Functional characterization of alternatively spliced GSN in head and neck squamous cell carcinoma.

Abstract

We have recently performed the characterization of alternative splicing events (ASEs) in head and neck squamous cell carcinoma (HNSCC), which allows dysregulation of protein expression common for cancer cells. Such analysis demonstrated a high ASE prevalence amongst tumor samples, including tumor-specific alternative splicing in the GSN gene. In vitro studies confirmed that overall expression of either ASE-GSN or wild-type GSN (WT-GSN) isoform inversely correlated with cell proliferation, whereas the high ratio of ASE-GSN to WT-GSN correlated with increased cellular invasion. Additionally, a change in expression of either isoform caused compensatory changes in expression of the other isoform. Our results suggest that the overall expression and the balance between GSN isoforms are mediating factors in proliferation, while increased overall expression of ASE-GSN is specific to cancer tissues. As a result, we propose ASE-GSN can serve not only as a biomarker of disease and disease progression, but also as a neoantigen for HNSCC treatment, for which only a limited number of disease-specific targeted therapies currently exist.

BACKGROUND

Head and neck squamous cell carcinoma (HNSCC) is a disease with an insufficient number of targeted therapeutics and biomarkers currently used in clinical practice. The detection of alternative splicing events (ASE) allows for the discovery of both biomarkers and therapeutic targets.

INTRODUCTION:

Alternative splicing events allow for the formation of variant transcripts and increased protein functional diversity. (1, 2) Previous studies have shown that some cancers express increased transcript variants due to alternative splicing events. (3, 4) We hypothesized that alternative splicing might be one of the mechanisms by which dysregulation occurs in cancer, which may result in the oncogenic transformation of cells (altered cell proliferation, apoptosis, invasion, and migration). To explore this hypothesis, our group has recently performed a whole-genome evaluation of ASEs in Head and Neck Squamous Cell Carcinoma (HNSCC), a disease that is characterized by abundant gene expression dysregulation. (5) Limited advancement has been made toward predicting HNSCC tumor recurrence, aggression, and improving the identification of targeted and disease-specific treatment. (6, 7)

HNSCC is treated by surgical resection, radiation therapy, chemotherapy (Cetuximab), or molecular targeted therapy (Pembrolizumab/Nivolumab). Cetuximab is a monoclonal antibody against EGFR, (8) Pembrolizumab, (9) and Nivolumab (10) are monoclonal antibodies targeting PD1, (7, 11) the only targeted therapeutics approved for all Head and Neck patients. These therapies offer greater curative power with fewer adverse effects than traditional regimens, but current targeted therapeutics still involve serious adverse reactions in almost half of patients receiving treatment. (10) Further investigation is needed to identify therapeutic options with fewer adverse effects on patient health. ASEs have the potential to serve not only as cancer-specific biomarkers, but also as neoantigens for novel targeted therapeutics.

In our recent paper, RNA-Seq, a next-generation sequencing technology, was used to define the whole-genome distribution of ASE in HNSCC. (2) Using Outlier Gene Set Analysis (OGSA), (12, 13) we detected and validated 109 significant ASEs. The GSN (gelsolin) gene, which codes for an actin-binding protein that plays a role in PI3K/AKT/mTOR pathway, was detected as one of the most abundant ASEs in this cohort (present in 34% of HNSCC samples). (2, 14) Furthermore, ASE in GSN was detected in the pan-cancer analysis of RNA-Seq from TCGA as leading alternative splicing in HNSCC, along with ASE detected in TNC and CAB39L. (15) ASE in GSN was also detected in breast invasive carcinoma and kidney chromophobe. (15) Additionally, downregulation of GSN has been shown to downregulate apoptosis, a hallmark of cancer cells. (16–19) From these data, we hypothesized that alternative splicing in GSN causes malignant cell transformation, but no functional analysis for the alternatively-spliced GSN was ever performed before our work. Herein, we demonstrate that different GSN transcripts (WT and ASE) affect cellular carcinogenic characteristics. Furthermore, the abundance of ASE-GSN in cancer cases indicates that utilization of ASE-GSN as neoantigen has a direct clinical application and significance not only for head and neck cancers, but also for other tumor types.

MATERIALS AND METHODS:

Human subject research

All human samples were obtained from previously published collections in accordance with Institutional Review Board Guidelines. (2) No new human samples were collected for this study.

RNA-Seq data and processing

Primary RNA-Seq data from Johns Hopkins patient samples were retrieved, normalized, and analyzed as described in our previous publications. (2) ASE-junction expression was detected as RPM (reads per million), and WT-junction expression was detected as RSEM (RNA-Seq by expectation-maximization) using the available RNA-Seq data. (2, 14)

Cell lines and cell culture

Human HNSCC cell lines UM-SCC-22B and SCC61 were provided by Dr. Thomas Carey from the University of Michigan and Ralph Weichselbaum from the University of Chicago, respectively. These cells were used for functional evaluation of the ASE-GSN, due to their increased cellular mobility, needed for the completion of cell invasion assays. Each cell line was authenticated using a Short Tandem Repeat (STR) Identifier kit (Applied Biosystems). UM-SCC-22B cells were cultured in DMEM (Invitrogen) High Glucose supplemented with 10% Fetal Bovine Serum (Gemini) mixed with X1 Penicillin and Streptomycin (Corning). SCC61 cell line was cultured in DMEM/F12 (Invitrogen) supplemented with 10% Fetal Bovine Serum mixed with Penicillin and Streptomycin. All cultured cell growth occurred in a 5 % CO₂ incubator at 37° C. Cell lines enumerated in Supplemental Table 1 were analyzed for ASE-GSN and WT-GSN expression using available RNA from our previous publication. (20)

Transient transfections

Ectopic expression

The ectopic expression of human WT-GSN and ASE-GSN was performed using commercial (WT-GSN, RC214871) and custom (ASE-GSN) plasmids with mammalian cDNA expressing pCMV6-Entry Vector (OriGene). Since ASE-GSN has an early stop codon in frame encoded in the intronic sequence (Supplemental Fig 1), the ASE-GSN plasmid contains the GSN sequence before this stop codon. Plasmid transfections were performed in parallel with empty pCMV6-Entry Vectors (21) in Opti-MEM (Invitrogen) culture medium using X-tremeGENE 9 DNA Transfection Reagent (Roche) per manufacturer instructions. The transfection efficiency and the level of the ectopic gene and protein expression were monitored by qRT-PCR and Western Blotting, respectively.

RNA interference experiments

ON-TARGETplus Human GSN pool (L-007775–00-0005) of four siRNA targeting sequences CUGUUGAGGUAUUGCCUAA, GCUAAGCGGUACAUCGAGA, GCACUGAACUCCAACGAUG, and GAACGGAAAUCUGCAGUAU (Dharmacon) was used to downregulate the expression of WT-GSN. It was challenging to design siRNA for efficient downregulation of ASE-GSN. Out of four custom designed ON-TARGETplus Human ASE-GSN siRNA, two of these downregulated ASE-GSN expression. These siRNA were labeled siRNA1 (targeting sequence GACCAGAUCUCCAGGCACAUU) and siRNA2 (targeting sequence GGCAGGGGAUGGUGAAUGAUU) (Dharmacon). Transfection of pooled or single siRNAs was performed in Opti-MEM (Invitrogen) using RNAiMAX Lipofectamine Reagent (Invitrogen) in parallel with ON-TARGETplus Pool (D-001810–10) controls. The transfection efficiency and the level of the endogenous gene and protein expression were monitored by qRT-PCR and Western Blotting, respectively.

Cell proliferation

The cellular growth-monitoring experiments were performed in 96-well plates with five independent wells per experiment. The cells were incubated for one hour in 100 μL of 1:10 diluted Alamar Blue (Bio-Rad) (22) and Opti-MEM Media (Invitrogen). Fluorescence was observed by exciting at 530 nm and detecting emission at 590 nm on Spectramax M2.

Invasion Assay

Matrigel invasion assays (23) were performed in 6.5mm Transwell® with 8.0μm Pore Polycarbonate Membrane Inserts (Corning) covered by 100 μL of diluted Matrigel (Corning) in serum-free DMEM High Glucose medium (Invitrogen) set up in 37° C overnight. Before plating cells on the top of solidified Matrigel, cells were washed three times with cold, serum-free DMEM High Glucose medium (Invitrogen). Cells were chemo-attracted to invade through the Matrigel and the membrane toward DMEM High Glucose medium (Invitrogen) supplemented with 10% Fetal Bovine Serum (Gemini) plated in the outer chamber. After 48 hours at 37° C, Transwell plates were fixed in 4% formaldehyde, stained with 0.5% crystal violet, and washed in DI water. Stained plates were left to air dry at room temperature for 24 hours and placed on microscopy slides. Slides were photographed and analyzed using ImageJ (https://imagej.net/) software using the Threshold and Measure tools to match covered area visually.

Quantitative Real-Time PCR (qRT-PCR) analysis of RNA

RNA extractions and purifications were performed using Qiazol and RNeasy Kit (Qiagen). Reverse transcription was performed with MultiScribe™ Reverse Transcription kit (Invitrogen). RNA Expression was determined using Taqman quantitative real-time PCR using 0.6% Platinum® Taq DNA Polymerase (Invitrogen), 2% ROX Reference Dye (Invitrogen), 0.2 mM of dNTPs (our lab), 0.6 μM of each primer, and 0.33 μM of probe per reaction on 25 ng/μL of DNA template with the following primer-probe sets: Forward Primer GGAAGCCCATGATCATCTACAA, Reverse Primer ACAAAGGCATCGTTGGAGT, and probe GCAATACCTCAACAGCCCGGGT were used for detection of WT-GSN. Forward Primer CTGAAACCTCCCAGCTCAAT, Reverse Primer ACAAAGGCATCGTTGGAGT, and probe GCAATACCCCGTCATTCACCAT were used for detection of endogenous ASE-GSN. Forward Primer CCTGTCCAGAGCCGTGT, Reverse Primer ATTGAGCTGGGAGGTTTCAG, and probe AGGACCTGGAAATTACCTCAACAGCCC were used for detection of the ectopic ASE-GSN plasmid. All assays were quantified in triplicate against a GAPDH control 20X Gene Expression Assay (Invitrogen) using the 2^-ΔΔCT method. (24–26)

Western Blotting

Protein extraction was performed with 10% RIPA Lysis Buffer (Upstate) with Complete Mini (Roche) and PhosphStop (Roche) inhibitors. Proteins were developed in Pierce 660 nm Protein Assay (Thermo Scientific, 22660) and quantified by absorption at 660 nm Spectramax M2 against bovine serum albumin (Sigma-Aldrich) standard curve. SDS-PAGE was performed on Criterion XT Precast Gel (Bio-Rad) per manufacturer’s instructions. Blotting performed with Gelsolin Rabbit mAb (D9W8Y, Cell Signaling Technology [CST]), and GAPDH XP® Rabbit mAb (D16H11, CST) primary antibodies. Secondary antibody Anti-rabbit IgG, HRP-linked Antibody (7074S, CST) was used with Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare). Notably, D9W8Y Gelsolin antibody bound to N-terminus of the protein and almost exclusively detected full-length WT-GSN due to its abundant expression relative to other GSN isoforms. Since the endogenous expression of all other GSN isoforms, including ASE-GSN, is much lower, they were not detected using Western Blotting. D9W8Y Gelsolin antibody detected ectopically expressed ASE-GSN in knock-in experiments. There is no other high-quality GSN antibody on the market. The Western Blotting images were quantified with ImageJ (https://imagej.net/) software, by creating density histograms and measuring the area under the curve with the magic wand tool. Antibody binding to GSN was calculated relative to GAPDH.

Statistical Analysis

Outlier Gene Set Analysis - OGSA

OGSA p-value was calculated using a Fisher exact test comparing the number of outliers occurring in tumors with normal as it was previously done in Guo et al (2)

Student t-test

Student t-test was used to compare cells with and without treatment for experiments performed in triplicates or pentaplicates. P-values below 0.05 were considered significant.

RESULTS:

High rate of GSN ASEs in primary HNSCC tumor samples

Our recent analysis of ASE distribution in HNSCC samples using RNA-Seq data revealed 109 ASE junctions with high specificity to tumor samples. (2) Among these tumorspecific splice variants, a recurring splice junction was identified within the GSN gene at a noncanonical chr9: 124089070−124089586 junction (Fig. 1). Of the 46 HNSCC samples, 18 (39%) were outliers with respect to normalized expression of the GSN alternative splice junctions and no outliers were found in the 25 normal control samples (OGSA p=0.0015, Fig 1A). The preferential expression of ASE-GSN in HNSCC tumor samples was previously reported for the TCGA (The Cancer Genome Atlas) cohort. (2, 15) In opposite, the average WT-GSN expression in tumor and normal samples was not significantly different by OGSA (Fig 1B).

Figure 1. Discovery of ASE-GSN Expression in HNSCC samples.

(A) The box plot of the RPM expression of the alternative GSN junction (chr9:124089070 − 124089586) within the discovery cohort of 25 normal samples and 46 HNSCC samples. (2) The samples further illustrated in panel C are identified by triangle shape. Out of the 46 discovery cohort HNSCC samples, 18 outliers were discovered among normalized GSN junction expression in tumors, and no outliers were found in normal samples determined by OSGA, indicating significant outlier expression in cancerous phenotype (p=0.0015). (B) The boxplot of the logarithmic expression of the total GSN RSEM gene shows that total GSN expression is tumor and normals are not significantly different (OGSA p-value=1). The samples further illustrated in panel C are identified by triangle shape. (C) Integrative Genomic Viewer (IGV) visualization shows the alternative splicing event found between RefSeq Exon 14 and 15 of the GSN transcript, which produced the extended Exon 14 (red frame) expression in primary tumor samples, but not in normal controls. Notably, the tumor-specific extended Exon 14 has TGA stop codon in the frame, which causes the production of the truncated ASE-GSN protein. (D) The 3D crystal structure of GSN (29) with six GEL (G) domains, with G5 and G6 to be lost (red) due to the expression of the truncated ASE-GSN. (E) The InterPro© analysis of the ASE-GSN (top, n=613 aa, including n=33 unique aa coded from the intronic DNA, red frame) and WT-GSN (bottom, n=782 aa) peptides revealed that the WT-GSN protein contains two extra GEL domains (red) not expressed in the truncated ASE-GSN protein.

Molecular structure effects of GSN alternative splicing

The full-length WT-GSN is composed of 17 exons expressed from Chromosome 9 (Supplemental Fig 2). (27) The non-canonical ASE junction between 124089070−124089586 bp results in a 110 bp insertion between exons 14 and 15, found primarily in tumor samples (Fig 1C). This cancer-specific insert codes for 33 extra amino acids and contains a TGA stop codon after first 99 bp (Supplemental Fig 1), which leads to the expression of a truncated ASE-GSN protein. The full-length WT-GSN protein, which consists of 6 gelsolin domains G1, G2, G3, G4, G5, and G6 that fold to form the active tertiary structure of the protein, was visualized with InterPro© analysis.(28–30) (Fig 1D, bottom) WT-GSN folds upon itself by a Ca²⁺ dependent latch between G2 and G6, which locks the protein in a closed, inactive conformation. (29, 31, 32) (Fig 1D) The tumor-specific ASE-mRNA encodes only for the G1, G2, G3, and G4 gelsolin domains (Fig 1E, top), while losing the G5 and G6 subunits, where the latter is critical for the Ca²⁺ latch (Fig 1D).

Total GSN expression affects cell proliferation and viability

To determine the translational value of acquired data to the clinical application, the functional effects of GSN alternative splicing on cellular function were verified by complimentary in vitro techniques, such as proliferation and invasion assays. First, expression of ASE-GSN in HNSCC cell lines was cross-validated in 19 Head and Neck Tumor cell lines and three oral keratinocyte cell lines. (Supplemental Table 1) We have not detected any statistical significance by OGSA in either ASE-GSN or WT-GSN expression (Supplemental Fig 3). From this panel of HNSCC cell lines, UM-SCC-22B and SCC61 cells were chosen for functional evaluation due to their high invasive potential.

Ectopic knock-in expression of both WT-GSN and ASE-GSN isoforms in UM-SCC-22B (t-test p_WT=0.05 and p_ASE=0.0001, Supplemental Fig 4A and 4B, as well as 4C, 4D, and 4I) and SCC61 (p_WT=0.02 and p_ASE=0.00001, Supplemental Fig 4E and 4F, as well as 4G, 4H, and 4I) decreased proliferation of UM-SCC-22B (Fig 2A) and SCC61 (Fig 2B) cells over time. Conversely, siRNA downregulation of both WT-GSN and ASE-GSN in UM-SCC-22B (p_pool=0.005, p_si1=0.0002, and p_si2=0.04, Supplemental Fig 5A and 5B, as well as 5C and 5G) and SCC61 (p_pool=0.00001, p_si1=0.003, and p_si2=0.25, Supplemental Fig 5D and 5E, as well as 5F and 5G) increased cell proliferation over time in UM-SCC-22B (Fig 2C) and SCC61 (Fig 2D) cells. Notably, ASE-GSN siRNA2 did not work well for SCC61 (p=0.25, Supplemental Fig 5E) due to some undetected genomic or transcriptomic variability. ASE-GSN siRNA1 worked better for both UM-SCC-22B and SCC61 (Supplemental Fig 5B and 5E). These data suggest that overall levels of total GSN played a role in the regulation of cellular viability and proliferation related to carcinogenesis.

Figure 2. Cell proliferation in the presence and absence of ASE-GSN or WT-GSN expression.

The ectopic knock-in of ASE-GSN (dotted line) and WT-GSN (dashed line) isoform significantly decreased proliferation of UM-SCC-22B (A) and SCC61 (B) cells. The siRNA-mediated decrease of ASE-GSN (dotted line and dash-dot line) and WT-GSN (dashed line) isoforms significantly increased the proliferation of UM-SCC-22B (C) and SCC61 (D) cells. Error bars indicate normalized standard error calculated for each experiment. Significance determined by t-test comparison of 72-hour time-points normalize to starting emission between samples and controls.

ASE-GSN expression modulates the metastatic potential of cancer cells

Ectopic knock-in expression of ASE-GSN in UM-SCC-22B (p=0.0001, Supplemental Fig 4B, as well as 4I) and SCC61 (p=0.00001, Supplemental Fig 4F, as well as 4I) increased cellular invasion of UM-SCC-22B (p=0.0881, Fig 3A) and SCC61 cells (p=0.0460, Fig 3B). In contrast, ectopic expression of WT-GSN in UM-SCC-22B (p=0.05, Supplemental Fig 4A, as well as 4I) and SCC61 (p=0.02, Supplemental Fig 4E, as well as 4I) decreased cellular invasion of UM-SCC-22B (p=0.0258, Fig 3C) and SCC61 (p=0.33, Fig 3D) HNSCC cell lines. Conversely, siRNA downregulation of ASE-GSN expression in UM-SCC-22B (p_si1=0.0002 and p_si2=0.04, Supplemental Fig 5B) and SCC61 (p_si1=0.003 and p_si2=0.25, Supplemental Fig 5E) decreased cellular invasion of UM-SCC-22B (p_si1= 0.0232 and p_si2= 0.0092, Fig 4A) and SCC61 (p_si1= 0.0289 and p_si2= 0.5209, Fig 4B) cells. Notably, low efficiency of ASE-GSN siRNA2 in SCC61 (p=0.25, Supplemental Fig 5E) can explain non-significant invasion changes (p=0.5, Figure 4B) in this cell line. Conditional siRNA downregulation of WT-GSN expression in UM-SCC-22B (p_pool=0.005, Supplemental Fig 5A, as well as 5C and 5G) and SCC61 (p_pool=0.00001, Supplemental Fig 5D, as well as 5F and 5G), which was confirmed with strong concordance between RNA- and protein-based detection, increased cellular invasion of UMSCC-22B (p_pool= 0.0016, Fig 4C) and SCC61 (p_pool= 0.0021, Fig 4D). This data suggests that the effects of GSN on invasion may be isoform specific. Given the ability of GSN to sever actin filaments, it is possible that ASE-GSN may increase tumor invasion and metastatic potential.

Figure 3. Cell migration and invasion in the presence of ASE-GSN or WT-GSN expression.

The ectopic expression of ASE-GSN (white) increased the cellular invasion and migration of UM-SCC-22B (A, p=0.088) and SCC61 (B, p=0.046) cells. The ectopic expression of WT-GSN (gray) decreased cellular invasion and migration of UM-SCC-22B (C, p=0.026) and SCC61 (D, p=0.33) cells. Asterisks indicate p<0.05. Error bars indicate normalized standard error calculated for each experiment.

Figure 4. Cell migration and invasion in the absence of ASE-GSN or WT-GSN expression.

The siRNA-mediated decrease of ASE-GSN expression (white and striped bars) decreased the cellular invasion and migration of UM-SCC-22B (A, p=0.002, and p=0.009) and SCC61 (B, p=0.03, and p=0.5) cells. The siRNA-mediated decrease in WT-GSN expression significantly increased the cellular invasion and migration of UM-SCC-22B (C, p=0.0016) and SCC61 (D, p=0.002) cells.

DISCUSSION:

Prior literature has shown that GSN has multifunctional roles in actin remodeling, PI3K/mTOR/AKT signaling, and apoptosis. (16, 29, 31–33) Given the significance of these cellular functions to metastatic potential, and considering the 109 ASE-genes we previously found in HNSCC (2), we focused on the functional analysis of ASE-GSN. Prior research suggested a paradoxical role for GSN, in some instances increasing apoptosis and in others increasing metastasis. (34–38) The results of our study provide new evidence that may help clarify the influence of GSN on cellular survival, at least with respect to HNSCC. In this discussion, we propose a model that is consistent with our data and with prior research, which illustrates possible roles for both ASE-GSN and WT-GSN in metastatic and apoptotic processes.

Until now, the role of non-canonical GSN alternative splicing in tumor samples has remained uncharacterized. Literature does suggest the expression of other canonical GSN splice sites that result in translation of 19 isoforms of GSN, (27) but their correlation with disease status was not evaluated in our paper because they were not detected in our initial discovery cohort. The existence of differential expression of GSN suggests that GSN isoforms may play a role in regulating oncogenic behaviors, such as apoptosis and proliferation. The overall expression of both ASE-GSN and full-length WT-GSN isoforms have an inverse relationship with cell viability and proliferation (Fig 2), which is consistent with available literature that documents the role of GSN in apoptotic pathways. (31) GSN can either increase or decrease apoptosis, depending on the cellular conditions and the specific tissue or cell type in which it is expressed (Fig 5). Caspase-3 cleaves GSN between the 352^nd and 353^rd residues, resulting in C-GSN and N-GSN (with three gelsolin subunits each). N-GSN protein can cleave actin-independent of Ca²⁺, which induces apoptotic morphological changes within the cell. (17)

Figure 5. The proposed mechanism for the role of GSN isoform expression in apoptosis and metastasis.

Top Left: The full-length WT-GSN isoform contains a Ca²⁺ latch, which permits binding of WT-GSN to actin fiber only in the presence of Ca²⁺. Binding of WT-GSN is noncompetitive with DNase I enzyme and does not induce apoptosis, but allows for GSN-mediated actin severing necessary for invasion. Top Right: In the absence of Ca²⁺ the full-length WT-GSN isoform is unable to bind actin. DNase I remains bound to actin and thus apoptosis is not induced. Actin is not severed, and thus invasion is not induced (relative to data shown in Figures 3C, 3D, 4C, and 4D). Bottom Left: The ASE-GSN isoform does not have a Ca²⁺ latch due to its truncated structure. Therefore, it can bind actin-independent of Ca²⁺ ions. As with WTGSN, binding of ASE-GSN is non-competitive with DNase I. Thus, DNase I remains bound to actin and apoptosis is not induced. In this model, binding of ASE-GSN causes actin severing, allowing for cell invasion (relative to data shown in Figures 3A, 3B, 4A, and 4B). Middle: Both WT-GSN and ASE-GSN are cleaved by Caspase-3, resulting in an N-GSN (subunits 1–3) fragment and a C-GSN (subunits 4–6) fragment. N-GSN binds actin in a Ca²⁺ independent manner competitive with DNase I, resulting in faster cleavage of the actin cytoskeleton and increased DNA degradation, stimulating apoptosis (relative to data shown in Figure 2). Furthermore, C-GSN is associated with invasion prevention, but C-GSN is only present in WT-GSN cleavage. Thus, WT-GSN offers protection from invasion, not provided by ASE-GSN (relative to data shown in Figures 3 and 4). (16, 17, 28, 29, 31, 32)

Full-length GSN binds to actin non-competitively with DNase I, but N-GSN binds competitively, resulting in the increased cellular valence of DNase I, leading to enhanced apoptotic activity. (31) In our study, ectopic knock-in expression of either WT-GSN or ASE-GSN led to reduced cell proliferation over time. This observation suggests that the overall level of total GSN (WT or ASE) influences cell survival. One possibility, consistent with these observations, is that the overexpression of cellular GSN (ASE or WT) levels results in higher rates of cleavage of GSN to N-GSN, resulting in higher downstream levels of active DNase I, inducing Ca-independent apoptosis (Fig 5 middle). Invasion of cells through the basement membrane, an important hallmark of cancer cells, was increased with higher expression of ASEGSN (Fig 3A and 3B) and decreased with reduced expression of ASE-GSN (Fig 4A and 4B). Furthermore, increased expression of WT-GSN lowered invasion (Fig 3C and 3D) of cells and decreased expression of WT-GSN promotes invasion (Fig 4C and 4D). Given the important role of GSN as actin-regulating protein, any slight changes in localization or activity of GSN would significantly affect cell migration and metastatic potential. Indeed, current literature supports our observations; it has been seen that the Ca²⁺ dependency of GSN is largely dependent on the C-terminus and that the metastatic suppressor qualities of GSN are determined by the presence of this C-terminus. (32) The tertiary structure of GSN proposed by (28) and (31) contains a “latch” between the G2 and G6 subunits that is dependent upon Ca²⁺ for release and activation of GSN actin binding (Fig 5 top). The G1, G2, and G3 subunits necessary for actin severing are still preserved in ASE-GSN. (16, 30, 31)

We propose a model in which ASE-GSN does not contain this latch and is thus able to bind to actin in a Ca²⁺ independent manner (Fig 5 bottom) while maintaining non-competitive binding with DNase I, unbalancing pathways critical in normal cell physiology. (17, 29) The difference in ability to unlatch could result in equal effects of ASE-GSN and WT-GSN upon cell proliferation and survival, but cause increased actin severing and metastasis when there is a high ASE/WT GSN ratio (Fig 5).

Although various sources have described the mechanism by which GSN binds actin, GSN has a seemingly paradoxical relationship with the PI3K/AKT/mTOR pathway. GSN has been previously shown to dysregulate the PI3K/AKT/mTOR pathway, bind actin monomers in a Ca²⁺-dependent manner, and cause severing of actin fibers. (16, 29, 32) A study by Ma et al. demonstrated that GSN upregulated pAKT, promoting survival in osteosarcoma, but another study by Yeh et al. demonstrated that GSN downregulated pAKT, causing death in hypoxic cardiomyoblast cells. (39, 40) Our data establish that alternative splicing of the GSN gene has potential to resolve this apparent contradiction.

In surgically resected non-small cell lung cancer, a loss of GSN expression has been observed by IHC. (34) In contrast, GSN expression in cervical carcinoma has been reported to be upregulated relative to the surrounding tissue. (35) Similarly, GSN expression in hepatocellular carcinoma has been found to be overexpressed relative to adjacent tissue. (36) An in vitro study by Zhuo et al. found that expression of GSN in colorectal cancer cells was necessary for invasion through matrigel. (37) However, overexpression of GSN in the NK lymphoma cell line YTS has been reported to increase apoptosis and decrease cell proliferation and invasion. (38)

Considering the inconsistencies in the prior literature, our observations support a model in which ASE-GSN promotes invasion, whereas WT-GSN promotes either apoptosis or invasion, dependent on overall levels of GSN expression. However, from our observations in this study, we cannot be certain that ASE-GSN promotes metastasis in vivo. Alternative explanations are possible for the upregulation of ASE-GSN that we observed in HNSCC. For example, the existence of ASE-GSN might be in some way related to alterations in tensile stress acting upon cell membranes within the tumor microenvironment (41). Future research could explore, through an in vivo xenograft model, whether ASE-GSN indeed promotes metastasis.

Additional study is also needed to elucidate the link between GSN isoforms and AK-Tregulated apoptosis, migration, and proliferation. Our proposed model requires further investigation into the effects of Ca²⁺ on isoform-specific actin-severing activity. Future studies of GSN should include analysis of the effects of calcium availability on downstream isoform-specific regulation. Further investigation is necessary to elucidate whether GSN effects on the PI3K/AKT/mTOR pathway are playing a role in GSN-mediated cell death. Prospectively, it would be interesting to identify different isoforms of GSN, using modern long-range RNA-Seq and computational technologies, such as CLASS2 (42, 43) or SUPPA (44). Their role in clinical outcome can be further evaluated in a larger clinical cohort.

Validation of alternative splicing in oncogenesis could improve both diagnosis and treatment of HNSCC. Over the past 30 years, there has been limited change in the treatment protocol for HNSCC. A better understanding of alternative splicing in cancer may open new avenues for future disease treatment such as generating new cancer-specific immunotherapeutics targeted against alternatively spliced proteins, neoantigens for immunotherapeutics, or biomarkers for detection and prognostication of disease. Specifically, if ASE-GSN protein were shown to be carcinogenic, monoclonal antibody drugs could be developed to target the unique epitope of ASE-GSN isoform. Due to the unique sequence of ASE-GSN predominately expressed in the tumor (outlier detection rates 39%), but not in normal samples (specificity 100%), this ASE-GSN junction may be used for detection purposes. Moreover, ASE-GSN or other cancer-specific ASEs can be detected by IHC analysis of pathology samples during cancer diagnostics. Currently, Human papillomavirus (HPV) in situ hybridization and immunostaining of surrogate HPV+ marker protein p16 are the only clinical marker of HNSCC. (45) Our data suggest that, because ASE outliers are found in 39% of cases, ASE might have 39% sensitivity with ultimate near 100% specificity for detection of cancerous tissues. This observation suggests that ASEs could help bolster specificity during HNSCC testing. Furthermore, using this method of detection would identify specific subsets of cancers with distinct biology, which may facilitate specifically targeted treatment regimens. This work will build a basis for the development of ASE-specific cancer biomarkers, which can potentially be used for primary and secondary detection and cancer surveillance using biopsy and biofluids, directly characterizing tumors without reliance on surrogate protein techniques.

TRANSLATIONAL SIGNIFICANCE.

We have recently detected the landscape of alternative splicing events in head and neck squamous cell carcinoma. In this work, we prove the oncogenic relevance of alternative splicing in Gelsolin through detailed functional validation. This data has direct clinical applications, where ASE-GSN can be used as a biomarker and a therapeutic target for HNSCC.