Head and neck squamous cell carcinoma (HNSCC) describes tumors that originate from the epithelium of the oral cavity, pharynx, and larynx. They represent the sixth most prevalent cancer type globally with 932,000 cases and a mortality of 467,000 deaths every year as of 2020 [1]. Risk factors primarily include smoking and excessive alcohol consumption, as well as aging, diet, oral hygiene, and family history. Additionally, infection with human papilloma virus (HPV), a circular DNA virus that targets the epithelial cells of the mucosa, plays an important role in the development of HNSCC. In general, treatment options include surgery, radio-, chemo- and immunotherapy [2]. Besides this, novel therapeutic approaches based on molecular tumor markers were developed like the targeting of epidermal growth factor receptor (EGFR), a highly upregulated receptor in HNSCC associated with poor prognosis. Additionally, as an immunotherapeutic approach, the checkpoint inhibitor pembrolizumab has been approved by the Food and Drug Administration (FDA) in 2019 as a first-line treatment for metastatic or recurrent HNSCC [3]. Despite recent advances in HNSCC therapeutic options, the mortality rate is still around 50%, which indicates the need for more effective therapies and diagnostic as well as disease monitoring tools. For this, reliable biomarkers should be defined that can predict the onset and progression of HNSCC or even the therapeutic outcome. Moreover, based on biomarker expression personalized therapeutic options could be developed and applied. However, presumably because of the high heterogeneity in HNSCC, in contrast to other tumor entities, a robust biomarker panel for HNSCC has not been defined so far, as applicable in clinical settings [4].

There are different ways to obtain information about the genetic changes of the tumor. Although regarded as the gold standard for tumor diagnosis, tissue biopsy suffers from several limitations. Firstly, tissue biopsy is invasive in nature and can potentially lead to complications, causing limitations in the reproducibility and repeatability of the diagnosis. On top of that, tumor heterogeneity impedes the ability to obtain spatially and temporally accurate information about the tumors [5]. For this reason, recent efforts have been made to find new methods for cancer diagnosis and screening options.

Liquid biopsy describes the analysis of biomarkers released by tumors into body fluids, such as blood, urine, or saliva. These biomarkers could be cells, nucleic acids, or proteins. Liquid biopsy offers a minimally invasive procedure, which is more easily applicable and could be repeated more frequently, offering a more real-time and dynamic snapshot of the tumor, which combats the spatial and temporal limitation of tissue biopsy [6]. This enables clinicians not only to utilize liquid biopsy as an effective diagnostic tool but also as an instrument for therapy and tumor progression monitoring. The advantage of analyzing genetic markers also offers a potential approach for personalized treatment [7].

Information about potentially existing tumors could be obtained from several sources through liquid biopsy, like circulating tumor cells (CTCs), circulating cell-free tumor DNA (ctDNA), and exosomes [8–10]. Their analyses were suggested as sensitive and specific for cancer prediction, diagnosis, prognosis, and monitoring over time [11–13] including HNSCC [14–18]. Thus, a recent study on 21 patients with locally advanced oropharyngeal carcinoma treated with neoadjuvant chemotherapy suggested that the variation of CTC numbers might reflect the effectiveness of treatment [19]. Also, the potential of ctDNA as a biomarker for detecting minimal residual disease and recurrence in HNSCC was demonstrated recently [20]. Additionally, the analyses of circulating exosomes emerged as a potential tool for HNSCC diagnosis [21] and several studies correlated proteins identified in patients’ exosomes with the activity of tumor disease and therapy responses [21, 22].

However, with regard to head and neck tumors, all these types of liquid biopsies are far from routine clinical use because, despite many advantages over standard diagnostics, they have some limitations. Over the last years, another form of liquid biopsy has received considerable attention—the analyses of so-called “tumor-educated platelets (TEP)”. Platelets interact with tumor cells leading to platelet degranulation, known as tumor cell-induced platelet aggregation. The release of their storage products supports tumor growth, angiogenesis, and tumor spread [23].

Cancer-platelet interaction is a two-way communication path. Not only does platelet activation promote tumor progression, but also tumor cells affect platelets, specifically, their RNA profile. The idea of platelets having a distinct RNA profile in cancer patients compared with healthy controls was first introduced in 2010 [24]. In this study, Calverley et al. [24] demonstrated that platelets of metastatic lung cancer patients exhibit 200 differentially expressed genes compared to healthy controls. In 2011, Nilsson et al. [25] described the transfer of tumor cell RNA into platelets, leading to the formation of so-called TEP. Based on platelet RNA profiles of 283 patients, Best et al. [26] developed an algorithm that distinguished healthy and tumor patients (TPs) with an accuracy of 96%. In this study, six tumor entities (non-small cell lung cancer, colorectal carcinoma, breast carcinoma, hepatobiliary carcinoma, glioblastoma, and pancreatic carcinoma) have been differentiated.

In recent years, great efforts have been made to define TEP-derived biomarkers for the diagnosis of a wide variety of tumors for different cancer entities. Thus, applying particle-swarm optimization-enhanced algorithms, Best et al. [27] defined a biomarker panel allowing the detection of early- and late-stage non-small cell lung cancer with an accuracy of 81% and 88%, respectively. Sheng et al. [28] identified a biomarker panel of 48 genes for the diagnostic of the same tumor entity with an accuracy of 89%, whereas Xing et al. [29] identified ITGA2B as a diagnostic and prognostic TEP-based marker for non-small cell lung cancer, just to mention some of the TEP-RNA-sequencing (RNA-seq) work on this cancer entity. Additionally, TEP-derived RNA panels, determined again by swarm intelligence, were applied to identify glioblastoma patients with an accuracy of 95% [30]. TEP-RNA-seq-derived biomarkers were also identified for sarcoma [31], colorectal [32], and liver cancer [33–36]. Moreover, various clinical studies have confirmed the analysis of TEP as a robust form of liquid biopsy and other clinical trials are currently recruiting. The study described here was initiated because of the advantages of the analysis of TEP compared to other forms of liquid biopsy and the lack of data on TEP in HNSCC.

The present work represents a pilot study with the aim to investigate whether the analysis of platelet RNA is a diagnostic option also for HNSCC. Since differentially existing messenger RNA (mRNA) in platelets of HNSCC patients were observed, the present study provides the basis for designing further prospective studies considering cohort size, robust inclusion criteria, in- or exclusion of certain tumor sites as well as HPV status.

The current knowledge of thrombocyte biology, especially in the context of cancer, still leaves many unanswered questions. In addition, the mechanism responsible for changes in the RNA profile in platelets of TPs is not completely understood. However, the quite big array of publications over recent years describing differences in mRNA expression between HDs and patients of different tumor entities is not to be neglected. Also, the comparative analysis of the RNA profiles of platelets of HDs vs. HNSCC TPs revealed 426 differentially existing RNA. Among them, 254 “tumor-related RNA”, coding for epithelial markers, and, importantly, for tumor-associated genes were identified. Particularly, cytokeratin 5 and 14 are characteristic elements of the squamous epithelium [67], the origin of HNSCC. Additionally, the presence of RNA, which encodes oncogenes or tumor suppressor genes that are characteristically expressed in HNSCC, like SOX2 [84], CRYAB [90], and PTHLH [91], further suggests transmission of tumor RNA into platelets [25, 81]. Therefore, the existence of epithelial and tumor marker RNA in TPs’ platelets may cause by the previous observations that tumor cells transfer their RNA into platelets [25].

Furthermore, 41 non-coding RNA forms were significantly differentially enriched in platelets derived from TPs vs. HD platelets, which could be explained by the fact that some of these forms are also poly-adenylated and therefore addressed by this approach [103]. Also, in the study of Best et al. [26], 20 differentially expressed non-coding RNAs were detected. The significantly altered existence of long non-coding RNA in platelets of TPs was also confirmed by Luo et al. [104]. In the present work, as well as in Best et al. [26], among others, the long intergenic non-protein coding RNA 152 (LINC00152) was abundantly found in TPs’ platelets, a long non-coding RNA expressed in many types of cancer, including oral squamous cell carcinoma [105, 106], acting as a potent oncogene involved in cancer cell proliferation, invasiveness, and motility.

Together, the presence of characteristics for cancer cells non-coding RNA as well as of RNA coding for epithelial and cancer-specific genes could indicate a transfer of RNA from nuclear tumor cells. However, these data should be interpreted with caution as the existence of platelet-coated CTC and an association of tumor-derived extracellular vesicles with platelets have been described [107, 108]. Therefore, contamination of platelet RNA with RNA from other tumor-derived sources cannot be excluded.

Despite the small cohort size, qRT-PCR results could validate significantly differentially existing mRNAs for 2 out of the 7 examined genes, namely ZNF750 and COL1A1. These are very promising findings: whereas ZNF750 is a lineage-specific tumor suppressor in squamous cell carcinoma [49, 100–102], COL1A1 is associated with many tumor entities including oral and pancreatic cancer with prognostic relevance [109–111]. Additionally, in gastric cancer, COL1A1 is highly expressed and associated with cancer cell invasion, metastasis, and poor prognosis [112, 113]. COL1A is also a driver of oncogenesis in hepatocellular carcinoma [114]. Importantly, COL1A1 was recently identified as a candidate biomarker for HNSCC [99] and associated with cancer stem cell signature in HNSCC [115]. Thus, COL1A1 may serve as a possible diagnostic as well as a prognostic biomarker not only for HNSCC but also for other solid tumor entities. It, therefore, seems reasonable to implement the molecular COL1A1 characterization as a biomarker in the routine clinical/pathological molecular HNSCC diagnostic.

Although non-significant, many genes demonstrate a clear trend, where platelets of TPs show elevated mRNA levels compared to those of healthy individuals, this could be seen in almost all cases in varying degrees, except for FAT1. Reasons for non-significant results could be variations in individual patients or the small sample sizes, which is a limitation of the present study. Although sample size calculations according to Lin et al. [37] were performed before starting the here described pilot study, not all hypotheses could be checked with the estimated sample size. However, it is remarkable that, despite this limitation, the abundance of tumor-related RNA could be verified in a second, equally small cohort.

After taking a closer look at the individual results, two HDs appeared to behave like a TP (cohort 1: HD_3; cohort 2: HD_12), whose platelets were showing substantial “tumor-related mRNA” abundance. There was an attempt to acquire a more recent blood sample in order to investigate whether any malignancies are present now, but unfortunately, the donors have not agreed. While no further details on the health status of HD_3 is available, the HD HD_12 was a chronic tonsillitis patient, 22 years old of age at the time of the blood draw (5 years ago), with a tobacco and alcohol consumption history. Since the results obtained from platelet RNA of HD_12 were suspicious, it was hypothesized that this individual had a pathological condition influencing platelet physiology. This is likely since changes in gene expression profile, as well as platelet count in patients of chronic tonsillitis and other diseases, were observed [116–118].

However, based on an HNSCC-causing history in the case of HD_12, the possibility cannot be excluded, that with the method applied here, a pre-cancerogenic processes in that person classified as a HD was identified. To prove this assumption, prospective studies should be conducted analyzing platelet mRNA content routinely and the patients should be observed for years.

Additionally, one TP (TP_8) could be considered a HD. One could assume that with early tumor stage less tumor RNA can be taken up by platelets. However, this association turned out to be inconsistent in the here performed analyses. But certain drugs inhibit the function of platelets which could affect the interaction of tumor cells and platelets [119], and consequently the amount of transferred mRNA.

Statistical procedures exist to detect outliers [120]. However, the data presented here stem from a small sample size and most do not follow a normal distribution, so it was not possible to reliably apply these methods. However, in order to determine sample sizes for this pilot study we followed the suggestions of Lin et al. [37].

The first and fundamental question this work attempted to answer was whether a transfer of tumor RNA to platelets actually takes place since platelets from head and neck cancer patients have not been analyzed so far. And this study was able to answer this question positively using this small cohort. Additionally, applying this study design, a set of RNA that was specified as “tumor-related” in a second, again small cohort, was identified as at least enriched in patient platelets. Precisely because the data were collected using small cohorts, these findings show the great possible diagnostic potential of platelet analysis in head and neck cancer patients.

A strength of the present study is the inclusion of tumor RNA into the sequencing analyses and not just comparing platelets of TPs with that of HDs. Doing so, it was revealed that differentially existing RNA in platelets is not exclusively caused by the transfer from tumor to platelets. This observation has to consider for further studies on TEP. There are many factors influencing the platelet transcriptome even under non-pathological conditions. According to Simon et al. [121], the mRNA profiles of platelets even differ depending on the age of the patients, their ethnicity, or their gender. As already mentioned, different diseases can cause transcriptomic changes in platelets [116–118]. Moreover, there is a certain platelet transcriptome variance observed even in HDs, probably depending on the platelet subpopulation [122, 123]. All these factors could contribute to individual variations and to the initial observation of the present study that not all differentially existing RNA observed in platelets of TPs are characterized by a gain of RNA. Thus, there is an urgent need for robust biomarkers allowing the identification of defined diseases, including HNSCC.

Despite the limitations of this work, the presented results are very promising, because so far, there is a lack of effective screening options for the diagnostics and therapy monitoring of HNSCCs [124] and liquid biopsy poses a great potential for the future of cancer diagnostics. The results presented here provide a basis for a biomarker panel that shows a potential ability for platelets to distinguish between healthy and TPs in HNSCC tumors with a simple blood draw.

With the help of future research, the analysis of platelet RNA in head and neck TPs could generate a powerful method for monitoring tumorigenesis at any time and in real-time. As was suggested also for other cancer entities, tumor-specific therapies could be applied in a targeted manner and modified with respect to response rate or resistance development [6, 34, 125].