Comprehensive Analysis of Expression, Prognosis and Immune Inltrates for Kinesin Superfamily Members in Human Pancreatic Adenocarcinoma

Kinesin superfamily (KIFs) has a long-reported signicant inuence on the initiation, development, and progress of pancreatic adenocarcinoma (PAAD). However, the expression level of different KIFs in PAAD and its relationship with the prognosis and immune inltration in patients with PAAD have not been fully elucidated. Comprehensive bioinformatics analyses were done using data from UCSC XENA data hubs, TCGA database(cid:0)Oncomine databases, GEO datasets, GEPIA, GTEx, The human protein atlas, Kaplan Meier plotter, cBioPortal, STRING and KEGG database. Then, the relationship between KIFs expression and tumor immune inltrates was studied by using the TIMER database.

PAAD is a kind of extremely malignant tumor of the digestive system that is di cult to diagnose and treat [2]. Due to the di culty of early diagnosis, high surgical mortality, poor treatment outcome, the 5-year survival rate of PAAD patients is only 5% [3]. Although PAAD treatment has been signi cantly improved in recent decades, most patients are diagnosed with cancer only in the middle and late stages because the early symptoms of PAAD patients are not obvious, and there is a lack of biomarkers for early diagnosis. It is not effective for surgical resection and ultimately leads to high mortality in patients with PAAD [4].
Therefore, it is urgent to nd effective biomarkers for early diagnosis and prognosis of PAAD and develop new strategies for targeted therapies for PAAD.
Kinesin superfamily proteins (KIFs) were rst identi ed by Vale et al. in 1985[5]. KIFs are a group of proteins featured to be microtubule-based motors and functioned as intracellular transporters that directionally transport various cargos, including organelles, protein complexes and mRNAs, along microtubules in an adenosine triphosphate (ATP)-dependent way and played crucial roles in cellular morphogenesis and fundamental biology, like mitosis and meiosis [6][7][8]. There are 45 KIFs discovered and identi ed in human, among which several family members were demonstrated varied functions in tumor pathobiology [9]. KIF11 was identi ed as a molecular target that shuttles between the proliferation and invasion of glioblastoma [10]. KIF20A peptide-based immunotherapy for cancer treatment was demonstrated availability and putative e cacy with promiscuous T-H-cell epitopes derived from KIF20A identi ed in solid tumor tissue and distinguished KIF20A-speci c TH1-cell responses were found in patients with HNMT receiving immunotherapy [11]. Microarray data analyses revealed the highly transactivated status of KIF4A in non-small cell lung cancer and targeting KIF4A might hold a promise for the development of anticancer drugs and cancer vaccines as well as a prognostic biomarker in the clinic [12]. Many researches were done highlighting the importance of KIFs in various aspects of breast cancer [13]. Given the essential roles of KIFs reported in cancer, KIF-targeting cancer therapies were highly expected to be of great e cacy. It is worth noting that several KIF-inhibitors were invented and tested in clinical trials. Ispinesib, a KIF11-targeted inhibitor, was the rst KIF-inhibitor that was evaluated both safety and e cacy in breast cancer in phase I clinical study [14]. Other KIF-targeted drugs further tested in various cancers by clinical trials including KIF11 inhibitors (litronesib [15,16], lanesib [17][18][19], SB-743921 [20], AZD4877 [21]), KIF5C inhibitors (Lidocaine and Tetracaine [22]) and KIFC1 inhibitors (AZ82 and SR31527 [23,24]). However, despite numerous researches done, the prognostic and therapeutic value of all KIFs remains uncorroborated. Considering the intricate functions of KIFs in mitosis, singling out any particular KIFs may not be an e cient way to ful ll the therapeutic capacity of KIFs. At the same time, common regulatory network of all KIFs is little known, which may give new insight into the limited therapeutic e cacy shown in clinical trials and provide putative drug target by mutually regulating KIFs in cancer.

Construction Of 5-kifs-based Risk Score
Given the expression pro le and survival analyses of all KIFs, 9 abnormal expression KIFs (KIF2C, KIF4A, KIF11, KIF15, KIF18B, KIF20A, KIF20B, KIF23, KIF21B) that demonstrated signi cant prognostic value in PAAD were enrolled for LASSO regression to construct a KIFs-based risk score for prediction of OS in PAAD. According to the cv t plot, the minimal of lambda value was seen in 5, indicating a 5-KIFs-based risk score model the best for both accuracy and simplicity (Fig. 8a). Therefore, only KIF15, KIF20A, KIF20B, KIF21BA, KIF23 were included (Fig. 8b) and a 5-KIFs-based risk score was generated as below: Rs: risk score; Exp(X): the expression level of gene X Validations of the 5-KIFs-based risk score done by Kaplan-Meier plots demonstrated good prognostic value in predicting OS of PAAD, with higher risk score signi cantly correlated with worse outcomes in both TCGA-PAAD data (P = 0.00076) and dataset (GSE62452) from GEO(P = 0.031) (Fig. 8c).

Multivariate Analysis And Nomogram
To evaluate the prognostic value of the risk score, we performed multivariate analyses on all demographic and clinical parameters (age, family history of cancer, history of chronic pancreatitis, history of diabetes, maximum tumor dimension, tumor grade, number of positive lymphnodes, radiation therapy, TNM stage and risk score) in TCGA dataset. Then we selected several indicators (age, number of positive lymphnodes, risk score) with signi cant differences for next step. A nomogram predicting 1-year and 3year OS of PAAD patients was constructed combining age, number of positive lymphnodes and LASSO regression generated 5-KIFs risk score using R package "rms" [46,47]. Assessment of predictive accuracy was done by calibration plot [48].

Genetic Alteration, Enrichment of co-expression genes, Strings, GO and KEGG Analysis
The genetic alteration was analyzed by cBioPortal (www.cbioportal.org) [49]. The genomic map includes data for putative copy-number alterations (CNA) from GISTIC, mRNA expression z-scores and mutations. Co-expression genes correlated with KIFs were enriched using GEPIA by TCGA-PAAD data. STRINGS (www.string-db.org) is an online analysis tool that collects, scores and integrates all publicly available sources of protein-protein interaction (PPI) data and supplements it with calculations and predictions [50]. In this study, we performed a PPI network analysis on co-expression genes of KIFs to explore their interactions. The intersection of co-expressed genes was put into GO, KEGG and transcription factors enrichment using R package "ClusterPro ler" [51]. Enrichment results were represented as bubble plot using R package "ggplot" [52].
The functions of the 5-KIFs and the genes signi cantly associated with alterations were predicted by analyzing gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) in the Database for Annotation, Visualization and Integrated Discovery (DAVID) (https://david.ncifcrf.gov/). GO enrichment analysis predicted the functional roles of target host genes based on three aspects, including biological processes, cellular components, and molecular functions. We found that GO:0140014 (Mitotic Nuclear Division), GO:0048285 (Organelle Fission), GO: 0000070 (Mitotic Sister Chromatid Segregation), and GO: 0000280 (Nuclear Division) were signi cantly regulated by the KIFs alterations in PAAD ( Figure  11A, B, C).
KEGG analysis can de ne the pathways related to the functions of KIFs alterations and the frequently altered neighbor genes; 5 pathways signi cantly related to the functions of KIFs alterations in PAAD were found through KEGG analysis ( Figure 11D). Among these pathways, hsa04110: Cell cycle, hsa04114: Oocyte Meiosis, hsa04914: Progesterone−Mediated Oocyte Maturation, has04115: P53 Signaling Pathway and has04218: Cellular Senescence were involved in the tumorigenesis and pathogenesis of PAAD.

The Relationship Between KIFs Expression Levels and Immune In ltration Levels in PAAD
TIMER online analysis tool is used to evaluate the relationship between the transcription level of KIFs and the level of immune in ltration in PAAD. It was found that KIFs are involved in immune cell in ltration, which affect the clinical outcome of PAAD patients. The analysis results are shown in Figure 12 (Table 2). Table 2 The cox proportional hazard model of the 5-KIFs and B immune cells in PAAD (TIMER).

Discussion
The KIF proteins participate in multiple normal cellular biological activities including mitosis and intracellular transport of vesicles and organelles [6]. Increasing evidence indicates kinesin proteins play critical roles in the genesis and development of human cancers. Some kinesin proteins are associated with malignancy as well as drug resistance of solid tumor. Thus, targeting KIF therapy seems to be a promising anticancer strategy. Therefore, comprehensive bioinformatics analyses were done in our study using data from multi-dataset to explore the expression, prognosis as well as Immune in ltrates of kinesin superfamily in PAAD patients. A total of 24 differentially expressed KIFs were identi ed between tumor tissue and normal tissue with 1 (KIF1A) downregulated and 23 KIF2A, KIF2C, KIF3A, KIF3B, KIF3C,  KIF4A, KIF5B, KIF7, KIF9, KIF10, KIF11, KIF13A, KIF13B, KIF15, KIF16B, KIF18B, KIF20A, KIF20B, KIF21B, KIF22, KIF23, KIF26B) overexpressed. Survival analyses revealed 9 abnormal expressed KIFs (KIF2C, KIF4A, KIF11, KIF15, KIF18B, KIF20A, KIF20B, KIF21B, KIF23) signi cantly correlated with worse OS and RFS of PAAD patients, indicating e cient biomarkers for predicting the prognosis of PAAD. Further analyses were done with a 5-KIFs-based risk score generated by LASSO regression, a nomogram was constructed with elements selected by multivariate survival analysis and an accurate predictive e cacy was validated by another dataset.
KIF15 plays an important role in several kind of tumors. As reported, KIF15 is required for maintenance of spindle bipolarity and is the breast cancer tumor antigen [25]. In our group's early study, we reported for the rst time that KIF15 promotes pancreatic cancer cell proliferation by promoting G1/S phase transition via regulating the MEK-ERK signaling pathway[26].
KIF20A has been documented to accumulate in the midzone of the spindle during anaphase and to the cleavage furrow and midbody during telophase [27]. It has been reported that KIF20A is overexpressed in human pancreatic ductal adenocarcinoma (PDAC) based on cDNA microarray analyses and that KIF20A and DLG5, which collaborate in the cytoplasm but not in the midzone of the spindle, are likely to be involved in pancreatic carcinogenesis [28]. KIF20A functions as part of the intracellular tra cking machinery for DLG5, transporting it to membrane sites in PDAC cells in which DLG5 can interact with other proteins including β-catenin[28]. Therefore, it is possible that the intracellular tra cking function of KIF20A is highly important for pancreatic carcinogenesis.
KIF20B is strongly overexpressed in bladder cancer tissues and downregulation of endogenous KIF20B leads to cytokinesis defect [29]. It's also reported that in multiple cancer cells, knockdown of KIF20B not only dramatically inhibits tumor cell growth, but also causes mitotic arrest, senescence and postmitotic apoptosis [30].
KIF21B is not only a classic kinesin protein but also a regulator of microtubule dynamics [31]. In vitro reconstitution studies show that KIF21B increases the microtubule growth rate and catastrophe frequency. Moreover, the puri ed protein surprisingly associates primarily with depolymerizing microtubule and microtubule ends [32]. A study published in 2017 con rmed that KIF21B is a potential microtubule-pausing factor [33]. Microtubules and kinesins play roles in intercellular signal transduction, transport, malignant tumorigenesis, and tumor progression, invasion, and metastasis [34,35].
KIF23 is essential for the completion of cytokinesis in midbody formation and for recruitment to the spindle midzone/midbody by a chromosomal passenger protein [36]. In a study, it discovered the high expression of KIF23 was associated with poor prognosis by immunohistochemical analysis of tumor sections and clinical data from patients with PDAC [37]. They also found that knockdown of KIF23 decreased proliferation in PANC-1 and BxPC-3 PDAC cells, and suppression of KIF23 in PANC-1 cells inhibited tumor growth in nude mice [37]. Furthermore, the mutation rate (6%) of the 5-KIFs in PAAD is low, which indicates the expression of the 5-KIFs is stable relatively. Then we analyzed the function of the 5-KIFs and its neighboring proteins by GO enrichment analysis and KEGG pathway enrichment. Studies have found that the functions of these proteins are mainly related to the Cell cycle, Oocyte Meiosis, Progesterone − Mediated Oocyte Maturation, P53 Signaling Pathway and Cellular Senescence, and some of these pathways are signi cantly associated with the occurrence and development of PAAD. These data indicate that differentially expressed KIFs in PAAD are potential targets for drug therapy.
Another important result of this study is that the transcription levels of the 5-KIFs is closely correlated with various levels of immune in ltration in PAAD. The correlation between the expression of the 5-KIFs and the marker genes of PAAD immune cells indicates that KIFs may be involved in the regulation of PAAD tumor immunity.
In conclusion, KIFs not only can be used as a prognostic indicator of patients with PAAD, but also re ect their immune status. These ndings may help better study the molecular basis of PAAD and may help create more appropriate prognostic tools for PAAD and facilitate the development of new immunotherapeutic. However, more experimental studies are needed to con rm our results, thereby promoting the clinical application of KIFs as a prognostic indicator or immunotherapy target in PAAD.

Conclusion
Our research comprehensively illustrates the expression of KIFs in PAAD. We also built a risk score model of PAAD patients based on the TCGA database and it can still accurately predict the prognosis of PAAD patients in independent GEO data set. And we found that KIFs are involved in immune cell in ltration. The cox regression of several KIFs and immune cell can also affect the clinical outcome of PAAD patients. In summary, our study may provide new choices of prognostic biomarkers and immunotherapy targets in PAAD patients.

Gepia Dataset And Oncomine Databases Analysis
Gene Expression Pro ling Interactive Analysis (GEPIA, www.gepia.cancer-pku.cn) is a newly developed interactive web server that analyzes RNA sequencing expression data from 9,736 tumors and 8,587 normal samples from the Cancer Genome Atlas (TCGA) and Genotypic Tissue Expression (GTEx) projects using standard processing pipelines (www.gepia.cancer-pku.cn/). GEPIA can provide customizable functions, including differential expression analysis, patient survival analysis, cancer type staging, cancer pathological staging, correlation analysis, similar gene detection and dimensionality reduction analysis [39]. Oncomine gene expression array dataset (www.oncomine.org) is a publicly accessible online cancer microarray database for analyzing the KIFs transcription levels in different cancers [40]. Student's t test was used to compare the transcription levels of KIFs in clinical cancer specimens with that in normal controls. The cut-off of P-value was de ned as 0.05.

The Human Protein Atlas Analysis
The Human Protein Atlas(HPA, https://www.proteinatlas.org/) is a Swedish-based program initiated in 2003 with the aim to map all the human proteins in cells, tissues and organs using an integration of various omics technologies, including antibody-based imaging, mass spectrometry-based proteomics, transcriptomics and systems biology [41]. We can nd the expression levels of KIFs in PAAD tissues and its adjacent tissues.

The Kaplan-meier Plotter Analysis
The Kaplan-Meier plotter is an online database (www.kmplot.com) that contains gene expression data and survival information for clinical cancer patients [42,43]. We used this online tool to assess the prognostic value of KIFs mRNA expression in PAAD patients and analyzed the OS and RFS of patients with PAAD. The database divides patient samples into high expression groups and low expression groups according to the median values of KIFs mRNA expression and validates them by Kaplan-Meier survival curve. Information on number of patients, median values of mRNA expression, 95% con dence interval (CI), hazard ratio (HR), and P-value can be found on the Kaplan-Meier plotter web page. P-value < 0.05 was considered as statistically signi cant.

Lasso Regression
LASSO Cox regression is a widely used method for high-dimensional predictors selection [44]. In this study, TCGA-PAAD data were used to construct a risk score of KIFs for the prediction of OS. R package "glmnet" was used to execute LASSO Cox regression model analysis [45]. Cv t plot was drawn, and the minimum lambda value was used as cutoff. The predictive model was validated in TCGA-PAAD and GEO datasets using KM plotter as previously described.
Timer Analysis TIMER (www.cistrome.shinyapps.io) is a convenient and accurate online analysis tool that can infer the abundance of tumor-in ltrating immune cells from gene expression pro les and evaluate their clinical impact [53]. In this study, we used TIMER to assess the correlation between KIFs expression levels and immune cell in ltration. And we also assess the correlation between clinical outcomes and immune cell in ltration and KIFs expression.

Statistics
For all the analyses done above, a P-value < 0.05 was considered statistically signi cant except for speci cally mentioned.

Declarations
Founding: This study was funded by the National Natural Science Foundation of China (81871965).

Consent for publication: Not applicable.
Author contributions Zehao Chen conceived the project and wrote the manuscript. Jie Wang, Jian Xu and Wenjie Zhu participated in data analysis. Jie Wang, Jian Xu and Wenjie Zhu participated in discussion and language editing. Jianxin Jiang reviewed the manuscript. All authors contributed to the article and approved the submitted version.
Competing interests: The research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest. The authors declare that they have no competing interests.