Development of a KRAS-Associated Metabolic Risk Model for Prognostic Prediction in Pancreatic Cancer

doi:10.21203/rs.3.rs-98504/v1

Download PDF

Primary research

Development of a KRAS-Associated Metabolic Risk Model for Prognostic Prediction in Pancreatic Cancer

https://doi.org/10.21203/rs.3.rs-98504/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: KRAS was reported to affect some metabolic genes and promote metabolic reprogramming in solid tumors. However, there is no comprehensive analysis to explore KRAS associated metabolic signature or risk model for Pancreatic cancer (PC).

Methods: In current study, multiple bioinformatics analyses were used to identify differentially expressed metabolic genes based on KRAS mutation status in PC. Then we developed and validated a prognostic risk model based on the selected KRAS-associated metabolic genes. Besides, we explored the association of the risk model and the metabolic characteristics as well as Gemcitabine associated chemoresistance in PC.

Results: 6 KRAS-associated metabolic genes (i.e. CYP2S1, GPX3, FTCD, ENPP2, UGT1A10, and XDH) were selected and were enrolled to establish a prognostic risk model. The prognostic model had a high C-index of 0.733 for overall survival (OS) in the TCGA pancreatic cancer database. The area under the curve (AUC) values of 1- and 3-year survival were both greater than 0.70. Then the risk model was validated in two GEO datasets and also presented a satisfactory discrimination and calibration performance. Further, we found that the expression of some KRAS-driven glycolysis associated genes (PKM, GLUT1, HK2, and LDHA) and Gemcitabine associated chemoresistance genes (i.e. CDA and RMM2) were significantly up-regulated in high-risk PC patients evaluated by the risk model.

Conclusions: We constructed a risk model based on 6 KRAS associated metabolic genes, which predicts patients' survival with high accuracy and reflects tumor metabolic characteristics and Gemcitabine associated chemoresistance in PC.

Cancer Biology

Risk Model

Pancreatic Cancer

Metabolism

KRAS

Prognosis

Chemoresistance

Pancreatic cancer (PC) is one of the most malignant cancers, which leads to 4.5% of all cancer-related deaths globally.[1] The vital causes for the poor prognosis are the highly aggressive phenotype and early cancer recurrence and metastasis following surgical treatment.[2] Despite some advances in the management of PC in recent years, very few major breakthroughs for effective biomarkers nor treatment strategies have emerged.

Almost all PC patients carry at least one of the four frequently-mutated driver genes, which are oncogene KRAS and the tumor suppressors TP53, SMAD4 and CDKN2A.[3] KRAS, the most frequently mutated oncogene in cancer especially in PC, was reported to rewire metabolism to support tumor growth.[4] Several studies had shown that KRAS promoted metabolic reprogramming through the enhance of glucose metabolism, differential channeling of glucose intermediates, reprogramming glutamine metabolism, increasing autophagy, and macropinocytosis.[5, 6] The metabolic reprogramming and fibrotic stroma of PC had been considered as a barrier for cytotoxic drugs delivery to cancer cells, therefore contributing to chemoresistance and treatment failure.[7, 8] Although some KRAS-associated genes were found to regulate cancer cell metabolism, the mechanism was still remain to be clarified and there was no comprehensive analysis to explore KRAS associated metabolic signature or risk model for PC.

In current study, multiple bioinformatics analyses were used to identify differentially expressed metabolic genes based on KRAS mutation status in PC. Then a prognostic model was constructed based on the selected KRAS-associated metabolic genes. Moreover, we explored the association of the risk model and the metabolic characteristics along with the Gemcitabine associated chemoresistance in PC.

Datasets and data acquisition

The gene expression data recorded based on Fragments Per Kilobase per Million (FPKM) and clinical information for 178 PC samples were obtained from the Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/repository) database up to September 2020. The KRAS mutation data of TCGA PC cohort were downloaded from cBioportal (http://www.cbioportal.org/).[9] Among these 178 PC samples, 167 PC samples with RNA-sequencing data and KRAS mutation information were subjected to subsequent analysis. GSE57495 data set based on GPL9115 (including 63 PC samples) and GSE79668 data set based on GPL11154 (including 51 PC samples) were downloaded from Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) database for validation. Besides, two other GEO datasets, GSE140077 based on GPL20795 (including 1 PC cell lines and 6 samples) and GSE106336 based on GPL18573 (including 1 PC cell line and 6 samples) were used for further study in Gemcitabine resistance of PC. A total of 1724 metabolic genes were obtained from the metabolic pathway-related gene lists of “c2.cp.kegg.v7.1.symbols.gmt" in gene set enrichment analysis (GSEA, https://www.gsea-msigdb.org/gsea/index.jsp).[10] Datasets above for PC are publicly obtainable and ethics approval was not needed.

Differential expression analysis

We compared 58 PC samples without KRAS mutations (KRAS WT) and 109 PC samples with KRAS mutations (KRAS MUT) in TCGA PC cohort to identify differentially expressed metabolic genes (DEGs) using the R package limma. The thresholds were |log2-fold change (FC)|>1.5 and FDR<0.05.

Establishment and validation of the KRAS-associated metabolic risk score system

With the help of R packages survival, survminer, and parallel, the univariate Cox proportional regression was performed to evaluate the association between expression of DEGs and patients’ overall survival (OS) in TCGA cohort. DEGs with p-value < 0.01 were considered as statistically significant prognosis-associated DEGs (PAGs). Then, a multivariate Cox proportional regression model was established through cycle computation to determine the model with highest discriminated ability for OS of PC. The KRAS-associated metabolic risk model was established based on a linear combination of the expression values of the PAGs and the multivariate Cox regression coefficients were used as the weight. For the external validation cohort for the risk score system, the GSE57495 and GSE79668 datasets were analyzed. Patients in separate datasets were divided into a high-risk group and a low-risk group using the median cut-off of the risk score. To evaluate the performance of model, the R packages pROC were used to evaluate the discrimination of the risk model in TCGA and GEO datasets through area under the curve (AUC) of the receiver operating characteristic (ROC) curve of 1- and 3-year survival. With the help of R package rms, Harrell’s concordance index (C-index) of the risk score system was calculated by a bootstrap approach with 1000 resamples. The calibration curves were used to assess the consistency between model-predicted and observed survival. Then the log-rank tests and Kaplan-Meier (KM) analyses were performed using the survival R package between the high-risk and low-risk group to assess the predictive ability of the prognostic model. We calculated the risk scores in different groups based on clinicopathological parameters (such as American Joint Committee on Cancer (AJCC) stage, histologic grade, diabetes status) in TCGA PC cohort by using Wilcoxon’s test to evaluate whether the risk model reflects PC progression.

Functional enrichment analysis

Gene set enrichment analyses (GSEA, Version3.0, http://software.broadinstitude.org/gsea/) was performed to determine how the metabolic pathways and relevant metabolic pathway-related genes differed between PC samples of KRAS WT and KRAS MUT in TCGA cohort. An annotated gene set file (c2.cp.kegg.v7.1.symbols.gmt) was used as the reference gene set and a gene set was considered as an enriched group when nominal p-value < 0.05. DEGs and co-expressed genes of PAGs screened from cBioPortal database were integrated to DAVID 6.7 (https://david-d.ncifcrf.gov/) separately to perform Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis.[11] Results were visualized by using R package ggplot2.

Association between the risk model with metabolic characteristics of PC

Several KRAS-driven metabolic targets which took part in metabolic reprogramming of PC were selected to explore relationships between the risk model with metabolic characteristics of PC.[12] The mRNA expressions levels of these promising KRA-driven metabolic genes were compared in high-risk and low-risk groups by using Student's t-test.

Association between the risk model and Gemcitabine chemoresistance of PC

In order to evaluate whether the risk model reflects Gemcitabine metabolism-associated chemoresistance of PC, metabolic risk scores were calculated in 2 parental and GEM-resistant cell lines (CFPAC-1 and HPAFII) in GSE140077 and GSE106336 datasets. Then several Gemcitabine metabolism-associated chemoresistance genes of PC were selected to explore relationships between the risk model with Gemcitabine chemoresistance of PC.[13] The mRNA expressions levels of these promising KRAS-driven metabolic genes were compared in high-risk and low-risk groups by using Wilcoxon's test.

Statistical analyses

All statistical analyses were performed using R software Version 4.0.1 (https://www.r-project.org/) and SPSS software Version 24.0 (SPSS, Inc., Chicago, IL, USA). If not specified above, a p-value < 0.05 was considered statistically significant.

Association between metabolic phenotype and KRAS mutations in PC

It’s shown in FireBrowse Database (http://www.firebrowse.org/) that KRAS mutation was the most common type of mutation in PC, followed by mutation frequency of TP53, MAMLD1, MAGEC1, CDKN2A, and SMAD4 (Figure 1A). Next, we utilized mRNA expression data and corresponding clinical information of PC samples in TCGA and cBioportal to investigate metabolic processes linked to KRAS mutation status. GSEA was conducted to determine the difference of metabolic pathways between 58 PC samples of KRAS WT and 109 PC samples of KRAS MUT. The results showed that 4 metabolic biological processes were significantly enriched in KRAS MUT group, which were Glycine Serine and Threonine Metabolism (NES=1.63, size=31 and p-value<0.05), Tryptophan Metabolism (NES=1.48, size=40 and p-value<0.05), Type II Diabetes Mellitus (NES=1.41, size=47 and p-value<0.05), and Primary Bile Acid Biosynthesis (NES=1.67, size=16 and p-value<0.05) (Figure 1B).

Identification and enrichment analysis of differentially expressed metabolic genes based on KRAS mutation status

We performed differential expression analyses between 58 PC samples of KRAS WT with 109 PC samples of KRAS MUT in TCGA PC cohort. Among 1724 metabolic genes, 54 genes were significant differentially expressed and considered as DEGs (|log2-fold change (FC)|>1.5 and FDR<0.05) (Figure 2A, B). Then these 54 DEGs were integrated to DAVID 6.7 to perform GO analysis and KEGG pathway analysis. GO analysis showed that the top 10 highly enriched functions in the metabolic process were “Oxidation Reduction, Carboxylic Acid Catabolic Process, Hormone Metabolic Process, Alcohol Catabolic Process, Glucose Metabolic Process, Lipid Catabolic Process, Cellular Amino Acid Catabolic Process, Vitamin Metabolic Process, Steroid Metabolic Process, and Glutamine Family Amino Acid Metabolic Process” (Figure 2C). In KEGG analysis (Figure 2D), DEGs was found to be mainly enriched in “Adipocytokine Signaling Pathway, Retinol Metabolism, Metabolism of Xenobiotics by Cytochrome P450, Type II Diabetes Mellitus, Drug Metabolism, Endocytosis, and Fatty Acid Metabolism”.

Construction and validation of a KRAS-associated metabolic risk model

We performed univariate Cox proportional regression and found that 21 out of 54 DEGs were significantly related to OS and considered as prognosis-associated DEGs (PAGs) (Figure 3A). Further, with the help of the R packages survival, a multivariate Cox regression analysis for these 21 PAGs was performed and it revealed that 6 of them were independently associated with OS (Figure 3B). Therefore, a 6-PAGs based KRAS-associated metabolic risk model was developed by weighting the normalized expression of these 6 PAGs multiplied by corresponding coefficients derived from the multivariate Cox regression analysis: risk score = (-0.29328 * normalized expression of CYP2S1) + (-0.45614 * normalized expression of GPX3) + (-0.57665 * normalized expression of FTCD) + (0.54899 * normalized expression of ENPP2) + (0.19472 * normalized expression of UGT1A10) + (0.34727 * normalized expression of XDH) (Table 1). Patients were divided into high-risk and low-risk groups by using the median cut-off value of the risk score (Figure 3C).

Table 1

6 KRAS-associated metabolic genes to establish the risk model
Genes	coef	HR	HR.95L	HR.95H	P-value
CYP2S1	-0.29328	0.74582	0.60422	0.92060	0.00633
GPX3	-0.45614	0.63372	0.47754	0.84098	0.00158
FTCD	-0.57665	0.56178	0.44136	0.71504	2.80E-06
ENPP2	0.54899	1.73150	1.2754	2.3506	0.000432
UGT1A10	0.19472	1.21498	1.0496	1.4064	0.00911
XDH	0.34727	1.41520	1.1314	1.7702	0.00236

The predictive accuracy of the risk model was further assessed through the ROC and C-index analysis. The results showed the AUC of the risk model for OS in TCGA cohort was 0.773 at 1 years and 0.704 at 3 years (Figure 4A). Besides, C-index of the risk model in TCGA cohort was 0.733 (95% CI, 0.675 - 0.761). The calibration curves of the risk model matched well, which indicated that it could accurately predict the 1- and 3‐year OS in TCGA cohort (Figure 4B). GSE57495 and GSE79668 datasets were used for the external validation for the risk model. The AUC for 1‐, and 3‐year OS in GSE57495 datasets were 0.627, and 0.698 (Figure 4A), while the AUC for 1‐, and 3‐year OS are 0.727 and 0.617 in GSE79668 datasets (Figure 4E). The C-index of the risk model in GSE57495 and GSE79668 datasets were 0.683 (95% CI, 0.655 - 0.723) and 0.625 (95% CI, 0.598 - 0.675). And the corresponding 1-year and 3-year calibration curves are respectively shown in Figure 4D and 4F.

Association between the risk model with patients’ survival and clinicopathological characteristics in PC

To further evaluate the prognostic power of the risk model, Kaplan-Meier analyses were performed and we found that all patients in high-risk group had a shorter OS (p-valve=4.234e-07) and Disease-free survival (DFS) (p-valve=1.83e-06) than those in low-risk group in TCGA cohort (Figure 5A-B). Besides, similar results for KM analyses of OS were observed in GSE57495 dataset (p-value=8.637e-04) and GSE79668 dataset (p-valve=1.833e-02) (Figure 5C-D). Further we calculated the risk scores in different groups based on clinicopathological characteristics in TCGA PC cohort and the results showed that patients who had advanced stage, higher histologic grade or diabetes history had higher risk scores (p-valve<0.05) (Figure 5E). The data in Figure 4 and Figure 5 suggest the KRAS associated metabolic risk model has effective value in predicting patients' survival and is associated with advanced tumor characteristics.

Association between the risk model with metabolic characteristics of PC

We next explored the activities of 6 PAGs (i.e. CYP2S1, GPX3, FTCD, ENPP2, UGT1A10, and XDH) by analyzing its potential metabolic pathways in PC. The co-expression analyses for 6 PAGs were performed by using cBioPortal dataset (Spearman’s correlated coefficient>0.6 or <−0.6, p-valve<0.05). We found 131 co-expression genes for CYP2S1, 163 co-expression genes for ENPP2, 387 co-expression genes for FTCD, 521 co-expression genes for GPX3, 189 co-expression genes for UGT1A10, and 213 co-expression genes for XDH, all of which were enrolled into DAVID 6.7 and subjected to functional and pathway enrichment analyses.

The potential metabolic pathways involved were shown in Figure 6A. CYP2S1 and its neighboring genes were mainly enriched in “Oxidation-reduction process, Lipid metabolic process, Protein glycosylation, Lysophospholipase activity, Steroid metabolic process, Regulation of glucuronidation, and Response to insulin stimulus”. GPX3 may act a vital role in “Lipid metabolic process, Glutamate receptor activity, and Regulation of insulin secretion”. FTCD may play an important role in “Glucose metabolic process and Regulation of insulin secretion”. ENPP2 may be involved in “L-amino acid transport, Glutamate receptor activity, Regulation of insulin secretion, and Response to insulin stimulus”. UGT1A10 was mainly associated with “Lipid metabolic process, Steroid metabolic process, Regulation of glucuronidation, CoA succinyl transferase activity, Endocytosis” and XDH was involved in “Protein glycosylation, Phagocytosis, and Endocytosis”.

To further explore the potential metabolic targets influenced by our risk model, we compared the expression levels of several KRAS-driven metabolic genes between high-risk with low-risk group in TCGA cohort. We found higher mRNA expressions of PKM (p=1.86e-05), GLUT1 (p=3.168e-05), HK2 (p=3.056e-04), LDHA (p=2.989e-06), and VDR (p=0.012) in the high-risk group (Figure 6B).

Association between the risk model with Gemcitabine chemoresistance in PC

We found that 5 of 6 PAGs were enriched in “Drug metabolism or Response to drug” in Functional and pathway enrichment analyses (Figure 7A). In order to explore their relationship with Gemcitabine associated chemoresistance in PC, we calculated metabolic risk scores in 2 parental and Gemcitabine-resistant cell lines (CFPAC-1 and HPAFII) in GSE140077 and GSE106336 datasets (Figure 7B). The results revealed that Gemcitabine-resistant group had significantly higher metabolic risk scores than parental group (p-valve<0.001).

To investigate how the risk model leads to Gemcitabine resistance, we compared the expression of previous reported genes regulating Gemcitabine drug metabolism and efficacy (i.e. CDA, DCK, hENT1, hCNT1, RMM1, RMM2) between high‐risk with low-risk group in TCGA cohort. The results showed that the mRNA levels of CDA (p-valve=0.001) and RMM2 (p-valve=2.517e-12) were significantly up-regulated in the high-risk group (Figure 7C).

As the most frequently mutated oncogene in PC, KRAS and its downstream pathways affect several cellular processes including cell proliferation, migration, metabolism and autophagy in PC.[14] Small molecule drug (Sotorasib) that specifically and irreversibly inhibits KRAS had passed phase 1 clinical trial, showing a promising therapeutic prospect.[15] Studies have revealed oncogenic KRAS rewired metabolism to favor a more anabolic state and therefore promote tumorigenesis and progression of PC.[4] However, the molecular mechanisms were still not clear and it’s of great significance to conduct a profound study on KRAS-associated metabolic genes in PC.

In current study, we found out 6 differentially expressed metabolic genes based on KRAS mutation status, including CYP2S1, GPX3, FTCD, ENPP2, UGT1A10, and XDH. Then we developed a significant KRAS-associated metabolic risk model for prognostic prediction in PC. Further, the risk model was validated in two external cohorts (GSE57495 and GSE79668 datasets), which showed a stably high prognostic value for PC. These results indicate our model is accurate in predicting patients' survival, which may be helpful in designing personalized therapy targeting metabolism reprogramming. In the risk model, high expression of CYP2S1, GPX3 and FTCD were correlated with better prognoses, while high expressions of ENPP2, UGT1A10 and XDH were correlated with poor prognoses in PC. As a member of Cytochrome P450 (CYPs), CYP2S1 was known to biotransformate some exogenous and endogenous compounds including drugs, fatty acids, and cholesterols.[16] Study showed that expressions of 8 CYPs were changed in CYP2S1-depleted cells and all of them were involved in lipid biotransformation. Indeed, CYP2S1 affected the metabolism of arachidonic acid (AA) and linoleic acid (LA) to modulate BEAS-2B cells growth.[17] In consistent with their results, our data also found CYP2S1 was enriched in Lipid metabolic process. It will be interesting to investigate the detailed mechanism by which CYP2S1 regulate lipid metabolism in PC in future study. As a member of glutathione peroxidase family (GPXs), GPX3 reduces glutathione to catalyze the reduction of hydrogen peroxide, hydroperoxides, and lipid hydroperoxides.[18] Low GPXs activity was found associated with enhanced lipid peroxidation in metastatic cancers, indicating that loss of GPX3 may promote systemic oxidative stress.[19] Studies had reported that low expression of GPX3 was associated with poor prognosis and chemoresistance in several tumor types.[18] Our results also proved that it’s associated with a good outcome and may involve in “Lipid metabolic process and Glutamate receptor activity” in PC. ENPP2, which encodes an ecto-lysophospholipase D called Autotaxin (ATX), was found significantly increased in several types of cancer including PC.[20–22] Overwhelming evidences revealed that the ATX/lysophosphatidate (LPA) signaling axis serves key roles in energy metabolism regulation and obesity control, dysregulation of which could cause inflammation and tumorigenesis.[23, 24] Studies have indicated that ATX/LPA axis promoted DNA synthesis, proliferation, and invasion in pancreatic cancer cells via ERK1/2 and Rho pathways.[25, 26] Over-expression of ATX was also proven to led to elevated tumorigenesis and invasiveness compared with control groups in RAS-mutated NIH3T3 murine fibroblasts.[27] Further studies are needed to explore the relationship and underlying regulation mechanism of KRAS mutation and ENPP2/ATX in PC tumorigenesis and metabolism reprogramming.

UGT1A10 was an extrahepatic phase II metabolizing enzyme that expressed highly in numerous target areas for tobacco-induced cancers, including the upper aerodigestive tract.[28] UGT1A10 were elevated in a CPT-11/SN-38-resistant cell line and responsible for SN-38 glucuronidation, which was one of the mechanisms associated with irinotecan hydrochloride/7-ethyl-10-hydroaxycamptothecin (CPT-11/SN-38) resistance in lung cancer.[29] FTCD was found significantly down-regulated in hepatocellular carcinoma (HCC) and serve as a diagnostic biomarker for HCC.[30] Recently, FTCD was found to be associated with the sensitivity of chemotherapeutic drug methotrexate by using a CRISPR-Cas9-based screen.[31] XDH, a rate-limiting enzyme to catalyse the final steps of purine metabolism, was found significantly decreased and serves useful predictor of poor prognoses in several cancer types.[32] Uric acid, which was transformed by XDH, was found to modulate tumor cell sensitivity to the antimetabolite 5-FU, one of the most commonly used anticancer drugs in the clinic.[33] Our study consistently suggested that UGT1A10, FTCD, and XDH were involved in “Drug metabolism or Response to drug” in PC, but their specific roles in metabolic reprogramming and chemoresistance in PC still remain to be studied.

Gemcitabine based chemotherapy is a major treatment for PC patients with or without surgical resection. It has been reported Gemcitabine drug metabolism was affected by some genes or metabolites, which included drug transporters (i.e. hEN1, and hCNT1), activating and inactivating enzymes (i.e. CDA, and DCK) and competitive substrates to active metabolites (i.e. RRM1, and RRM2).[13] Cytidine deaminase (CDA) induced the deamination of dFdC to dFdU, leading to inactivation of gemcitabine.[34] It had been confirmed that CDA expression was correlated with OS in PC and several in vitro studies revealed that over-expression of CDA led to gemcitabine resistance, while loss of CDA recovered gemcitabine sensitivity.[35, 36] RRM2 was involved in the activity of Ribonucleotide Reductase (RR), which was a rate-limiting enzyme of DNA synthesis. It has been demonstrated that up-regulation of RRM2 led to gemcitabine chemoresistance in PC cells and human PC xenografts in mice.[37] Besides, expression level of RRM2 was correlated inversely with OS in gemcitabine-treated PC patients in clinical study.[38] In current study, CDA and RRM2 were up-regulated in high-risk group evaluated by the risk model. Taken these together, the risk model may reflect the possibility of gemcitabine resistance, which would help oncologist choose appropriate chemotherapy regimen for PC.

There are a few limitations of our study. First, our study is mainly based on bioinformatic analysis and more experimental studies are needed to investigate how KRAS may regulate cancer cell metabolism in PC. Second, the established model needs to be further validated in prospective clinical studies.

In conclusion, we constructed and validated a prognostic model based on the KRAS-associated metabolic genes in PC. The prognostic model reflects tumor metabolic characteristics and Gemcitabine associated chemoresistance, which may have important value in aiding personalized therapy.

PC-Pancreatic Cancer; OS - Overall Survival; AUC - Area Under the Curve; FPKM - Fragments Per Kilobase per Million; TCGA - the Cancer Genome Atlas; GEO - Gene Expression Omnibus; GSEA - Gene Set Enrichment Analysis; DEGs - Differentially Expressed Genes; PAGs - Prognosis-associated DEGs; ROC - Receiver Operating Characteristic; AJCC - American Joint Committee on Cancer; GO - Gene Ontology; KEGG - Kyoto Encyclopedia of Genes and Genomes.

Ethics approval and consent to participate

This article did not contain any studies with human participants or animals performed by any of the authors. Therefore, local ethics approval was not needed.

Availability of data and materials

The data that support the findings of this study are available from TCGA (https://portal.gdc.cancer.gov/repository) , GEO (GEO, https://www.ncbi.nlm.nih.gov/geo/) and GSEA databases (https://www.gsea-msigdb.org/gsea/index.jsp).

Conflicts of interest

The authors declare that they have no potential conflicts of interest.

Funding

This study was supported by National Natural Science Foundation of China (Project NO.: 82072635, 81702783 and 81672475) and Guangdong Medical Science and Technology Fund (Project NO.: 201707010323).

Authors’ contributions

Conceptualization, Zuyi Ma, Shanzhou Huang, Chuanzhao Zhang and Baohua Hou; Data curation, Zuyi Ma, Zixuan Zhou, Hongkai Zhuang; Formal analysis, Zuyi Ma, Zhenchong Li; Methodology, Zuyi Ma; Software, Zhenchong Li, Bowen Huang, Yuanfeng Gong, Chunsheng Liu, Yiping Zou, Zehao Zheng and LinLing Yang; Validation, Zuyi Ma, Zixuan Zhou and Shanzhou Huang; Visualization, Zuyi Ma; Writing-review, Zuyi Ma, Shanzhou Huang, Chuanzhao Zhang and Baohua Hou.

Acknowledgements

We would like to thank everyone who participated in this study.

Kamisawa T, Wood LD, Itoi T, Takaori K: Pancreatic cancer. LANCET 2016, 388(10039):73-85.
Singhi AD, Koay EJ, Chari ST, Maitra A: Early Detection of Pancreatic Cancer: Opportunities and Challenges. GASTROENTEROLOGY 2019, 156(7):2024-2040.
Cicenas J, Kvederaviciute K, Meskinyte I, Meskinyte-Kausiliene E, Skeberdyte A, Cicenas J: KRAS, TP53, CDKN2A, SMAD4, BRCA1, and BRCA2 Mutations in Pancreatic Cancer. Cancers (Basel) 2017, 9(5).
Tape CJ, Ling S, Dimitriadi M, McMahon KM, Worboys JD, Leong HS, Norrie IC, Miller CJ, Poulogiannis G, Lauffenburger DA et al: Oncogenic KRAS Regulates Tumor Cell Signaling via Stromal Reciprocation. CELL 2016, 165(4):910-920.
White E: Exploiting the bad eating habits of Ras-driven cancers. Genes Dev 2013, 27(19):2065-2071.
Kimmelman AC: Metabolic Dependencies in RAS-Driven Cancers. CLIN CANCER RES 2015, 21(8):1828-1834.
Dimou A, Syrigos KN, Saif MW: Overcoming the stromal barrier: technologies to optimize drug delivery in pancreatic cancer. THER ADV MED ONCOL 2012, 4(5):271-279.
Lin J, Xia L, Liang J, Han Y, Wang H, Oyang L, Tan S, Tian Y, Rao S, Chen X et al: The roles of glucose metabolic reprogramming in chemo- and radio-resistance. J Exp Clin Cancer Res 2019, 38(1):218.
Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, Sun Y, Jacobsen A, Sinha R, Larsson E et al: Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. SCI SIGNAL 2013, 6(269):l1.
Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES et al: Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 2005, 102(43):15545-15550.
Huang DW, Sherman BT, Lempicki RA: Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. NAT PROTOC 2009, 4(1):44-57.
Kawada K, Toda K, Sakai Y: Targeting metabolic reprogramming in KRAS-driven cancers. INT J CLIN ONCOL 2017, 22(4):651-659.
Amrutkar M, Gladhaug IP: Pancreatic Cancer Chemoresistance to Gemcitabine. Cancers (Basel) 2017, 9(11).
Lanfredini S, Thapa A, O'Neill E: RAS in pancreatic cancer. Biochem Soc Trans 2019, 47(4):961-972.
Hong, D S, Fakih, M G, Strickler, J H, Desai, J, Durm, G A et al: KRAS(G12C) Inhibition with Sotorasib in Advanced Solid Tumors.N Engl J Med 2020, 383(13): p. 1207-1217.
Yan P, Eng OC, Yu CJ: A Review on the Expression and Metabolic Features of Orphan Human Cytochrome P450 2S1 (CYP2S1). CURR DRUG METAB 2018, 19(11):917-929.
Madanayake TW, Lindquist IE, Devitt NP, Mudge J, Rowland AM: A transcriptomic approach to elucidate the physiological significance of human cytochrome P450 2S1 in bronchial epithelial cells. BMC GENOMICS 2013, 14:833.
Chang C, Worley BL, Phaëton R, Hempel N: Extracellular Glutathione Peroxidase GPx3 and Its Role in Cancer. Cancers (Basel) 2020, 12(8).
Guven M, Ozturk B, Sayal A, Ozeturk A, Ulutin T: Lipid peroxidation and antioxidant system in the blood of cancerous patients with metastasis. Cancer Biochem Biophys 1999, 17(1-2):155-162.
Lee SC, Fujiwara Y, Liu J, Yue J, Shimizu Y, Norman DD, Wang Y, Tsukahara R, Szabo E, Patil R et al: Autotaxin and LPA1 and LPA5 receptors exert disparate functions in tumor cells versus the host tissue microenvironment in melanoma invasion and metastasis. MOL CANCER RES 2015, 13(1):174-185.
Wu JM, Xu Y, Skill NJ, Sheng H, Zhao Z, Yu M, Saxena R, Maluccio MA: Autotaxin expression and its connection with the TNF-alpha-NF-kappaB axis in human hepatocellular carcinoma. MOL CANCER 2010, 9:71.
Kadekar S, Silins I, Korhonen A, Dreij K, Al-Anati L, Högberg J, Stenius U: Exocrine pancreatic carcinogenesis and autotaxin expression. PLOS ONE 2012, 7(8):e43209.
Barbayianni E, Kaffe E, Aidinis V, Kokotos G: Autotaxin, a secreted lysophospholipase D, as a promising therapeutic target in chronic inflammation and cancer. PROG LIPID RES 2015, 58:76-96.
Benesch MG, Ko YM, McMullen TP, Brindley DN: Autotaxin in the crosshairs: taking aim at cancer and other inflammatory conditions. FEBS LETT 2014, 588(16):2712-2727.
Ryder NM, Guha S, Hines OJ, Reber HA, Rozengurt E: G protein-coupled receptor signaling in human ductal pancreatic cancer cells: neurotensin responsiveness and mitogenic stimulation. J CELL PHYSIOL 2001, 186(1):53-64.
Takiguchi S, Nishino Y, Inoue K, Ikeda M, Kataoka Y, Matsusue K, Nishiyama K, Iguchi H: The bisphosphonate incadronate inhibits intraperitoneal dissemination in an in vivo pancreatic cancer model. ONCOL REP 2012, 28(1):111-116.
Liu S, Umezu-Goto M, Murph M, Lu Y, Liu W, Zhang F, Yu S, Stephens LC, Cui X, Murrow G et al: Expression of autotaxin and lysophosphatidic acid receptors increases mammary tumorigenesis, invasion, and metastases. CANCER CELL 2009, 15(6):539-550.
Balliet RM, Chen G, Dellinger RW, Lazarus P: UDP-glucuronosyltransferase 1A10: activity against the tobacco-specific nitrosamine, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol, and a potential role for a novel UGT1A10 promoter deletion polymorphism in cancer susceptibility. DRUG METAB DISPOS 2010, 38(3):484-490.
Oguri T, Takahashi T, Miyazaki M, Isobe T, Kohno N, Mackenzie PI, Fujiwara Y: UGT1A10 is responsible for SN-38 glucuronidation and its expression in human lung cancers. ANTICANCER RES 2004, 24(5A):2893-2896.
Chen J, Chen Z, Huang Z, Yu H, Li Y, Huang W: Formiminotransferase Cyclodeaminase Suppresses Hepatocellular Carcinoma by Modulating Cell Apoptosis, DNA Damage, and Phosphatidylinositol 3-Kinases (PI3K)/Akt Signaling Pathway. Med Sci Monit 2019, 25:4474-4484.
Kanarek N, Keys HR, Cantor JR, Lewis CA, Chan SH, Kunchok T, Abu-Remaileh M, Freinkman E, Schweitzer LD, Sabatini DM: Histidine catabolism is a major determinant of methotrexate sensitivity. NATURE 2018, 559(7715):632-636.
Saidak Z, Louandre C, Dahmani S, Sauzay C, Guedda S, Chauffert B, Chatelain D, Ceballos-Picot I, Galmiche A: A pan-cancer study of the transcriptional regulation of uricogenesis in human tumours: pathological and pharmacological correlates. Biosci Rep 2018, 38(5).
Cantor JR, Abu-Remaileh M, Kanarek N, Freinkman E, Gao X, Louissaint AJ, Lewis CA, Sabatini DM: Physiologic Medium Rewires Cellular Metabolism and Reveals Uric Acid as an Endogenous Inhibitor of UMP Synthase. CELL 2017, 169(2):258-272.
Shipley LA, Brown TJ, Cornpropst JD, Hamilton M, Daniels WD, Culp HW: Metabolism and disposition of gemcitabine, and oncolytic deoxycytidine analog, in mice, rats, and dogs. DRUG METAB DISPOS 1992, 20(6):849-855.
Weizman N, Krelin Y, Shabtay-Orbach A, Amit M, Binenbaum Y, Wong RJ, Gil Z: Macrophages mediate gemcitabine resistance of pancreatic adenocarcinoma by upregulating cytidine deaminase. ONCOGENE 2014, 33(29):3812-3819.
Maréchal R, Bachet JB, Mackey JR, Dalban C, Demetter P, Graham K, Couvelard A, Svrcek M, Bardier-Dupas A, Hammel P et al: Levels of gemcitabine transport and metabolism proteins predict survival times of patients treated with gemcitabine for pancreatic adenocarcinoma. GASTROENTEROLOGY 2012, 143(3):664-674.
Jonckheere N, Skrypek N, Merlin J, Dessein AF, Dumont P, Leteurtre E, Harris A, Desseyn JL, Susini C, Frénois F et al: The mucin MUC4 and its membrane partner ErbB2 regulate biological properties of human CAPAN-2 pancreatic cancer cells via different signalling pathways. PLOS ONE 2012, 7(2):e32232.
Nakai Y, Isayama H, Sasaki T, Sasahira N, Tsujino T, Toda N, Kogure H, Matsubara S, Ito Y, Togawa O et al: A multicentre randomised phase II trial of gemcitabine alone vs gemcitabine and S-1 combination therapy in advanced pancreatic cancer: GEMSAP study. Br J Cancer 2012, 106(12):1934-1939.

Download PDF

Version 1

posted

You are reading this latest preprint version

Development of a KRAS-Associated Metabolic Risk Model for Prognostic Prediction in Pancreatic Cancer

Status:

Version 1

Abstract

Figures

Introduction

Materials And Methods

Datasets and data acquisition

Differential expression analysis

Establishment and validation of the KRAS-associated metabolic risk score system

Functional enrichment analysis

Association between the risk model with metabolic characteristics of PC

Association between the risk model and Gemcitabine chemoresistance of PC

Statistical analyses

Results

Association between metabolic phenotype and KRAS mutations in PC

Identification and enrichment analysis of differentially expressed metabolic genes based on KRAS mutation status

Construction and validation of a KRAS-associated metabolic risk model

Association between the risk model with patients’ survival and clinicopathological characteristics in PC

Association between the risk model with metabolic characteristics of PC

Association between the risk model with Gemcitabine chemoresistance in PC

Discussion

Abbreviations

Declarations

Ethics approval and consent to participate

Availability of data and materials

Conflicts of interest

Funding

Authors’ contributions

Acknowledgements

References

Status:

Version 1