Introduction

Head and neck cancers (HNC) have one of the highest incidences worldwide, reaching approximately 932,000 cases in 2020 [1]. The disease takes its burden, especially in low- and middle-income countries where patients are often diagnosed with tumors in advanced stages [2, 3]. In 2016, 67% of HNC cases and 82% of HNC-related deaths came from these regions [4]. This group of tumors originates from the mucosal epithelium of the upper aerodigestive tract, where it develops in three primary subsites — oral cavity, pharynx, and larynx. Roughly 90% of HNC are squamous cell carcinomas (HNSCC) [5]. The primary risk factors for developing this malignancy are long-term, excessive tobacco use and/or alcohol consumption, as well as environmental carcinogens and human papillomavirus (HPV) infection [6, 7]. The first ones are predominantly linked with tumor formation in the oral and larynx localization, whereas the latter is primarily associated with the oropharynx [8]. Interestingly, undergoing HPV infection was correlated with more favorable outcomes [7]. There is an urgent need for vaccination programs, especially among young men from developing countries, who are increasingly diagnosed with HNSCC [7]. Patients are proposed with therapeutic approaches consisting of chemo- and radiotherapy, surgical resection, and biological medicaments, including immunotherapeutics [6]. However, for late-diagnosed patients with advanced disease, even aggressive multimodal treatment does not bring long-term positive results. Belated detection, tumor heterogeneity, and metastasis are the main reasons why the mortality rate of HNSCC in the first five years from diagnosis is approximately 50% [6, 9]. Necessity to improve the diagnosis process and find more personalized treatments for cancer patients led scientists to search for potential markers or therapeutic targets among some of the most crucial and universal regulators of cell biology — non-coding RNAs (ncRNAs).

One of the most abundant ncRNA subgroups consists of microRNAs (miRNAs) - small (approximately 22 nucleotides long), endogenous ncRNAs with great power. These molecules are crucial for post-transcriptional regulation of gene expression, which is required for maintaining cell homeostasis and normal organism development. Over 60% of human protein-coding genes have conserved one or more miRNA-binding sites, which, combined with numerous non-conserved sites, indicate possible miRNA-based control of most of those transcripts [10]. Inhibition of target messenger RNA (mRNA) expression is caused by miRNA interaction with its 3’ untranslated region (3’ — UTR) [11]. Dysregulation of this complex molecular machinery is associated with many diseases, including cancers [11, 12]. That is why it is vital to explore the role and possible diagnostic or therapeutic potential of ncRNA.

This review aims to discuss the current knowledge regarding miRNAs in HNSCC and elucidate their possible future usage as molecules in the personalized medicine service.

miRNA biogenesis

The canonical pathway of miRNA biogenesis begins in the nucleus with transcription of miRNA genes, which in the vast majority are located in introns of non-coding or coding transcripts. Sequences residing in the introns of protein-coding genes usually share promoters with the host. On the other hand, intergenic sequences are transcribed separately from the host although controlled by their promoters [13, 14]. Each locus encodes two mature molecules — one on the 5’ and the second on the 3’ strand. miRNA genes are transcribed by polymerase II (Pol II), which results in the formation of long primary transcripts (pri-miRNA) containing both future miRNAs enclosed in the local stem-loop structure [11]. This process is regulated by transcription factors associated with Pol II and various epigenetic regulators [15]. The pri-miRNA maturation process is initiated by the nuclear RNase III — Drosha and its cofactor DGCR8, which together form the Microprocessor complex that can cleave molecules in a precisely defined position [16]. Complex recognizes pri-miRNA structure consisting of terminal loop, approximately 33-35 bp long stem, as well as single-stranded tails and crops RNA specifically 11 bp from the basal junction (junction of the single-stranded tails turning into stem) and 22 bp from the apical junction (junction below the terminal loop), creating a hairpin-structured RNA called — pre-miRNA [17, 18]. Both described junctions are crucial for determining the split site [11, 17, 18]; however, the mechanism of interaction between the Microprocessor and pri-miRNA is still unknown.

Subsequently, protein exportin 5 (EXP5) and GTP-binding nuclear protein Ran-GTP form a “baseball mitt”-like transport complex to export pre-miRNA to the cytoplasm for the final steps of the maturation process [11, 19, 20]. Translocation through the nuclear pore requires GTP hydrolysis, which leads to complex disassembly [20]. Released into the cytoplasm pre-miRNA is then cleaved close to the terminal loop by RNase III-type endonuclease, called Dicer, resulting in the creation of a short RNA duplex [21]. Said enzyme consists of an N-terminal helicase domain, the PIWI-AGO-ZWILLE (PAZ) domain, two double-stranded RNA (dsRNA) binding domains (DUF283, dsRBD), and a catalytic center formed by two C-terminal RNase III domains [22–25]. The PAZ domain has two spatial pockets, which can bind simultaneously a 5’ end and a two-nucleotide-long overhang at the 3’ end of the pre-miRNA [24, 25]. Nevertheless, the 5’ end binds to the enzyme only when it is thermodynamically unstable due to the lack of strong guanine and cytidine base pairing [24]. The region between the PAZ and RNase III domains may act as a “molecular ruler” [22, 23], which ensures that Dicer cleaves processed RNA 21-25 nucleotides from the 3’ end and 22 nucleotides from the 5’ end [22, 24]. This process generates small dsRNA, which is then loaded to the Argonaute family protein (AGO) in an ATP-dependent process and becomes the RNA-induced silencing complex (RISC) [11]. The process of miRNA biogenesis is presented on Figure 1.

Figure 1. Canonical and non-canonical pathways of microRNA (miRNA) biogenesis. Short RNA transcripts named primary-miRNA transcripts (pri-miRNAs), mitrons, and short hairpin RNAs (shRNAs) shared common and distinct biogenesis steps. Pri-miRNAs and shRNAs are cleaved by the Microprocessor complex in contrast to mitrons, which are trimmed by spliceosomes. All cropped RNAs are stabilized by exportin 5 (EXP5), which, together with Ran-GTP, transfers RNA transcripts from the nucleus to the cytoplasm. Next, pre-miRNAs are cleaved by the Dicer enzyme and, as a mature miRNA, loaded into the Argonaute family protein (AGO) complex in canonical biogenesis. In the case of non-canonical pathways, the Dicer-dependent step is omitted, and pre-miRNAs are loaded into AGO. Finally, in both cases, a functional complex is created with the ability to interact with different types of RNA molecules, creating an RNA interference (RNAi) phenomenon

In humans, there are four AGO proteins (AGO1–4) that can interact with duplex RNA and induce target mRNA degradation causing its translational repression, however, only AGO2 has the ability to cleave perfectly matched transcripts [26]. These proteins contain 4 functional domains — N-terminal domain, PAZ domain, MID domain, and PIWI domain, which are required to bind dsRNA and determine which strand will be further processed. The transcript with high thermodynamic stability and lack of uracil at the beginning of the 5’ end isn’t preferentially attached to the AGO, resulting in its unwinding from the remaining strand (the “guide strand”) and degradation [11, 27]. The above leads to a strong bias toward the “guide strand” generating an imbalance in the mature miRNAs expression levels as well as biological activity. These molecules tend to be more active and abundant than the ones encoded on the “passenger strand”. The selection of RNA strands may vary in different tissues due to a phenomenon called “arm switching”, which can be generated by alternative Drosha processing [28]. Additionally, these molecular species could be co-expressed in various tissues and display different modulatory roles in tumorigenic processes independently and/or cooperatively [29]. miRNA enclosed in the RISC complex is highly stable because AGO protects both its’ ends [30]. That is why small RNAs need to be unloaded to enable exonucleases to access their termini. The process of mature miRNA disassembly from AGO isn’t fully described yet.

Interestingly, approximately 1% of conserved miRNAs are produced in processes deviating from the canonical pathway. For example, molecules residing in short introns called ‘mirtrons’ after co-expression with their host genes don’t undergo Drosha/DGCR8 cleaving but are trimmed by spliceosome instead, and then exported directly into the cytoplasm (Fig. 1)[13]. Experiments studying cells deficient in DGCR8, Drosha, or Dicer showed that some of the small RNAs could be generated in Microprocessor-independent or Dicer-independent (e.g. shRNA, Fig. 1) ways [11, 31]. This suggests the need for evolutionary flexibility in the biogenesis process of these short RNA molecules.

Function and interactions with molecules

miRNAs are known for their wide scope of functions and the great variety of molecules they interact with [10–12]. In general, these small RNAs bind with the specific sequence at the 3’ UTR of mRNAs leading to its translational repression, causing deadenylation, decapping, and, consequently, degradation [32]. However, miRNAs could also bind to the 5’UTR and coding region of target mRNA, which results in silencing of the gene expression. They can also interact with the promoter, which on the other hand, may induce the transcription [33].

Gene silencing can be promoted by the minimal miRNA-induced silencing complex (miRISC), which is the AGO with the loaded “guide strand” [34]. Target mRNA has miRNA response elements (MREs), whose sequence is to a different extent complementary to a particular short RNA, conditioning the choice of one of the two possible paths. The first one is activated when miRNA:MRE complementarity is 100% and leads to target cleavage by AGO2 endonuclease activity [35], while the second pathway occurs when miRNA:MRE binding contains mismatches and uses the ability that all AGO proteins have – to mediate RNA interference, causing translational inhibition and target degradation [33, 35, 36]. miRNA can affect transcriptional and post-transcriptional gene regulation by shuttling from cytoplasm to nucleus through importin-8 or exportin-1 in the form of miRISC via interaction with TNRC6A protein [37].

Principally, functional miRNA:MRE interaction happens via the miRNA “seed region” located at the 5’ end and consisting of nucleotides 2–8 [33, 36]. One miRNA can interact with numerous mRNAs, and one mRNA can have many MREs incorporated into its sequence [38]. On the cellular level, the relation between all molecules of a particular small RNA and the total number of their available response elements is described by the term “MRE load”. This relation can be manipulated by elevating the level of a specific gene or a competing endogenous RNA (ceRNA), e.g. long non-coding RNA (lncRNA) or circular RNA (circRNA), causing sequestration of miRNA, as well as whole miRISC, from their target transcript leading to its derepression [39]. Additionally, MREs have different affinity degrees – ones with the higher affinity will bind miRNAs for a longer time and lead to mRNAs’ increased sensitivity for post-transcriptional repression. The above process might provide a stable gene expression by controlling the high mRNA ratio and reducing the expression noise [38]. Interestingly, the study by Tian et al. determined that miRNA-mediated regulation is more strict for tumor suppressor genes than oncogenic ones [40].

In recent years many reports emphasized the relevance of lncRNA and miRNA, as well as their interactions in cancer biology, treatment, and diagnosis [41-48]. Apart from acting as a “molecular sponge” sequestering miRNA, lncRNAs can compete with them by binding to target mRNAs and preventing transcriptional repression. Additionally, some lncRNAs can be processed into miRNAs, which is another argument corroborating the fascinating joint work of these ncRNAs to control gene expression through numerous complex post-transcriptional mechanisms [41–43, 48].

miRNAs potential — biomarkers or therapeutic targets?

High mortality rates and rising incidence of HNSCC dictate the necessity for developing therapy tailored to each patient. The first step to personalized medicine is to discover therapeutic targets, affecting processes crucial for carcinogenesis or further tumor progression, and diagnostic biomarkers, which could determine not only the occurrence of the disease but also assess individuals’ response to the applied treatment. miRNAs can be used as potential biomarkers and therapeutic targets in HNSCC, Figure 2.

Figure 2. miRNAs as potential biomarkers and therapeutic targets in head and neck squamous cell carcinomas (HNSCC)

Promising biomarkers

A growing number of studies describe miRNAs’ potential in diagnostics, emphasizing their specificity, stability of extracellular RNAs in various biological fluids, their relatively concise isolation process from acquired material, and simple analysis of obtained results [46, 49–53]. Importantly, these molecules can be successfully acquired via “liquid biopsy”, which is a non-invasive alternative to traditional tumor sampling [51]. Umu et al. determined that serum contains significant amounts of ncRNAs, and approximately 45.7% is miRNA [54]. Biopsy often doesn’t fully reflect tumor heterogeneity [55] or can’t be obtained at all, i.e., when neoadjuvant therapy reduces the volume of the neoplastic lesion to an undetectable size [56]. Developing a specific panel of diagnostic, prognostic, or predictive biomarkers would be an innovative approach leading to improvement of the HNSCC detection rate, patients’ life quality, and expectancy.

Head and neck squamous cell carcinoma (HNSCC)

Fadhil et al. described two miRNAs with significant specificity and sensitivity to distinguish between patients and healthy individuals. let-7a-5p and miR-3928-5p were significantly downregulated in cancer patients’ saliva compared to controls, and their expression levels could be associated with the clinical T stage. Interestingly, low let-7a-5p levels were linked with the more advanced stage as well as the presence of lymph node metastasis, suggesting its prognostic properties [57]. The panel of salivary-based biomarkers will be a useful non-invasive tool for HNSCC prophylaxis and early detection. The study by Ganci et al. focused on determining whether tumor, peritumoral and healthy tissue samples collected from HNSCC patients differ in gene expression alterations. Despite the lack of histopathological evidence, peritumoral mucosa often presents an abnormal molecular landscape that could predict recurrence development. Out of 35 deregulated miRNAs found in the peritumoral tissue, authors identified a signature consisting of miR-21-3p, miR-21-5p, miR-429, and miR-96-5p, which upregulation was linked with significantly shorter local recurrence-free survival (RFS) [58]. Furthermore, Vahabi et al. examined the last of mentioned miRNAs and discovered that it additionally correlates with the most frequently detectable genetic alteration – mutation in TP53, as well as increases HNSCC cell lines’ resistance to chemotherapeutics and irradiation [59]. Hudcova et al. studied the expression profile of miR-29c-3p, miR-200b-5p, and miR-375-3p in tumor and paired healthy margin tissue samples. The first molecule was downregulated in cancer, and its low levels were associated with worse overall survival (OS) and less favorable outcome. Decreased expression of miR-200b-5p was linked with advanced tumor grade, and its reduction in adjacent healthy tissue was observed in patients with nodal metastasis. These molecules seem to have prognostic properties worth further exploring. Finally, miR-375-3p was significantly upregulated in control samples compared to cancer tissue, and ROC analysis indicated its sensitivity and specificity as a biomarker reaching 87.5% and 65%, respectively, which implies the great ability to distinguish tumor samples from normal tissue [60].

Esophageal squamous cell carcinoma (ESCC)

Qi and Fan described miR-18a-5p as an interesting potential diagnostic molecule in the ESCC. Its high expression was linked with more advanced disease and worse OS. The ROC analysis indicated that miRNAs’ sensitivity and specificity as peripheral blood-based biomarkers was 80.7% and 71.9%, respectively. Moreover, the authors have emphasized the promising prognostic and therapeutic potential of the interplay between miR-18a-5p and ATM in ESCC [61]. A different molecule that might be helpful in ESCC diagnosis is miR-20b-5p. The authors demonstrated that serum levels of said miRNA could distinguish controls from cancer patients with 87.1% specificity and 76.3% sensitivity. In addition, its expression levels were positively associated with lymph node metastasis (LNM), advanced stages of ESCC, and poor survival rates [62]. These findings were corroborated by Xue et al. study, which indicated that miR-20b-5p expression level might predict LNM in patients with T1 stage of ESCC better than lymphovascular invasion, invasion depth and tumor differentiation grade, indicating its tremendous prediction potential [63].

Laryngeal squamous cell carcinoma (LSCC)

The study by Lucas Grzelczyk et al. revealed three miRNAs with exceptional diagnostic potential. Based on let-7a, miR-31, and miR-33 serum levels, patients could be distinguished from healthy individuals with nearly 100% accuracy [64]. A biomarker panel consisting of these molecules could become a helpful tool in LSCC prophylaxis. Whereas Hu et al. proposed the molecular signature of two other molecules — miR-21 and miR-375, which may serve as detection biomarkers. Authors indicated that, except for the diagnostic properties of mentioned miRNAs interplay, miR-21 alone could be an independent patients’ OS predictor [65]. Wang et al. compared miR-155 expression levels in serum and tissue samples from LSCC patients. miRNA derived from both sample types had significant diagnostic potential, especially in the early stages of the disease. However, miR-155 detected in tumor specimens manifested higher accuracy compared to serum, not only in the whole patient group but also in individuals at the early stages of cancer development [66]. The above serves as a promising basis for further investigation, which could lead to the creation of a routine screening panel for LSCC patients.

Nasopharyngeal carcinoma (NPC)

Ye et al. analyzed exosomes present in serum samples from NPC patients and controls. Obtained results indicated that vesicles acquired from the first group are significantly more enriched in miR-24-3p. Additionally, the high expression level of said miRNA was correlated with shorter disease-free survival (DFS), implying its potential diagnostic value [67]. Wen et al. identified two miRNA-based molecular signatures crucial for NPC diagnosis. The first combination consisted of 8 miRNAs: miR-4465-3p, miR-4433-5p, miR-3935, miR-188-5p, miR-513b, miR-3196, miR-1908, and miR-4284, and enabled to distinguish healthy individuals from NPC patients with 88.9% specificity and 86.1% sensitivity. The second one was the combination of 16 miRNAs: miR-296-5p, miR-361-3p, miE-4665-3p, miR-4439, miR-155-5p, miR-5091, miR-4706, miR-4436b-5p, miR-4284, miR-1224-3p, miR-4740, miR-425-5p, miR-1973, miR-513b, miR-1908, and miR-1280, and allowed to differentiate NPC patients from patients with other head and neck cancers and controls with 72.2% specificity and 94.4% sensitivity. These 2 signatures could become powerful biomarker panels in the future [68]. Recently, Wei et al. identified molecules with substantial diagnostic potential, miR-34a and Ki67 protein. This miRNA was proven to be downregulated in the cancer tissue compared to the healthy samples, and its decreased levels were associated with more advanced disease (stage III and IV in TNM staging) and bone metastasis. The diagnostic value of both molecules was examined with the ROC analysis, which indicated that joint detection exceeded the specificity and sensitivity of miR-34a and Ki67 alone, reaching 82.9% and 73.3%, respectively. Furthermore, survival analysis determined that patients with low expression of studied miRNA and high levels of Ki67 had significantly better OS than the group with the opposite expression pattern. Authors underlined not only diagnostic but also prognostic properties of the above signature in NPC [69].

Oral squamous cell carcinoma (OSCC)

Scholtz et al. described a promising molecular signature consisting of three miRNAs, which could distinguish healthy individuals from OSCC patients with a high accuracy. The combination of miR-31-5p, miR-345, and miR-424-3p detected from saliva samples could become a very precise non-invasive diagnostic tool [70]. Interestingly, Lee et al. determined that miR-769-3p could be strongly associated with specific subtypes of OSCC. Its high expression levels were linked with better treatment response, more favorable OS, and less invasive phenotype of the tumor [71]. Whereas low levels of miR-126 were associated with poor outcomes and shorter DFS in cancer patients [72]. Analysis of miR-99a presence in serum samples demonstrated that individuals with OSCC had significantly lower expression levels than the control group and could be distinguished with specificity and sensitivity reaching 83.6% and 80.2%, respectively. Furthermore, decreased miRNA amounts were associated with shorter OS and RFS as well as more advanced TNM stage and histological grade. Analysis of said miRNAs level in blood samples obtained from patients before and after surgery indicated post-treatment increase in the studied molecule level. The authors emphasized that miR-99a could serve as an independent prognostic biomarker of OSCC [73].

Tongue squamous cell carcinoma (TSCC)

Chen et al. proposed a diagnostic signature consisting of 3 miRNAs: miR-21, miR-139-5p, and miR-486-3p, which had exceptional potential in TSCC detection. These molecules differentiated tumors from healthy tissue with specificity reaching 86.7% and 100% sensitivity. However, it is worth mentioning that these results require validation on a larger cohort [74]. Interestingly, an earlier study by Duz et al. also underlined the diagnostic properties of miR-139-5p. This miRNA was downregulated in saliva from cancer patients compared to controls and post-operative TSCC individuals. Detection of miR-139-5p allowed it to differentiate these groups with significant accuracy, which suggests its possible role as a reliable non-invasive TSCC biomarker [75]. Another molecule with high diagnostic accuracy is miR-196a-5p, whose upregulation was identified in TSCC tissue samples. Moreover, its high levels were associated with shorter lymph node metastasis-free survival and early stages of the disease, confirming its importance for cancer screening and prophylaxis [76].

Although much work remains to be done in the field of miRNA-based diagnostic tools, there are many tremendous candidates which might become useful biomarker panels in the future. It is worth mentioning that a few such assays are currently in use, e.g. for distinguishing molecular subtypes of breast cancer [77], forms of lung malignancies [78], or even identifying primary tissue of origin for different tumor types [79]. Among the molecules described above, let-7a, miR-21, miR-29c-3p, miR-31, miR-96-5p, miR-375, miR-429, and miR-3928-5p, can be considered the most universal for detecting the HNSCC tumor group. However, it should be emphasized that most of the miRNAs proposed in this review are currently being investigated for association and possible diagnostic or therapeutic use in more than one type of cancer. More analysis needs to be done to precisely identify molecules specific to a given malignancy, or on the other hand, propose a panel that detects cancer in general. Potential diagnostic biomarkers and their role in the HNSCC are summarized in Table 1.

Beneficial therapeutic targets

miRNAs affect numerous different molecules modulating a variety of pathways and cellular processes, which can lead to astounding therapeutic results as well as severe off-target consequences. Carefully designed small RNA-based drugs have the potential to target not only the multitude of proteins that aren’t influenced by any common medicaments but also many compensatory mechanisms participating in the development of therapy resistance [80]. However, the above is inevitably linked with the so-called “too many targets for miRNA effect” (TMTME), which leads to a wide range of difficult-to-predict results. Among others, that is why miRNA-based therapeutics don’t enter phase III of clinical trials [81]. On the other hand, bioinformatic tools predicting potential targets and interactions become more and more accurate, giving us hope for designing a new drug which would revolutionize modern medicine.

HNSCC

One such future therapeutic molecule might be miR-31 described by Lu et al. This molecule is upregulated in cancer compared to normal samples, and it is directly reducing the expression level of ARID1A, a known inhibitor of Nanog/OCT4/Sox2 stemness factors and epithelial cell adhesion molecules (EpCAM) activity. miR-31/ARID1A interplay promotes tumorigenesis and could be correlated with a worse prognosis. The authors suggested that therapeutic induction of ARID1A expression is a valuable approach for combating HNSCC [82]. Furthermore, the aforementioned miR-96-5p, a promising predictor of local recurrence, was considered a valuable asset in novel therapeutic approach design. Its high levels induced cell migration, strengthened the resistance to irradiation and cisplatin-based chemotherapy, as well as negatively regulated its direct target PTEN, the primary inhibitor of the PI3K/AKT/mTOR pathway. Repression of miR-96-5p expression or induced upregulation of PTEN might be a beneficial new strategy for individuals with a high risk of local HNSCC recurrence [59]. Chen et al. described different miRNA with oncogenic properties of miR-18a-5p. It promoted a malignant phenotype in cancer cell lines via SORBS2 expression inhibition, resulting in the induction of cell migration and proliferation as well as suppression of apoptosis. Forced increase in SORBS2 expression hindered miR-18a-5p negative impact and implied that therapeutic modulation of the described axis could provide interesting new directions in the design of personal medicine [83]. These reports and the aforementioned diagnostic properties in ESCC make miR-18a a fascinating aspect of oncology research [84].

On the other hand, Hauser et al. found that miR-128 might be a promising tumor suppressor. Its direct negative regulation of PAIP2, BAG-2, H3F3B, BMI-1, and BAX mRNA level was associated with the proliferation inhibition and promotion of apoptosis. Authors implied that miR-128 has a significant clinical value and its artificial induction might be beneficial for the patients [85]. Nohata et al. showed that miR-1 modulated the expression of TAGLN2, a protein with oncogenic properties described in different cancers. Induction of miRNA caused a decrease in tumor growth and invasiveness, proving the importance of further studies regarding its possible therapeutic potential [86].

ESCC

miR-20b-5p that was mentioned before is also a potential therapeutic target. Yu et al. showed that its high expression promoted growth, colony formation, and invasion process in cancer cell lines through negative regulation of RB1 and TP53INP1. Nude mice after injection with cells overexpressing miR-20b-5p developed a stronger metastasis burden than the control group, emphasizing miRNAs’ oncogenic character. A therapeutic approach focusing on reducing miR-20b-5p expression level would be a valuable addition to protocols of personalized oncology treatment [62]. Nourmohammadi et al. described an interesting strategy for diminishing the negative impact of miR-200c, a known tumorigenic miRNA. Reduction of miRNAs’ expression level through repressing EZH2 led to a decrease in EMT-related proteins, such as N-cadherin, Zeb2, or Vimentin. This strategy might substantially improve patients’ survival rates by attenuating ESCC malignant phenotype [87]. Another miRNA that could significantly repress the process of metastasis is miR-27b-3p. It exerts tumor suppressor function by targeting Nrf2 and inhibiting N-cadherin, Vimentin, and Claudin-1 expression levels. Upregulation of miR-27b-3p could be beneficial for patients with advanced, metastatic stages of tumor development [88].

LSCC

Niu et al. described miR-154 as one of the potential therapeutic targets in LSCC. Its low expression levels were associated with an unfavorable prognosis and manifestation of a malignant phenotype. Induced overexpression caused downregulation of GALNT7, resulting in decreased proliferation, clonogenicity, and migration of cancer cell lines, as well as promotion of cell cycle arrest. Strategy for miR-154/GALNT7 regulation could provide beneficial cancer treatment [89]. The study by Li et al. showed that miR-744-3p was significantly upregulated in LSCC patients compared to controls and negatively correlated with PTEN and PDCD4 levels. High expression of of miRNA was linked with cervical LNM and increased invasion, which resulted from MMP-9 overexpression via activation of AKT and NF-κB. Reducing miR-744-3p could attenuate the LSCC malignancy [90]. A recent article by Tu et al. indicated that miR-129-5p directly regulates OTX1 oncogene. Overexpression of said miRNA resulted in OTX1 reduction and, consequently, a decrease in migration and invasion ability of cancer cell lines. A therapeutic approach modulating the described axis could substantially reduce the metastatic character of LSCC [91].

NPC

Ou et al. determined that miR-21 promoted proliferation and suppressed apoptosis in NPC cell lines. It exerts an oncogenic function modulating AKT phosphorylation through regulation of PTEN expression level. Authors suggest the significant therapeutic potential of studied miRNA [92]. Xu et al. demonstrated that miR-375 negatively regulates PDK1 expression, inhibiting tumor progression. Enhancement of PDK1 expression diminishes miRNas’ antitumor effect in NPC cell lines via PI3K/AKT axis, promoting proliferation and migration, and inhibiting apoptosis. Modulating miR-375/PDK1/PI3K/AKT axis might be a promising treatment strategy for NPC patients [93]. A similar interplay between miR-375 and USP1 was described by the same authors a year later. However, in the case of miR-375/USP1/PI3K/AKT axis, the oncogenic effect of USP1 overexpression could be reversed by the administration of selective PI3K inhibitor (S2739) [94]. The above indicates the possibility of restoring the positive miR-375 effect in NPC through multiple molecular strategies. A different miRNA with substantial therapeutic potential is miR-24a-3p. It was significantly enriched in serum exosomes of NPC patients compared to controls, leading to repression of FGF11, which affected the proliferation and differentiation of various T cells. This mechanism supporting cancer immune evasion is still unclear, however, obtained results lay a valuable foundation for designing an antitumor exosome-based therapy in the future [67].

OSCC

Henson et al. demonstrated that simultaneous overexpression of miR-100 and miR-125b affects the growth and development of OSCC cell lines. Moreover, this modification caused alteration in expression levels of numerous direct and non-direct target genes, which were associated with crucial cellular processes. An artificial increase in miR-100 expression level might diminish OSCC radioresistance, emphasizing tremendous therapeutic potential [95]. Sasahira et al. determined that miR-126 is a promising biomarker as well as an interesting target. The significant association of its low expression level with tumor progression, VEGF-A-related angio- and lymphangiogenesis, and unfavorable prognosis of OSCC patients suggest its potential, especially in advanced cases. Further research regarding miR-126 can inspire new treatment approaches [72]. Recently Zhao et al. presented the miR-617/SERPINE axis, which artificial regulation might be beneficial for patients. Mentioned miRNA was downregulated in OSCC, and its low expression level was linked with advanced stages of tumor development. Its induction led to a reduction of SERPINE1 expression and diminished its oncogenic effect on the proliferation, viability, and apoptosis of cancer cell lines [96].

TSCC

Liu et al. indicated that miR-222 is strongly associated with the migration and invasion ability of TSCC. Identified mechanism of this carcinogenic effect primarily depended on MMP1 expression regulation which happened directly through miRNA targeting MMP1 mRNA, or indirectly via modulating SOD2 levels [97]. Moreover, miR-222 negatively modulated ABCG2 expression, which led to enhanced sensitivity to cisplatin [98]. Therapeutic induction of the miR-222/ABCG2 axis might be profitable for patients burdened with metastatic and chemoresistant cases of TSCC. A different miRNA with substantial therapeutic potential is miR-27b whose direct target for ITGA5 mRNA was positively associated with poor prognosis, EMT, and advanced stages of the disease. Stimulated high expression of miRNA resulted in suppression of malignant phenotype and could provide another valuable strategy for combating metastatic TSCC [99].

There is a long way from finding the appropriate target to design the suitable drug form, dose, and route of administration. Then comes the last, most crucial step — clinical trials — a stage where only a few molecules are currently at and where many potentially promising miRNAs have failed so far [100]. The list of miRNA-based therapeutic targets in the HNSCC is presented in Table 2.

Conclusions

The above numerous examples underline the potential of miRNAs both in the field of diagnostics and oncological therapy. Unfortunately, along with the wide range of functions comes the challenge of designing a drug modulating the desired pathway without disastrous off-target effects. Lack of standardization, burdensome drug delivery design, and difficult-to-predict results of TMTME are primary reasons why currently, miRNA-based therapy attempts are succeeding mostly in the cell line models and then stop at the early stages of clinical trials [81,100]. Although the above data favors miRNAs utility as biomarkers, we are convinced that finding the suitable molecule, whether with diagnostic or therapeutic potential, will be a significant step towards the oncology treatment of tomorrow - early administered personalized therapy with minimal side effects that will significantly improve the quality and extend the lives of HNSCC patients.

Conflict of interest

Concerning the publication of an article titled: “miRNA in head and neck squamous cell carcinomas: promising but still distant future of personalized oncology” written by J.K.-M., K.G., T.K., K.L., and I.M., the authors declare:

Funding

Article was supported by National Science Centre — PRELUDIUM 20 grant No.: 2021/41N/NZ5/02966 (J.K.-M.s’ contribution) and by the Greater Poland Cancer Centre — internal grant No.: 5/12/2022/PGN/WCO/011 (TKs’ and KLs’ contribution).

Acknowledgment

J.K.-M. received a Ph.D. program scholarship at the time of writing this manuscript from Adam Mickiewicz University in Poznan, and KG received a scholarship at the time of writing this manuscript from the European Union POWER Ph.D. program. We would like to thank all the institutions supporting our scientific work.

References

Table 1. Potential diagnostic biomarkers in the group of head and neck squamous cell carcinoma (HNSCC)