Syntaxin binding protein 4 as a novel therapeutic target and an effective biomarker for better treatment selection in lung squamous cell carcinoma: A theranostic study


 Background

Lung squamous cell carcinoma (LSCC) remains a challenging disease to treat, and further improvements in prognosis are dependent upon the identification of LSCC-specific therapeutic biomarkers and/or targets. We previously found that Syntaxin Binding Protein 4 (STXBP4) plays a crucial role in lesion growth and, therefore, clinical outcomes in LSCC patients through regulation of tumor protein p63 (TP63) ubiquitination.
Methods

To clarify the potential of STXBP4 as a novel therapeutic target and/or a predictive biomarker of therapeutic response, we assessed its prognostic impact of the protein in 144 LSCC patients and examined whether its action pathway is distinct from those of currently used drugs in in vitro experiments including RNA-seq analysis through comparison with the other putative exploratory targets and/or markers.
Results

Kaplan–Meier analysis revealed that, along with ΔNp63 (an isoform of TP63) and vascular endothelial growth factor receptor 2 (VEGFR2), STXBP4 expression signified a worse prognosis in LSCC patients, both in terms of overall survival (OS, P = 0.002) and disease-free survival (DFS, P = 0.041). Multivariate analysis demonstrated STXBP4 to be an independent prognostic factor for poor DFS, while VEGFR2 (P < 0.001) and TP63 (ΔNp63, P < 0.05) were independent prognostic factors for poor OS. The action pathway of STXBP4 on suppression of TP63 (ΔNp63), which results in the and inhibits tumor growth, was unique: Ingenuity pathway analysis using the knowledge database and our RNA-seq analysis in human LSCC cell lines indicated that 35 pathways were activated or inactivated in association with STXBP4, but the action pathway of STXBP4 was distinct from those of other current drug targets: STXBP4, TP63 and KDR (VEGFR2 gene) formed a cluster independent from other target genes of tumor protein p53 (TP53), tubulin beta 3 (TUBB3), stathmin 1 (STMN1) and cluster of differentiation 274 (CD274: programmed cell death 1 ligand 1, PD-L1). STXBP4 itself appeared not to be a potent predictive marker of individual drug response, but we found that TP63, main action target of STXBP4, might be involved in drug resistance mechanisms of LSCC.
Conclusion

STXBP4 could afford a unique therapeutic target and a key to the development of precision medicine for LSCC patients.


Conclusion
STXBP4 could afford a unique therapeutic target and a key to the development of precision medicine for LSCC patients.

Background
Despite recent advances in therapeutics, lung squamous cell carcinoma (LSCC) remains a challenging disease to treat (1)(2)(3)(4). The advent of immunecheckpoint inhibitors along with several active target agents such as anti-angiogenic agents has altered LSCC treatment to some extent, but treatment options remain limited. The intractable patient characteristics at diagnosis; i.e., high rates of advanced stage, older age, and comorbidities, also remain a problematic issue in terms of treatment decision-making. To date, very few druggable mutations and active predictive biomarkers have been identi ed; thus, no LSCC-speci c target therapy has yet been established. The development of precision medicine with truly active target drugs is eagerly awaited (5)(6)(7)(8)(9).
We recently found that Syntaxin Binding Protein 4 (STXBP4) plays a crucial role in LSCC growth through regulation of ΔNp63 (an isoform of tumor protein 63, TP63) ubiquitination and is an independent prognostic factor signifying a worse outcome in LSCC patients (10,11). ΔNp63 is an isoform of TP63, a member of the TP53 family, and its expression is widely used as a highly speci c diagnostic marker for LSCC. STXBP4 binds to ΔNp63 and suppresses the anaphase-promoting complex/cyclosome (APC/C) complex-mediated proteolysis of ΔNp63. STXBP4 drives the oncogenic potential of ΔNp63α and may be a useful therapeutic target and/or marker for patients with LSCC.
These ndings encouraged us to clarify the potential in clinical application of STXBP4. In this study, we assessed whether STXBP4 is truly and signi cantly related to patient outcome and whether it can afford a unique druggable target through comparison with the other powerful prognostic biomarkers and molecular action networks of other key agents in LSCC treatment. Despite a lack of de nitive prognostic markers, we selected VEGFR2 (vascular endothelial growth factor receptor 2), TUBB3 (tubulin beta 3), and PD-L1 (programmed cell death 1 ligand 1), along with p53 (tumor protein p53), ΔNp63 and STMN1 (stathmin 1), as other putative exploratory markers. Their response to drugs strongly affects the prognosis of each patient. At present, taxane, antiangiogenesis inhibitors and immuno-checkpoint inhibitors are regarded as essential in the treatment of LSCC, the drug targets of which are TUBB3, VEGFR2, and PD-L1, respectively. Needless to say, the TP53 gene is a key factor in tumorigenesis and tumor resistance to therapy in lung cancer (5)(6)(7)(8)(9), and ΔNp63 is a putative diagnostic marker for LSCC (12). STMN1 (oncoprotein 18 and LAP18) has been suggested to be a potent predictive marker for a variety of cancers including LSCC (13)(14)(15)(16).
We further performed a genome-wide transcriptome analysis (RNA-seq) using next-generation sequencing (NGS) in 2 human LSCC cell lines, totally drugsensitive and -resistant cells, before and after treatment with key drugs, and assessed the modulation of each exploratory target to clarify its functional molecular network.
The percentage of STMN1 and TUBB3 staining was scored as follows: 1, ≤ 10%; 2, 11%-25%; 3, 26%-50%; and 4, ≥ 50%. The expression of VEGFR2 was considered positive only if distinct membrane staining was present, and was scored in the same manner as that used for STMN1 and TUBB3. For PD-L1, immunohistochemical staining was scored as 1, < 1%; 2, 1-5%; 3, 6-10%; 4, 11-25%; 5, 26-50%; and 6, > 50% of cells were positive. The tumors in which stained cancer cells were scored above 3 were de ned as demonstrating high expression, with those scored 1 and 2 de ned as demonstrating low expression. P53 microscopic examination of the nuclear reaction product was also undertaken and scored. P53 expression in > 10% of tumor cells was de ned as positive expression. The sections were evaluated under a light microscope in a blinded fashion by at least two of the authors.
Genome-wide transcriptome analysis (RNA-seq) Total RNA was prepared from cell lines LK-2 and RERF-LC-AI using NucleoSpin® RNA (Takara Bio Inc., Kusatsu, Shiga, Japan). The quality of the RNA was assessed by RNA integrity number (RIN) using the Agilent RNA6000 Pico Kit and the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). High-quality RNA samples alone (RNA integrity numbers > 7.0) were used for genome-wide transcriptome analysis (RNA-seq experiments). Library preparation was performed using the TruSeq Standard mRNA Sample Prep Kit (Illumina, San Diego, CA, USA) from 1 µg of total RNA, according to the manufacturer's protocol. The resulting libraries were subjected to paired-end sequencing using a NextSeq500 High Output v2 Kit and the Illumina NextSeq 500 system (43-base paired-end reads; Illumina). Data processing and analyses were performed using STAR v2.5.2b on the BaseSpace Sequencing Hub (Illumina). Brie y, reads were ltered, trimmed, and aligned against the UCSC human reference genome 19 (hg19) using a STAR pipeline. Normalization and differentially expressed genes were detected with TCC (Sun et al., BMC Bioinformatics, 2013) package of R software (R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/). Genes with a false-discovery rate (FDR)-adjusted p-value < 0.05 were de ned as being signi cantly modulated genes in LK-2 and RERF-LC-A1 cells. The networks and canonical pathways were generated through the use of IPA (QIAGEN Inc., https://www.qiagenbio-informatics.com/products/ingenuity-pathway-analysis).

Statistical analysis
Probability values (P value) < 0.05 indicated a statistically signi cant difference. The Fisher exact test was used to examine the association between two categorical variables. The correlation between drug sensitivity and gene expression value was analyzed using the parametric Pearson's product-moment correlation analysis. The correlation among target gene modulation and other modulations was analyzed using linear regression analysis. Follow-up for the 144 patients was conducted by reference to the patient medical records. The Kaplan-Meier method was used to estimate survival as a function of time, and differences in survival were analyzed by the Cox proportional hazards model. Multivariate analyses were performed using a "survival" package in R software (Cox proportional hazards model to identify independent prognostic factors: R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/). Hierarchical clustering was performed by "hclust" from the stats package in R software. The day of surgery was de ned as day 0 for measuring postoperative survival. OS was determined as the time from tumor resection to death from any cause. DFS was de ned as the time between tumor resection and rst disease progression or death. Statistical analysis was performed using R software.
A large-scale public database, The Cancer Genome Atlas (TCGA), was used to obtain data sets, for both gene expression and survival outcome, in 474 primary LSCC patients. Kaplan-Meier analysis of OS and relapse-free survival (RFS) using these data showed that TUBB3 expression alone was correlated with RFS when patients were tentatively classi ed into positive-and negative-expression groups according to the expression level in each tumor (cut off set as the median, p=0.001) [An additional table le shows this in more details (See Additional le 1)]. Despite the lack of statistical signi cance, the analysis also suggested some prognostic impact of 6 molecules except TP53; i.e., TP63 (P=0.072), TUBB3 (P=0.091), and STMN1 (P=0.052) in OS, and STXBP4 (P=0.076), KDR (VEGFR2, P=0.071), STMN1 (P=0.089) and CD274 (PD-L1, P=0.065) in RFS.
As a single layer of "omics" can only provide limited insights into biological signi cance, we performed immuno-histochemical analysis to elucidate the relevance of these 7 exploratory targets to patient outcome ( Figure 1). A total of 144 patients were enrolled in this study ( Table 1). None of the patients received any cancer treatment before the operation and the majority of patients were former or current smokers (97.9%).
Multivariate analyses indicated that the positive expression of VEGFR2 (P<0.001) and ΔNp63 (P<0.05) were independent prognostic factors for poor OS, with STXBP4 (P<0.001) alone being an independent prognostic factor for poor DFS in LSCC patients ( Table 2).

STXBP4 as a novel therapeutic target
The observed prognostic impact of STXBP4 and ΔNp63 con rmed previous results (10)(11)(12). We also found close relationships between VEGFR2, TUBB3, and STMN1 to patient outcome, which indicated the existence of some biological interactions between STXBP4 and these molecules. Ingenuity pathway analysis (IPA) using the knowledge database demonstrated that STXBP4 acts as an up-stream regulator of TP63 (ΔNp63) and subsequently of KDR (VEGFR2) via TP63, but the action pathway of STXBP4 was independent from those of the other 4 exploratory targets ( Figure 3A, B).
To con rm this, we performed in vitro experiments using human LSCC cell lines. According to the half-maximal inhibitory concentration (IC 50 ) published on the Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org), we rst chose 4 cell lines (LK-2, EBC-1, NCI-H520, and RERF-LC-AI), and then selected 2 cell lines as totally drug-sensitive (LK-2) and -resistant cells (RERF-LC-AI). The selection was based on a CCK8 assay to con rm the cellular sensitivities to cisplatin (CDDP), 5-uorouracil Exposure of cells to a drug causes a dynamic alteration in gene expression, and RNA-seq analysis following such drug treatment enables us to identify all the genes modulated together in response to the drug. VEGFR2 and TUBB3 are the drug action targets of Ramucirumab and TXT, respectively, and STMN1 has been suggested to be a marker of tumor resistance to taxanes (13)(14)(15)(16). LK-2 and RERF-LC-AI cells were treated with or without TXT and Ramucirumab in single and combination treatment settings, and then subjected to RNA-seq analysis. We selected genes highly correlated in terms of expression level with each target gene, and then performed hierarchical clustering of canonical pathways.
The analysis showed that STXBP4, TP63 and KDR (VEGFR2) formed a cluster independent from the other target genes [TP53, TUBB3, STMN1 and CD274 (PD-L1)], which was in accord with the ndings obtained in our previous studies (Figure 4) [An additional table le shows this in more details (See Additional le 4)] (10). Thirty-ve pathways were extracted as signi cantly (activation z-score >=2) activated or inactivated pathways in correlation with STXBP4.
Among them, the EIF2 signaling pathway, which plays a critical role in stress-related signals to regulate both global and speci c mRNA translation, was the most signi cantly activated [An additional table le shows this in more details (See Additional le 5)].
The action pathway of STXBP4 is distinct from those of other conventional drugs such as TXT and immuno-checkpoint inhibitors. The pathway is thought to suppress 2 prominent determinants of poor prognosis in LSCC patients, TP63 and VEGFR2, and possibly p53 as well.

STXBP4 as a possible predictive biomarker of individual therapeutic response
The observed correlations between STXBP4, ΔNp63, and VEGFR2 and clinical outcome, particularly the close correlation between STXBP4 and DFS, suggested that STXBP4 expression might afford a powerful predictive biomarker of individual response to current therapy. This hypothesis, however, cannot be directly veri ed due to the insu cient number of available coupled data related to clinical response and omics pro ling, even when a large-scale public clinical and genomic database was used.
Our in vitro experiments clari ed the relevance of each exploratory target to drug response at least in part. RNA-seq analysis revealed that CD274 (PD-L1) expression alone was signi cantly higher in the totally drug-resistant RERF-LC-AI cells as the base line [An additional table le shows this in more details (See Additional le 2)]. In the drug sensitive LK-2 cells, none of the drug treatments caused any signi cant changes in the expression levels of the 7 targets examined ( Table 3). In the resistant RERF-LC-AI cells, however, all of the drug treatments, single TXT, single Ramucirumab, and their combination, yielded a signi cant up-regulation in TP63 (representing ΔNp63) and a remarkable down-regulation in CD274. Ramucirumab also signi cantly increased STMN1 expression in the resistant cells. No changes in the expression levels of STXBP4, KDR (VEGFR2), or TUBB3 were observed, regardless of the cell lines and drug treatments examined.
These ndings suggested that the high-level expression of CD274 (PD-L1) is related to cellular drug resistance, at least in part, but could be partially downregulated by TXT and/or Ramucirumab. TP63 (ΔNp63) induction might be involved in the cellular resistance mechanisms of LSCC to TXT and/or Ramucirumab treatment, and the up-regulation of STMN1 could also participate in Ramucirumab resistance. These ndings may afford some help in the development of precision medicine for LSCC patients, with the optimal treatment for individual LSCC patients selected through expression analysis of CD274, TP63, and STMN1. STXBP4 is a potent prognostic marker in LSCC patients but not a powerful predictive marker of individual response to widely used current therapeutic drugs.

Discussion
Despite the advent of new treatment options, advanced and metastatic LSCCs remain di cult-to-treat malignancies. Extensive work is underway to expand the treatment options. Among the work in progress, druggable targets speci c to the disease and biomarkers for optimal treatment selection have been intensively researched to develop precision medicine with truly active target drugs (1)(2)(3)(4)(5)(6)(7)(8)(9)19). We have been involved in these researches and identi ed STXBP4 as a potent therapeutic target in LSCC by elucidating its biological function in the malignancy (9,10).
In this study, we con rmed the clinical prognostic impact of STXBP4 in 144 LSCC patients and rst suggested that TP63 (ΔNp63) induction might be involved in the cellular resistance mechanisms of the widely used current key drugs CDDP, TXT, and Ramucirumab. As STXBP4 has been shown to act as an up-stream regulator of TP63 (ΔNp63) (10,11), reduction in TP63 (ΔNp63) expression by STXBP4 might ameliorate tumor resistance to the current drug treatments. Additionally, IPA indicated that the action pathway of STXBP4 was independent from those of the other 4 targets examined in this study, with STXBP4, TP63 and KDR (VEGFR2) found to form a cluster independent from the other genes, TP53, TUBB3, STMN1 and CD274 (PD-L1).
LSCC is closely associated with smoking, which causes frequent genomic alterations and makes it cumbersome to identify de nitive therapeutic targets.
Several potent genomic alterations related to LSCC have been identi ed in a variety of genomic studies (5)(6)(7)(8)(9). Based on the hypothesis that squamous cell carcinoma may possess common biologic/molecular alterations regardless of the organ of origin, Schwaederle et al. proposed a new concept "squamousness" and identi ed frequent mutations in TP53 (64.5% of analyzed patients), PIK3CA (28.5%), CDKN2A (24.4%), SOX2 (17.7%), and CCND1 (15.8%) (8). Friederlaender et al. summarized recent NGS studies and listed ERBB mutations, FGFR1 ampli cations and PI3K missense mutations and ampli cations as further potentially actionable targets, and suggested that several key molecules involved in their related pathways such as PDGFR could be a therapeutic target in LSCC (9). Despite these encouraging ndings, very few druggable mutations have been identi ed.
The inhibition of STXBP4 results in the suppression of TP63 (ΔNp63), a p53 family protein, and inhibits tumor growth; however, the action pathway differs from those of current key agents (10,11). This study revealed that, among 35 pathways activated or inactivated is association with STXBP4, the EIF2 signaling pathway was the most signi cantly activated. eIF2β, a subunit of the heterotrimeric G protein EIF2 that functions as a transcription initiation factor, was recently reported to play a critical role in stress-related signals to regulate both global and speci c mRNA translation, and is highly up-regulated in lung cancer specimens on multi-omics levels (DNA, RNA, and protein) (18). STXBP4 could, therefore, represent an unprecedented and unique therapeutic target to improve LSCC treatment.
STXBP4 was shown to be a potent prognostic marker in LSCC patients but not to be a predictive marker of individual response to widely used current therapeutic drugs. Even so, this study may contribute to the development of better therapeutic strategies using current key drugs. Despite the limited in vitro data, our RNA-seq analysis showed that the high-level expression of CD274 (PD-L1) might be related to cellular drug resistance in LSCCs. The increased CD274 expression could be partially down-regulated by TXT and/or Ramucirumab treatments, but the up-regulation of STMN1 possibly participates in resistance to TXT and Ramucirumab. TP63 (ΔNp63) induction might also be involved in the cellular resistance mechanisms of LSCC to CDDP, TXT and Ramucirumab therapies. These ndings suggest that the optimal treatment for individual LSCC patients could be selected through expression analysis of CD274 (for immune-checkpoint inhibitor), STMN1 (for TXT and Ramucirumab), and TP63 (for all treatment failures including platinum agents). Along with that of STXBP4, the expression levels of VEGFR2 (p = 0.007) and TUBB3 (p = 0.077) were closely and potentially connected with DFS. For the selection of TXT and Ramucirumab, additional expression analysis of each target molecule; TUBB3 and KDR (VEGFR2), respectively, would be helpful.
These ndings are partially validated in clinical practice, although de nitive predictive markers for CDDP, taxane, antiangiogenetic inhibitors and immunecheckpoint inhibitors remain controversial. The use of the immune-checkpoint inhibitor (Pembrolizumab) for metastatic LSCC patients with tumors showing 50% or greater PD-L1 (CD274) is now widely recognized as a standard rst-line therapy (1)(2)(3)20). The putative predictive markers of Ramucirumab-based regimens remain unclear (21), but high-level STMN1 expression was demonstrated as a potent determinant of chemo-resistance and, thus, a poor prognosis in LSCC patients (13). The AKT/FOXM1/STMN1 pathway was indicated to drive resistance to tyrosine kinase inhibitors in advanced non-small cell lung cancer including LSCC (22). TP63 is an action target of STXBP4 (10)(11)(12). The development of STXBP4 inhibitors is considered to be key to the development of precision medicine with truly active target drugs for LSCC patients.

Conclusions
Herein, we demonstrated that STXBP4 could afford an unprecedented and unique therapeutic target to improve LSCC treatment. The development of STXBP4 inhibitors would not only expand treatment options but also lead to precision medicine guided by expression analysis of several key genes such as CD274 (for immune-checkpoint inhibitor), STMN1 (for TXT and Ramucirumab), and TP63 (for all treatment failures including platinum agents) in LSCC patients.

Consent to publication
Not applicable.

Availability of data and materials
The data of this study were derived from the The Cancer Genome Atlas (TCGA) and ArrayExpress, which were available respectively from: https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga and https://www.ebi.ac.uk/arrayexpress//experiments/E-MTAB-2706/. The datasets used and analysed during the current study are available from the corresponding author on reasonable request.    Figure 1 Representative immunohistochemical staining of STXBP4, TP63 (ΔNp63), p53, VEGFR2, TBB3, STMN1, and PD-L1. A total of 144 LSCC samples (formalinfixed and paraffin-embedded sections) were stained immunohistochemically (x200, scale bar 200µm), and classi ed into positive-and negative-expression groups according to the expression score evaluated by a semi-quantitative method as described in "Patients and Methods".