Abstract

Background

   Head and neck squamous cell carcinoma (HNSC) ranks as the sixth most
   common malignancy. The identification of highly specific and sensitive
   prognostic markers and potential drug targets can contribute to
   enhanced patient prognosis and individualized treatments. Heat shock
   proteins (HSPs) act as molecular chaperones and play a crucial role in
   maintaining cell homeostasis. Recently, research has indicated that
   HSPs also act as “evil chaperones” in cancer development.

Methods

   In this study, we assessed the expression of HSPs in HNSC patients
   using the ONCOMINE, GEPIA, and UALCAN databases. Mutations of HSP genes
   were also analysed using the cBioPortal database. Additionally, the
   expression levels of HSPs were verified using the Human Protein Altas
   (THPA) database.

Results

   We found that the expression levels of HSPH1, HSPD1, SERPINH1, HSPA4,
   and HSP90AA1 were significantly higher in tissues from HNSC patients
   compared with normal tissues. Moreover, HSPH1, HSPD1, SERPINH1, HSPA4
   and HSP90AA1 expressions were linked to disease progression. Survival
   analysis with the GEPIA and OncoLnc databases indicated that
   upregulation of HSPH1, HSPD1, SERPINH1, HSPA4 and HSP90AA1 was related
   to poor overall survival (OS).

Conclusion

   This study suggests that the HSPH1, HSPD1, SERPINH1, HSPA4 and HSP90AA1
   genes are potential clinical targets and prognostic biomarkers for
   patients with HNSC.

   Keywords: Biomarkers, Prognosis, Expression, Heat shock proteins, HNSC

Background

   Head and neck squamous cell carcinoma (HNSC) commonly occurs in the
   oral cavity, larynx, and pharynx and ranks as the sixth most common
   malignancy. In 2018, there was an estimated 43,000 HNSC-associated
   deaths worldwide and 835,000 new cases [[31]1, [32]2]. Unfortunately,
   diagnosis of HNSC is usually made at an advanced disease stage as the
   clinical symptoms of HNSC are not obvious during the early stage. As a
   result, the current 5-year survival rate is less than 65% [[33]3]. It
   is generally understood that the accumulation of genetic mutations in
   epithelial cells plays a key causal role in the development and
   progression of HNSC [[34]4]. Hence, the identification of highly
   specific and sensitive prognostic markers and potential drug targets
   can contribute to enhanced patient prognosis and individualised
   treatments.

   Heat shock proteins (HSPs) are genetically highly conserved proteins
   that act as molecular chaperones and play a crucial role in maintaining
   cell homeostasis [[35]5]. Aside from their cytoprotective effects,
   recent research has suggested that HSPs also act as “evil chaperones”
   in the development, progression, metastasis and drug resistance of
   cancers [[36]6, [37]7]. Therefore, HSPs have recently been proposed as
   potential therapeutic targets for tumor therapy [[38]8]. HSPH1 (also
   named HSP105), a member of the HSP70 superfamily, is a component of the
   β-catenin degradation complex. Previous studies have demonstrated that
   overexpression of HSPH1/HSP105 in various cancers is associated with
   increased levels of nuclear β-catenin protein and upregulation of Wnt
   target genes [[39]9]. HSPD1 is a molecular chaperone primarily
   localised in the mitochondrial matrix. It has been described as a
   potential prognostic and diagnostic biomarker for cancer. Recent
   studies have demonstrated that HSPD1 not only regulates the stability
   of survivin protein, but also regulates the mRNA expression of survivin
   [[40]10]. SERPINH1, also called heat shock protein 47 (HSP47), is a
   collagen specific molecular chaperone. Several studies have confirmed
   that SERPINH1 participates in numerous steps of collagen synthesis,
   blocking the aggregation of procollagen and inducing the hydroxylation
   of proline and lysine residues. Abnormal expression levels of SERPINH1
   are frequently found in a variety of cancers, including cervical, lung
   and gastric cancers [[41]11–[42]13]. Heat shock protein A4 (HSPA4), a
   member of the HSP110 family, is widely expressed in a variety of organs
   and can be induced under different conditions, including carcinogenic
   stress [[43]14–[44]16]. Recent studies have indicated that knockdown of
   HSPA4 can significantly reduce the migration, invasion, and
   transformation activities of tumor cells [[45]17]. Heat shock protein
   90α (Hsp90α) is the major cytosolic chaperone in eukaryotes. It is
   involved in cell protection and intracellular signaling transduction,
   controls intracellular homeostasis and assemblies of endoplasmic
   reticulum-secreted peptides, and regulates the translocation of
   proteins across the membranes of organelles after translation.
   Upregulated expression of Hsp90α is observed in a variety of cancer
   tissues, including liver, breast, and pancreatic cancers
   [[46]18–[47]20]. However, there is limited understanding of the
   underlying mechanisms and the unique roles of these genes in HNSC.

   Although some studies have reported dysregulated expression of HSPs in
   HNSC and have linked this to patients’ prognosis [[48]21, [49]22], the
   overall HSP expression profiles and the prognostic relevance of these
   expression profiles remain unknown. In the current study, we assessed
   the expression levels and mutations of HSPs in HNSC patients. The aim
   was to assess the potential functions, patterns of expression, and
   prognostic relevance of these genes in HNSC. To achieve this goal, we
   analysed large datasets available in various public databases.

Methods

Ethics statement

   Our study was approved by the Academic Committee of Huazhong University
   of Science and Technology, and conducted in accordance with the
   principles expressed in the Declaration of Helsinki. All datasets used
   in this study were extracted from online public databases or published
   literature, and thus, written informed consent was not required for the
   current data analysis.

ONCOMINE analysis

   The expression patterns of HSPs in different cancers can be assessed
   using the online ONCOMINE database ([50]https://www.oncomine.org).
   ONCOMINE is an integrated data-mining platform that store previously
   published or open-access cancer microarray data. The results were
   filtered by selecting HNSC vs. normal tissue.

GEPIA dataset

   GEPIA is an online database that incorporates gene expression data from
   TCGA and GTEx, bringing together 9736 tumor samples and 8587 normal
   controls. It allows for the assessment of differential expression
   profiles, patient outcomes, and various other analyses [[51]23].
   Through these public bioinformatics platforms, we are able to analyse
   the expression profiles of HSP genes in HNSC.

UALCAN dataset

   The UALCAN database ([52]http://ualcan.path.uab.edu/index.html) is a
   user-friendly interactive platform for facilitating tumor subgroup gene
   expression and survival analyses [[53]24]. Using the UALCAN database,
   we can evaluate the expression levels of HSP genes in HNSC and normal
   tissues based on tumor stages in the Cancer Genome Atlas (TCGA) HNSC
   datasets.

The Human Protein Atlas (THPA) database

   The Human Protein Atlas (THPA) ([54]https://www.proteinatlas.org) is a
   public database that can be used to validate the expression of target
   genes. It contains immunohistochemical expression data for near 20
   common kinds of cancers. In this study, we compared the protein
   expression levels of different HSPs between normal and HNSC tissues
   using immunohistochemistry images.

Prognostic analysis

   Using the GEPIA ([55]http://gepia.cancer-pku.cn) database and OncoLnc
   ([56]http://www.oncolnc.org) database, we analyzed the relationship
   between HSP gene expressions and overall survival in HNSC patients
   using a cox p value threshold of < 0.05.

TCGA data and cBioPortal

   The Cancer Genome Atlas (TCGA) database contains sequencing and
   clinicopathologic data for 30 cancer types [[57]25]. The Head and Neck
   Squamous Cell Carcinoma (TCGA, Provisional) was selected for further
   analysis of HSP genes using cBioPortal
   ([58]https://www.cbioportal.org). Genomic profiles, including mutations
   and copy-number alterations (CNA) were calculated using the
   cBioPortal’s online tool.

Protein–protein interaction (PPI) network construction

   In order to better understand the molecular mechanisms of HSPs in
   tumorigenesis, the PPI network for HSP family genes was constructed
   using STRING database. We choose a minimum interaction score of 0.4 as
   a cut-off when visualizing this PPI network.

Functional and pathway enrichment analysis

   The online STRING ([59]https://string-db.org/) tool provides
   investigators with systematic and comprehensive functional annotation
   tools to identify the biological meaning behind an extensive list of
   genes. In our study, Gene ontology (GO) analysis and Kyoto Encyclopedia
   of Genes and Genomes (KEGG) enrichment analyses were conducted for HSP
   family genes using STRING. The significance threshold was p < 0.05.

Results

The expression levels of HSPs in HNSC patients

   In order to explore the prognostic and potential therapeutic values of
   different HSP members in HNSC, we analyzed the mRNA and protein
   expressions of HSPs in HNSC patients using the ONCOMINE database. The
   expressions of five HSPs members (HSPH1, HSPD1, SERPINH1, HSPA4 and
   HSP90AA1) in more than 20 types of cancers were detected and compared
   with expressions in normal tissues using the ONCOMINE database. As were
   shown in Fig. [60]1, the expression levels of HSPs were significantly
   upregulated in HNSC patients. Interestingly, the expression levels of
   these five HSP genes were significantly upregulated in most tumors. In
   addition, we searched various HNSC datasets and found that these five
   genes showed significant upregulation across all datasets (Additional
   file [61]1: Fig. S1, Table [62]1). We further used the UALCAN.database
   to validate our findings. The results also showed that the expression
   levels of HSPH1, HSPD1, SERPINH1, HSPA4 and HSP90AA1 in HNSC were
   significantly increased (Additional file [63]2: Fig. S2).

Fig. 1.

   [64]Fig. 1
   [65]Open in a new tab

   The expression levels of HSPs in different types of cancers (ONCOMINE).
   The expression levels of HSPH1, HSPD1, Serpinh1, HSPA4 and HSP90AA1 in
   different types of cancers. Red: over-expression; Blue: down-regulated
   expression

Table 1.

   The significant changes of HSP associated genes expression levels
   between different types of Head and neck cancer and normal tissues
   (ONCOMINE database)
   Gene Types of head and neck cancer vs. normal Fold change p value T
   test Reference (PMID)
   HSPH1 Tongue carcinoma vs. normal 2.902 5.56E−7 6.528 Pyeon, 17510386
   Tongue squamous cell carcinoma vs. normal 2.136 7.85E−5 4.433 Ye,
   18254958
   Head and neck squamous cell carcinoma vs. normal 1.896 1.37E−5 5.107
   Ginos, 14729608
   Tall cell variant thyroid gland papillary carcinoma vs. normal 1.178
   3.22E−6 7.594 Giordano, 16609007
   Hypopharyngeal squamous cell carcinoma vs. normal 2.982 0.033 2.362
   Schlingemann, 16205657
   HSPD1 Salivary gland adenoid cystic carcinoma vs. normal 1.500 3.23E−4
   4.102 FriersonHF, 12368205
   Head and neck squamous cell carcinoma vs. normal 2.282 4.67E−7 6.773
   Ginos, 14729608
   Floor of the mouth carcinoma vs. normal 2.488 9.57E−7 7.144 Pyeon,
   17510386
   Nasopharyngeal carcinoma vs. normal 1.812 3.60E−5 5.246 Sengupta,
   16912175
   Tongue squamous cell carcinoma vs. normal  1.732 1.59E−5 4.785 Ye,
   18254958
   SERPINH1 Head and neck squamous cell carcinoma vs. normal 6.350 1.25E−6
   12.064 Cromer, 14676830
   Tongue squamous cell carcinoma vs. normal 3.148 1.09E−10 7.757
   Estilo,19138406
   Salivary gland adenoid cystic carcinoma vs. normal 308.933 3.74E−6
   7.691 FriersonHF, 12368205
   Oral cavity squamous cell carcinoma vs. normal 2.072 2.44E−9 7.125
   Peng,21853135
   Tongue squamous cell carcinoma vs. normal 2.022 2.05E−9 7.405 Talbot,
   15833835
   HSPA4 Thyroid gland oncocytic follicular carcinoma vs. normal 1.259
   1.32E−4 6.863 Giordano, 16609007
   Oral cavity carcinoma vs. normal 2.732 4.49E−6 6.212 Pyeon, 17510386
   Tongue squamous cell carcinoma vs. normal 1.478 1.74E−6 5.146 Talbot,
   15833835
   Follicular variant thyroid gland papillary carcinoma vs. normal 1.017
   4.02E−4 3.464 TCGA, 2013
   Oral cavity squamous cell carcinoma epithelia vs. normal 1.199 0.002
   3.391 Toruner, 15381369
   HSP90AA1 Floor of the mouth carcinoma vs. normal 3.010 1.40E−8 8.051
   Pyeon, 17510386
   Oral cavity carcinoma vs. normal 3.878 6.57E−9 8.959 Pyeon, 17510386
   Tongue carcinoma vs. normal 2.549 2.98E−7 6.220 Pyeon, 17510386
   Tonsillar carcinoma vs. normal 2.641 1.81E−6 6.050 Pyeon, 17510386
   Hypopharyngeal squamous cell carcinoma vs. normal 2.931 0.011 3.044
   Schlingemann, 16205657
   [66]Open in a new tab

   TCGA the Cancer Genome Atlas

Correlations between the expression levels of HSPs and the
clinicopathological parameters of HNSC patients

   GEPIA was used to compare HSPs between tumor and normal tissues. The
   findings revealed that the expression of all five genes was higher in
   HNSC samples than in normal control samples (Fig. [67]2a, b). We also
   analyzed HSPH1, HSPD1, SERPINH1, HSPA4, and HSP90AA1 expression as a
   function of the HNSC tumor stage. The results revealed a clear
   correlation between gene expression and tumor stage, with HNSC patients
   in more advanced stages tending to exhibit HSP expression levels
   (Fig. [68]3). After investigating the expression patterns of HSPH1,
   HSPD1, SERPINH1, HSPA4 and HSP90AA1 in HNSC, we examined HSP expression
   patterns in HNSC using THPA. The results confirmed that the protein
   levels of HSPH1, HSPD1, SERPINH1, HSPA4 and HSP90AA1 were elevated in
   HNSC samples relative to normal control samples (Fig. [69]4).

Fig. 2.

   [70]Fig. 2
   [71]Open in a new tab

   The expression profiles of HSPs in HNSC patients. The expression
   profile of HSPH1, HSPD1, Serpinh1, HSPA4 and HSP90AA1 in HNSC patients
   were analyzed by GEPIA (a) and UALCAN database (b); the p-value was set
   at 0.05

Fig. 3.

   [72]Fig. 3
   [73]Open in a new tab

   Correlations between expression levels of HSPs and tumor stage in HNSC
   patients. The relationships between expression levels of HSPH1, HSPD1,
   SERPINH1, HSPA4, and HSP90AA1 and tumor stages in HNSC patients were
   determined using the UALCAN database. The p-value was set at 0.05. The
   abscissa indicates the HNSC tumor stage and the ordinate indicates the
   expression levels of HSPs (gene expression ~ pathological stage)

Fig. 4.

   Fig. 4
   [74]Open in a new tab

   HSP genes were distinctively high expressed in HNSC tissues from Human
   Protein Atlas. HSP proteins were not expressed in normal head and neck
   tissues, whereas their high expressions were observed in HNSC tissues

Genetic alterations to HSP-associated genes and neighbor gene networks in
HNSC patients

   The frequencies of mutations of the HSPH1, HSPD1, SERPINH1, HSPA4 and
   HSP90AA1 genes in HNSC were assessed using cBioPortal. In total, 504
   patients from the Head and Neck Squamous Cell Carcinoma (TCGA,
   Provisional) were analyzed. According to this dataset, the percentages
   of genetic variations in HSPH1, HSPD1, SERPINH1, HSPA4 and HSP90AA1
   genes among HNSC patients ranged from 1.4 to 4% for individual genes
   (HSPH1, 1.4%; HSPD1, 2.6%; SERPINH1, 4%; HSPA4, 1.8%; HSP90AA1, 3%). A
   total of 62 (12%) samples exhibited gene set/pathway alterations, with
   mutations in 2.82% of analyzed gene sets (Fig. [75]5a).The cBioPortal
   online tool also allows Pearson correlation analysis of HSP expression
   data (RNA Seq V2 RSEM) in HNSC (TCGA, Provisional). The results
   revealed a significant negative correlation between HSPD1 and SERPINH1
   (Fig. [76]5b). Next, a PPI enrichment analysis, was then used to
   explore the relationships among these genes in HNSC. The PPI network
   was constructed by STRING. The results showed that several HSPs,
   including DNAJB1, HSPA1A, STIP1, HSPE1, HSPA8, HSPA9, HSF1, and
   HSP90AB1, were closely associated with HSPH1, HSPD1, SERPINH1, HSPA4
   and HSP90AA1 (Fig. [77]5c).

Fig. 5.

   [78]Fig. 5
   [79]Open in a new tab

   Expression of HSP genes and mutation analysis in HNSC. a Summary of
   alterations in HSPs. b Pearson correlation analysis of HSP family
   members. c Protein–protein interaction network of HSPs

The prognostic values of HSPs in HNSC

   We evaluated the prognostic significance of HSPH1, HSPD1, SERPINH1,
   HSPA4 and HSP90AA1 in all HNSC patients using Kaplan–Meier plots and
   the OncoLnc database. The results revealed that increased expression
   levels of HSPH1, HSPD1, SERPINH1, HSPA4 and HSP90AA1 were strongly
   associated with poor overall survival (Fig. [80]6). Thus, the results
   suggest that highly expressed HSPs (HSPH1, HSPD1, SERPINH1, HSPA4 and
   HSP90AA1) are prognostic factors for HNSC. Since HSPs are reportedly
   associated with tumor immunity [[81]26], we used the TIMER
   ([82]https://cistrome.shinyapps.io/timer/) database to investigate the
   relationships between the HSP expression levels and the levels of
   immune infiltration in HNSC. Unfortunately, we did not find any
   significant correlations between HSP gene expressions and immune
   infiltration levels (Additional file [83]3: Fig. S3).

Fig. 6.

   [84]Fig. 6
   [85]Open in a new tab

   The prognostic values of HSP family members in HNSC patients. The
   overall survival curves comparing HNSC patients with high (red) and low
   (blue) HSPH1, HSPD1, SERPINH1, HSPA4, and HSP90AA1 expression levels
   were plotted using the GEPIA (a) and OncoLnc databases (b) at the
   threshold p-value of < 0.05

Functional enrichment analysis of HSPs in HNSC patients

   The functions of these five genes were next explored through GO and
   KEGG analyses. GO analyses allow assessment of the biological process,
   molecular function, and cellular component annotations of genes of
   interest. The results revealed that HSPH1, HSPD1, SERPINH1, HSPA4 and
   HSP90AA1 as well as their neighbor genes, are primarily enriched for
   regulation of protein ubiquitination, cellular protein metabolic and
   immune system process, chaperone-mediated autophagy, regulation of
   apoptotic process, positive regulation of DNA metabolic process,
   protein targeting to mitochondrion, regulation of cellular response to
   stress and heat, and protein folding (Additional file [86]4: Fig S4A).
   Enriched molecular functions included transcription regulation by Hsp70
   protein binding, ATPase regulator activity, enzyme regulator activity,
   DNA polymerase binding, ubiquitin protein ligase binding, chaperone
   binding, ATP binding, protein binding, HSP binding, and unfolded
   protein binding (Additional file [87]4: Fig S4B). Cellular component
   annotations for these genes included mitochondrion, cytoplasmic vesicle
   lumen, intracellular organelle part, protein-containing complex,
   extracellular exosome, ficolin-1-rich granule lumen, cytosol,
   intracellular organelle lumen, cytoplasmic part, and chaperone complex
   (Additional file [88]4: Fig S4C). The KEGG pathways for these genes are
   shown in Table [89]2. Among these pathways, protein processing in the
   estrogen signaling pathway, endoplasmic reticulum, antigen processing
   and presentation, Prostate cancer, PI3K-Akt signaling pathway, NOD-like
   receptor signaling pathway, Epstein-Barr virus infection, MAPK
   signaling pathway, IL-17 signaling pathway, and Th17 cell
   differentiation were involved in tumor development and pathogenesis in
   HNSC (Additional file [90]5: Fig S5A and B).

Table 2.

   KEGG pathway analysis of HSP associated genes in HNSC
   Pathway ID Pathway name Gene count False discovery rate Genes
   hsa04141 Protein processing in endoplasmic reticulum 6 7.82E−07 DNAJB1,
   HSP90AA1, HSP90AB1, HSPA1A, HSPA8, HSPH1
   hsa04612 Antigen processing and presentation 5 7.82E−07 HSP90AA1,
   HSP90AB1, HSPA1A, HSPA4, HSPA8
   hsa04915 Estrogen signaling pathway 5 7.43E−06 AKT1, HSP90AA1,
   HSP90AB1, HSPA1A, HSPA8
   hsa05215 Prostate cancer 3 0.0017 AKT1, HSP90AA1, HSP90AB1
   hsa04151 PI3K-Akt signaling pathway 4 0.0033 AKT1, CDC37, HSP90AA1,
   HSP90AB1
   hsa04621 NOD-like receptor signaling pathway 3 0.0045 HSP90AA1,
   HSP90AB1, SUGT1
   hsa05169 Epstein-Barr virus infection 3 0.0061 AKT1, HSPA1A, HSPA8
   hsa04010 MAPK signaling pathway 3 0.0173 AKT1, HSPA1A, HSPA8
   hsa04657 IL-17 signaling pathway 2 0.0209 HSP90AA1, HSP90AB1
   hsa04659 Th17 cell differentiation 2 0.0242 HSP90AA1, HSP90AB1
   [91]Open in a new tab

Discussion

   Dysregulation of HSPs is common in cancer development. Research
   suggests that HSPs are essential for tumor cell proliferation and
   differentiation. Although rarely used as diagnostic biomarkers for
   cancer, the expression levels of HSPs may predict the development of
   various cancers. For example, there is considerable evidence
   demonstrating that the over-expression of HSP27 may confer poor
   prognosis in gastric, prostate, and liver cancers [[92]27–[93]29].
   Herein, we sought to investigate the expression pattern and prognostic
   values of different HSP members (HSPH1, HSPD1, SERPINH1, HSPA4 and
   HSP90AA1) in HNSC. These findings advance our current understanding of
   HNSC and may offer a means for improving treatment approaches and
   prognostic accuracy in HNSC patients.

   The primary role of HSPs in tumorigenesis involves the stabilisation of
   functions of mutated or aberrantly-expressed tumor-related genes. Thus,
   high expression of HSPs is a hallmark of many cancers. In addition,
   HSPs are released from cancer cells, influencing their properties and
   functions through receptor-mediated signaling [[94]30]. HSPH1 has been
   reported to be over-expressed in melanoma and colon cancer patients
   [[95]31, [96]32]. The upregulation of HSPH1 is implicated in
   hyperactivation of the Wnt signalling pathway. The transcription levels
   of Wnt signalling target genes are significantly downregulated in cell
   line models with HSPH1 inhibition [[97]9]. In our study, we confirmed
   that the expression of HSPH1 in HNSC tissue was significantly elevated
   compared with normal tissue. We also observed a significant correlation
   between the expression of HSPH1 and tumor stage in HNSC patients. High
   HSPH1 expression was associated with low overall survival in all HNSC
   patients.

   HSPD1 is a molecular chaperone that is primarily localized in the
   mitochondrial matrix [[98]33]. Recently, HSPD1 has been found in many
   extramitochondrial sites, including the extracellular surface, cell
   surface, intracellular vesicles, nucleus, extracellular fluid, and even
   the cytoplasm. HSPD1 has been recognized as a potential biomarker for
   tumor diagnosis and prognosis, especially in colorectal cancer. There
   is increasing evidence that HSPD1, especially intracellular HSPD1, is
   involved in the survival and metastasis of various tumors
   [[99]34–[100]36]. In our study, the ONCOMINE and TCGA datasets revealed
   that the expression of HSPD1 was higher in HNSC tissue than in normal
   tissues. High HSPD1 expression was significantly associated with poor
   overall survival among HNSC patients followed up for more than
   6000 days.

   SERPINH1 is a collagen-binding protein and has been identified to be a
   collagen-specific chaperon. It promotes the malignant behavior of
   cancer cells and predicts the metastatic activity of human cancer
   cells. SERPINH1 also plays an essential role in regulating the
   expression of extracellular matrix (ECM) proteins and fibronectin (FN);
   therefore, dysfunction of SERPINH1 stimulates abnormal expression of
   ECM proteins, which promotes epithelial-mesenehymaltransition (EMT)
   [[101]37]. Our data confirmed that dysregulation of SERPINH1 is closely
   related to the development and progression of HNSC, and further affects
   the prognosis of HNSC.

   HSPA4 is highly expressed in malignant tumor cells and is involved in
   tumor development and chemotherapy resistance, presumably due to its
   ability to inhibit tumor cell apoptosis [[102]38]. It has been reported
   that overexpression of HSPA4 can inhibit cell apoptosis and prevent
   activation of the caspase signaling pathway, which leads to
   accumulation of misfolded proteins, ROS, and DNA damage. HSP70 not only
   affects the apoptotic pathway, but also the autophagic pathway. It
   prevents the formation of autophagosomes by activating the mTOR
   pathway. Consistent with these results, Leu et al. found that the
   inhibition of HSP70 resulted in significant increases in the expression
   of LC3-II and the number of autophagosomes [[103]39]. In our study, it
   is clear that HSPA4 acts as an “evil chaperone” in HNSC. Its expression
   was significantly increased in HNSC patients and was associated with
   poor prognosis.

   HSP90AA1 acts as a highly conserved chaperone protein and participates
   in tumor cell differentiation, proliferation, and angiogenesis.
   Recently, it has been regarded as a promising target for specific
   cancer therapy [[104]40]. Hsp90AA1 is expressed in various cancers,
   including breast, colon, ovarian, lung and prostate cancers. This may
   be related to the involvement of HSP90AA1 in the regulation of
   apoptosis and signaling transduction triggered by growth factors, death
   receptors, and stress signals [[105]41]. HSP90AA1 inhibits the
   initiation of apoptosis by blocking the binding of caspase 9 to
   apoptotic protein 1 activator. Additionally, it promotes the formation
   of tumor cells by stabilizing mutant p53 complexes, thereby inhibiting
   the apoptosis of tumor cells [[106]38, [107]42]. In our systematic
   analyses of various databases, we found that HSP90AA1 was significantly
   upregulated in HNSC patients, and its expression level was
   significantly associated with the tumor stage in HNSC patients. As
   expected, high HSP90AA1 expression predicted poor overall survival
   among HNSC patients, implying an oncogenic role of HSP90AA1 in HNSC.

   In this study, we also explored the genetic alterations and potential
   functions of HSP family members. The percentages of genetic mutations
   in HSPs in HNSC varied from 1.4% to 4% for individual genes. At the
   same time, we constructed a network for these five genes and their
   neighbor genes. Functional analysis demonstrated that these five genes
   were primarily enriched in tumor-related signaling pathways, indicating
   that HSPH1, HSPD1, SERPINH1, HSPA4, and HSP90AA1 play crucial
   cancer-promoting roles in the development of head and neck cancer.

Conclusion

   In summary, our results indicate that HSPH1, HSPD1, SERPINH1, HSPA4,
   and HSP90AA1 are significantly upregulated in HNSC patients and their
   upregulation is negatively correlated with HNSC tumor stage. Based on
   the above findings, it is expected that HSPH1, HSPD1, SERPINH1, HSPA4,
   and HSP90AA1 could act as potential prognostic biomarkers and
   therapeutic targets for HNSC. Our research contributes to a better
   understanding of the pathogenesis of HNSC and may assist in the
   development of more effective targeted drugs for head and neck cancer.
   However, further mechanistic studies are needed to validate our
   findings and to promote clinical application of HSPs as prognostic or
   therapeutic targets in HNSC.

Supplementary information

   [108]12935_2020_1296_MOESM1_ESM.tif^ (451.3KB, tif)

   Additional file 1: Fig. S1. The expression levels of HSPs in HNSC
   (ONCOIME). The expression levels of HSPH1, HSPD1, Serpinh1, HSPA4 and
   HSP90AA1 in several head and neck cancer studies. Red: over-expression.
   The significance threshold was p < 0.05.
   [109]12935_2020_1296_MOESM2_ESM.tif^ (727.1KB, tif)

   Additional file 2: Fig. S2. Expressions of HSPs across TCGA cancers
   (with tumor and normal samples). The expression levels of HSPH1, HSPD1,
   Serpinh1, HSPA4 and HSP90AA1 in pan-cancer.
   [110]12935_2020_1296_MOESM3_ESM.tif^ (3MB, tif)

   Additional file 3: Fig. S3. The relationships between the HSP
   expression levels and the levels of immune infiltration in HNSC.
   [111]12935_2020_1296_MOESM4_ESM.tif^ (455.3KB, tif)

   Additional file 4: Fig. S4. Functional Enrichment Analysis of HSPs in
   patients with HNSC. GO enrichment analysis predicted the function of
   target genes from three aspects: biological processes (A), cellular
   components (B), and molecular functions (C).
   [112]12935_2020_1296_MOESM5_ESM.tif^ (6MB, tif)

   Additional file 5: Fig. S5. p53 signal pathway and cell cycle pathway
   regulated by HSPs in HNSC. The MAPK signal pathway (A) and PI3K-Akt
   signaling pathway (B) regulated by HSPs in HNSC are shown.

Acknowledgements