TGF-β-induced stromal CYR61 promotes resistance to gemcitabine in pancreatic ductal adenocarcinoma through downregulation of the nucleoside transporters hENT1 and hCNT3
TGF-β-induced stromal CYR61 promotes resistance to gemcitabine in pancreatic ductal adenocarcinoma through downregulation of the nucleoside transporters hENT1 and hCNT3
Rachel A.Hesler 2
Jennifer J.Huang 2
Mark D.Star 1
Victoria M.Trebosch 1
Alyssa G.Bernanke 1
Andrew B.Nixon 1
Shannon J.McCal 0
Rebekah R.White 3
Gerard C.Blob e 1 2
0 ,Department of Pathology
1 D,ivision of Medical Oncology, Department of Medicine
2 Department of Pharmacology and Cancer Biology
3 Department of Surgery, Duke University , B354 LSRC Research Drive, Box 91004, Durham, NC 27708 , USA
Pancreatic ductal adenocarcinoma (PDAC) is a lethal cancer in part due to inherent resistance to chemotherapy, including the first-line drug gemcitabine. Although low expression of the nucleoside transporters hENT1 and hCNT3 that mediate cellular uptake of gemcitabine has been linked to gemcitabine resistance, the mechanisms regulating their expression in the PDAC tumor microenvironment are largely unknown. Here, we report that the matricellular protein cysteine-rich angiogenic inducer 61 (CYR61) negatively regulates the nucleoside transporters hENT1 and hCNT3. CRISPR/Cas9-mediated knockout of CYR61 increased expression of hENT1 and hCNT3, increased cellular uptake of gemcitabine and sensitized PDAC cells to gemcitabine-induced apoptosis. In PDAC patient samples, expression ohfENT1 and hCNT3 negatively correlates with expression oCfYR61. We demonstrate that stromal pancreatic stellate cells (PSCs) are a source of CYR61 within the PDAC tumor microenvironment. Transforming growth factoβr(-TGF-β) induces the expression of CYR61 in PSCs through canonical TGFβ--ALK5-Smad2/3 signaling. Activation of TGFβ- signaling or expression of CYR61 in PSCs promotes resistance to gemcitabine in PDAC cells in ain vitro co-culture assay. Our results identify CYR61 as a TGβF-i-nduced stromal-derived factor that regulates gemcitabine sensitivity in PDAC and suggest that targeting CYR61 may improve chemotherapy response in PDAC patients.
PDAC is the fourth leading cause of cancer death in the United recently, in combination with nab-paclitaxel. Gemcitabine is -uti
States, with more than 4000 patient deaths per year 1(). lized in first- and second-line treatment for locally advanced and
Moreover, PDAC is projected to become the second leading cause metastatic PDAC, as well as adjuvant therapy for these patients.
of cancer death by 2030 due to a rising incidence and the lack Incorporation of gemcitabine into DNA results in masked-chain
of improvement in survival compared with other cancers 2(). termination, which stops DNA synthesis and induces apoptosis
PDAC has one of the lowest 5-year survival rates at 61%), (under- of the cell 3(). Although gemcitabine is one of the most
comscoring the need for better treatment options. Gemcitabine is a monly used treatments for PDAC, as a single agent it prolongs
nucleoside pyrimidine analog that has long been the backbone median survival by just over a month and is not effective for
of chemotherapy for PDAC, both as a single agent, and more all patients4(). Attempts to enhance gemcitabine efficacy with
Abbreviations pyruvate and 10% fetal bovine serum (FBS). MiaPaCa-2 cells were
grown in DMEM with 1mM sodium pyruvate, 10% FBS and 2.5%
α-SMA α-smooth muscle actin horse serum. CFPAC-1 cells were grown in Iscove’s Modified
CA-ALK5 constitutively active ALK5 Dulbecco’s Medium (IMDM) with 10% FBS. BxPC3 cells were
CYR61 cysteine-rich angiogenic inducer 61 grown in RPMI-1640 media containing 1mM sodium pyruvate,
dCK deoxycytidine kinase 10 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic
ECM extracellular matrix acid) and 10% FBS. Conditioned media (CM) from cells was
conEMT epithelial-to-mesenchymal transition centrated by centrifugation using an Amicon Ultra-15 cellulose
hCNT3 human concentrative nucleoside transporter-3 filter with a molecular weight cutoff ofkD3a from Millipore
hENT1 human equilibrative nucleoside transporter-1 (Billerica, MA). Chemical inhibitors against ALK5 (SB431542), p38
NBMPR S-(4-Nitrobenzyl)-6-thioinosine MAPK (SB203580) and PI3K (LY294002) were purchased from Cell
PDAC pancreatic ductal adenocarcinoma Signaling Technology (Danvers, MA) and dissolved in DMSO.
TGFPDX patient-derived xenograft β1 ligand was purchased from R&D Systems (Minneapolis, MN).
PSC pancreatic stellate cells Gemcitabine (2, 2- difluoro-2-deoxycytidine) was purchased from
TGF-β transforming growth factoβr- the Duke Hospital Pharmacy Store Room. The hENT1 inhibitor
S-(4-Nitrobenzyl)-6-thioinosine (NBMPR) was purchased from
targeted agents or other cytotoxic agents, with the exception ofSigma-Aldrich (St Louis, MO) and dissolved in DMSO.
nab-paclitaxel, have had limited success5().
Because gemcitabine is hydrophilic, it must be transported Adenovirus
through the hydrophobic cell membrane by transmembrane HA-tagged constitutively active ALK5 adenovirus (HA-ALTK2054D)
nucleoside transporters. The equilibrative nucleoside tr-ans was provided by Dr Carlos Arteaga (Vanderbilt Universit2y1))(.
port family mediates bidirectional transport of nucleosides The luciferase control and mouse CYR61 adenoviruses were
across the plasma membrane along the concentration gradient, provided by Dr Brahim Chaqour (SUNY Downstate) 2(
whereas the concentrative nucleoside transport family conc-en Adenoviruses were generated and purified using the Adeno-X
trates nucleosides in the cell by coupling transport with -cati Maxi Purification Kit from Clontech (Mountain View, CA).
). Human equilibrative nucleoside transporter-1 (hENT1) Adenovirus titer was determined using the Adeno-X Rapid Titer
and human concentrative nucleoside transporter-3 (hCNT3) Kit from Clontech, and cells were infected at the indicated m-ul
both have important roles in the cellular uptake of the nu-cle tiplicity of infection (MOI).
oside analog gemcitabine 8(). Consistent with this role, PDAC
patients with low expression of hENT1 and hCNT3 have si-g Lentivirus
nificantly worse survival after gemcitabine treatment compared Lentivirus CRISPR constructs targeting hCYR61, rSmad2 and
with patients with high hENT1 and hCNT3 expression9(–12). rSmad3 were made using the LentiCRISPRv2 vector (Addgene
Although hENT1 expression is currently being evaluated as a Plasmid 52961) following the GeCKO protocol24(,25). Briefly, the
biomarker to predict patient response to gemcitabine13(), the lentiCRISPRv2 vector was digested by BsmB1 and de-phosphory-l
molecular mechanisms regulating hENT1 and hCNT3 expre-s ated by CIP alkaline phosphatase. sgRNA target sequences were
sion in the PDAC tumor microenvironment are largely unknown. designed using the GeCKO library 2(
) (sequences listed in
Recent studies suggest that epithelial-to-mesenchymal tra-nsi Supplementary Table S1, available atCarcinogenesis Online), and
tion (EMT) (14) and ErbB2 expression (15) negatively regulate the synthesized oligos were annealed and phosphorylated using
hENT1 and hCNT3 expression, but further studies are needed T4 polynucleotide kinase.The annealed sgRNA target sequence-oli
to identify mechanisms that regulate their expression in PDAC gos were ligated into the digested lentiCRISPRv2 backbone using T4
cells in the context of the tumor microenvironment. Here, we DNA ligase.The ligated DNA was transformed into One Shot Stabl3
investigate factors regulating hENT1 and hCNT3 expression in competent cells and selected on LB-Amp plates. Each construct
the PDAC tumor microenvironment. was sequenced to verify correct incorporation of the sgRNA target
sequence into the lentiCRISPRv2 vector. To generate lentivirus for
Methods and materials each lentiCRISPRv2 construct, a 1c0m dish of 293T cells was tran-s
fected with 4.5µ g of the respective lentiCRISPRv2 construct along
Cell culture and reagents with 2.25 µg PAX2, 0.75 µg pMD2.G and 18 µl Xtremegene. Media
PANC1, MiaPaCa-2, BxPC3, CFPAC-1 and 293T cells were obtained was changed on the 293T cells the morning after transfection. At
from American Type Culture Collection (Manassas, VA) and were 48 and 72h later, the 293T media containing lentivirus was h-ar
verified by Short Tandem Repeat analysis. After verification, cells vested and filtered through a 0.4µ5m cellulose filter. The media
were cultured for <1 month before being frozen, and all exp-eri was applied to MiaPaCa-2, PANC1 or LTC-14 cells with 6µ g/ml
ments were performed with <6 months of culturing. L3.6p cells polybrene. Stably infected cells were selected usingµg2/ ml
purowere provided by Dr Isaiah Fidler (MD Anderson)1(
). RLT-PSC mycin. Single cell clones were isolated for MiaPaCa-2 and PANC1
human pancreatic stellate cells (PSCs) were provided by Dr Ralf hCYR61 CRISPR to achieve knockout of expression. For rSmad2 and
Jesenofsky (University of Heidelberg)1(7); HPSC-T human PSCs rSmad3, the CRISPR/Cas9 vectors were stably introduced, and the
were provided by Dr Rosa Hwang (MD Anderson)1(
); LTC-14 rat bulk populations of cells with partial knockdown were used.
PSCs were provided by Dr Gisele Sparman (University Hospital of
Rostock) (19) and imPSC mouse PSCs were provided by Dr Raul Western blotting
Urrutia (Mayo Clinic)2(0). Both human and murine PSCs were Total cell lysates were harvested, boiled in sample buffer, separated
obtained directly from the labs that isolated the cells and wereby sodium dodecyl sulfate-polyacrylamide gel electrophoresis,
functionally validated by their expression patterns in the i-ndi transferred onto nitrocellulose membranes, blocked in 5% milk
cated studies. All cells were grown at 37°C at 5%2.CPOANC1, in Tris-buffered saline and incubated overnight with the primary
L3.6p, LTC-14, HPSC-T, RLT-PSC and imPSC cells were grown in antibody of interest in 5% bovine serum albumin (BSA) in
TrisDulbecco’s Modified Eagle Medium (DMEM) with 1mM sodium buffered saline (TBS)/0.1% TWEEN. Quantification was performed
Microarray and RNAseq dataset analysis
[3H]gemcitabine (16.32 µg/ml, 16.2 Ci/mmol) was purchased
from Moravek Biochemicals (Brea, CA). Cells were incubated in
transport buffer (20mM Tris/HCl, 3mM K 2HPO4, 5mM glucose,
130mM NaCl, 1mM MgCl2-6H2O and 2mM CaCl2) as previously CYR61 ELISA
described (30). Cells were plated in 12 well plates at 105000 cells/ The CYR61 ELISA kit was purchased from R&D Systems (DCYR10),
well (MiaPaCa-2) or 9000 cells/well (PANC1). The following day, and the ELISA was performed according to kit instructions. All
cells were rinsed in transport buffer then incubated with n10M0 patient serum samples were de-identified, and informed
con[3H]gemcitabine in transport buffer for m2in (MiaPaCa-2) or 30s sent was received. The study was conducted with approval of
(PANC1). When indicated, cells were pretreated for m10in with the Duke IRB. Serum was obtained from 5 cc blood at the time
doses of the hENT1 inhibitor NBMPR or DMSO control in transport of a diagnostic blood draw from subjects with confirmed PDAC.
buffer, and NBMPR or DMSO was included in 100nM
[3H]gemcitabine incubation. After incubation, cells were rinsed three times In vitro co-culture assay
with transport buffer containingµ5M NBMPR to inhibit efflux of LTC-14 or imPSC pancreatic stellate cells were infected with
[3H]gemcitabine. Cells were lysed in 1% (v/v) Triton-X-100, and adenovirus at indicated MOIs. After 24 h infection, PSCs were
protein concentration was determined using a bicinchoninic acid washed with phosphate-buffered saline, and media was replaced
(BCA) assay from Thermo Scientific (Waltham, MA). Cell lysates to start collecting CM. After 24 h, PSC CM was harvested an-d fil
were added to Ultima Gold from Perkin Elmer (Waltham, MA) tered through a 0.45 μM cellulose filter then applied to PDAC
and cell-associated radioactivity in counts per minute (CPM) was cells. After 24 h incubation, PSC CM was refreshed, and PDAC
determined using a liquid scintillation counte3rH. ][Gemcitabine cells were then treated with gemcitabine for 48 h at indicated
transport was calculated by normalizing CPM to protein con-cen doses. Adherent and floating PDAC cells were collected for w-est
tration for each well. Each condition was performed in triplicate, ern blot analysis of cleaved caspase 3 levels.
and the experiment was repeated three times.
All statistical analyses were conducted with GraphPad Prism -soft
RNA was extracted using the Quick-RNA™ MiniPrep kit from ware. For all experiments, significance was set aPt < 0.05. Allin
Zymo Research (Irvine, CA) according to kit instructions. Five vitro experiments were analyzed using parametric statistics
[twohundred nanograms of RNA was reverse transcribed using the sided t test or analysis of variance (ANOVA) with indicatepdost
iScript cDNA Synthesis Kit from BioRad (Hercules, CA) following hoc test] and expressed as the mean ± SEM. Microarray expression
data and ELISA on serum samples were analyzed using nonpa-r in cancer and negatively correlates withhENT1 and hCNT3
ametric statistics (Mann–WhitneyU, Wilcoxon matched pairs expression. Additionally, CYR61 is a secreted matricellular -pro
signed rank test or Kruskal–Wallis global test). Linear regressiontein that can be targeted using a neutralizing antibody, which
was performed on microarray data with theR2 value,P value and indicates it has the potential to be targeted clinically. CYR61 is
slope for the line of best fit reported for each comparison. Survival a member of the CCN family of matricellular proteins, which
curves were analyzed with log-rank statistics. includes connective tissue growth factor (CTGF) and nephrob-las
toma overexpressed (NOV). The CCN family regulates diverse
cell behaviors in a context-specific manner, primarily through
Results interacting with integrins and heparin sulfate proteoglycans to
activate downstream signaling31().
CYR61 promotes chemoresistance by negatively The mRNA expression of SLC29A1 (hENT1) and SLC28A3
regulating gemcitabine transport through hENT1 (hCNT3) negatively correlated witChYR61 mRNA expression
and hCNT3 in PDAC patient samples (Figures 1A and B), indicating that
To identify potential regulators of hENT1 and hCNT3 in the CYR61 may play a role in suppressing expression of the
nuclePDAC microenvironment, we analyzed a publically available oside transporters that mediate cellular uptake of gemcitabine
microarray dataset of PDAC tumor samples2(
) to identify genes in the PDAC tumor microenvironment.CYR61 expression did
whose expression significantly correlated with expression of not significantly correlate with the expression of other nu-cle
hENT1 (SLC29A1) and hCNT3 (SLC28A3) and whose expression oside transporters in PDAC patient samplesS(upplementary
is significantly altered in PDAC tumor samples compared with Figure 1B–F, available atCarcinogenesis Online), suggesting
normal adjacent tissueS(upplementary Figure 1A, available at specific regulation of hENT1 and hCNT3. To examine whether
Carcinogenesis Online). We identified 25 genes whose expres- CYR61 negatively regulated hENT1 and hCNT3 expression, we
sion significantly correlated with bothhENT1 and hCNT3 and used CRISPR/Cas9 technology to knockout CYR61 expression
whose expression is significantly altered in pancreatic cancer in two PDAC cell lines with high CYR61 expression. We co-n
(Supplementary Table 3, available atCarcinogenesis Online). We firmed that CRISPR knockout decreased the soluble secreted
were particularly interested in investigating cysteine-rich a-ngio CYR61 present in the CM (Supplementary Figure 2A and B,
genic inducer 61 (CYR61) becauseCYR61 expression is increased available atCarcinogenesis Online). CRISPR-mediated knockout
of CYR61 significantly increased hENT1 and hCNT3 expre-s and blunting the effects of gemcitabine-mediated downreg- u
sion in PANC1 cells F(igure 1C; Supplementary Figure 2C, lation S(upplementary Figure 2G, available atCarcinogenesis
available atCarcinogenesis Online). Knockout of CYR61 also Online).
significantly increased hENT1 expression in MiaPaCa-2 cells, To determine whether the CYR61 knockout-mediated
and increased hCNT3 expression, albeit with larger increases increases in hENT1 and hCNT3 in PDAC cells resulted in
in CRISPR 2 cells F(igure 1D, Supplementary Figure 2D, avail- higher cellular uptake of gemcitabine, we performed gemc-it
able atCarcinogenesis Online). In a reciprocal manner, aden-o abine transport assays using radiolabele3dH-gemcitabine as
virus-mediated overexpression of CYR61 in BxPC3 and CFPAC previously described (30). Increasing doses of the hENT1
specells, which have low basal CYR61 expression, significantly cific inhibitor NBMPR dramatically decreased the levels of
decreased hENT1 expression (Figure 1E and F, Supplementary 3H-gemcitabine transported into the cell, showing specificity
Figure 2E and F, available atCarcinogenesis Online). All cell lines of the assay and supporting hENT1 as the major gemcitabine
had low basal expression of hCNT3 as previously reported for transporter in MiaPaCa-2 and PANC1 cellsFi(gure 2A).
CRISPRin vitro cell culture condition1s4(), so overexpression of CYR61 mediated knockout of CYR61 significantly increased the amount
in BxPC3 and CFPAC cells was not able to further decrease of 3H-gemcitabine transported into PANC1 F(igure 2B) and
these low basal levels of hCNT3 expression. In PANC1 cells, MiaPaCa-2 cellsF(igure 2C). These data indicate that the upreg-u
treatment with gemcitabine for 4h8 induced downregulation lation of hENT1 and hCNT3 following CRISPR knockout of CYR61
of hENT1, with knockout of CYR61 increasing hENT1 levels results in enhanced cellular uptake of gemcitabine.
While nucleotide transporters are an important mechanism for
regulating entry of gemcitabine into PDAC cells, there are several
additional mechanisms that regulate resistance to gemcitabine
in PDAC. Deoxycytidine kinase (dCK) phosphorylates gemcitabine
to its active form once it enters the cytoplasm, and low expr-es
CYR61 expression is increased in PDAC sion of dCK is associated with worse survival after gemcitabine
Bioinformatic analysis of a microarray dataset demonstrated treatment 9(,10). Additionally, high expression of the subunits of
that CYR61 expression was increased in PDAC samples com- the enzyme ribonucleotide reductase (RRM1 and RRM2), which
pared with matched normal adjacent tissueF(igure 3A and B), catalyzes the conversion of ribonucleotides to deoxynucleosides,
supporting increased CYR61 expression in PDAC, consistent is associated with gemcitabine resistance in patients33(–35).
with a prior report 3(
). Further analysis demonstrated that The expression of ATP-binding cassette transporters, which act
patients with familial PanIN precursor lesions27() have an as drug efflux pumps, is also associated with drug resistance in
intermediate level ofCYR61 (Figure 3C). Although these asses-s PDAC (36). The drug efflux pumps ABCB1 (MDR1/P-glycoprotein)
ments at the mRNA level were suggestive, serum protein l-ev and ABCC1 (MRP1) have been linked to the resistance of PDAC
els of CYR61 in PDAC patients have not been investigated. Here cells to gemcitabine3(7,38). CYR61 has been previously reported to
we demonstrate that CYR61 protein expression is significantly regulate expression of the drug efflux pump MDR1/P-glycoprotein
elevated in the serum of PDAC patients, with a mean expression in renal cell carcinoma3(9), but it has not been studied in PDAC.
We examined the effect of CYR61 on the expression of these fa-c perhaps the most prominent of all epithelial cancers. This
tors associated with gemcitabine resistance. Either overexp-res stroma contains PSCs, which are the predominant cells respo-n
sion of CYR61 or CRISPR-mediated knockdown of CYR61 did not sible for secretion of the extracellular matrix (ECM) components
alter the expression levels of MDR1 or dCK at the protein level that comprise the fibrotic stroma4(
). The microarray
data(Supplementary Figure 4A–C, available atCarcinogenesis Online). set used to examine CYR61 expression analyzed whole-tissue
Additionally, there was no consistent or significant effect of tumor samples that include both cancer and stromal cel2ls6)(,
CYR61 expression on the mRNA level ofRRM1, RRM2 or ABCC1 suggesting that the PSCs might be a source of CYR61. To det-er
(Supplementary Figure 4D–F, available atCarcinogenesis Online). mine whether PSCs within the tumor microenvironment secrete
Moreover, in PDAC patient samples, the level oCfYR61 did not CYR61, we examined RNAseq data that analyzed gene expre-s
significantly correlate with expression of these gemcitabine sion in PDAC tumors as well as three cell population isolated
resistance factors S(upplementary Figure 5A–E, available at from the tumor: PSCs, tumor epithelial cells grown in
patientCarcinogenesis Online). These data suggest that CYR61 functions derived xenografts (PDXs) and tumor epithelial cells growninin
to mediate resistance to gemcitabine largely through its effects vitro cell culture2(9). Isolated PSCs, identified asα-SMA positive,
on the nucleotide transporters hENT1 and hCNT3. vimentin positive and EpCam negative (29), expressed
significantly higher levels oCfYR61 compared with human tumor epi
Pancreatic stellate cells are a source of CYR61 in the thelial cells in patient-derived xenograftsinorvitro cell culture
PDAC tumor microenvironment (Figure 4A). PDAC samples expressed an intermediate amount,
PDAC is characterized by an abundant fibrotic stroma that can suggesting that the CYR61 from these samples is derived partly
comprise up to 80% of the tumor volume (40), making this stroma from stromal cells present in the samples. Isolated tumor
epithelial cells grownin vitro also had higher CYR61 expression of CYR61 by TGF-β was a direct effect and not via induction of
than tumor cells grown in PDXFi(gure 4A), suggesting that some other growth factorsFi(gure 5Eand F).
PDAC epithelial cells may express higher levels of CYR61 iinn TGF-β induces activation of canonical Smad signaling in
vitro cell culture conditions as compensation for the lack of s-tro PSCs, but TGF-β also induces activation of several noncanonical
mal-derived factors that are presenint vivo. signaling pathways, including p38 MAPK and PI3K-Akt signaling
To investigate expression of CYR61 at the protein level, we (Figure 5G; Supplementary Figure 8A, available atCarcinogenesis
performed IHC staining for CYR61 on human PDAC tissue. PSCs Online). To determine which downstream signaling pathways
in the tumor microenvironment were identified by staining were important for TGF-β-induced CYR61 expression in PSCs,
for α-SMA on consecutive slides, andα-SMA staining in mus- we pretreated PSCs with kinase inhibitors against ALK5, p38
cular arterial wall and duodenal smooth muscle (muscu-la MAPK and PI3K and examined the effect on TGF-β-induced
ris propria) overlying head of pancreas was used as a positive CYR61 expression. Treatment with the ALK5 inhibitor, but not
control S(upplementary Figure 6A, available atCarcinogenesis the p38 MAPK or PI3K-Akt inhibitor, blocked TGFβ- induced
Online). IHC on human PDAC samples demonstrated co-local-i CYR61 expression (Figure 5H; Supplementary Figure 8B,
availzation of CYR61 staining with PSCs labeled bαy-SMA (Figure 4B; able atCarcinogenesis Online). We confirmed that these inhibitors
Supplementary Figure 6, available atCarcinogenesis Online), indi- blocked activation of downstream targets in the LTC-14 PSCs
cating that PSCs are a source of CYR61 in the tumor micro-en (Supplementary Figure 8C–E, available atCarcinogenesis Online).
vironment. Consistent with a previous report32(), most PDAC Further, CRISPR-mediated knockdown of Smad2 and Smad3 in
epithelial cells also stained positive for CYR61Fig(ure 4B; the LTC-14 PCSs decreased TGF-β-induced CYR61 expression
Supplementary Figure 6, available atCarcinogenesis Online), (Figure 5I), with the level of knockdown of both Smads cor-re
although some PDAC epithelial cells showed weak CYR61 sta-in lating with the loss of TGβF-induction. In addition, expression
ing (Supplementary Figure 6D and E, available atCarcinogenesis of CA-ALK5 induced CYR61 expression in LTC-14 and imPSC
Online). α-SMA negative fibroblasts near acinar cells did not cells S(upplementary Figure 8F and G, available atCarcinogenesis
stain positive for CYR61 S(upplementary Figure 6H and I, avail- Online). These results suggest that TGβFi-nduces CYR61
expresable atCarcinogenesis Online). sion in PSCs through canonical TGFβ--ALK5-Smad signaling.
In addition, we evaluated expression of CYR61 in five PDAC TGF-β did not induce CYR61 in Smad4-null cells (CFPAC
cell lines and two human PSC cell lines, HPSC-T18() and RLT-PSC and BxPC3) or TGF-β nonresponsive MiaPaCa-2 cells 4(
(17), which were isolated from PDAC and chronic pancreatitis (Supplementary Figure 9A and B, available atCarcinogenesis
samples, respectively. We validated the identity of PSCs by co-n Online). Human PSC cell lines HPSC-T and RLT-PSC had high
firming expression of the PSC-specific markers α-SMA, vimen - basal CYR61 expression (Figure 4C), likely because they were
tin, collagen α11 and desmin ( Supplementary Figure 7, available isolated from PDAC and chronic pancreatitis samples. However,
at Carcinogenesis Online). The mesenchymal PSCs expressed ELISA data demonstrate that activation of TGβF-signaling by
high levels of the ECM protein fibronectin, whereas most PDAC expression of a constitutively active version of the
TGβFr-ecepcells expressed higher levels of the epithelial marker E-cadherin tor ALK5 (CA-ALK5) induced even higher secretion of CYR61
(Figure 4C). The PSC cell lines had higher CYR61 expression than (Supplementary Figure 9B). Although CYR61 expression
sigthe majority of PDAC cell linesFi(gure 4C). Interestingly, the nificantly correlated with expression of the Hippo target gene
PANC1 cell line also had high expression of CYR61Fi(gure 4C). AXL in PDAC samples ( Supplementary Figure 10A, available at
Consistent with a role for CYR61 in gemcitabine resistance, Carcinogenesis Online), activation of Hippo signaling in LTC-14
pancreatic cancer cell lines that express higher levels of CYR61 PSCs through expression of constitutively active YAP (YAP5SA)
were more resistant to gemcitabine-induced apoptosisin vitro only moderately induced CYR61 expression relative to
TGβF(Figure 4Dand E). treatmentS( upplementary Figure 10B, available atCarcinogenesis
TGF-β signaling induces CYR61 expression in PSCs
in the PDAC tumor microenvironment TGF-β-Induced CYR61 promotes gemcitabine
CYR61 has been demonstrated to be regulated by both the tra-ns resistance in an in vitro co-culture assay
forming growth factorβ- (TGF-β) signaling and Hippo-YAP/TAZ To examine the role of TGF-β-induced CYR61 in PSCs on
gemsignaling pathways 4(
). TGF-β ligand expression is elevated citabine-induced apoptosis of cancer cells, we established ainn
in PDAC, and patients with high levels of TGFβ-1 ligand in their vivo co-culture assay where PDAC cells were treated with CM
serum have a significantly worse prognosis4(
). However, muta- from PSCs with activated TGFβ- signaling F(igure 6A). To exam
tions that inactive the canonical TGβF--Smad signaling pathway ine the role of stromal CYR61, LTC-14 PSCs were infected with
are common in PDAC, with around 55% of PDAC patients having adenoviruses to express CYR61 or CA-ALK5 with a luciferase
inactivating mutations inSMAD4 (47). Therefore, elevated TGFβ- adenovirus used as a controlF(igure 6B). We verified that act-i
may negatively affect PDAC progression or therapy response in vation of TGF-β signaling in LTC-14 PSCs through expression
part through stromal cells with intaScMtAD4, including PSCs. of CA-ALK5 releases soluble CYR61 into the CMFi(gure 6C). CM
Consistent with this hypothesis, in the microarray dataset that from PSCs with CYR61 expression or CA-ALK5 protected CFPAC
analyzed gene expression in whole-tissue PDAC samplesC,YR61 cells from gemcitabine-induced apoptosis as shown by reduced
expression significantly correlated with expression of tThGeFB1 levels of cleaved caspase 3 F(igure 6D). Similar results were
ligand and also with the well-established TGβFt-arget genes obtained in the BxPC3 cell lineSu(pplementary Figure 10C). In
SERPINE1 (PAI-1) and SMAD7 (Figure 5A–C). Moreover, in the rat contrast, expression of constitutively active YAP in LTC-14 PSCs
PSC cell line LTC-141(9) and the mouse PSC cell line imPSC 2(0), only weakly induced CYR61 and did not affect the
gemcitabinewhich were isolated from normal pancreas and have low basal induced apoptosis of PDAC cells in ourni vitro co-culture model
CYR61 expression, TGF-β induced CYR61 expression in a dose- (Supplementary Figure 10D, available atCarcinogenesis Online),
dependent fashion (Figure 5D). TGF-β induced CYR61 protein suggesting that TGFβ- signaling was the primary pathway
expression as early as 6h post treatment and mRNA expression inducing CYR61 expression in PSCs. Finally,in silico analysis of
as early as 3h post treatment, suggesting that the induction whole-tissue PDAC samples demonstrated thatTGFB1 ligand
expression negatively correlates with expression oSfLC29A1 role in promoting gemcitabine resistance through downregu-la
(hENT1) and SLC28A3 (hCNT3) (Figure 6E and F), suggesting that tion of the nucleoside transporters hENT1 and hCNT3.
TGF-β plays a role in regulating these nucleoside transporters We have demonstrated that TGFβ- strongly induces CYR61
in vivo. expression in PSCs. A recent phase II clinical trial suggested
that addition of the TGFβ- inhibitor galsunisertib to gemc-it
Discussion abine led to improved overall and progression-free survival in
PDAC patients compared with gemcitabine alone 5(
). Due to
Chemotherapy resistance is a major clinical problem in PDAC. the pleotropic homeostatic functions of TGFβ-, global inhibition
Even one of the commonly used first-line agents, gemcitabine, of TGF-β signaling does have the potential to have side effects.
has a very low response rate and only modestly prolongs s-ur Therefore, understanding the specific downstream effectors of
vival. Therefore, it is important to understand the cellular m-ech TGF-β signaling in PDAC is important for development of future
anisms that regulate resistance to gemcitabine. Our results therapies. Our results indicate that stromal TGβFs-ignaling
prodemonstrate that CYR61 promotes resistance to gemcitabine motes resistance to gemcitabine in PDAC cells via induction of
predominantly by modulating the levels of the nucleoside tra-ns CYR61. TGF-β is a major driver of EMT (54), and EMT has recently
porters that mediate cellular uptake of gemcitabine. The role of been shown to regulate gemcitabine resistance and expression
the stroma in therapy resistance is an emerging area of interest of hENT1 and hCNT3 in PDAC (14). Whether additional
TGFβ-in PDAC with recent genetic mouse models and clinical trials induced genes promote gemcitabine resistance through reg-u
that target PSCs having conflicting results, with both positivelation of these nucleoside transporters, and whether CYR61
and negative effects on cancer progression and response to ge-m mediates the effects of TGF-β on EMT in PDAC (32), remains to
citabine (49–52). Understanding what aspects of PSCs promote be established.
therapy resistance and identifying signaling mechanisms that CYR61 is a member of the CCN family of matricellular pr-o
regulate this are important to effectively target the stroma. Hereteins, which includes CTGF and NOV. Targeting the CCN fa-m
we demonstrate that stromal-derived CYR61 has an important ily member CTGF in combination with gemcitabine in PDAC
has shown promise in a preclinical mouse model (55), and the Acknowledgements
CTGF neutralizing antibody FG-3019 is being tested in an on-go
ing clinical trial56(). However, although CTGF and CYR61 are The authors thank Tam How for assistance with adenovirus
structurally similar, the FG-3019 neutralizing antibody does preparation, Angela Gaviglio for help with microarray dataset
not interact with CYR6157(). The current results regarding the analysis and Melissa Hector-Greene for assistance with de-vel
role of CYR61 in gemcitabine resistance provide a rationale for opment of CRISPR/Cas9 constructs.
inhibiting CYR61 or both CTGF and CYR61 in combination with Conflict of Interest Statement: None declared.
gemcitabine (and nab-paclitaxel) in PDAC patients. Although
our study is the first to identify a role for CYR61 in nucleoside References
transporter expression and gemcitabine resistance, CYR61 has
recently been implicated in therapy resistance in the contexts of
several other cancers, including renal cell carcinom39a),(breast
cancer (58,59), acute myeloid leukemia (60) and ovarian cancer
(61), which indicates that CYR61 may be a promising target to
improve efficacy of chemotherapy.
In summary, here we have identified that the matricellular
protein CYR61 is induced by the TGF-β-ALK5-Smad signaling
pathway in PSCs in the tumor microenvironment, where it ne-g
atively regulates the expression of the nucleoside transporters
hENT1 and hCNT3 in PDAC to mediate resistance to gemcitabine.
10. Maréchal, R. et al. (2012) Levels of gemcitabine transport and meta-bo 35. Nakahira, S. et al. (2007) Involvement of ribonucleotide reductase M1
lism proteins predict survival times of patients treated with gem-cit subunit overexpression in gemcitabine resistance of human pancr-e
abine for pancreatic adenocarcinoma. Gastroenterology, 143, 664–674.e1. atic cancer. Int. J. Cancer, 120, 1355–1363.
11. Liu, Z.Q. et al. (2014) Prognostic value of human equilibrative nucl-eo 36. König, J. et al. (2005) Expression and localization of human multidrug
side transporter1 in pancreatic cancer receiving gemcitabin-based resistance protein (ABCC) family members in pancreatic carcinoma.
chemotherapy: a meta-analysis. PLoS One, 9, e87103. Int. J. Cancer, 115, 359–367.
12. Maréchal, R. et al. (2009) Human equilibrative nucleoside transporter 37. Nath, S. et al. (2013) MUC1 induces drug resistance in pancreatic cancer
1 and human concentrative nucleoside transporter 3 predict survival cells via upregulation of multidrug resistance genes. Oncogenesis, 2, e51.
after adjuvant gemcitabine therapy in resected pancreatic adenoca-rci 38. Zhang, W. et al. (2013) Emodin sensitizes the gemcitabine-resistant cell
noma. Clin. Cancer Res., 15, 2913–2919. line Bxpc-3/Gem to gemcitabine via downregulation of NFκ-B and its
13. Elnaggar, M. et al. (2012) Molecular targets of gemcitabine action: regulated targets. Int. J. Oncol., 42, 1189–1196.
rationale for development of novel drugs and drug combinations. Curr. 39. Long Q.Z. et al. (2013) Interaction of CCN1 with alphavbeta3 integrin
Pharm. Des., 18, 2811–2829. induces P-glycoprotein and confers vinblastine resistance in renal cell
14. Zheng, X. et al. (2015) Epithelial-to-mesenchymal transition is disp-en carcinoma cells. Anticancer Drugs, 24, 810–817.
sable for metastasis but induces chemoresistance in pancreatic c-an 40. Erkan M. et al. (2012) The impact of the activated stroma on panc-re
cer. Nature, 527, 525–530. atic ductal adenocarcinoma biology and therapy resistance. Curr. Mol.
15. Skrypek, N. et al. (2015) The oncogenic receptor ErbB2 modulates gem- Med., 12, 288–303.
citabine and irinotecan/SN-38 chemoresistance of human pancreatic 41. Bachem, M.G. et al. (2005) Pancreatic carcinoma cells induce fibrosis by
cancer cells via hCNT1 transporter and multidrug-resistance as-soci stimulating proliferation and matrix synthesis of stellate cells. Ga-stro
ated protein MRP-2. Oncotarget, 6, 10853–10867. enterology, 128, 907–921.
16. Bruns, C.J. et al. (1999)In vivo selection and characterization of me-ta 42. Yen, T.W. et al. (2002) Myofibroblasts are responsible for the desmopl-as
static variants from human pancreatic adenocarcinoma by using tic reaction surrounding human pancreatic carcinomas. Surgery, 131,
orthotopic implantation in nude mice. Neoplasia, 1, 50–62. 129–134.
17. Jesnowski, R. et al. (2005) Immortalization of pancreatic stellate cells 43. Apte, M.V. et al. (2004) Desmoplastic reaction in pancreatic cancer: role
as an in vitro model of pancreatic fibrosis: deactivation is induced by of pancreatic stellate cells. Pancreas, 29, 179–187.
matrigel and N-acetylcysteine. Lab. Invest., 85, 1276–1291. 44. Jun, J.I. et al. (2011) Taking aim at the extracellular matrix: CCN proteins
18. Hwang, R.F. et al. (2008) Cancer-associated stromal fibroblasts promote as emerging therapeutic targets. Nat. Rev. Drug Discov., 10, 945–963.
pancreatic tumor progression. Cancer Res., 68, 918–926. 45. Bartholin, L. et al. (2007)The human Cyr61 gene is a transcriptional target of
19. Sparmann, G. et al. (2004) Generation and characterization of imm-or transforming growth factor beta in cancer cells. Cancer Lett., 246, 230–236.
talized rat pancreatic stellate cells. Am. J. Physiol. Gastrointest. Liver46. Javle, M. et al. (2014) Biomarkers of TGFβ- signaling pathway and
progPhysiol., 287, G211–G219. nosis of pancreatic cancer. PLoS One, 9, e85942.
20. Mathison, A. et al. (2010) Pancreatic stellate cell models for transcriptional47. Hahn, S.A. et al. (1996) DPC4, a candidate tumor suppressor gene at
studies of desmoplasia-associated genes. Pancreatology, 10, 505–516. human chromosome 18q21.1. Science, 271, 350–353.
21. Ueda, Y. et al. (2004) Overexpression of HER2 (erbB2) in human breast 48. Simeone, D.M. et al. (2000) Disruption of TGFbeta signaling pathways in
epithelial cells unmasks transforming growth factor beta-induced cell human pancreatic cancer cells. Ann. Surg., 232, 73–80.
motility. J. Biol. Chem., 279, 24505–24513. 49. Rhim, A.D. et al. (2014) Stromal elements act to restrain, rather than
22. Liu, H. et al. (2008) Cysteine-rich protein 61 and connective tissue support, pancreatic ductal adenocarcinoma. Cancer Cell, 25, 735–747.
growth factor induce deadhesion and anoikis of retinal pericytes. 50. Sherman, M.H. et al. (2014) Vitamin D receptor-mediated stromal
Endocrinology, 149, 1666–1677. reprogramming suppresses pancreatitis and enhances pancreatic ca-n
23. Hasan, A. et al. (2011) The matricellular protein cysteine-rich protein cer therapy. Cell, 159, 80–93.
61 (CCN1/Cyr61) enhances physiological adaptation of retinal vessels 51. Özdemir, B.C. et al. (2014) Depletion of carcinoma-associated fibr-o
and reduces pathological neovascularization associated with ischemic blasts and fibrosis induces immunosuppression and accelerates pa-n
retinopathy. J. Biol. Chem., 286, 9542–9554. creas cancer with reduced survival. Cancer Cell, 25, 719–734.
24. Sanjana, N.E. et al. (2014) Improved vectors and genome-wide libraries 52. Provenzano, P.P. et al. (2012) Enzymatic targeting of the stroma ablates
for CRISPR screening. Nat. Methods, 11, 783–784. physical barriers to treatment of pancreatic ductal adenocarcinoma.
25. Shalem, O. et al. (2014) Genome-scale CRISPR-Cas9 knockout screening Cancer Cell, 21, 418–429.
in human cells. Science, 343, 84–87. 53. Melisi, D. et al. (2016) A phase II, double-blind study of
26. Badea, L. et al. (2008) Combined gene expression analysis of whole-t-is galunisertib+gemcitabine (GG) vs gemcitabine+placebo (GP) in patients
sue and microdissected pancreatic ductal adenocarcinoma identifies (pts) with unresectable pancreatic cancer (PC). 2016 ASCO Annual
genes specifically overexpressed in tumor epithelia. Hepatogastro-en Meeting. J Clin Oncol, 34, abstract 4019.
terology, 55, 2016–2027. 54. Lamouille, S. et al. (2014) Molecular mechanisms of epithelial-mese-n
27. Crnogorac-Jurcevic, T. et al. (2013) Molecular analysis of precursor chymal transition. Nat. Rev. Mol. Cell Biol., 15, 178–196.
lesions in familial pancreatic cancer. PLoS One, 8, e54830. 55. Neesse, A. et al. (2013) CTGF antagonism with mAb FG-3019 enhances
28. Scarlett, C.J. et al. (2011) Precursor lesions in pancreatic cancer: -mor chemotherapy response without increasing drug delivery in murine
phological and molecular pathology. Pathology, 43, 183–200. ductal pancreas cancer. Proc. Natl. Acad. Sci. USA, 110, 12325–12330.
29. Moffitt, R.A. et al. (2015) Virtual microdissection identifies distinct 56. Dimou, A. et al. (2013) Novel agents in the treatment of pancreatic a-de
tumor- and stroma-specific subtypes of pancreatic ductal adenocar-ci nocarcinoma. J. Pancreas, 14, 138–140.
noma. Nat. Genet., 47, 1168–1178. 57. Wang, Q. et al. (2011) Cooperative interaction of CTGF and TGβF-in
30. Mackey, J.R. et al. (1998) Functional nucleoside transporters are animal models of fibrotic disease. Fibrogenesis Tissue Repair, 4, 4.
required for gemcitabine influx and manifestation of toxicity in cancer 58. Lin, M.T. et al. (2004) Cyr61 expression confers resistance to apoptosis
cell lines. Cancer Res., 58, 4349–4357. in breast cancer MCF-7 cells by a mechanism of NF-kappaB-dependent
31. Leask, A. et al. (2006) All in the CCN family: essential matricellular -sign XIAP up-regulation. J. Biol. Chem., 279, 24015–24023.
aling modulators emerge from the bunker. J. Cell Sci., 119, 4803–4810. 59. Menendez, J.A. et al. (2005) A novel CYR61-triggered ‘CYR61-alph-av
32. Haque, I. et al. (2011) Cyr61/CCN1 signaling is critical for epithelial- beta3 integrin loop’ regulates breast cancer cell survival and ch-emo
mesenchymal transition and stemness and promotes pancreatic ca-r sensitivity through activation of ERK1/ERK2 MAPK signaling pathway.
cinogenesis. Mol. Cancer, 10, 8. Oncogene, 24, 761–779.
33. Duxbury, M.S. et al. (2004) RNA interference targeting the M2 subunit of 60. Long, X. et al. (2015) Stromal CYR61 confers resistance to mitoxantrone
ribonucleotide reductase enhances pancreatic adenocarcinoma c-he via spleen tyrosine kinase activation in human acute myeloid leuk-ae
mosensitivity to gemcitabine. Oncogene, 23, 1539–1548. mia. Br. J. Haematol., 170, 704–718.
34. Xie, H. et al. (2013) Predictive and prognostic roles of ribonucleotide 61. Lee, K.B. et al. (2012) CYR61 controls p53 and NFκ-B expression through
reductase M1 in resectable pancreatic adenocarcinoma. Cancer, 119, PI3K/Akt/mTOR pathways in carboplatin-induced ovarian cancer cells.
173–181. Cancer Lett., 315, 86–95.
1. Siegel , R. et al. ( 2014 ) Cancer statistics, 2014 . CA: A Cancer J. Clin., 64 , 9 - 29 .
2. Rahib , L. et al. ( 2014 ) Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid, liver, and pancreas cancers in the United States . Cancer Res. , 74 , 2913 - 2921 .
3. de Sousa Cavalcante , L. et al. ( 2014 ) Gemcitabine: metabolism and molecular mechanisms of action, sensitivity and chemoresistance in pancreatic cancer . Eur. J. Pharmacol ., 741 , 8 - 16 .
4. Burris , H.A. 3rd et al. ( 1997 ) Improvements in survival and clinical b-en efit with gemcitabine as first-line therapy for patients with advanced pancreas cancer: a randomized trial . J. Clin. Oncol. , 15 , 2403 - 2413 .
5. Garrido-Laguna , I. et al. ( 2015 ) Pancreatic cancer: from state-of-the-art treatments to promising novel therapies . Nat. Rev. Clin. Oncol. , 12 , 319 - 334 .
6. Gray , J.H. et al. ( 2004 ) The concentrative nucleoside transporter family, Supplementary material SLC28 . Pflugers. Arch., 447 , 728 - 734 . Supplementary Figures S1-S10 and Tables S1-S4 can be found at 7 . Baldwin , S.A. et al. ( 2004 ) The equilibrative nucleoside transporter f-am http://carcin .oxfordjournals.org/ ily, SLC29. Pflugers. Arch., 447 , 735 - 743 .
8. Paproski , R.J. et al. ( 2013 ) Human concentrative nucleoside transporter 3 transfection with ultrasound and microbubbles in nucleoside tr-ans Funding port deficient HEK293 cells greatly increases gemcitabine uptake . PLoS One , 8 , e56423. National Institutes of Health (F31-CA180602 to R.A .H., 9 . Fujita, H. et al. ( 2010 ) Gene expression levels as predictive markers F30-CA196162 to J.J.H. , R01-CA135006 to G.C.B., R01-CA136786 to of outcome in pancreatic cancer after gemcitabine-based adjuvant G.C.B. and R21-CA198287 to R.R .W.). chemotherapy. Neoplasia , 12 , 807 - 817 .