Expression of Connexins 37, 40 and 45, Pannexin 1 and Vimentin in Laryngeal Squamous Cell Carcinomas

Approximately 60% of patients with squamous cell carcinoma (LSCC) have regional occult metastatic disease/distant metastases at the time of diagnosis, putting them at higher risk for disease progression. Therefore, biomarkers are needed for early prognostic purpose. The aim of this study was to analyze the expression pattern of connexins (Cx) 37, 40 and 45, pannexin1 (Panx1) and vimentin in LSCC and correlate with tumor grade (G) and outcome. Methods: Thirty-four patients who underwent (hemi-)laryngectomy and regional lymphadenectomy due to LSCC from 2017 to 2018 in University Hospital Split, Croatia, were studied. Samples of tumor tissue and adjacent normal mucosa embedded in paraffin blocks were stained using the immunofluorescence method and were semi-quantitatively analyzed. Results: The expression of Cx37, Cx40, and Panx1 differed between cancer and adjacent normal mucosa and between histological grades, being the highest in well-differentiated (G1) cancer and low/absent in poorly differentiated (G3) cancer (all p < 0.05). The expression of vimentin was the highest in G3 cancer. Expression of Cx45 was generally weak/absent, with no significant difference between cancer and the controls or between grades. Lower Panx1 and higher vimentin expression were found to be prognostic factors for regional metastatic disease. Lower Cx37 and 40 expressions were present in patients with disease recurrence after the three-year follow-up period. Conclusion: Cx37 and Cx40, Panx1, and vimentin have the potential to be used as prognostic biomarkers for LSCC.


Introduction
Intercellular communication is a key factor in cellular function and tissue homeostasis and is controlled at intracellular, extracellular, and intercellular levels [1,2]. Intercellular communication is mainly mediated by gap junctions (GJs) at the cell membrane, formed by the interaction of two hemichannels (HCs) [3,4]. Each HC, also known as a connexon, is composed of six connexins (Cxs), which are defined as tetraspanning integral membrane proteins [5]. In humans, 21 connexin subtypes have been characterized, and while some of them are expressed in different tissue types, most Cxs exhibit a tissue-specific expression pattern [2,6]. They are named according to their molecular weight, from the smallest (Cx23) to the largest (Cx62) [7]. GJs enable the intercellular transport of ions and small and hydrophilic molecules with molecular weights less than 1-1.5 kilodaltons [2]. In addition, connexins provide a communication pathway between the intracellular and with distant metastases were excluded from the study. Paraffin blocks of laryngeal tumor tissue and adjacent normal mucosa were obtained from the Department of Pathology, Cytology and Forensic Medicine of the same institution, and clinical data were collected from the hospital records. For each patient, the following data of interest were recorded: age, sex, tumor location, type of surgery, tumor size, TNM staging, lymphovascular and perineural invasion and pathohisto-logical grade (Figure 1). Immunofluorescence staining was performed at the Department of Anatomy, Histology and Embryology, University of Split, School of Medicine. For every patient, both malignant and normal laryngeal tissues were discriminated by an experienced pathologist and simultaneous staining was performed. The expression of connexin 37, connexin 40, connexin 45, pannexin-1 and vimentin was analyzed in the tumor samples in comparison to the adjacent healthy tissue by two independent investigators.

Subjects and Study Protocol
Thirty-two patients who underwent partial or total laryngectomy for LSCC at the Department of Otolaryngology and Head and Neck Surgery, University Hospital of Split, Split, Croatia, between 1 January 2017 and 31 December 2018, were included in this retrospective study. Patients who had undergone primary oncologic treatment and patients with distant metastases were excluded from the study. Paraffin blocks of laryngeal tumor tissue and adjacent normal mucosa were obtained from the Department of Pathology, Cytology and Forensic Medicine of the same institution, and clinical data were collected from the hospital records. For each patient, the following data of interest were recorded: age, sex, tumor location, type of surgery, tumor size, TNM staging, lymphovascular and perineural invasion and pathohisto-logical grade (Figure 1). Immunofluorescence staining was performed at the Department of Anatomy, Histology and Embryology, University of Split, School of Medicine. For every patient, both malignant and normal laryngeal tissues were discriminated by an experienced pathologist and simultaneous staining was performed. The expression of connexin 37, connexin 40, connexin 45, pannexin-1 and vimentin was analyzed in the tumor samples in comparison to the adjacent healthy tissue by two independent investigators.

Immunofluorescence
Five µm thin slides were cut from the paraffin blocks, deparaffinized in xylene, dehydrated through grades of alcohol, washed in distilled water, boiled for 20' in sodium citrate buffer (pH = 7) at 95 °C in a water steamer, then gradually cooled to room temperature and rinsed with phosphate-buffered saline (PBS). Protein blocking buffer (ab64226, Abcam, Cambridge, UK) was then applied for 30 min to prevent non-specific staining.

Immunofluorescence
Five µm thin slides were cut from the paraffin blocks, deparaffinized in xylene, dehydrated through grades of alcohol, washed in distilled water, boiled for 20 in sodium citrate buffer (pH = 7) at 95 • C in a water steamer, then gradually cooled to room temperature and rinsed with phosphate-buffered saline (PBS). Protein blocking buffer (ab64226, Abcam, Cambridge, UK) was then applied for 30 min to prevent non-specific staining. The primary antibodies (Table 1) were then applied and incubated overnight in a humidity chamber. The next day, after washing in PBS, the samples were rinsed with PBS, before incubation with the appropriate secondary antibodies (Table 1) for 60 min. After washing in PBS, DAPI (4 ,6-diamidino-2-phenylindole) was applied for 2 blue nuclear staining, and the slides were washed and coverslipped (Immu-Mount, Thermo Shandon, Pittsburgh, PA, USA). No staining was observed when the primary antibodies were omitted from the immunofluorescence protocol.
Statistical analysis was performed using the National Council for the Social Studies (NCSS) 2022 software (Kaysville, Utah, USA). Statistical significance was set at p < 0.05, and all confidence intervals (CI) were at the 95% level. Spearman's correlation coefficient (Rho) was used to calculate the degree of association. Analysis of the statistical significance of the differences in several numeric variables was performed with the Kruskal-Wallis test and analysis of the differences between the two groups was performed with the Mann-Whitney and Dunn tests. The Chi2 test was used for the analysis of the categorical variables. A logistic regression model was performed to determine the independent prognostic factors. Receiver-operating characteristic curves (ROC) were constructed to calculate the cut-off values that were significant for both recurrence and overall survival.

Clinical and Pathological Characteristics of the Study Sample
Among the 32 patients with LSCC, 29 (91%) males and 3 females (9%) with a mean age of 68 (min-max: 43-84) years. Patients were divided into three groups according to the degree of tumor differentiation (G1, G2; G3). There were 10 (31%), 12 (38%), and 10 (31%) patients in each group, respectively. The clinical and pathohistological characteristics are summarized in Table 2. Patients with different tumor grades did not differ by age, sex, or tumor volume (all p > 0.05). Lymphovascular invasion was more common in poorly differentiated LSCC patients (G3) (p = 0.049).

Alteration of Expression of Cx45
As shown in Figures 3 and 5, the expression of Cx45 was also membranous, but absent/weak in the tumor and control, with values of 0.83 (IQR: 0.37-1.16) and 1 (IQR: 1-1.14), respectively. No significant difference was found between the tumors and controls or between tumor grades (p = 0.373 and p = 0.670, respectively). There was no correlation with the expression of vimentin (p = 0.342), Cx40 (p = 0.881), or Panx1 (p = 0.708).

Alteration of Expression of Cx45
As shown in Figures 3 and 5, the expression of Cx45 was also membranous, but absent/weak in the tumor and control, with values of 0.83 (IQR: 0.37-1.16) and 1 (IQR: 1-1.14), respectively. No significant difference was found between the tumors and controls or between tumor grades (p = 0.373 and p = 0.670, respectively). There was no correlation with the expression of vimentin (p = 0.342), Cx40 (p = 0.881), or Panx1 (p = 0.708). Genes 2023, 14, x FOR PEER REVIEW 9 of 16

Expression Pattern of Vimentin
The expression of vimentin in normal laryngeal mucosa was negative (Figures 3 and  7). Homogeneous cytoplasmic expression was found in poorly differentiated laryngeal cells. It was the lowest in G1, with a median value of 2.2 (IQR: 1.8-3.83), and the highest in G3, with a median value of 8.6 (IQR: 8.515-10.2). A statistically significant difference was observed between vimentin expression when comparing G1 vs G2 (median 5.6; IQR: 3.48-8.36), where p = 0.035, Z = −2.111, and when comparing G1 vs G3, when p = 0.006, Z = −2.761, respectively. There was no difference in vimentin expression between G2 and G3, where p = 0.094. The analysis of vimentin expression in the control and tumor tissues showed a statistically significant difference for all tumor grades (all p < 0.05). In contrast

Expression Pattern of Vimentin
The expression of vimentin in normal laryngeal mucosa was negative (Figures 3 and 7).
Conversely, the expression of Cx45, Panx1, and vimentin was higher in cases without PNI (all p < 0.001). Tumor location was not associated with Cxs or Panx-1 (all p > 0.05). Vimentin expression was the highest in transglottic tumors (p = 0.029) and higher in supraglottic compared with glottic tumors.
3.8. Diagnostic and Prognostic Role of Cx37, Cx40, Cx45, Panx1 and Vimentin All patients were without distant metastases at the time of diagnosis. N status was analyzed according to the presence (N+ including N1-N3) or absence (N0) of regional lymph node metastases. There was no difference in connexin expression according to N status (all p > 0.05). The Panx-1 IRS median was significantly higher in patients with a negative neck status compared to metastatic neck disease, with values of 4.5 and 2 (p = 0.045). Vimentin expression was higher in patients with a positive N status (median 7.65 versus 3.83), where p = 0.048. According to the logistic regression model, Panx-1 was an independent prognostic factor for regional metastatic disease in LSCC (p = 0.049, 95% CI: 0.563-0.980, OR: 0.76; regression coefficient −0.271). A higher histological grade, higher T and positive lymphovascular invasion were also proven to be risk factors for positive neck disease (all p < 0.05). In addition, positive neck disease was an independent risk factor for disease recurrence and mortality (p = 0.021 and p < 0.001, respectively). The expression of Cx37, as well as Cx40, was stronger in patients who had recurrent disease in a follow-up period of 3 years (p = 0.044 and p = 0.024, respectively). The median IRS values of Cx37 and Cx40 were 3.25 (IQR: 2.06-4.53) and 0.57 (IRS: 0.44-4.78). In patients without a recurrence medians of the Cx37 and Cx40 expression were 5.5 (IRS: 4.67-8.83) and 6 (IRS: 3.23-7.89), respectively.

Discussion
In our study, we investigated the expression of various connexins and Panx-1 as possible prognostic markers in LSCC. The expression of Cx37, Cx40, Panx-1, and vimentin differed significantly between the epithelium of normal laryngeal mucosa and the atypical epithelial cells of LSCC and was related to the histological grade and biological behavior of the cancer. The main function of connexins and Panx1 is to exchange information between cells [15]. Cross-talk between tumor cells and their environment enables cancer cells to reprogram the tumor microenvironment and migrate to distant sites, which promotes angiogenesis, immune surveillance, EMT, invasion, spread, and resistance to therapy [1]. To our knowledge, this is the first study to investigate the immunohistochemical expression of pannexin-1, connexin 37, 40, and 45 in human LSCC and laryngeal tissue, compared with vimentin and pathological parameters and outcomes. The highest staining intensity of hemichannel proteins was observed in well-differentiated tumors and gradually decreased, until they completely disappeared in poorly differentiated/undifferentiated LSCC. Vimentin is an intermediate type III filament, a structural protein of the cytoskeleton and viscoelastic scaffold [26]. Under pathological conditions, vimentin modulates genes for EMT inducers, such as Snail, Slug, Twist, and ZEB1/2, as well as epigenetic factors [26]. Vimentin suppresses cell differentiation, regulates pluripotent potential, and increases the stemness of cancer cells, promoting tumor invasiveness and causing resistance to therapy [26,27]. According to Usman et al., vimentin is the "heart" of EMT-mediated metastasis [26]. Recent evidence on the role of vimentin in LSCC has shown that vimentin plays an important negative prognostic role in LSCC, regardless of the TNM stage and histological grade of the tumor [28].
In contrast to vimentin, the possible involvement of Cxs and Panxs in LSCC progression remains unexplored, and the elucidation of their control mechanisms offers an opportunity to target cancer behavior [7]. To date, connexins 26, 30, 31.1, 32, and 43 have been studied in LSCC [29]. Immunohistochemical results have shown no change in the expression of Cx26, 30, 32, and 43 during laryngeal epithelial carcinogenesis [30], but Cx26 and Cx40 were found to play a role in tumor suppression [29,31]. Many studies have reported a correlation between a decrease in connexin expression and an increase in histological grade or dislocation of connexin within the cell [32]. The most important question remains whether connexins have tumor-suppressive or tumor-promoting effects [9]. Experts in this field suggest that altered connexin activity may be a unique feature of cancer [32]. The evaluation of connexins in cancer involves the study of GJ-related, hemichannel, and non-channel functions in the context of cell progression and checkpoint control, cancer cell proliferation and differentiation, metabolic reprogramming, angiogenesis, cancer microenvironments, tissue invasion and metastasis, and various cancer pathologies and therapeutic potential [9]. In a review paper, Jiang summarized the current knowledge on the tumor suppressive role of connexins in the early stages of cancer progression or in primary tumors, while noting how they play an opposite role in late or advanced cancers and metastases [9]. Moreover, Mulkearns-Hubert in his review article highlights the tumor suppressive or pro-tumorigenic effect on various human malignancies [32].
According to our results, the expression of Cx37, Cx40, and Panx-1 might be abolished in less differentiated cancer. We hypothesize that the lower expression of Cx37, Cx40, and Panx-1 and the higher expression of vimentin may be associated with a more aggressive tumor phenotype, possibly by regulating growth and tumorigenic properties. Moreover, this may confirm the hypothesis of the tumor-suppressive activity of Panx-1 reported in the C6 glioma cell line, adenocarcinoma of gallbladder and basal and squamous cell carcinoma of skin [9]. The proposed antitumorigenic mechanisms of Panx1 reduce cell proliferation, cell motility, anchorage-independent growth, and tumor growth in experimental animals [33]. The immunohistochemical study of Panx-1 with Ki-67 in human gallbladder and gallbladder adenocarcinomas showed that Panx1 expression negatively correlated with proliferation in gallbladder carcinomas [34]. In non-melanoma human skin cancers, the downregulation of Panx1 may indicate its protective function against keratinocyte transformation [35]. In contrast, most of the published work suggests that the amplification or upregulation of Panx1 plays a pro-tumorigenic role [23]. Panx1 expression positively correlates with the occurrence or progression of melanoma, hepatocellular carcinoma, breast cancer, colorectal cancer, and hematologic malignancies [9]. Pannexins play distinct roles in different subtypes of cancer cells. Cells that express mutant Panx1 have a survival advantage in metastasis because ATP released through the pannexin channel acts on purinergic receptors to suppress apoptosis and reduce cell death, suggesting the alteration of tumor cell proliferation and disease progression through transmembrane signaling of solutes, such as nucleotides and Ca 2+ [35].
Risk factors associated with an unfavorable outcome in LSCC include male gender, malnutrition, fitness status, alcohol and tobacco abuse, supraglottic location, perineural and lymphatic invasion, and a higher pathological stage and histological grade [36]. Thus, patients with supraglottic SCC have the highest frequency of regional metastases, both clinically visible and occult, due to its rich network of lymphatic vessels. Despite the availability of higher resolution laryngoscopes and a greater number of radiological images, clinical staging is insufficient, as 7-37% of patients with a clinical N0 stage still have occult metastatic lymph nodes [29]. Therefore, the identification of new biomarkers as disease predictors is needed [29]. Despite the promising results of several previously used biomarkers, such as E-cadherin, Ki-67, and Bcl-2, prospective studies are needed in systematic reviews to confirm the results in large numbers and laryngeal subsites [37][38][39]. Neither clinical, radiological nor histological evaluation have been included in the standard guidelines for the diagnosis and prognosis of LSCC [11,40].
In our study, the change in the expression pattern of Cx37, Cx40, pannexin-1, and vimentin was not only associated with different pathological stages (suggesting their possible role in tumor suppression), but also had a significant diagnostic/prognostic role. Indeed, lower Panx-1 expression was an independent prognostic factor for regional metastatic disease at diagnosis. Neck nodal status is a modifiable risk factor, which means that depending on a patient's Panx-1 status, elective neck dissection may lead to a better prognosis. In addition, low or absent Cx37 and Cx40 expression in a 3-year follow-up period proved to be a marker for disease recurrence. This may be particularly important in the modern oncological era with more organ-preserving approaches in the removal of all tumors, including LSCC, with the main goal of maintaining/improving patients' quality of life. This approach contributes to the possibility of overlooking occult metastatic disease, and thus an unfavorable outcome.
The beneficial or detrimental role of connexin and pannexin-1 in different cancer types remains controversial. Further research is needed to confirm whether the low expression or absence of connexin and pannexin-1 in undifferentiated LSCC, which have the worst prognosis, is a consequence or cause of the pathophysiological cascade, and why they exhibit opposite biological behavior in different tumors or even in different cell lines of the same tumor type, as reported in gliomas [9].
The main drawback of our study is the relatively small number of sample cohorts. Future studies, using other protein indices associated with the occurrence of squamous cell carcinoma, larger cohorts and in vitro approaches are needed in order to determine the exact mechanism of action between different grades of LSCC.

Conclusions
The recently discovered Pannexin-1 and the family of connexins have different roles in physiological and pathological conditions that have not yet been fully elucidated. According to studies, connexins and Panx1 are involved in the development of many human cancers. In our study, the expression of Cx37, Cx40, and Panx1 was investigated for the first time in surgically removed squamous cell carcinoma of the larynx by immunofluorescence method. We found a difference in the expression of Cx37 and Panx1 in cancer and adjacent normal laryngeal mucosa and between different histological grades of cancer. The intensity of staining was lowest in poorly differentiated LSCC (G3) and correlated with cancer recurrence after 3 years. These results support the idea that loss of cellular communication may be involved in the dedifferentiation of LSCC and suggest their potential utility as diagnostic and prognostic markers; further research is needed.