ERK2-regulated TIMP1 Induces Hyperproliferation of K-Ras^G12D-Transformed Pancreatic Ductal Cells^,

Abstract

Pancreatic ductal adenocarcinoma (PDAC) commonly contains a mutation in K-Ras^G12D and is characterized by a desmoplastic reaction composed of deregulated, proliferating cells embedded in an abnormal extracellular matrix (ECM). Our previous observations imply that inhibiting the mitogen-activated protein kinase (MAPK)-extracellular signal-regulated kinase (ERK2) kinase signal pathway reverses a matrix metalloproteinase 1-specific invasive phenotype. Here, we investigated the specific genes downstream of MAPK-ERK2 responsible for the hyperproliferative abilities of human and murine primary ductal epithelial cells (PDCs) within an ECM. Compared with control, DNA synthesis and total cell proliferation was significantly increased in human PDCs harboring the PDAC common p53, Rb/p16^INK4a, and K-Ras^G12D mutations. Both of these effects were readily reversed following small-molecule inhibition or lentiviral silencing of ERK2. Microarray analysis of PDCs in three-dimensional (3D) culture revealed a unique, MAPK-influenced gene signature downstream of K-Ras^G12D. Unbiased hierarchical analysis permitted filtration of tissue inhibitor of matrix metalloproteinase 1 (TIMP1). Pancreatic cells isolated from Pdx1-Cre; LSL-K-ras^G12D/+-mutated mice exhibit increased TIMP1 RNA transcription compared to wild-type littermate controls. Analyses of both 3D, in vitro human K-Ras^G12D PDCs and data mining of publicly annotated human pancreatic data sets correlatively indicate increased levels of TIMP1 RNA. While silencing TIMP1 did not significantly effect PDC proliferation, exogenous addition of human recombinant TIMP1 significantly increased proliferation but only in transformed K-Ras^G12D PDCs in 3D. Overall, TIMP1 is an upregulated gene product and a proliferative inducer of K-Ras^G12D-mutated PDCs through the ERK2 signaling pathway.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is derived from an initial, injury-induced, or K-Ras-induced acinar-to-ductal metaplasia followed by a stepwise pancreatic intraepithelial neoplastic progression (PanIN) of the cells lining the ducts of the exocrine pancreas [1,2]. PDAC is characterized by a highly proliferative, desmoplastic microenvironment consisting of epithelial and stromal hyperplasia and abnormal extra-cellular matrix (ECM) assembly [3]. More than 90% of PDACs harbor a K-Ras^G12D mutation that inhibits GTPase functionality and initiates constitutively active downstream kinase signaling [4,5]. PDAC activates a variety of downstream signaling pathways that contribute to cancer growth, motility, invasion, and metastasis [6]. Recently, we described a distinct K-Ras^G12D-mediated invasive signaling pathway in cultured pancreatic ductal epithelial cells (PDCs) that specifically involves the mitogen-activated protein kinase (MAPK)-extracellular signal-regulated kinase (ERK2) signaling pathway. Specifically, activation of the ERK2 isoform results in the up-regulation of a variety of oncogenic genes, including matrix metalloproteinases (MMPs) and tissue inhibitor of matrix metalloproteinases (TIMPs), which is evident only when cells are cultured in three-dimensional (3D) ECM-derived hydrogels [7].

TIMPs are 22- to 28-kDa protease inhibitory proteins that regulate the proteolytic activity of MMPs by sterically binding to and inhibiting activation of their catalytic zinc moiety [8]. In normal tissue, a proper balance is maintained between TIMPs and MMPs, which enables controlled cellular remodeling of the ECM [9]. Disruption of this balance leads to numerous pathologies, including inflammatory arthritis, tumor growth, and metastasis [9,10]. The conserved TIMP family is currently comprised of four subtypes (TIMP1, TIMP2, TIMP3, and TIMP4) all sharing several gene sequence homologies as well as structural protein similarities. Specifically, the N-terminal region is capable of binding to the catalytic zinc site of MMPs, a side chain that inserts into specific MMP clefts, while the C-terminal region enables receptor-ligand signaling [11]. Interestingly, only TIMP1 resides on the X chromosome and is the only member to have its first exon transcribed but not translated [12]. Some activated transcription factors, such as Ets family members, bind to exon/intron 1 upstream of the translational start site and regulate TIMP1 expression. Increased TIMP1 expression is observed in fetal ductal morphogenesis and growth as well as during active epithelial proliferation of pubertal mammary ducts. Conversely, homozygous TIMP1^-/- knockout mice exhibit alterations of the terminal ducts and their basement membrane [13].

PDAC mouse models along with human biopsies, serum, and stroma exhibit increased gene expression and secretion of TIMP1, which may be due to elevated levels of transforming growth factor-β, interleukin, and leukemia inhibitory factor [12,14]. TIMP1 expression levels have been used diagnostically as indicators of poor outcome prognosis in both mammary and pancreatic carcinoma [15,16]. However, it is unclear whether TIMP1 up-regulation is a secondary response to a neoplastic increase in MMP expression or if TIMP1 itself harbors pro-tumorigenic, anti-apoptotic features [5,9,17–19,25]. While a majority of TIMP1 expression studies have been carried out in two-dimensional (2D) in vitro environments, recent evidence suggests that its tumorigenic properties are only evident in 3D culture [18,20–22].

Our results indicate that only constitutively active K-Ras^G12D-mutated PDCs exhibit significant TIMP1 RNA and protein increases through the MAPK-ERK2 pathway. Further, elevated levels of TIMP1 expression are seen in human in vitro models incorporating a 3D ECM as well as in vivo K-Ras^G12D mouse models. Lastly, the exogenous addition of human recombinant TIMP1 (rTIMP1) sufficiently increases cellular aggregate growth and DNA synthesis but only in 3D E6/E7/Ras PDCs.

Materials and Methods

Human Cell Lines

The isogenic set of immortalized primary human PDCs was a generous gift from Dr Michel Ouellette (University of Nebraska) [23]. All populations were maintained in conventional 2D cultures in T-75 flasks in a humidified incubator (95% air/5% CO₂) at 37°C and cultured in a pancreatic-specific growth medium [complete pancreatic medium (CPM)]: four parts low-glucose Dulbecco's modified Eagle's medium (Cellgro, Manassas, VA) to one part M3 base culture medium (INCELL, San Antonio, TX) supplemented with 5% FBS + 1% penicillin-streptomycin. Cells were passaged every 48 hours by 0.05% trypsin detachment over 5 minutes, as previously described [7].

Murine Pancreatic Ductal Cell Isolation

Single-cell suspensions of primary isolates from the mouse pancreas were prepared as described previously [24,25]. Briefly, a mixed cell population was resuspended and divided into several fractions each containing 10⁶ cells with one fraction kept as a pre-sorting sample. Fluorescein-labeled Dolichos Biflorus Agglutinin (DBA) lectin or biotin-labeled DBA (Vector Laboratories, Inc, Burlingame, CA), respectively, was added and the cells were washed and pelleted. Anti-fluorescein isothiocyanate or streptavidin-coated nanobeads (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany), respectively, were added and incubated before magnetic separation was performed using magnetic-activated cell sorting (MACS) columns (Miltenyi Biotec GmbH), according to the manufacturer's protocol. RNA from each of the three fractions [DBA positive (+), DBA negative (-), and pre-sorting] was isolated using the RNAqueous-Micro Kit (Ambion, Inc, Austin, TX).

3D Cell Culture Assays

3D cultures of PDCs were established as previously published [7]. Briefly, 200 µl of growth factor reduced (GFR) Matrigel (BD Biosciences, San Jose, CA, #354230) was spread within a 12-well tissue culture plate at a concentration of 10.5 mg/ml and incubated at 37°C for 30 minutes. The polymerized matrix had an elastic modulus of approximately 5 to 8 kPa as determined by rheometry (Bohlin CS, Lund, Sweden). The elastic modulus of the native (healthy) human pancreas is ∼500 Pa, which is similar to that of normal breast tissue (∼600 Pa) and liver (640 Pa), as determined by atomic force microscopy and rheometry [7]. To further decrease the elastic modulus of our matrix construct to levels approximating modulus of native pancreas, another 50 µl of GFR Matrigel was added to the center of the already coated well and solidified for another 30 minutes at 37°C to obtain an average elastic modulus of 480 ± 63 Pa at the center of each well. Cells growing in log phase were trypsinized and resuspended at a concentration of 10,000 cells/ml in culture medium, which consisted of a 50:50 mixture of 3D media (Dulbecco's modified Eagle's medium/F12, 2% horse serum, 100 µg/ml hydrocortisone, 1 µg/ml cholera toxin, 10 µg/ml insulin, and 1% penicillin-streptomycin) and CPM (see above), which was then supplemented with 2% vol/vol of GFR Matrigel. Two milliliters of this cell suspension was added to each well. PDCs were cultured from 1 to 8 days with a medium change every fourth day.

Cell Proliferation Analysis

DNA synthesis of PDCs cultured in 2D and 3D cultures was measured using the Invitrogen Click-iT EdU Assay (C10337). Briefly, 0.2x10⁶ PDCs were plated the night before analysis in 2D supplemented with CPM or 3D supplemented with CPM and 2% GFR Matrigel in a 12-well plate. A 1x EdU solution was added to each well for 3 hours, the medium was removed, and the cells were fixed with 3.7% formaldehyde in phosphate-buffered saline (PBS) for 15 minutes. The wells were rinsed two times with 3% BSA in PBS and permeabilized with 0.5% Triton X-100 in PBS for 20 minutes at room temperature. A 1x Click-iT reaction cocktail was prepared by sequential addition of reaction buffer, copper sulfate (CuSO₄), Alexa Fluor 488 azide, and reaction buffer additive, as per the manufacturer's instructions. Following removal of the permeabilization solution, the wells were washed two times for 5 minutes each with 3% BSA in PBS and 500 µl of the Click-iT reaction cocktail was added to each well and incubated for 30 minutes at room temperature in the dark. Afterward, the wells were washed again with 3% BSA in PBS, and the cells were counterstained with a 1:2000 dilution of 4′,6-diamidino-2-phenylindole (DAPI) in PBS for 40 minutes at room temperature. The wells were finally washed two times with PBS and five random fields were imaged using a Nikon TE-2000 fluorescence microscope. A first image was taken using the DAPI filter cube followed by EdU imaging using a 488/520 nm filter cube. Each individual image was pseudocolored with blue or green ImageJ filters to provide the correlative wavelength observed for each fluorescently tagged cell. These images were then analyzed in ImageJ using the automatic counting subroutine macro to filter cells not proliferating (blue filter only) to those proliferating (blue + green filters combined). The EdU-positive cells were divided by the number of DAPI-positive cells to obtain the fraction undergoing active DNA synthesis.

Pharmacologic Inhibition Experiments

The MAPK/ERK kinase (MEK) inhibitor bis[amino[(2-aminophenyl) thio]methylene]butane-dinitrile (UO126; Calbiochem, Gibbstown, NJ) was obtained from Invitrogen (Carlsbad, CA), dissolved in DMSO stored at -20°C, and used at final concentrations of 5 and 10 µMto verify concentration dependence. The specific ERK2 inhibitor 3-(2-aminoethyl)-5-((4-ethoxyphenyl)methylene)-2,4-thiazolidine hydroxychloride (AEMT; Calbiochem, 328006) was stored at -20°C and used at a 20 µM final concentration.

RNA Isolation

RNA was isolated from 3D cultures of human PDCs grown in CPM on GFR Matrigel by trypsinization as previously described [7]. Briefly, 5 ml of freshly thawed 0.05% trypsin, supplemented with a 1:100 dilution of a phosphatase and protease inhibitor cocktail (Thermo, Rockford, IL; #78440), was added to each well containing Matrigel and cells. After incubation at 37°C for 30 minutes, the cells and gels were scraped from the plates, pipetted into 15-ml conical tubes, and spun at 1200 rpm for 15 minutes at room temperature. The cell pellets were lysed using the RNeasy Mini Kit (Qiagen, Valencia, CA) as per the manufacturer's protocols. RNA integrity was determined by electrophoresis on a 1% agarose gel (distilled water, 10x 3-(N-morpholino)propanesulfonic acid buffer, and 37% formaldehyde) following the manufacturer's protocol (Ambion, Inc).

TIMP1 ELISA

Short hairpin RNA (shRNA) lentiviral particles, small molecule inhibitors, or DMSO (vehicle control) was added to 3D PDCs cultured in CPM. After 72 hours, the amount of TIMP1 protein in the cell culture medium was determined using the Human TIMP1 ELISA Kit (RayBiotech, Norcross, GA; Catalogue No. ELH-TIMP1-001) following the manufacturer's instructions. In brief, aliquots of the cell culture supernatant were diluted 100-fold with a proprietary buffer containing animal serum and 0.09% sodium azide. A standard curve of human rTIMP1 protein was established as per the manufacturer's instructions. HRP-conjugated TIMP1-specific antibody from the ELISA kit was then added at a 100 µl volume (as per the manufacturer's instructions) to each well of a 96-well plate (in three replicates) and incubated for 2.5 hours at room temperature. The wells were washed four times with the proprietary wash solution. One hundred microliters of biotinylated antibody was added to each well and incubated at room temperature for 1 hour. Next, 100 µl of a prepared streptavidin solution was added to each well and incubated at room temperature for 1 hour with gentle shaking on a laboratory shaker (Belly Dancer). Lastly, the chromogenic HRP substrate 3,3′,5,5′-tetramethylbenzidine was added to each well. After 30 minutes, the reaction was stopped with 50 µl of sulfuric acid and the plate was scanned at 480 nm in a Synergy BioTek4 microplate reader (Winooski, VT). The mean absorbance of each set of triplicate standards, controls, and samples was calculated and adjusted to the optical density of the “zero/background control” standard. Protein concentration in each sample was calculated from a best-fit line of a plot of absorbance versus standard concentration.

Microarray Analysis

Focused human microarrays (SABiosciences, Valencia, CA; PAHS-013Z) were processed and analyzed as previously described [7]. For RNA isolation, the cell pellets were lysed using the Qiagen RNeasy Mini Kit as per the manufacturer's protocols. RNA integrity was determined by electrophoresis on a 1% agarose gel (distilled water, 10x 3-(N-morpholino)propanesulfonic acid buffer, and 37% formaldehyde) following the manufacturer's protocol (Ambion, Inc). RNA concentration was determined using a NanoDrop 2000. To eliminate genomic DNA contamination, 5 µg of RNA was first dissolved in 50 µl of the Qiagen GE buffer and incubated at 42°C for 5 minutes. One microgram of the purified RNA from each experimental condition was reverse transcribed to cDNA using the Qiagen First Strand Kit reaction assay on an Eppendorf Mastercycler Pro S (Hauppauge, NY). Next, 50 µl of a mix of reverse transcriptase buffer, primers, deoxynucleotide triphosphates (dNTPs), and reverse transcriptase was added, incubated at 42°C for 15 minutes, and heat inactivated at 95°C for 5 minutes. Finally, 455 µl of nuclease-free water was added for a total volume of 555 µl. For each reaction, 5 µl of this cDNA was added to 20 µl of Qiagen Mastermix and this cocktail was added to a 96-well SABiosciences Extracellular Matrix Microarray Plate (PAHS-013Z; SABiosciences). The cDNA plate was processed on an aluminum block Eppendorf ep Realplex II calibrated with an SABiosciences “A” Plate using a 520 nm filter and a SYBR Green probe. The polymerase chain reaction (PCR) program entailed one heat activation cycle at 95°C followed by 40 cycles of PCR (95°C x 15 seconds, 55°C x 15 seconds, 60°C x 20 seconds) at a 35% ramp rate. Quantitative reverse transcription- PCR (qRT-PCR) was followed by a melting curve analysis (95°C x 15 seconds, 60°C x 15 seconds, at a 0.4°C heating rate with a final inactivation at 95°C x 15 seconds) to verify purity by constant linear dissociation over time. Bioinformatics analysis of the PCR data was carried out using the Web-based RT² Profiler PCR Array Data Analyzer (SABiosciences) after logarithmic transformation of C_t amplification values against internal controls, while the threshold was consistently set by internal program controls. Data are expressed as ΔΔC_t as well as fold changes with ±2-fold considered as significant.

qPCR of Human Cells

Total RNA from both E6/E7 and E6/E7/Ras PDC lines was isolated from 3D cultures using the Qiagen RNeasy Mini Kit and reverse transcribed to cDNA as above. Real-time qPCR analysis of TIMP1 gene expression was performed with the Applied Biosystems (Carlsbad, CA) human TaqMan Gene Expression Assay (Hs00171558_m1) in an Eppendorf Mastercycler ep Realplex II. For the PCR reaction, we used the TaqMan Universal PCR Mastermix in a volume of 25 µl following the manufacturer's protocol (Applied Biosystems) with glyceraldehyde 3-phosphate dehydrogenase (GAPDH; Hs99999905_m1) as an invariant housekeeping gene.

qPCR of Murine Cells

RNA from sorted magnetic-assisted cell sorting/fluorescence-assisted cell sorting (MACS/FACS) murine cells was isolated using the RNAqueous-Micro Kit (Ambion, Inc). One microgram of RNA was primed using random hexamers and transcribed into cDNA (TaqMan Reverse Transcription Reagents; Applied Biosystems, Branchburg, NJ). SYBR Green was detected and quantitative analysis was performed on the StepOnePlus System (Applied Biosystems, Branchburg, NJ). Primer sequences for TIMP1 are given as follows: forward, 5′-CTCAAGACCTATAGTGCTGGC-3′; reverse, 5′-CAAAGTGACGGCTCTGGTAG-3′. P < .05 was statistically significant (Mann-Whitney-Wilcoxon test). Error bars represent the SEM.

Gene Silencing

PDCs were plated on GFR Matrigel in a 24-well plate 24 hours before viral infection at ∼25% confluence. Five hundred microliters of CPM supplemented with 2% Matrigel was added to each well and incubated overnight. The next day, the medium was aspirated and the cells were refed with 400 µl of CPM. The cationic polymer, Polybrene, was added to each well at 4 µg/ml to increase infection efficiency and incubated for 4 hours (Sigma-Aldrich, St Louis, MO; AL-118). Pools of three to five target-specific 19- to 25-nucleotide siRNA particles directed against TIMP1 (Santa Cruz Biotechnology, Dallas, TX; 29505-V) and copGFP control viral particles (Santa Cruz Biotechnology; 134220) were added to 2D and 3D cultures of PDCs grown on GFR Matrigel at a volume of 80 µl (approximately 1 x 10⁵ infectious particles). The plates were incubated overnight. The next day, 500 µl of fresh CPM was added to each well without Polybrene for another 24 hours. This medium was collected and frozen for TIMP1 ELISA studies, while the cDNA Fastlane Isolation Protocol (Qiagen; 215011) was followed to isolate the cellular RNA for immediate PCR. The lysates were transferred to RNase-free tubes and incubated at 75°C for 5 minutes before qRT-PCR as above.

Animal Procedures

All procedures were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania (Protocol No. 802868, No. 803699, and No. 804209). The Pdx1cre and LSL-K-ras^G12D/+ mice were used for studies [26,27].

Statistical Analysis

Unless stated otherwise, all experiments were run in triplicate and repeated independently at least three times. Where possible, the data are expressed as averages ± SEM. Significance was based on P < .05 and P < .01 and clarified in each figure. Significance between two data sets was determined using a two-tailed Student's t test, while differences between multiple experimental groups were determined by two-way analysis of variance (ANOVA) with a Bonferroni post hoc test. Microarray data sets were initially filtered for genes with more than a two-fold change in ΔΔC_t values. Internal microarray controls were assessed across all samples for equal variance before comparison. Significant increases in gene expression were determined by the Pffafl method, which normalizes expression values to endogenous levels of GAPDH to untreated control cells.

Results

Mutant K-Ras^G12D Increases Human PDC Proliferation

We previously demonstrated that an invasive phenotype in vitro is only observed in K-Ras^G12D-mutated PDCs cultured in a 3D basement membrane ECM [7]. We therefore hypothesized that 3D culture will similarly reveal differences in the proliferative capacity of the isogenic, progressively transformed PDCs [18,20]. For these studies, we used the defined set of primary human PDCs that incorporate a stepwise retroviral transfection of human telomerase reverse transcriptase (hTERT), the addition of human papillomavirus E6 and E7 proteins, and a final glycine to aspartate mutation in the 12th position of K-Ras that imparts transformative and epithelial-to-mesenchymal characteristics (Figure 1A)[7].

Figure 1.

K-Ras^G12D-mutated PDCs reveal increased proliferation in 3D. (A) Schematic of the transduction of primary human PDCs with human telomerase (hTERT), the human papillomavirus proteins E6 and E7, and a K-Ras^G12D mutation to yield three separate clone lines with a common genetic background. (B) Each clone was cultured either in 2D (left panels, x20) or 3D ECM (right panels, x10) for 72 hours and evaluated for proliferative and morphologic characteristics. Scale bars, 50 µm. (C) Total cell numbers for each individual clone was quantified after 72 hours in 3D culture at 12-hour intervals. At each time point, cell numbers were determined as the average count of five individual fields of three separate cultures of each individual clone. Data are presented as means ± SEM. *P < .05, **P< .01. (D) Two-way ANOVA through Tukey B Test for Contrasts on Ordered Means of each clone confirms significant differences in proliferation in 3D at 48 hours (left graph) and 72 hours (right graph). (E) The total number of cells per aggregate was quantified for each PDC clone after 72 hours in 3D culture. Ten separate aggregates of each clone from three different experiments were averaged. *P< .05. Data are presented as means ± SEM.

The proliferative rates of cell cycle-deregulated E6/E7 and E6/E7/Ras clones were indistinguishable in 2D despite the additional K-Ras^G12D mutation (Figure W1). As expected following p53 and Rb/p16 inhibition by introduction of E6/E7, both clones (E6/E7 and E6/E7/Ras) exhibited increased proliferation in 3D when compared to the hTERT clone (Figure 1C). Conversely, 3D culture could readily distinguish between E6/E7 and E6/E7/Ras PDCs, in terms of their invasive morphology (invadapodia, disorganized aggregation; Figure 1B) and rates of proliferation (Figure 1C), although both of these parameters were indistinguishable in 2D. The population doubling times of immortalized hTERT and E6/E7 PDCs in 3D were approximately 48 and 24 hours, respectively, while the doubling time of the K-Ras^G12D was shortened to 12 hours (Figure 1C; two-way ANOVA significance at 72 hours). The two non-K-Ras^G12D mutated clones (hTERT and E6/E7) formed rounded non-invasive pancreatospheres composed of 3 to 15 individual cells within 48 hours (Figure 1B, right panels). Although the organization of the pancreatospheres was similar between the hTERT and E6/E7 PDC clones, the E6/E7 PDCs contained three to eight more cells per aggregate, in line with their altered cell cycle deregulation and increased proliferative potential (Figure 1D). Remarkably, addition of a K-Ras^G12D mutation caused severe disruption of the rounded epithelial morphology. Not only were rounded pancreatospheres no longer recognizable, these invasive aggregates contained five times more cells per aggregate and the cells appeared more flattened, elongated, and mesenchymal-appearing in phenotype (Figure 1, B and D).

ERK2 Regulates Mutant K-Ras^G12D-Directed PDC Proliferation

E6/E7/Ras PDC invasion is specifically regulated by the K-Ras-MAPK/ERK2 pathway but only in 3D culture [7]. Therefore, we tested the hypothesis that inhibition of this particular pathway is sufficient to also significantly reduce PDC proliferation. MEK1/2 and ERK2-specific small molecule inhibitors (UO126 and AEMT, respectively) were added to E6/E7/Ras PDC 3D cultures and DNA synthesis was assessed by EdU incorporation (Figure 2A, green fluorescence) [28]. As seen in Figure 2B, MEK1/2 inhibition by UO126 reduced the number of proliferating cells in the disorganized cell aggregates by approximately 40%, while ERK2-specific inactivation by AEMT decreased the number of EdU-positive nuclei by an additional 10% (approximately 50% inhibition).

Figure 2.

E6/E7/Ras PDCs exhibit decreased proliferation after ERK2 inhibition. (A) K-Ras^G12D-mutated PDCs were cultured in 3D for 24 hours. Then, either DMSO, the MEK1/2 inhibitor UO126, or the ERK2-specific inhibitor AEMT was added for an additional 24 hours. At that point, each well was pulse-labeled with a bromodeoxyuridine analog (EdU, green) for 6 hours as a marker of DNA synthesis and cell proliferation. All nuclei were subsequently labeled (DAPI, blue). Images were observed at x20 magnification. Scale bar, 100 µm. (B) Inhibition of MEK1/2 and ERK2 by either UO126 or AEMT caused a significant decrease in the proliferation of the oncogenic PDCs after quantifying EdU incorporation compared to total cell numbers (EdU/DAPI). Data are normalized to the proliferation of control untreated cells and presented as means ± SEM. *P< .05, **P< .01. (C) K-Ras^G12D-mutated PDCs were infected with lentivirus harboring scrambled or shRNA constructs directed against ERK1 and ERK2 before 3D culture. The total ERK1 and ERK2 protein after RNA silencing was determined by Western blot analysis and normalized to GAPDH loading control. Ratios reflect densitometric protein comparisons between E6/E7/Ras infection with vehicle control, scramble, sh-ERK1, and sh-ERK2. For details, see Materials and Methods section. (D) E6/E7/Ras PDCs were either left untreated (control) or transfected with scrambled, ERK1, or ERK2-specific sh-constructs and cultured in 3D for 72 hours before proliferation analysis. A percentage of cell numbers compared to control E6/E7/Ras cells was calculated in five different fields of view between three experiments. Data are presented as means ± SEM. *P< .05. (E) Phase contrast micrographs of E6/E7/Ras PDCs treated as in D and cultured in both 2D (top panels) and 3D (bottom panels) for 72 hours before observation. Original magnification, x20. Scale bars, 50 µm.

To further implicate ERK2 as the specific isoform controlling the proliferation of mutated K-Ras^G12D cells, we infected PDCs with shRNA lentiviral constructs capable of silencing ERK1 or ERK2, respectively. After verifying >90% transfection efficiency by constitutive green fluorescence protein (GFP) expression (data not shown), we observed that total ERK1 or ERK2 protein expression was decreased for both constructs, respectively, by at least 75%, 48 hours post-silencing (Figure 2C). Subsequently, we cultured each silenced clone in both 2D and 3D for another 24 hours (72 hours total post-silencing) and measured proliferation by counting the total number of cells. As seen in Figure 2D, proliferation of E6/E7/Ras PDCs in 3D was significantly reduced only upon silencing of ERK2 but not ERK1. Additionally, both the number of invadapodia and number of cells per aggregate were decreased in E6/E7/Ras PDCs, again only when ERK2 was silenced, yielding cultures that were morphologically reminiscent of E6/E7 and hTERT 3D pancreatospheres (Figures 1B and 2E, bottom panels). The proliferation of K-Ras^G12D-mutated PDCs cultured in 2D was also decreased with ERK2 silencing but not with ERK1 silencing. The rate of proliferation of these sh-ERK2-treated “transformed” cells was similar to that of the “normal” hTERT clone (Figures 1B and 2E, top panels). These data indicate that the K-Ras^G12D mutation increases the proliferation of PDCs through an ERK2-specific MAPK pathway. Further, ERK2-specific antagonists, both small molecule inhibitors and RNAi, will abolish the hyperproliferation induced by K-Ras^G12D and degenerate the invadapodia of mutated PDCs, yielding more organized pancreatospheres in 3D.

TIMP1 Is Upregulated by the Constitutive K-Ras-ERK2 Signaling Axis

Because both invasion and cell proliferation appear regulated by the K-Ras^G12D-ERK2 pathway in 3D culture, we next attempted to identify proliferation-related genes that were affected downstream of this pathway. Using a focused microarray, we compared the expression profiles of 84 ECM-related genes in 3D cultures of E6/E7 PDCs and E6/E7/Ras PDC s in the presence of two different concentrations of the MAPK inhibitor (MAPKi), UO126 to understand dose dependent expression patterns (Figure 3A). Initially, we filtered the data for those genes that were upregulated by more than two-fold in the K-Ras^G12D-mutated cells (Figure 3B, cluster 1, top bar, red boxes). Next, we identified genes of that filtered subset that were subsequently downregulated by more than two-fold upon exposure to increasing concentrations of the MAPKi (Figure 1B, cluster 1, top bar, yellow (5 µM) and green (10 µM) boxes). Of the 84 genes analyzed, we identified 12 genes that were under the direct control of mutant K-Ras^G12D through MAPK (Figure 3C and Table 1). For each of these 12 genes, we calculated the ratio of expression levels (adjusted to the group's overall sum) following the addition of K-Ras^G12D. The resulting data are depicted in a pie chart where the gene's relative share of the pie is reflective of its K-Ras^G12D-induced log-fold change. All of these genes have previously been implicated in human and mouse PDAC transformation and metastasis [5,29,30]. Of the genes listed in Figure 3C, only the promoter for TIMP1 has published binding sites relevant for transcription factors (e.g., PEA3, AP1, and Ets) that could relay K-Ras^G12D-mutated signals through ERK2 in PDCs [11,12,20,31].

Figure 3.

TIMP1 is directly upregulated by the 3D K-Ras^G12D-ERK2 signaling axis. (A) Schematic detailing genetic/pharmacologic manipulations of the cells used for subsequent microarray analyses. (B) Heat chart of gene expression signatures produced by unbiased hierarchical clustering of PDCs under various conditions. Top blue boxes, E6/E7 PDCs; red boxes, E6/E7/Ras PDCs; yellow boxes, E6/E7/Ras PDCs + 5 µM UO126; green boxes, E6/E7/Ras PDCs + 10 µM UO126. (C) Twelve genes with more than a two-fold change and regulated specifically by the K-Ras^G12D-ERK2 axis were extracted from the gene signatures and distributed by relative increases in gene expression after addition of K-Ras^G12D within the pie chart. (D) Volcano plot of 84 genes distributed by P value and fold change. Horizontal blue line represents a significant P value threshold of .5 and induced by K-Ras^G12D. MMP1 and TIMP1 are labeled. (E) Volcano plot of the same 84 genes after K-Ras^G12D up-regulation of specific genes followed by ERK2 inhibition of a subset of this pool of genes by P value and fold change. Those now significantly downregulated based on MAPKi alone are above the P value > .05 threshold. MMP1 and TIMP1 are labeled. (F) Increasing concentrations of UO126 were added to 3D E6/E7/Ras cultures and RNA was isolated for qRT-PCR analysis. MAPKi significantly decreased TIMP1 expression in a dose-dependent manner. Data are presented as means ± SEM. *P< .05. No significance between MAPKi concentrations.

Table 1.

Official Gene Name of Analyzed RNA for Each Gene Symbol.

Genes determined to be significantly changed after microarray analysis were set into this table to coordinate each gene's symbol with its official name.

Comparison of significantly upregulated genes in E6/E7/Ras versus E6/E7 PDCs by volcano plot analysis confirms significant gene up-regulation of TIMP1 and, as previously published, MMP1, signifying explicit control by the MAPK pathway (Figure 3D) [7]. More importantly, when the MAPK pathway was inhibited with either 5 or 10 µM of UO126, the most significantly downregulated genes in the K-Ras- mutated cells were again MMP1 and TIMP1, emphasizing that their expression is under the direct influence of MAPK (Figure 3E). Validation of the array data by qPCR indicates that TIMP1 mRNA expression is reduced when the K-Ras^G12D-ERK pathway is inhibited by UO126 (Figure 3F).

We subsequently used an ELISA-based assay to measure the concentration of secreted TIMP1 protein under various 3D culture conditions (Figure 4A). In line with the gene expression data, TIMP1 secretion from E6/E7/Ras PDCs was approximately five-fold higher than that from untransformed E6/E7 PDCs (∼12 pg/ml increased to ∼57 pg/ml) after normalization of total cell protein lysates (Figure 4B). As expected for a receptor tyrosine kinase-uncoupled GTPase, epidermal growth factor (EGF) addition had no statistically significant effect on TIMP1 expression in K-Ras^G12D PDCs. By contrast, inhibition of ERK1/2 phosphorylation by UO126 and sh-ERK2 (but not sh-ERK1) significantly reduced TIMP1 secretion from E6/E7/Ras PDCs to levels similar to those seen in the E6/E7 cell culture supernatants. Taken together, these data suggest that the elevated TIMP1 expression at the level of both gene and gene product is mediated specifically through the K-Ras^G12D-MAPK-ERK2 axis.

Figure 4.

TIMP1 secretion is specifically regulated by ERK2 but is not necessary for basal proliferation. (A) E6/E7 and E6/E7/Ras PDCs plus 10 ng/ml EGF, 10 µM UO126, or 80 µl of sh-ERK1 or sh-ERK2 lentiviral particles were cultured in 3D for 72 hours before analyzing TIMP1 levels in aliquots of supernatant. Protein levels were quantified based on a TIMP1 standard curve. Data are presented as means ± SEM. **P< .01. (B) E6/E7/Ras PDCs were infected with five pooled lentiviral TIMP1 shRNA constructs and cultured in 3D for 72 hours. RNA was isolated and qRT-PCR analysis determined a 70% reduction in TIMP1 transcript expression. Data are presented as means ± SEM. *P< .05. (C) Total number of E6/E7/Ras PDCs 72 hours post-TIMP1 silencing was determined in five fields of view between three repeated experiments and compared to control. Data are presented as means ± SEM. NS, not significant. (D) No significant differences were observed in the cell morphology or total numbers of control E6/E7/Ras PDCs versus TIMP1 silenced. Images observed at x40 magnification. Scale bars, 100 µm.

TIMP1 Increases Proliferation of K-Ras^G12D-mutated PDCs

TIMP1 has been shown to increase proliferation of epithelial and stellate cells [20,32]. Recent evidence suggests that the proliferative augmentation of TIMP1 can only be recognized in 3D cell culture [18]. To establish TIMP1 as a functional modulator of aberrant proliferation of PDCs harboring mutated K-Ras^G12D, we transduced these cells with TIMP1-specific shRNA reducing TIMP1 expression by ∼70% (Figure 4B). Paradoxically, neither 2D nor 3D culture of TIMP1-silenced K-Ras^G12D PDCs showed a significant decrease in cell numbers or a morphologic reversal to pancreatospheres, similar to mammosphere assays (Figure 4, C and D) [18].

As increased TIMP1 expression is observed in fetal ductal morphogenesis and growth as well as during active epithelial proliferation of pubertal mammary ducts, we hypothesized that TIMP1 may initiate its growth promoting effects only at higher concentrations that are found in deregulated, frank PDAC [33,34]. To determine the in vivo TIMP1 RNA levels within a native pancreatic microenvironment, we compared TIMP1 gene expression levels in mice with a K-Ras^G12D mutation (Pdx1-Cre;LSL-K-ras^G12D/+) to appropriate Pdx1-Cre;K-ras^+/+ control mice. A majority of Pdx1-Cre;LSL-K-ras^G12D/+ mice develop PanIN lesions between 10 and 20 weeks of age, while a subset develop PDAC [35]. PDCs were separated from all other cells in primary isolates of mixed populations of pancreatic cells by ferromagnetic isolation of DBA+ cells, which is specific for ductal structures including ADM and PanIN lesions (Sox 9, Keratin19, HNF1β positive) [2]. RNA purified from the DBA+ cell fraction of Pdx1-Cre;LSL-K-ras^G12D/+ mice contained 12-fold higher levels of TIMP1 when compared to control animals (Figure 5).

Figure 5.

TIMP1 is significantly increased in primary pancreatic cells isolated from K-Ras^G12D-mutated mouse and in sera from human PDAC patients. (A) Graphic representation of the process for pancreatic ductal cell sorting (ductal vs non-ductal compartments) from Pdx1-Cre;LSL-K-ras^G12D/+ and Pdx1-Cre;K-ras^+/+ control mice before RNA analysis. For details, see Materials and Methods section. (B) Graph showing relative mTIMP1 expressed in the primary DBA+ ductal epithelial cells isolated from K-ras^G12D+ mice compared to that of wild-type mice. Data are presented as means of triplicates ± SEM. (C) The Oncomine public database was mined for TIMP1 gene up-regulation in six separate RNA microarray analyses over the last decade of multiple human samples of pancreatic cancer, specifically ductal adenocarcinoma. Multivariate analysis of the data sets confirms that the TIMP1 gene is significantly increased in human pancreatic cancer samples with a P value < .000173. Oncomine (Compendia Bioscience, Ann Arbor, MI) was used for analysis and visualization.

We determined that E6/E7/Ras PDCs secreted approximately 57 pg/ml TIMP1 protein over 72 hours by ELISA. This concentration of human rTIMP1 protein was unable to significantly increase the proliferation of neither E6/E7 nor E6/E7/Ras PDCs in 3D when added for 24, 48, or 72 hours (data not shown). This would indicate that independent levels of E6/E7/Ras-secreted TIMP1 is not sufficient to induce proliferation and that additional exogenous sources added by the non-ductal compartment may be necessary to observe effects on ductal cells. In an effort to mimic the increased in vivo RNA and protein levels of TIMP1 in both the ductal and non-ductal compartments, a minimal three-fold increased concentration (∼200 pg/ml) was added to both E6/E7 and E6/E7/Ras PDCs in 3D for 24, 48, and 72 hours. As seen in Figure 6A, addition of this concentration of rTIMP1 promoted significant proliferation of K-Ras^G12D-mutated PDCs (especially pronounced at 72 hours) but did not result in a statically significant increase in the cell numbers in the untransformed E6/E7 PDCs.

Figure 6.

Exogenous TIMP1 induces E6/E7/Ras proliferation. (A) E6/E7 and E6/E7/Ras PDCs were cultured in 3D for 24 hours and then were cultured in the presence of either DMSO (vehicle) or human rTIMP1 and analyzed at 24, 48, and 72 hours. Data (average cell numbers/field of view) are presented as means ± SEM. *P< .05, **P < .01. (B) Control E6/E7/Ras PDCs with and without sh-TIMP1 and/or rTIMP1 were cultured in 3D for 72 hours before EdU (green) pulse labeling for 6 hours. Cultures were fixed and nuclei were labeled with DAPI (blue) before analysis. Images at x4 magnification. Scale bars, 100 µm. (C) The number of proliferating E6/E7/Ras PDCs in the presence of the MEK/ERK2 inhibitors UO126 and AEMT, sh-TIMP1, sh-TIMP1 plus rTIMP1, and control plus rTIMP1 were quantified. By normalizing the nuclei labeled with EdU to the total number of DAPI-stained nuclei, the percent proliferating under each experimental condition, normalized to the proliferation of non-treated control cells, was assessed. Data are presented as means ± SEM. *P< .05, **P< .01.

To further define the influence of K-Ras^G12D mutations on TIMP1 secretion and proliferation, we performed rescue experiments using shRNA directed against TIMP1 mRNA with or without the addition of exogenous rTIMP1 and determined cell proliferation by assessing the number of EdU-positive cells (Figure 6B). As seen in the quantitative evaluation of these micrographs, addition of sh-TIMP1 had no observable effect on E6/E7/Ras DNA synthesis compared to vehicle control, which was contrary to the addition of either a MEK or ERK2 inhibitor (Figure 6C). However, the addition of rTIMP1 after TIMP1 silencing caused a small but statistically significant increase in DNA synthesis. Importantly, the addition of human rTIMP1 to vehicle-only K-Ras^G12D-mutated PDCs caused a robust increase in DNA synthesis, underscoring TIMP1's ability to induce proliferation over basal levels only at pathologically higher concentrations (Figure 6C).

Discussion

Here, we provide evidence that a K-Ras^G12D mutation activates ERK2-specific signaling that increases TIMP1 expression and secretion from an isogenic series of human PDCs, an increase also readily seen in primary pancreatic cells directly processed in vivo from K-Ras^G12D - mutated mice. This induction of TIMP1 sufficiently provokes aberrant hyperplasia in our human PDC culture model but only in the K-Ras^G12D-mutated cells cultured in 3D. Importantly, normal MAPK signaling dynamics must be shifted toward constitutively activated MEK-ERK2 for hyperproliferation to occur [36–38]. Recent evidence suggests that this MAPK signaling pathway can only be recapitulated in epithelial cells embedded in a 3D ECM matrix, providing a more “in vivo-like” aggregate architecture [7,18,39,40].

When compared to the immortalized parental hTERT PDCs, increased proliferation of both E6/E7 and E6/E7/Ras cells is to be expected due to the inactivation of p53/Rb and p16^INK4a by the human papilloma-virus proteins E6 and E7, a fact readily seen even in 2D [23]. Despite a deregulated cell cycle, E6/E7 PDCs still attach to laminin through integrin α6 and organize into “pancreatospheres” with a center void of cells, similar to hTERT PDCs [7]. It is only with the addition of aK-Ras^G12D mutation and only in 3D culture that an invasive phenotype results in the characteristic invadapodia of epithelial-to-mesenchymal transition [7,18]. In extending our previous findings, we now report that only K-Ras^G12D-mutated PDCs 1) increase TIMP1 secretion both in vitro and in vivo and 2) exhibit significant hyperproliferation and non-ductal architecture in 3D when exposed to high concentrations of exogenous TIMP1. These observations suggest that constitutive K-Ras^G12D signaling leads to the up-regulation of a variety of downstream genes that stimulate both invasion and hyperproliferation in an ERK2-dependent manner (Figure 2).

K-Ras regulates cell cycle progression as well as proliferative gene expression through the MAPK pathway [7,41,42]. DNA synthesis (as defined by EdU incorporation) and total cell counts in 3D cultured E6/E7/Ras PDCs were greatly diminished after MAPK inhibition with either UO126 (having an affinity for MEK1 over MEK2 at 5–10 µM) or the ERK2-specific inhibitor (AEMT; Figure 2). Preferential MEK1 inhibition prohibits accumulation of ERK2 in the nucleus, thereby inhibiting proliferation [43]. AEMT permits ERK2 phosphorylation by MEK but abolishes its downstream kinase activity, thereby inhibiting activation of ERK2-specific proliferation-associated genes [28]. This specific feature of phosphorylated ERK2 is only apparent in 3D cultured E6/E7/Ras PDCs and its activity positively correlates with enhanced cell proliferation (Figures 1 and 2). As in vivo data suggest that ERK1 overexpression inhibits Ras-induced subcutaneous tumor proliferation in nude mice, it follows that ERK2 overexpression may be a more favorable driver of Ras-regulated proliferation as we have shown in our mutated pancreatic cells [7,28,38]. In support of this notion, silencing of ERK2, and not ERK1, reduces E6/E7/Ras PDC hyperproliferation by approximately 65% in 3D (Figure 2). Although ERK1 and ERK2 share both primary RNA sequence and tertiary protein structure, ERK2's hinge region is modulated differently than ERK1 allowing different interdomain interactions [44]. Interestingly, when cultured in 2D, ERK2-silenced E6/E7/Ras PDCs exhibit a similar rate of proliferation as hTERT-transfected “normal” PDCs (Figure 2). As E6 and E7 proteins readily upregulate upstream MAPK pathway signaling in addition to oncogenic Ras, it follows that ERK2 pathway inhibition either reverses the unrestrained cell cycle progression caused by p53/Rb and p16^INK4a inactivation in PDCs or that it directs the K-Ras-MAPK signaling pathway toward ERK1 away from ERK2 proliferation induction [38]. It is important to note that while we have shown increased DNA synthesis due to the K-Ras^G12D mutation here in PDCs, increased cell survival due to AKT can also contribute to the total cell mass of neoplasms [45]. By targeting uniquely ERK2-dependent transformation capacities in addition to AKT, novel MAPK isoform-specific inhibitors that could bypass toxicity induced by the less selective inhibition of upstream effector molecules (i.e., B-Raf and MEK) while reducing the overall tumor burden might be developed [7,39,44,46].

Having discovered a constitutive K-Ras^G12D-driven, ERK2-mediated hyperproliferation axis, we used the gene signature of E6/E7 PDCs as a “non-transformed” baseline. This threshold distinguished downstream gene expression changes due to only the K-Ras^G12D mutation in E6/ E7/Ras PDCs and pinpointed specific genes that were directly regulated by MEK-ERK2 (Figure 3). Unbiased hierarchical analysis of our micro-array data identified 12 K-Ras^G12D-ERK2-regulated genes that were associated with potential epithelial-to-mesenchymal transition, ECM molecules, and proteases. Of those genes with proliferation-inducing potential, we focused on TIMP1 and validated its enhanced expression by qPCR. Indeed, analysis of tissue biopsy data in the Oncomine database confirms that TIMP1 is consistently overexpressed by at least twofold in multiple data sets of PDAC patients with a significant P value of at least .0002 (Figure 5C). Further, TIMP1 is increased within the sera of pancreatic cancer patients, both of which implicate involvement in proliferation and biomarker potentiality [47,48].

Although originally named for its ability to inhibit MMPs, TIMP1 is a dual function molecule theorized to be capable of growth-promoting activities at its C terminus, independent of its N-terminal endopeptidase inhibition characteristic [49]. Interestingly, TIMP1 has gene sequence homology to erythropoietin and has been shown to induce growth in a variety of cell types including breast, skin, and liver cells [18,20,50]. Although the concentrations of TIMP1 found in vivo are less than necessary to promote erythropoietin-like JAK2 proliferative signaling, TIMP1 may still retain the ability to promote proliferation through cellular membrane binding [18,21]. Indeed, prior studies demonstrate that high TIMP1 expression can induce in vivo tumor growth and alter the abundance of its MMP1 target depending on the dimensionality of the model [7,18,51]. Analysis of the MMP1 and TIMP1 promoter gene sequences reveals that both share putative consensus binding sites for a number of ERK-regulated transcription factors. Adjacent AP1, PEA3, and Ets regions 180 bp upstream of these gene expression sites may be involved in simultaneous transcriptional activation [12,31,41]. Thus, if ERK2 were to activate various transcription factors such as early growth response or ets transactivation variance, these factors could have a greater tendency to transcribe MMP1 and TIMP1, thus providing potential pretranscriptional and intertranscriptional targets for pharmacological inhibition of invasion and proliferation [7,52].

In accordance with ERK2-activated transcription of the TIMP1 gene, ELISA revealed increased TIMP1 protein secretion in PDCs harboring a K-Ras^G12D mutation (Figure 4). Importantly, addition of exogenous EGF did not further increase TIMP1 secretion, presumably due to the tyrosine kinase-independent activity of constitutive K-Ras^G12D signaling [4]. Pharmacological and RNAi inhibition of ERK2, but not ERK1, reduced the secretion of TIMP1, underscoring its regulation by this particular ERK isoform. Whether ERK2 silencing decreases both transcript and protein abundance or instead prevents extracellular TIMP1 secretion through an alternate mechanism has yet to be determined. TIMP1 silenced E6/E7/Ras PDCs exhibited minimal changes in 3D morphology or proliferation when compared to non-silenced controls similar to previous studies, underscoring that K-Ras^G12D must sufficiently induce higher microenvironment TIMP1 concentrations to yield increased proliferation (Figure 4) [31,53].

Indeed, primary cells isolated from the ductal pancreatic compartment of Pdx1-Cre;LSL-K-ras^G12D/+ mice express approximately 12-fold higher levels of TIMP1 RNA than control cells from “normal” mice (Figure 5). As others have found that the majority of TIMP1 is derived from the ductal compartment and secreted into the extracellular milieu, it is logical that its expression increased in the DBA+ PDC fraction of K-ras^G12D/+ mice pancreata [10]. Paralleling this finding and that elevated TIMP1 levels are found in clinical patients' sera, a minimal threefold addition of human rTIMP1 increased EdU incorporation into the nuclei of PDCs cultured in 3D but, again, only if they harbored a K-Ras^G12D mutation (Figure 6) [47,48]. This observation suggests that 1) the proliferative effect of TIMP1 overexpression requires an active K-Ras^G12D mutation in PDCs and cannot induce proliferative signaling otherwise and 2) the addition of non-ductal derived and secreted TIMP1 is necessary to raise the microenvironment concentration over a non-proliferative threshold. As overexpression but not silencing had an appreciable effect on cell proliferation, we surmise that TIMP1 may initiate its growth factor-like activity only at concentrations high enough to saturate its primary MMP inhibitory abilities, shifting the MMP-TIMP balance and enhancing extracellular membrane binding [9,54]. Interestingly, addition of exogenous rTIMP1 did not affect the proliferation of E6/E7 PDCs, which harbor an intact EGF receptor-dependent K-Ras signaling axis. As mutated K-Ras^G12D PDCs are independent of EGF receptor activation, these observations suggest that TIMP1 is exerting its proliferative effects by activating either a yet to be identified K-Ras^G12D upregulated receptor tyrosine kinase molecule or other putative cellular targets, such as hepatocyte growth factor or the CD63 receptor [50,55].

In non-transformed cells, TIMP1 inhibits cellular proliferation due to its repression of Wnt/β-catenin signaling; this inhibitory effect seems to be suppressed in transformed cells harboring K-Ras mutations because mutant K-Ras constitutively stimulates Wnt signaling and stabilizes β-catenin [54,56,57]. Further, although in vivo TIMP1 over-expression has been found to reduce the proliferation of pancreatic cancer cells injected within a murine peritoneum, decreased proliferation is only observed when TIMP1 is measured acutely [58]. When TIMP1 is overexpressed 24 hours after pancreatic cancer cell introduction, it actually increases the disseminated tumor volume, in line with our primary murine and human PDC 3D models (at 72 hours). These studies may explain why TIMP1 is capable of initiating proliferation in only K-Ras^G12D PDCs and only in 3D: TIMP1 harbors a protumorigenic effect after K-Ras^G12D-mutated PDCs have established unrestricted mechanical associations with the 3D ECM [7].

We conclude that K-Ras^G12D PDCs overexpress TIMP1 through an ERK2 signaling axis in a 3D ECM environment. The resulting increase in TIMP1 protein secretion leads to hyperproliferation of only K-Ras^G12D-mutated PDCs. Implicitly and as a caveat, therapeutic utilization of exogenous TIMP1 to reduce MMP-directed metastases in K-Ras^G12D-mutated pancreatic cancer may invariably stimulate proliferation of PDCs.

Abbreviations

AEMT

3-(2-aminoethyl)-5-((4-ethoxyphenyl)methylene)-2,4-thiazolidine hydroxychloride

ECM

extracellular matrix

GFR

growth factor reduced

PDAC

pancreatic ductal adenocarcinoma

PDCs

pancreatic ductal epithelial cells

MMP

matrix metalloproteinase

TIMP

tissue inhibitor of matrix metalloproteinase