancers other than HNC, they successfully present strategies to combat the challenges associated with bioinformatics analysis.

Recently, a review published on the applications of single cell RNA sequencing in the field of otolaryngology, self-analyzed the single cell RNA seq data of HNC patients taken from the study by Puram et al. The analysis gave following findings that were relevant for clinicians 1) The scRNA-Seq data not only distinguished the disease causing cells from native tissue but also revealed the heterogeneity within diseased tissue samples. 2) Malignant cells from 10 HNC patients, when mixed, formed patient specific clusters i.e. with the cells of their original native tissue only. This suggested that clonal evolution is unique to each patient, and therefore the treatment strategy needs to be personalized. 3) Cells from the tumor microenvironment (TME) were also profiled along with malignant cells. However, these were not found to be clustering on patient-specific basis but rather on a cell-type basis. These cells could thus represent shared disease pathogenesis between all HNC patients that can be targeted using a similar therapy. 4) Rare cell types like stem cells, progenitor cells, CD4+ T-regulatory cells or exhausted T-cells were also identified from TME. These helped in understanding the disease maintenance, immune evasion and decreased efficacy of immune therapies. 5) Most importantly, the cell type specific biomarkers can be identified by investigating gene expression in heterogenous cell clusters detected by scRNA-Seq. For example, Puram et al. identified partial-EMT signature detected in a subset of malignant cells which was also present in existing bulk RNA-Seq tumor data. Such identifications can enable clinicians to determine the risks of nodal dissections on the basis of signatures indicating risk of metastasis. The prognostic signatures predicting survival, metastasis, chemoresistance can vary patient to patient. Such signatures can also be identified as markers to monitor drug response, emergence of resistance etc. before and after treatment. 6) Looking for genetic targets of FDA-approved drugs or small molecules in clusters of malignant sub-populations or TME cells can help identifying new druggable targets. A new database called Pharos describes 20,000 gene/protein targets and the drugs molecules available which can be further repurposed for use in HNC treatment (67).

Some bioinformatics studies have gone beyond biomarker discovery and cellular heterogeneity. Few researchers have used appropriate bioinformatics tools in creating and maintaining the tumor ecosystem’s spatial organization. Researchers have found that partial-EMT (P-EMT) cells were loosely arranged, and positioned in between malignant cells and CAFs. The study attributed the compactness of HNC tumor architecture to the expression of CD63 (68). Studies by Ligorio et al. (69) and Wagner et al. (70) in pancreatic and breast cancer respectively, have highlighted the need to utilize single cell separation method (SCS methods) with preserved spatial information, to gain insights into the role of intercellular interactions.

Another study by Navin et al. elucidated the tumor evolution process in breast cancer through sequencing of 100 single cells and revealed 3 distinct clonal sub-populations that represent sequential expansions. Contrasting to the gradual models of tumor progression their data indicated that tumors grow by punctuated clonal expansions. The study was performed on breast cancer and its liver metastases (71). More such studies on HNC will help in developing an understanding of the temporal progression of tumor heterogeneity. In response to systemic therapy, the issue of recurrence of tumor and overall temporal dynamics are other issues of transcription data analysis that heavily rely on suitable computational strategies, which are still under development.

The scRNA-Seq is a stride towards personalized medicine, but is still daunted by several challenges. Lack of large cohorts of scRNA-Seq data from human patient samples, high costs, user-friendliness, and tissue preservation are some of the major issues. The use of scRNA-Seq on individual patient tumors for drug selection is now feasible but more studies are still needed to establish personalized drug selection and drug repurposing using scRNA-Seq results for improved patient outcomes.

The cost of scRNA-Seq varies based on the chosen methodology, and hence depends on the cost of equipment, reagents, and sequencing. The costs of isolation and sequencing per cell have dropped significantly, but the throughput of sequencing machines has also increased, so the cost per run with more cells still remains high. Most of the platforms are available only in science laboratories and require a large investment and planning to procure for hospital use. In addition to cost, analysis of scRNA-Seq data requires basic bioinformatics knowledge and coding skills. Furthermore, standardization of different pipelines is also required for clinical use.

Tissue preservation is a major issue because of its fragility and cell viability. Currently, the use of frozen tissue samples or methanol fixed tissues for scRNA-Seq platforms is in its infancy. However, a few other options to aid tissue preservation are available and includes, temporary tissue stabilization buffers that can preserve cells for sequencing for 48 hours.

Generally, single nucleus sequencing (sNuc-seq) usually involves tissue disruption and cell lysis, carried out in cold conditions, followed by centrifugation and separation of the nuclei from the debris. It minimizes the skewing effect of degraded mRNA or cell-stress response genes on the data. Cell lysis in sNUC-Seq allows for potentially more efficient cell type delineation that includes for even the most interdigitated cell types. These advantages potentially make sNuc-Seq a better alternative to SCRNA-Seq. strategy.

Single cell DNA (scDNA) sequencing is focused mainly on the copy number variations (CNVs) and identification of single-nucleotide variations (SNVs). These are the driving forces in biological processes which cause genomic heterogeneity and thus necessitate study of the cell at an individual level. The whole genome wide analysis of HNC identified mutations in many gene families, but the most significant percentage of mutations were observed in the NOTCH gene family (72–74), especially NOTCH1. NOTCH and many other known oncogenes, including cyclin E, MYC, and JUN are targets of FBXW7, a ubiquitin ligase. FBXW7 is known to be mutated in 4.7% of cancers of HNC (74). Apart from this, more than 60% of mutations were observed in serine/phosphatidylinositol 3-kinase (PI3K) pathway genes such as PTEN and PI3KCA (75, 76). In fact, this is the most commonly affected pathway in HNC, and a more aggressive form of the disease can be attributed to multiple mutations in this pathway (77). Approximately 8-23% of HNCs possess mutation in PTEN that causes down-regulation and constitutive activation of threonine-specific protein kinase Akt and mammalian target of rapamycin (mTOR) (74, 78). It increases the susceptibility of the oral epithelium to carcinogens. The genome analysis in HPV positive HNSSC tumor showed mutations in PI3KCA gene leading to an increase in mTOR activity rather than Akt phosphorylation and hence helps explains the better efficacy of dual inhibitors against PI3K/mTOR (79). Interestingly, p53 was not found expressed in HNC tumors with PTEN downregulation, implying the exclusion of p53 gene mutation (80).

The Epidermal Growth Factor Receptor (EGFR), a receptor tyrosine kinase (RTK) gene found upregulated in 80% of the patients suffering with HNC. The EGFR on activation causes cellular proliferation via either RAS/RAF/MAPK pathway, JAK/STAT, or PI3K/AKT/mTOR axis. Its over-expression in many HNSSC tumors is correlated to poor prognosis (81). In nearly 20% of HNC, oxidative stress genes are altered by mutation or variation in copy number. NRF2 (encoded by the NFE2L2 locus) is a transcription factor that activates a cellular antioxidant response. It is overexpressed in 90% of the tumors leading to poor prognosis (82). Elevated NRF2 levels are shown to cause chemoresistance in a variety of cancer cell lines that is reversible with siRNA inhibition of NRF2 (83)⁠. Several chromatin-related genes in HNC viz, MLL2 (a histone methyltransferase), NSD1 (another histone methyltransferase), EP300 (a histone acetyltransferase) and FAT1 were also found to be repeatedly mutated in 19%, 10%, 7% and 23% of tumors respectively (84, 85). A recent study on HNSCC patients assessed the prognostic value of altered immune gene expression using a cohort of 96 patients (86). The expression of 46 immune-related genes was analyzed and, 4-1BB, IDO1, OX40L, GITR, FOXP3 were found significantly overexpressed along with PD-1, TIGIT, and CTLA-4. Almost half of the immune related genes had deregulated mRNA levels. The study assessed that a combination of high OX40-L and low PD-1 mRNA levels, high PDGFRB, and low CD3E mRNA levels are associated with increased tumor recurrence. While CD8A was observed to be associated with poor prognosis, the increased expression of PD-1 was associated with a good prognosis. These findings offer a therapeutic strategy in the treatment of HNSCC through the application of a combination of immune checkpoint inhibitors. Genetic alterations due to tobacco and betel quid chewing were also reported in oral cancer patients (87). These included i) single nucleotide polymorphism (SNPs) with non-synonymous type variations such as in FAT1& 2, TP53, NOTCH2, Cadherin 3 (CDH3), and ATM; ii) synonymous type variations in Adenomatous Polyposis Coli gene (APC) (a tumor suppressor gene) and IL12B (cytokine gene). SNPs were also observed in non-coding regions, located in or near EGFR, STAT5B, Cyclin dependent kinase 5 (CDK5), and a protooncogene, MYCL1 ( Figure 4 ). Sayans et al. (88) analyzed 528 tumors of HNSCC subset in TCGA database and found 3491 deregulated genes. The somatic copy number alteration analysis showed CDKN2A, CDKN2B, PPFIA1, FADD, and ANO1 as the most altered HNSCC genes. At the same time, genes with the most somatic mutations were TP53, TTN, FAT1 and, MUC16. Another relevant result from the study was the mutual exclusivity pattern found between TP53 and PIK3CA mutations. The difference in expression profiles between different studies i.e., the heterogeneity in the results could be attributed to the nature of the cancer.

One of the recent applications of transcriptomics in cancer is the study of the cellular heterogeneity in tumor towards better understanding to achieve precision treatment. HPV positive HNC is a vital cancer type and has been identified with different gene expression patterns compared to HPV negative HNC. Transcriptomic data analysis between HPV positive and negative tumors provided important insights into the expression profiles (76, 89).Activated receptor (RTKs)-RAS-PI3K pathways and inactivated TP53 and CDKN2A in HPV-negative tumors were observed. In HPV-positive tumors, PIK3CA, FGFR3, and E2F1 were found to be activated while TP53 and RB1 were inactivated by viral oncoproteins E6 and E7 respectively. PI3K activation in HNC is reported by either of these mechanisms, receptor- tyrosine kinases, such as EGFR or mutation occurring in PI3K catalytic subunit, p110α (encoded by PIK3CA gene). Mutations often target one of two hotspot locations in the kinase or in helical domain, thereby promoting constitutive signaling through the pathway (90). Yu et al. (91) reported results from a network-based meta-analysis, identifying the biological signatures of HNC in pathways like integrin signaling, tight-junction regulation, antigen presentation, chemokine signaling, leucocyte extravasation, and vascular endothelial growth factor (VEGF) signaling.

Another transcriptomics study in HNC suggested the upregulation of genes involved in digestion and remodeling of the ECM, such as matrix metalloproteinases (MMP) 1-3, 9, 10, 13, urokinase plasminogen activator (uPA), Integrin alpha (ITGA) 3 and ITGA5. Both neoplastic and stromal cells secrete MMPs that digest certain components of the ECM (92) and promote cell migration and metastases in early stages of tumorigenesis (93, 94). Overexpression and activation of MMPs is critical in cancer progression and the pro-MMP-9/NGAL complex has been identified as a potential prognostic marker (95). A related study on a cohort of 145 oral cancer patients exhibited high levels of MMP2 in severe patients when compared to non-severe oral cancer patients. High levels of CD276 and low levels of CXCL10 and STAT1 were also observed to be associated with reduced overall survival. However, when compared MMP2 appeared to be a superior and independent prognostic marker (96).

The upregulation of interleukin (IL) 8, chemokine C-X-C ligand 1 (CXCL1), CD28, CD3D, CD4, IL-18, and IL-2 is observed in chemotaxis and lymphocyte activation while downregulation of MHC 1 &2 are hallmarks of invasive HNCs. Also, upregulation of VEGF and interleukin-8 (IL-8) connoted tumor cell angiogenesis, while EGFR, STAT-3, PI3K, and NOTCH upregulation influenced signal transduction pathways (97).

One hundred forty-six novel miRNAs expressed in HNC have been identified; but expression patterns among smokers and non-smokers remained undistinguishable. The three novel miRNAs significantly associated with HPV status, were mapped to chromosome 12 between genes Keratin 6C (KRT6C) and KRT6B (98).

Puram et al. have reclassified HNC into three malignant subtypes: classical, basal-mesenchymal and atypical. Single-cell transcriptomics from 18 HNC patients identified p-EMT as an independent predictor of grade, metastasis and critical pathological features (60). They performed the scRNA-seq analysis by considering 6,000 single cells from eighteen HNC patients containing five sets of matched primary tumor and lymph node metastases. The significant finding of the study was to distinguish among non-malignant (3363) and malignant (2215) cells on the basis of copy-number variations (CNVs) and epithelial cells where stromal and immune cells were excluded (60, 99, 100). Clinical and genomic meta-analysis of multicohort HNSCC gene expression profile has clearly demonstrated that HPV⁺ and HPV^- HNSCCs are not only derived from tissues of different anatomical regions, but also present with different mutation profiles, molecular characteristics, immune landscapes, and clinical prognosis. Cell lines and primary cells of HNC have been explored at single-cell transcriptomics (60, 101). The datasets have significantly improved the identification of distinct cells which are highly tumorigenic in nature in the HNC ecosystem. In the pool of cells, including malignant and non-malignant type, intra-tumoral variations at cell cycle, partial-EMT, proliferation, hypoxia-related genes have been observed. In this context, scRNA-seq is becoming a reliable technique for exploring HNC heterogeneity both at the genetic and functional levels. All the tumor influencing factors, such as circulating tumor cells (CTCs), immune cells, cancer stem cells (CSCs), present within, or in surroundings are investigated to gain clarity at a single cell level.

The scRNA-seq data may be used for understanding the drug response, as well as, drug resistance in individual HNC patients. The cetuximab-treated and untreated HNC cells yielded heterogeneous expressions of TFAP2A and EMT during the early stage of treatments, indicating onset of resistance. The expression variation analysis (EVA) analysis of scRNA-seq data suggests that cetuximab treatment increases cell heterogeneity, leading to evolution of different clonal cells with differentially activated pathways, thereby preventing EGFR inhibition (102).

A comprehensive multi-omics, single-cell analysis was performed in HNC cell lines by Kagohara et al., to identify responses to cetuximab, an anti-EGFR drug (102). It was observed that hundreds of genes altered their expression pattern as a response to the drug within 5 days of treatment. scRNA seq analysis identified onset of resistance following changes in various signaling pathways including regulation of receptor tyrosine kinases by Transcription Factor AP-2 (TFAP2A) and epithelial-to-mesenchymal transition (EMT) pathway. Different squamous cell carcinoma cell lines exhibited cell type dependent differential expression of TFAP2A and Vimentin (VIM) genes that corroborates inter cell line heterogeneity. The available HNC data bases provide clinical and genomic information on HNC cell systems (102–104). A holistic HNdb database curates all major omics data and literature on HNC-related genes (105). This database has laid the foundation for identification of possible biomarkers and development of HNC personalized medicine. It is interesting to note that a few genes are common in genomics, transcriptomics and scRNA-seq analysis of the HNC ( Figure 4 ). These finding have stemmed from the independent studies. Therefore, it is imperative to perform integrated multi-omics studies and visualize molecular linkages using systems medicine for paving a way for personalized medicine.

The CSCs are responsible for failures of cancer therapeutics, drug resistance, and tumor recurrence. The single-cell transcriptomic data from salivary gland squamous cell carcinoma reported luminal and basal epithelial cells, as well as, small populations of CSCs. Overall, the study indicated that the process of tumorigenesis followed ‘gain-of-function’ by β-catenin and ‘loss-of–function’ by Bmpr1A mutations in basal cells, EMT markers expression, and activated Wnt signaling in CSCs of luminal cells (106).

In order to minimize variables arising from HNC intra-tumor heterogeneity, analysis of differentially expressed proteins have been strategized. Bhat et al. (107) identified 286 biomolecules, having relevance in HNC. A few of these included i) insulin like growth factor binding protein (IGFBP) ii) downstream signaling components ERK, COX2, STAT, PFN2, EPCAM, SERPINH1, MCM2, iii) genes involved in prolactin signaling iv) angiogenesis v) DNA repair genes using integrated transcriptomics and proteomics approach. It has been reported that the ERK, COX2 and STAT1 proteins are important in progression and development of chemo resistance in HNC. Hence, these may be potential targets for effective therapy (108, 109). The saliva serves as a source for identification of bio-markers in cancer, and its proteomic analysis is considered to be a promising tool for HNC diagnosis; for example, over-expression of PLUNC and zinc-alpha-2-glycoprotein (110). To better understand the process of tumor progression and to make detection of cancer with precision, a technical triad of laser microdissection, protein chip technology and immunohistochemistry have been employed to identify the tumor relevant biomarkers. This study encompasses the protein profiling of calgranulin A and calgranulin B which are implicated in cancer pathology. Thus, such combinatorial approaches open up the possibility towards accurate prediction of metastasizing ability of a cell population (111, 112).

The proteins like Hsp90, VIM and keratin are already established bio-markers and drug targets while prelamin-A/C and PGAM1, have been recently suggested as potential markers (113). Bohnenberger et al. (114) identified distinct proteomic profiles between lung metastasis of HNC (metHNC) and squamous cell lung carcinoma (SQCLC). On classifying 51 squamous cell lung tumors, as either primary SQCLC or metHNC using proteomic approaches, 518 proteins with significantly different expression levels in HNC and SQCLC were identified. These proteins belonged to pathways involved in (i) vesicle transport, (ii) glycosylation, or (iii) RNA-processing. The FAM83H expression generally upregulated in cancers, was correlated to poor prognosis in HNC as well (115). The locoregional recurrence after chemotherapy (platinum-based concurrent chemoradiation) frequently occurs in HNC patients. It was observed that the intra-tumoral heterogeneity is linked to clonal evolution, and it is actually responsible for cisdiamminedichloridoplatinum (II) (CDDP) resistance in HNC (115). Niehr and co-workers (116) have applied targeted next-generation sequencing, fluorescence in situ hybridization, microarray-based transcriptome, and mass spectrometry-based phosphor-proteome analysis to elucidate the molecular basis of CDDP resistance. This resistance was observed to be associated with aneuploidy of chromosome 17, increased TP53 copy-number, overexpression of the gain-of-function (GOF) mutant variant p53R248L and increased activity of the PI3K–AKT–mTOR pathway, which were also considered as molecular targets for treatment optimization (116). Furthermore, label-free profiling of proteins in oral cancer has been performed by relative quantitation and employing nano-UPLC-Q-TOF ion mobility mass spectrometry hence, enabling rapid and simultaneous identification of multiple cancer biomarkers (117). This approach appears to have promising implications on tumor diagnosis. Single cell proteomics approach has encouraged system-wide protein profiling, direct assessment of immune cell health and tumor–immune interactions. This further helped augmenting evaluation of immunotherapy (118). Moreover, profiling of every single individual cell appears to indicate its role in tumor progression and molecular basis of the disease (119). The p53 tumor suppressor proteins have been counted in single colorectal cancer cells with 88% accuracy using the MAC chip (microfluidic antibody capture) (120). However, MAC chip utility in HNC is yet to be established. Multiplexing of protein markers at single-cell level using immunofluorescence methods have also been applied. However, single cell proteomics methods are in developing state and the proteome coverage is smaller in comparison to single-cell transcriptomics. In the context of precision medicine, integrating the protein based prognostic biomarkers is emerging as a supporting strategy for the treatment of cancer patients.

Most head and neck cancers expressing elevated levels of desmoglein 3 (DSG3) metastasize to the neck lymph nodes. The IHC and H&E reports may not always detect DSG3 during the initial metastasis process when metastatic lesions are less than 2mm in size. The use of sensitive methods like RT-PCR, scRNA-seq, and next-generation sequencing (NGS) is costlier and time consuming. Measuring the protein expression of tumor metastasis marker during the earlier phase of cell growth at the single-cell level for therapeutics provides additional advantages. The 3D printed microfluidics immune-array has a 10,000-fold higher sensitivity, which is superior to ELISA. This does not even requires any sorting experiments prior detection of proteins from a single cell. Not only it detects DSG-3, VEGF-A, and VEGF-C at lower concentrations, but its automated operations also provide results at a fast pace and lower cost. In addition to delivering information about HNC, it also quickly reproduces the results with minimal errors (121).

A comprehensive analysis of metabolites or metabolomic study is cardinal to cancer pathology as metabolome is a summary manifestation of all the other upstream omic profiles (122).

In a tissue metabolite profiling of HNC, 41 out of 109 metabolites screened were observed to be higher in tumorous versus non-tumorous tissues, while 15 appeared lower. Serum levels of glycolytic pathway metabolites increased (glucose, fructose, tagatose etc.), while that of several amino acids for example, lysine decreased significantly. Conversely, in tissue samples the glycolytic pathway metabolites decreased, and amino acids (valine, phenylalanine, threonine etc.) increased in tumorous versus non-tumorous tissues (122). Since, cancer cells depend more on aerobic glycolysis rather than oxidative phosphorylation for energy, and also use glutamine as major source of energy, they deplete glucose in hypo-vascular microenvironment. Also, amino acid levels are higher due to degradation of ECM in tumors. Another study showed the increased levels of polyamines in saliva of oral cancer patients in comparison to that of other cancer types. The choline to creatinine ratio revealed oral cancer specific elevation. In addition to this, 28 metabolites that accurately differentiate oral cancers from control samples were also identified. However, oral cancer may have higher impact on the metabolite composition of saliva in comparison to other cancers simply because of its location. Therefore, to confirm this a concurrent and comparative metabolic profile from saliva, blood and cancer tissue is warranted to confirm the oral cancer specific role of choline-creatinine ratio (123). Additional conformation was derived from another serum based study of 25 metabolites, of which 7 metabolites (leucine, isoleucine, taurine, valine, choline, tryptophan and cadaverine) were manifested in both the studies. Altered levels of urea and 3-hydroxybutyric acid were also reported for the first time in the later study (124).

A study by Wei (125), identified a signature panel of salivary metabolites (phenylalanine, valine, γ-aminobutyric acid, n-eicosanoic acid and lactic acid) whose levels were significantly altered in oral squamous cell carcinomas (OSCC). Hence these could potentially be used as biomarkers to distinguish between healthy and disease physiologies (125). While increase in lactic acid is simply explained by Warburg effect in glycolysis, valine and other amino acids are found significantly to be decreased presumably due to increased metabolic utilization. Increased ketone bodies, abnormal lipolysis, TCA cycle and amino acid metabolism have been reported in blood serum from OSCC patients (126). Patients with disease relapse exhibited increase in glucose, ribose, fructose, and tagatose with decrease in lysine, hippurate, trans-4-hydroxy-L-proline, and 4-hydroxymandelate in serum samples. A GC-MS based serum screening of OSCC revealed differences in 38 metabolites at pre-operative levels in comparison to healthy individuals. Furthermore, a comparison of pre-operative and post-operative metabolite profiles yielded significant differences in 32 metabolites. Seven potential biomarker candidates were found, i glyceric acid, lauric acid, N-acetyl-L-aspartic acid, ornithine, heptadecanoate, serine and asparagines. The sensitivity and specificity of biomarker pairs were assessed as 94.4% and 82.8% for ornithine+asparagine, 88.8% and 85.7% ornithine+glyceric acid, 88.8% and 97.1% ornithine+N-acetyl-L-aspartic acid, and 88.8% and 82.8% for ornithine+serine; endorsing their potential in early detection and stage identification in OSCC (127). An increase in choline compounds in OSCC implies its significant role in cancer feedback cell signaling. These increased choline levels renders it as a potential biomarker for cancer cell proliferation, survival and malignancy (128). Decreased levels of PUFA and creatine, and increased levels of amino acids and glutathione, were also observed in a study in tissues through proton high-resolution magic angle spinning magnetic resonance (HR-MAS MR) (129).

The significant data on HNC metabolomics, is hindered by differences in detection and analytical methods. In addition, the inherent heterogeneity in HNC has obstructed the identification of an accurate biomarker for its early detection (130). Studies based on single cell analysis have shown significant differences from average pattern in bulk samples. Most metabolic changes in single malignant cells are not captured through bulk measurements as they tend to underestimate the highly complicated cellular composition of bulk samples. Though there is a universal upregulation of metabolic pathways, the over-expressions of certain genes (for example, OXPHOS i.e. oxidative phosphorylation pathway genes) are evidenced only at single cell level. Their absence at the bulk level is credited to the probable fallout of bulk measurements, enmeshed in the complexity of tumor composition. Differential expression from bulk level is also observed in genes involved in Vitamin b₆ metabolism, lysine degradation, synthesis of aromatic amino acids, drug metabolism through cytochrome P450, degradation of fatty acids, oxidative phosphorylation, TCA cycle etc. However, where the expression at single cell and bulk level is different in purine metabolism, it was found similar in pyrimidine metabolism. Twenty-four out of fifty-six pathways show similar patterns of up-regulation or downregulation upon comparison between single malignant cells and bulk tumors, while 25 pathways that were reported downregulated through bulk tumor analysis were found upregulated on single cell level (131). Figure 5 represents the major metabolic pathways upregulated in single cell and bulk tumor analyses. The inter-section in the Venn is indicative of pathways similarly upregulated or downregulated in both. The major cause of heterogeneity is the variations in mitochondrial metabolic activity (TCA cycle and Oxidative phosphorylation). Also, the metabolic features of immune and stromal cell sub-types were found distinct when the mean expression level of genes within these pathways were compared. Therefore, more single cell-based studies are required to not only gain better insights but also eliminate existing discrepancies, and to help identify different metabolic phenotypes in cell sub-populations.

Notably, intra tumor heterogeneity is the most significant hurdle in developing effective anticancer drugs, as targeted drugs and chemotherapy are effective until the development of drug resistance (133, 134). Tools like single-cell pharmacokinetic imaging have emerged as a powerful means to elucidate the mechanism of drug resistance in the tumor that may help overcoming the resistance (135). Characterization of cancer heterogeneity in epigenomic sub-populations appears to be relevant as cancer evolution, drug sensitivity, etc. are necessarily impacted by epigenetic alterations. This can be achieved using single-cell technology but is viable only at an early stage of cancer. In this context the degree of single cell chromatin accessibility also constitutes a significant challenge (136).

The epigenetic modifications are known to control programmed developmental changes and the ability of the genome to register, signal and perpetuate environmental cues (132). In order to sustain the inheritance of gene expression and biological functions, epigenetic mechanisms are linked to the transmission of cell lineage and phenotype from progenitor to progeny. These modifications are now known to be transmitted to the progeny cells with the epigenetic marks or genome bookmarking by transcription factors and other gene regulatory proteins (137, 138). The deviation from the transmission of normal epigenetic marking is suggested to be relevant not only in cell differentiation but also in the onset of several diseases, including cancer. In this context, some other vital chemical modifications altering chromatin states and subsequent gene expression patterns include DNA methylation, histone modifications, small non-coding RNAs, and chromatin remodeling factors. This is currently a subject of intensive study.

Protein kinases are involved in cancer metabolism and have been the second most important drug targets in the pharmaceutical industry after G-protein-coupled receptors. The crystal structures of kinase-inhibitor complexes of different families have been determined, these include (i) receptor tyrosine kinase (EGFR, HER2), vascular endothelial growth factor receptor (VEGFR), JAK2, JAK3, Syk, ZAP-70, Tie2, EGFR, V-EGFR, FGFR) (ii) non-receptor tyrosine kinase (Bcr-Abl) NOTCH1, Janus kinase (JAK) (iii) serine-Threonine kinase (Clk, Dyrk, Chk1, IKK2, CDK1, CDK2, PLK, JNK3, GSK3, mTOR, p38 MAPK, PKB) (iii) Rho-kinase (iv) Cyclin-dependent kinases (179). In these structures, active and inactive state of the protein kinases, ATPase pocket, point mutations, catalytic and non-catalytic domains of the kinases have been used as targets by kinase inhibitors and provided the mechanism of inhibition. A well-documented crystallographic analysis of the cAbl kinase domain with Gleevec inhibitor revealed locking of the protein kinase’s inactive conformation (180). The conformational flexibility and stability of protein kinases are central to their inhibition and subsequent drug designing strategies. IIdentifying the diagnostic significance of p38 isoforms in HNC and the subsequent design of specific peptide inhibitors against p38a MAPK aims to contribute to anti-kinase drug development, and the expanding expertise offers optimistic prospects for future cancer treatment.

G protein-coupled receptors (GPCRs) are involved in signaling pathways and can elicit both cytostatic and cytotoxic effects. Four of the GPCRs, (i) galanin receptor type 1 (GALR1) (ii) GALR2, (iii) tachykinin receptor type 1 (TACR1), and (iv) somatostatin receptor type 1 (SST1) are the most studied and promising therapeutic target in a wide variety of cancer. GALR1 & 2 both inhibit cell proliferation and apoptosis of HNC cells. GALR1 act through ERK1/2-mediated activation of cell cycle control proteins such as p27, p57, and suppression of cyclin D1 protein. In p53 mutant HNC cells, GALR2 was found to have anti-proliferative and pro-apoptotic effects (190). The significant reduced disease-free survival and a higher recurrence rate is associated with hypermethylation of GALR1, GALR2, TACR1, and SST1. Methylation of GAL, TAC, and SST and its investigation as potential prognostic markers in HNC has already been discussed in Epigenomics of HNC.

The chemotherapeutic agents such as, afatinib, poziotinib, vandetanib, nintedanib, gefitnib, erlotinib, lapatinib, dacomitinib, alpelisib, PX-866 are the multitargeted inhibitors of protein kinases that regulate Ras/Raf/MEK/ERK/PI3K signaling pathways (193). The immunotherapeutic approaches such as specific antibodies targeting tumor, cytokines, cancer vaccines, and immune-modulating agents are other cancer treatment strategies, discussed below and in Tables 6 and 7 .

A continuous flow of new molecules, explicitly targeting the upcoming biomarkers, is required as few of the promising agents have failed to show desired results in clinical trials. These include inhibitors of PI3K and mTOR pathways, e.g., px-866, an inhibitor of PI3K that binds to ATP catalytic site (194). Another antiproliferative and immunosuppressive drug sirolimus (extracted from streptomyces bacteria) demonstrated critical challenges in the form of poor bioavailability and long half-life in patients leading to frequent monitoring of the drug (195). Thus, substituting the drug with its better analogs with improved pharmacokinetic properties seems desirable.

A study by our group on the effects of a combination of two drugs against HNC showed that a combination of resveratrol and quercetin improved cytotoxicity and altered gene expression in oral cancer cells (196). The above combination of drugs was found to modify the epigenetic markers by downregulating histone deacetylases such as HDAC1, HDAC3, and HDAC8.

The cetuximab is an approved targeted therapeutic against HNC. It is accredited for first-line use with platinum-based chemotherapy: the chemotherapy plus cetuximab appreciably extended basic survival compared to chemotherapy alone. Significant improvements were visible within the progression-free survival and goal response prices. In a retrospective analysis of the trial, the enhancements observed with cetuximab were regarded on par with tumors being HPV positive against tumor being HPV negative (197). The single-cell analysis following treatment with cetuximab to different squamous carcinoma cell lines identified a heterogeneous cell population (198). Resistance to cetuximab appeared to be cell-type-specific which was attributed to altered gene expression of TFAP2A and EMT. However, resistance to cetuximab was found to be very common in HNC. Various evading mechanisms such as mutations in receptors may act in accordance to restore original oncogene dependence. A gain in copy number of target genes is another factor that counteracts the action of inhibitors. It has been found that altered copy number by amplification of chromosome 7p11.2 which encompasses EGFR gene, causes various cases of changes in EGFR activation in HNC (199). Gillison et al. (200) observed that with HPV-positive oropharyngeal squamous cell carcinoma (OPSCC) patients, cetuximab and radiotherapy demonstrated an inferior overall survival when compared with radiotherapy plus cisplatin.

The cisplatin plus fluorouracil treatments were given to 657 patients in the SPECTRUM phase III trial, with or without panitumumab, another monoclonal antibody targeting the EGFR receptor (201). A statistically non-significant trend indicated increased overall survival with the addition of panitumumab. As with the EXTREME trial using cetuximab, there was slightly more toxicity in the panitumumab arm than in the control arm. In the phase II trial (202) however, a comparison was made to identify the efficacy of panitumumab plus radiotherapy with chemoradiotherapy groups in locally advanced HNC patients. In the combined study, the efficacy of panitumumab was found to be inferior to cisplatin. It cannot be considered as its substitute for the treatment of unresected stage III–IVb HNC.

Larotrectinib is another type of targeted therapy that does not target specific cancer types but focuses on specific genetic changes in neurotrophic tropomyosin receptor kinase (NTRK) genes. This uncommon genetic change was found in head and neck cancer. NTRK is observed to be highly expressed in aggressive cancer and is used as a predictive biomarker and drug target (203). This FDA- approved TRK inhibitor is used for tumor-agnostic treatment after pembrolizumab (204).

Gene therapy is another targeted approach used to treat a variety of cancers, including head and neck cancer. The p53, a tumor suppressor gene, is mutated in over 50% of all types of cancers in humans. It plays a critical role in suppressing malignancy. Thus, restoration of functional wild-type p53 gene appears as a promising strategy for cancer treatment. The commonly used p53 stimulants are advexin, gendicine, ONYX-015, and H-101 (205).

Monoclonal antibodies play a major role in the treatment of HNSCC. Monoclonal antibodies, such as antibody-drug conjugate to cytotoxic agents (206), are used to target particular cell surface proteins conferring tumor specificity by identifying selective targets. Clinically useful agents that target cell surface proteins in HNC such as AVID100 for EGFR; BAY1129980 for C4.4a; IMMU-132 target for TROP-2 antigen, and tisotumab vedotin are being developed and investigated (207–210).

The other approved targeted antibodies have been developed against specific targets such as CTLA-4 and programmed cell death protein 1(PD-1) that can stimulate co-stimulatory signals. The later one includes the agonistic mAb against OX-40 such as tavolimab and CD137 (urelumab, utomilumab) (211) or toll-like receptor 8 (TLR-8) agonist (motolimod) that mimics the viral ssRNA, the natural ligand of TLR8 and enhances immune response (212). The motolimod plus cetuximab plus was found to be safe in a phase I trial metastatic HNC patients (https://clinicaltrials.gov/ct2/show/NCT04272333) (213). The mAbs, pembrolizumab, and nivolumab are the approved PD-1 inhibitors. These have shown lasting responses in many cancers and were rapidly expanded for use in HNC treatments (214).

Another mAb that targets the EGFR domain, prevents a change in its conformation required for its activation. A randomized phase III trial with zalutumumab has failed to meet its endpoint of improved overall survival or no disease-specific survival and thus was suspended for further development (215). Other antibodies that targets different pathways/receptors and are under evaluation in head and neck cancer clinical trials. These include DLL/Notch pathway, FGF/FGF-R, HER2, TROP2 protein and VEGF/VEGF-R pathway and are discussed underneath.

Angiogenesis is important process in tumor growth and metastasis. The first-in-class anti-angiogenic mAb directed against ligand is bevacizumab that targets VEGF. Bevacizumab, has shown some evidence of activity combined with platinum-based chemotherapy. However, bevacizumab does not have a role in managing advanced or metastatic HNC outside of a clinical trial setting. Combining bevacizumab with chemotherapy in the first line of treatment of advanced metastatic HNC showed enhanced response rate and increased toxicity. In the E1305 trial, 403 patients without prior systemic therapy for advanced HNC were randomly assigned to platinum-based chemotherapy with or without bevacizumab (216). Thus, cisplatin with either fluorouracil or docetaxel or carboplatin with either fluorouracil or docetaxel were used. The primary endpoint was overall improved survival.

There has been substantial progress in the development of mAbs targeting FGFR pathway. Trastuzumab, a mAb targeting HER2, binds to domain IV of HER2 and blocks the homo-dimerization. In a phase II clinical trial study the effectiveness of trastuzumab on patients with advanced/metastatic salivary gland cancer was conducted, however no result is posted till date (NCT00004163) (217) ( Table 6 )

The transmembrane glycoprotein Trop2 is involved in several cell signaling pathways and is upregulated in a variety of cancers, including HNC. The overexpression of Trop-2 is associated with poor disease-free and overall survival in several solid tumors. IMMU-132 (hRS7-SN38 or Sacituzumab govitecan) is an Antibody Drug Conjugate (ADC) that target Trop-2. It consists of an antibody, hRS7 linked to SN38. SN38 is the active metabolite of irinotecan. The preclinical data demonstrated 136-fold more SN-38 delivery by IMMU-132 to a xenograft mouse model than irinotecan with lower toxicity, including lesser cases of severe diarrhea than irinotecan alone. IMMU-132 is under phase I/II clinical trials for evaluation of the safety and efficacy in patients of HNC (NCT03964727 & NCT01631552) (218–220). Table 7 enlists clinical trials with combination of small molecules along with immunogens.

Small molecules have emerged as an important class of targeting agents that target multiple TK. The well-known molecular targets which have shown promising results are EGFR, EGFR TK and VEGF/VEGFR inhibitors and protein kinases or PI3K.

Gefitinib and erlotinib are the most common EGFR TKIs that are being used in clinical studies (phase II) for treatment of HNSCC. Lapatinib is another TKI that targets ErbB1/ErB2. The phase II study of lapatinib plus chemoradiotherapy in HNSCC has showed beneficial effect in HPV negative tumors (221). The lapatinib plus capecitabine combination has demonstrated best activity in the metastatic/recurrent HNSCC. Afatinib is an irreversible TKI that blocks the signaling originating from ErbB family. It is also used in other cancers with high EGFR mutations. In the stage III & IVa HNSCC, it is evaluated as adjuvant following radiotherapy (222).

Sorafenib, sunitinib and vandetanib are small molecules that targets VEGF (223). Sorafenib also act as radiosensitizer of HNSCC cells (224). Other VEGF inhibitors in clinical trials for treating HNSCC are linifanib, axitinib, pazopanib and nilotinib (225).

Several studies in vitro and in vivo demonstrated that temsirolimus, an mTOR inhibitor, inhibits proliferation of HNC. A study with HNSCC cell lines demonstrated beneficial effect of mTOR inhibitors plus cetuximab in the treatment of tumor with low EGFR expression or those that acquired resistance due to cetuximab/cisplatin (226). However, in other studies, temsirolimus failed to demonstrate significant changes in patients with advanced malignancies due to toxicity and subsequent death of patients. Everolimus, another mTOR inhibitor demonstrated antitumor effect in phase II clinical trial in patients with advanced HNSCC (NCT01111058) (227). The other small molecules that targets BCR-Abl kinase and are under clinical trial for treatment of HNSCC include imatinib, dasatinib, nilotinib, ponatinib inhibitors (193).

Although for the last several years, a large number of small molecules are being scrutinized against HNC ( Table 6 ) with diverse heterocyclic structures, still a preferred specific and effective pharmacophore is yet to be assigned by drug development scientists. Current treatment is still associated with significant toxicities and includes chemotherapy mainly with platinum compounds, radiation, surgery, and a few targeted treatments. The scarcity of highly efficient drugs prompts researchers to identify novel targets for single-agent or for combined therapy.