Predicting drug response of cancer stem cells using gene signatures

Cancer stem cells are known to exhibit exquisite sensitivity to the small molecule inhibitor, salinomycin, however there is a lack of mechanistic understanding of its precise mode of action. We reasoned that molecular insights gained from experimental studies can be used to derive a gene expression signature to predict the response of cancer cells to salinomycin, and further aid in identifying groups of patients or tumor types that would benefit the most from salinomycin therapy.

We observed that salinomycin treatment of cells was associated with disruption of the KRAS nanoscale membrane organization by altering the distribution of phosphatidyl serine (PS), eventually leading to decreased signalling output from KRAS nanoclusters due to reduced effector recruitment. Moreover, overexpression of caveolin decreased the sensitivity of cells to salinomycin by affecting the membrane organization (see publication III for details). This suggests that gene expression state of known modulators of KRAS nanoscale membrane organization can influence the drug response. To gain further insights into the gene expression signature associated with KRAS nanoscale membrane organization, we utilized the ESTOOLS database (180) to find genes that are correlated with its known modulators. Based on the 13 genes that were identified, the gene expression signature classified embryonic stem cells separately from the fibroblasts, suggesting that the signature was also associated with stemness property (Figure 10).

55

Figure 10: Clustering of embryonic stem cells (ESCs) and fibroblasts based on the gene expression of known modulators and correlated genes (VIM, ITGA5 and CAV2). Gene expression data of Metaset 1 from ESTOOLS database is presented as heatmap.

Figure 11: Unsupervised clustering of selected cancer cell lines identified as ESC-like and fibroblast-like based on correlation with the KRAS nanoclustering associated gene expression signature. Gene expression data from CCLE and ESTOOLs were quantile-normalized and scaled.

To study the stemness property associated with KRAS nanoclustering in cancer cells, we further identified cancer cell lines that were correlated with gene expression signature of the 13 genes in ESCs and fibroblasts. As expected, we observed that the ESC-like cell lines clustered with the ESCs and the fibroblast-like cell lines clustered with the fibroblasts (Figure 11).

Drug sensitivity profiling revealed that the ESC-like cell lines were more sensitive to salinomycin whereas less responsive to staurosporine compared to fibroblast-like cell lines, suggesting that the gene-expression

KRASNRAS Cell_Type ESCs Fibroblasts 68101214 KRAS NRAS HRAS NPM1 NCL EGFR LGALS1 LGALS3 CAV1 CAV2 PTRF ITGA5 VIM

Cell_Type

KRAS NRAS HRAS NPM1 NCL EGFR LGALS1 LGALS3 CAV1 CAV2 PTRF ITGA5 VIM

Cell_Type

Co− clustering of rank product selected cancer cell lines with ESCs and Fibroblasts for 13 genes

SET1_ESC.ES2SET1_ESC.ES4SET2_Fibroblast.HS27Fib.1SET2_Fibroblast.HS27Fib.2SET2_ESC.fibroblastSET2_FibroblastSET2_Fibroblast.SC33AFibSET2_Fibroblast.SC33BFibSET2_Fibroblast.HS27Fib.3SET2_Fibroblast.HS27Fib.5SET2_Fibroblast.HS27Fib.4SET2_Fibroblast.SC30AFibSET2_Fibroblast.SC30BFibSET1_hTFC.rep3SET1_hTFC.rep1SET1_hTFC.rep2SET1_H1.OGN.hES.cellsSET1_HFF1SET1_Foreskin.fibroblast.rep1SET1_Foreskin.fibroblast.rep2SET1_Fibroblast.5.SBDSSET1_Fibroblast.1.MBDSET1_Fibroblast.3.fetal.skinSET1_Fibroblast.6.fetal.lungSET1_Parental.foreskin.fibroblast.rep2SET1_Fibroblast.2.GDSET1_Fibroblast.4.PDSET1_Fibroblast.NHDF3535p61.mRNASET1_Fibroblast.PDBSET1_FibroblastSET1_HDFSET1_Parent.fibroblast.for.2.4SET1_Parent.fibroblast.for.1.8.and.3.2SET1_Fibroblast.with.ctrl.vector.18.daysSET1_Fibroblast.rep2SiTayeb_Fibroblast.rep2SET1_Fibroblast.rep3SiTayeb_Fibroblast.rep3SET1_Fibroblast.rep1SiTayeb_Fibroblast.rep1SET1_Fibroblast.1SET1_D551.fibroblast.rep1SET1_D551.fibroblast.rep2SET1_D551.fibroblast.rep3SET1_WI38.fibroblast.rep3SET1_WI38.fibroblast.rep1SET1_WI38.fibroblast.rep2SET1_dH1f.fibroblast.44SET1_dH1f.fibroblast.58SET1_MRC5.fibroblast.40SET1_dH1cf16.fibroblastSET1_BJ.fibroblast.rep1SET1_BJ.fibroblast.rep2SET1_BJ.fibroblast.rep3SET1_MRC5.fibroblast.59SET1_Fibroblast.SMASET1_Fibroblast.unaffectedSET1_Fibroblast.MBD.rep1SET1_BJ1.fibroblast.48ESC_42MGBAESC_A375ESC_U2OSFib_NMCG1Fib_TUHR4TKBFib_MOGGCCMFib_NCIH196Fib_GMS10Fib_YH13Fib_HS281TFib_HS863TFib_HS940TFib_HS604TFib_HS742TFib_NCIH1975Fib_OUMS27Fib_YD8Fib_HS870TFib_HS895TFib_HS737TFib_HS739TFib_HS606TFib_HS616TFib_HS688(A)TFib_OSRC2Fib_HS675TFib_HLFAFib_HS839TFib_HS618TFib_HS840TFib_U87MGFib_SNU489Fib_U343SET1_Pr

i.WIBR3.1.ESCSET1_Naive.WIBR3.2.ESCSET1_Naive.WIBR3.5.ESCSET1_Naive.WIBR3.1.ESCSET1_Naive.WIBR3.3.ESCSET1_ESC.ES2.rep2SET1_ESC.H1SET1_ESC.HS181.p50SET1_WIBR3.ESC.1SET1_WIBR3.ESCSET1_WIBR1.ESCSET1_WIBR2.ESCSET1_ESC.H9.2SET1_WIBR3.ESC.5.pO2SET1_CA2.ESC.Ctrl.DSET1_CA2.ESC.Ctrl.ESET1_CA1.ESC.Ctrl.HSET1_ESC.H9.3SET1_BJ1.fibroblast.47SET1_BGO1.ESCSET1_BGO1.ESC.mTESRSET1_BGO1.Nanog.tr

ansgenic.ESC.lineSET1_GFP..ESC.H1SET1_ESC.H1LSET1_ESC.H9SET1_ESC.H14ASET1_ESC.H7SET1_ESC.H13BSET1_HUES8.ESC.rep2SET1_HUES8.ESC.rep1SET1_HUES8.ESC.rep3SET1_ESC.H9.rep2.2SET1_ESC.H9.rep1.2SET1_ESC.H9.rep3.2SET1_CA1.ESC.Ctr

l.ISET1_ESC.CSES4p37.mRNASET1_ESC.rep2SET1_ESC.Cyt25.rep3SET1_ESC.Cyt25.rep1SET1_ESC.Cyt25.rep2SET1_ESC.HUES6.rep3SET1_ESC.HUES6.rep1SET1_ESC.HUES6.rep2SET1_H9.ESC.20..O2.rep1SET1_H9.ESC.4..O2.rep1SET1_ESC.H1.20..O2.rep1SET1_ESC.H1.4..O2.rep2SET1_ESC.rep1SET1_ESC.H1.rep1SET1_ESC.H1.rep2SET1_ESC.H9.rep1.3SET1_ESC.H9.rep2.3SET1_ESC.VUB03.DM1.rep3SET1_ESC.VUB03.DM1.rep2SET1_ESC.SA01.rep2SET1_ESC.SA01.rep2.1SET1_ESC.HS235SET1_ESC.VUB03.DM1.rep1SET1_ESC.HD83.p24.abnormal.VHLSET1_ESC.HD90.p8SET1_ESCHD129.p6.CFSET1_ESC.VUB01.rep2SET1_ESC.VUB01.rep2.1SET1_ESC.VUB01.rep3SET1_ESC.VUB01.rep3.1SET1_ESCs.H1.rep2SET1_ESCs.H1.rep3SET1_ESCs.H1.rep1SET1_ESC.VUB01.rep1SET1_ESC.VUB01.rep1.1SET1_ESC.T3.rep1SET1_ESC.T3.rep2SET1_ESC.T3.rep3SET1_ESC.SA01.rep1SET1_ESC.SA01.rep1.1SET1_ESC.SA01.rep3SET1_ESC.SA01.rep3.1SET1_ESC.BG01SET1_ESC.HSF1SET1_ESC.H9.1SET1_ESC.chHES.20SET1_ESC.H1.20..O2.rep2SET1_ESC.H9.rep3SET1_ESC.H9.rep1SET1_ESC.H9.rep2SiT

ayeb_ESC.H9.rep1SiTayeb_ESC.H9.rep2SET1_ESC.H9.rep3.1SiTayeb_ESC.H9.rep3SET1_ESC.H9.rep1.1SET1_ESC.H9.rep2.1SET1_H9.ESC.rep1SET1_H9.ESC.rep2SET1_H9.ESC.rep3SET2_ESC.H9.1.day.in.UM...4h.in.UM.GTA.rep1SET2_ESC.H9.1.day.in.UM...4h.in.UM.GTA.rep2SET2_ESC.H9.1.day.in.UM.rep2SET2_ESC.H9.1.day.in.UM...4h.in.MEF.CM.rep2SET2_ESC.H9.1.day.in.UM.rep1SET2_ESC.H9.1.day.in.UM...4h.in.MEF.CM.rep1SET2_ESC.H9...bFGFSET2_ESC.H9...bFGF.1SET2_ESC.H9..bFGFSET2_ESC.H9..bFGF.1ESC_SNU398SET2_ESC.Miz6.2SET2_ESC.Miz4.2SET2_ESC.Miz5.2SET2_ESC.Miz5.1SET2_ESC.Miz6.1SET2_ESC.BG02.1SET2_ESC.H1.1SET2_ESC.H1.2SET2_ESC.H1.in.N2B27.ActivinA.12h.rep2SET2_ESC.H1.in.N2B27.ActivinA.36h.rep1SET2_ESC.H1.in.N2B27.ActivinA.36h.rep2SET2_ESC.H1.in.N2B27.0h.rep1SET2_ESC.H1.in.N2B27.0h.rep2SET2_ESC.H1.in.N2B27.ActivinA.24h.rep2SET2_ESC.H1.in.N2B27.ActivinA.12h.rep1SET2_ESC.H1.in.N2B27.ActivinA.24h.rep1SET2_ESC.CyT25SET2_ESC.ES4SET2_ESC.ES2SET2_ESC.ES3SET2_ESC.H1.in.N2B27.rep1SET2_ESC.H1.in.N2B27.rep2SET2_ESC.H1.in.N2B27.BMP4.2hSET2_ESC.H1.in.N2B27.rep2.1SET2_ESC.H1.in.N2B27.rep1.1SET2_ESC.H1.in.N2B27.ActivinA.2hSET2_ESC.Miz4.1SET2_ESC.H9.10SET2_ESC.H9.11SET2_ESC.H9.P51SET2_ESC.H9.P91SET2_ESC.H9.P50SET2_ESC.H9.P92SET2_ESC.BG02.2SET2_ESC.H9.2SET2_ESC.H9.4SET2_ESC.BG03.3SET2_ESC.H9.1SET2_ESC.H9.3SET2_ESC.BG01SET2_ESC.CyT203SET2_ESC.H1.in.N2B27.ActivinA.72h.rep2SET2_ESC.H1.in.N2B27.ActivinA.72h.rep1SET2_ESC.H1.in.N2B27.ActivinA.96h.rep1SET2_ESC.H1.in.N2B27.ActivinA.96h.rep2SET2_ESC.H1.in.N2B27.ActivinA.48h.rep1SET2_ESC.H1.in.N2B27.ActivinA.48h.rep2SET2_ESC.BG03.1SET2_ESC.BG03.2SET2_ESC.BG03.4SET2_ESC.HSF6.1SET2_ESC.H9.rep2SET2_ESC.H9.rep1SET2_ESC.H9.rep3SET2_ESC.HUES21SET2_ESC.HUES20SET2_ESC.HUES13.1SET2_ESC.HUES13.2SET2_ESC.H9.5SET2_ESC.H9.6SET2_ESC.HSF6.2SET2_ESC.HUES9SET2_ESC.HUES22SET2_ESC.HUES7.1SET2_ESC.HUES7.2ESC_LN18ESC_SBC5ESC_KURAMOCHIESC_COLO741ESC_IGR1ESC_SKMEL2ESC_KMS26ESC_WM983BESC_CALU6ESC_MEWOESC_GCTESC_HS944TESC_U

ACC62ESC_L1236ESC_MOTN1 ESC_L540ESC_NUDHL1ESC_KMS21BMESC_L428ESC_KHM1BESC_SUPHD1ESC_G402ESC_HGC27ESC_MFE296ESC_HUTU80ESC_O

VSAHO

Co− clustering of rank product selected cancer cell lines with ESCs and Fibroblasts for 13 genes

SET2_Fibroblast

NCL EGFR HRAS NPM1 KRAS NRAS VIM ITGA5 PTRF CAV1 LGALS3 CAV2 LGALS1

−3

signature is capable of predicting the stemness property of cancer cell lines and also its response to a CSC inhibitor (Figure 12).

Figure 12: Drug response levels of ESC-like and fibroblast-like cancer cell lines to salinomycin and staurosporine. Logarithm of IC50 values were obtained from a drug dose response curve. p values were obtained by one-side Wilcoxon rank sum test.

We further hypothesized that patient tumor samples exhibiting the gene expression signature associated with stemness property should present differences in their clinical characteristics. To assess that, we performed correlation analysis to identify the patient tumor samples in The Cancer Genome Atlas (TCGA) dataset that were displaying ESC-like and fibroblast-like gene expression signature. Interestingly, we found that ESC-fibroblast-like patient samples were associated with lower survival probability, when compared to non-ESC like samples. As expected, fibroblast-like samples did not show the same difference (Figure 13), suggesting that patient tumors that are more cancer stem cell like are more aggressive.

ESC-like Fibroblast-like 7.5

8.0 8.5 9.0

log (IC50)

p = 0.03

ESC-like Fibroblast-like 4.5

5.0 5.5 6.0 6.5 7.0

log (IC50)

p = 0.025

Salinomycin Staurosporine

Figure 13: Survival analysis of patient tumor samples from TCGA, defined as ESC-like (rank correlation >= 0.6) and Non ESC-like (rank correlation <= 0.2) and Fibroblast-like like (rank correlation >= 0.6) and Non fibroblast-like (rank correlation <= 0.2) based on correlation with gene expression signature with ESCs and fibroblasts.

0 5 10 15 20 25 30

0.00.20.40.60.81.0

Survival plot for ESC like clinical samples

Time in years

Survival probability

Cox PH P−value = 2.3e−09 Cox PH HR = 2.228 [1.713−2.897]

Non−ESC ESC

0 5 10 15 20 25 30

Time in years 0.0

0.2 0.4 0.6 0.8 1.0

Survival probability

Non ESC-like ESC-like

0 5 10 15 20 25 30

0.00.20.40.60.81.0

Survival plot for Fibroblast like clinical samples

Time in years

Survival probability

Cox PH P−value = 0.486 Cox PH HR = 1.074 [0.878−1.313]

Non−fib Fib

0 5 10 15 20 25 30

Time in years 0.0

0.2 0.4 0.6 0.8 1.0

Survival probability

Non Fibroblast-like Fibroblast-like p = 0.486 p = 6.1 10^-7