Anoikis is an important programmed cell death process that prevents the re-adhesion and growth of the shed cells or attachment of the shed cells to an incorrect location during the body’s development [18]. However, in many cancers, anoikis resistance is recognized as a primary mechanism for tumor invasion and migration, metastasis, and treatment resistance [19,20,21]. HNSCC is an insidious onset and highly invasive neoplasm, wherein a majority of the patients were diagnosed with metastatic carcinoma. It showed a low 5-year OS rate because of the lack of an effective early diagnosis and drug resistance strategy [1]. Thus, the construction of anoikis-related predictive models may help in effectively guiding the prognosis and treatment of HNSCC patients.

Herein, we construct a risk model consisting of seven anoikis-related lncRNAs (AC015878.1, CYTOR, EMSLR, LINC01503, LINC02084, RAB11B-AS1, Z97200.1) based on the data presented by the Cox and LASSO regression analyses conducted for predicting the prognosis, immune response, immunotherapy and chemotherapy response for HNSCC patients. More specifically, AC015878.1 was seen to be a member of the stemness-related model for HNSCC [22], while EMSLR and Z97200.1 were seen to be important components of the prognostic signature used for bladder cancer and kidney renal clear cell carcinoma, respectively [23, 24]. EMSLR regulated the cell proliferation and differentiation by repressing the promoter activity of LncPRESS1 in lung cancer cell [25]. LINC01503 could promoted the proliferation, migration, and invasion in esophageal squamous cell carcinoma (ESCC) cell lines. It disrupted the interaction of EBP1 and the subunit of PI3K, and then increased the AKT signaling [26]. Furthermore, LINC02084 was used as a risk predictor in kidney renal clear cell carcinoma, colon cancer, and hepatocellular carcinoma [27,28,29]. Moreover, LINC01503 could facilitate cell migration, infiltration, and epithelial–mesenchymal transition in cholangiocarcinoma cells [30], whereas CYTOR was up-regulated and significantly associated with the poor prognosis of the cancer patients[31], In HNSCC, CYTOR inhibited cell apoptosis following treatment with the chemotherapeutic drug diamminedichloroplatinum (DDP) [32]. In addition, RAB11B-AS1 was observed to be important for metastasis and poor prognosis in tumor cells [33, 34]. These studies suggested that the 7 anoikis-related lncRNAs could be advantageous in the construction of prognostic models. Further analysis of these lncRNAs could present novel targets for developing effective strategies for tumor therapy.

This model helped in categorizing the HNSCC patients into both risk groups on the basis of their median risk scores. This prognostic anoikis-related lncRNA signature was seen to be a better discriminator for HNSCC patients compared to the whole genome, anoikis-related genes, and anoikis-related lncRNAs. Therefore, we conducted a comprehensive analysis and evaluation of the proposed risk model for forecasting the prognosis and OS of HNSCC patients. K–M curve analysis showed that low-risk patients showed significantly better OS, DSS, and PFS values than those displayed by high-risk patients. K–M analysis of the clinical subgroup characteristics stated that low-risk patients showed significantly higher OS values. Furthermore, the results of the Cox regression analyses highlighted the fact that the proposed risk model could serve as an independent prognosis-predictive indicator in HNSCC patients using the data acquired from the training, validation, and entire sets. The proposed risk score-based nomogram offered findings that validated its predictive value for HNSCC patients. These results highlighted the effective role played by the anoikis-related lncRNA signature in anticipating the prognosis and OS of patients, suggesting that the proposed risk model complemented the clinicopathological characterization methods.

The Food and Drug Administration (FDA) proposed the application of TMB as a clinical biomarker for determining the ICB response in solid tumors, however, very few studies determined the predictive power of TMB in HNSCC patients [35]. Therefore, we determined the link between TMB and the risk model in HNSCC tissues based on mutation data derived from TCGA. The TMB score was computed after dividing the total sum of somatic mutations by size of exomes, and the findings implied that the high-risk HNSCC patients showed a high number of somatic mutations in comparison to low-risk HNSCC patients, specifically for TP53. Loss of TP53 might influence the survival of tumors after radiation or chemotherapy and it could influence the patient’s prognosis [36]. HNSCC samples were categorized into the high- and low-TMB groups on the basis of their median TMB scores. A high TMB score was significantly related to poor outcomes, but it must be noted that low-risk HNSCC patients exhibited a better prognosis irrespective of their TMB score. This finding further indicated that the proposed risk model could act as an independent candidate for anticipating the prognosis of HNSCC patients.

TME contributes significantly to tumor progression, especially during tumor initiation and metastasis, since the infiltration of the immune cells around the malignant tissues was sensitive to detecting cancer cells and inhibiting their growth [37]. Earlier studies have noted that high immune cell infiltration levels and TME scores were linked to the good prognosis of many cancer patients [38,39,40]. The TME of HNSCC is distinguished by abnormal changes in immune cell populations, pro-inflammatory cytokines, and immune checkpoint genes [41].

We observed that high-risk HNSCC patients displayed a significantly low immune score in comparison to the low-risk score patients. Furthermore, the low-risk patients exhibited a significantly higher infiltration level of follicular helper T cells, plasma cells, CD8 T cells, resting dendritic cells, regulatory T cells, and resting mast cells, and they also showed an enrichment in the checkpoint, cytolytic activity, HLA, pro-inflammatory, T-cell co-stimulation, and T-cell co-inhibition pathways. In addition, the results of KEGG, GO, and GSEA functional analyses also validated the immune-linked pathways included in the proposed risk model, which enabled us to understand the probable role played by the risk model in anticipating the effect of immunotherapy treatment during clinical studies.

HNSCC is an immunosuppressive disease, however, the development of immunotherapy for HNSCC has progressed rapidly in the past few years [42, 43]. The FDA approved the application of several immune checkpoint inhibitors, such as anti-PD-1 or programmed cell death 1-ligand 1 (PD-L1) antibodies, which include nivolumab and pembrolizumab, durvalumab and atezolizumab for treating the recurrence/metastasis of HNSCC [44]. A few other immune therapies which included the CTLA4 and IDO-1 inhibitors were also evaluated for clinical practice [45]. Though the above treatment strategies showed significant efficacy, very few HNSCC patients benefitted from immunotherapy during clinical practice [46]. Hence, novel prognostic biomarkers need to be identified to determine the immunotherapy response for optimizing the therapeutic strategies. This study noted a significant increase in the IPS values for anti-PD1, anti-CTLA4, and the combined anti-PD1 and anti-CTLA4 immunotherapy in low-risk patients. The expression levels of key immunomodulator or inflammatory mediator ICB genes such as IDO1, CD8A, GZMB, GZMA, PRF1, LAG3, CTLA4, IFNG, and PDCD1 were significantly elevated in low-risk patients. All the findings suggested that low-risk HNSCC patients showed a high sensitivity to the immune checkpoint inhibitors.

Multimodal combination therapies that include radiotherapy, surgery, and chemotherapy can act as the primary treatment strategy for advanced HNSCC patients with poor prognosis owing to recurrence or metastases [45]. Since cisplatin was first introduced in the 1970s, there has been an advancement in the chemotherapeutic strategies for HNSCC patients [47]. Hence, several cytotoxic anti-cancer agents, such as taxane-based anticancer drugs, such as docetaxel and paclitaxel, were more conventionally used for HNSCC [1]. Combined treatment of docetaxel and cisplatin for advanced HNSCC showed a good response of 33–53% [48]. Based on the above data, we employed the pRRophetic algorithm to study the impact of the risk model on the response of four common anti-cancer agents, such as cisplatin, paclitaxel, gemcitabine, and docetaxel. A significantly low IC50 value was noted in the high-risk patients for gemcitabine, docetaxel, and cisplatin, which indicated that these patients were more sensitive to chemotherapy. The above findings offered a theoretical basis for formulating personalized treatment regimens for HNSCC. If this finding is validated in a large, multi-centre clinical trial, patients could be accurately stratified based on their risk scores, allowing physicians to tailor treatment strategies and make informed decisions about the use of anti-cancer agents. On the other hand, by identifying those individuals at higher risk of cancer progression or recurrence, physicians would be able to intervene more aggressively with targeted therapies, thereby increasing the chances of successful treatment outcomes.

Although there are some prognostic models for HNSCC, only three anoikis-related models have been reported [9, 49, 50]. Compared with these studies, we constructed the anoikis-related prognostic model from the perspective of lncRNA. Since lncRNA has been proven to have good application in biomarkers for many diseases, the anoikis-related lncRNA signature may be more capable of assessing the prognostic value in HNSCC. However, this study presented a few limitations. First, these sets may not represent the entire HNSCC patient population. Even though we integrated an additional multi-center set, this study remains an in-depth analysis of HNSCC samples from public databases using bioinformatics methods, it is not sufficient for use in clinical practice before further studies and experiments. Second, there could be some bias in the random allocation of samples into the training and validation sets. In addition, the mechanisms of seven prognostic anoikis-related lncRNAs in HNSCC require further investigation.

In conclusion, the risk model that was designed using seven prognostic anoikis-related lncRNAs could anticipate the prognosis of HNSCC patients and can be employed as a good independent predictive indicator for HNSCC patients. Furthermore, this risk model could help in developing immunotherapeutic and chemotherapeutic strategies for treating HNSCC patients. It can help the clinicians develop personalized and precise treatment strategies for HNSCC.