Beta 1 integrin signaling mediates pancreatic ductal adenocarcinoma resistance to MEK inhibition

Pancreatic cancer, one of the deadliest human malignancies, has a dismal 5-year survival rate of 9%. KRAS is the most commonly mutated gene in pancreatic cancer, but clinical agents that directly target mutant KRAS are not available. Several effector pathways are activated downstream of oncogenic Kras, including MAPK signaling. MAPK signaling can be inhibited by targeting MEK1/2; unfortunately, this approach has been largely ineffective in pancreatic cancer. Here, we set out to identify mechanisms of MEK inhibitor resistance in pancreatic cancer. We optimized the culture of pancreatic tumor 3D clusters that utilized Matrigel as a basement membrane mimetic. Pancreatic tumor 3D clusters recapitulated mutant KRAS dependency and recalcitrance to MEK inhibition. Treatment of the clusters with trametinib, a MEK inhibitor, had only a modest effect on these cultures. We observed that cells adjacent to the basement membrane mimetic Matrigel survived MEK inhibition, while the cells in the interior layers underwent apoptosis. Our findings suggested that basement membrane attachment provided survival signals. We thus targeted integrin β1, a mediator of extracellular matrix contact, and found that combined MEK and integrin β1 inhibition bypassed trametinib resistance. Our data support exploring integrin signaling inhibition as a component of combination therapy in pancreatic cancer.

www.nature.com/scientificreports/ ECM-adjacent cells when given an β1 integrin neutralizing antibody. Lastly, dual blockade of both MEK and β1 integrin significantly increased PDAC cell apoptosis compared to singular inhibition of MEK or β1 integrin.
These results indicate that β1 integrin plays an important role in mediating PDAC resistance to MEK inhibition.

Results
Establishing a 3D culture model of pancreatic cancer. The iKras*;p53* mouse model of pancreatic cancer mimics the progression of the human disease 12 . In this model, oncogenic Kras G12D (Kras*) expression is regulated by a tet-response element, while mutant p53 R172H is constitutively expressed in the pancreas, allowing for inducible and reversible expression of Kras* upon administration or removal of doxycycline (DOX), respectively (Fig. 1a). The generation of cell lines from primary tumors formed in iKras*;p53* pancreata was previously described 13 . Subsequently, iKras*;p53* PDAC cells were passaged and maintained in two-dimensional culture in presence of DOX to maintain expression of oncogenic Kras (Fig. 1b). Growth in 3D was achieved by trypsinizing and resuspending PDAC cells in single-cell-suspension with cell culture medium that contained DOX and solubilized Growth Factor Reduced (GFR) Matrigel, a basement membrane mimetic comprised of extracellular matrix (ECM) proteins: laminins, Type IV collagen (Col4), and entactin. Matrigel facilitates 3D proliferation and adhesion of cells, interactions that may regulate crucial aspects of cancer molecular pathogenesis in vivo 14 . Subsequently, PDAC cells in suspension were plated atop a layer of solidified GFR Matrigel. Approximately 6 days following plating, PDAC cells organized into 3D, discrete clusters that were fixed for histochemical analysis (Fig. 1c-e). To determine whether cells in this 3D assay recapitulated oncogenic Kras* dependency observed in vivo, cells were randomized into two experimental groups: DOX was either withheld or administered to the media at the time of plating to inactivate or activate oncogenic Kras* expression, respectively. Brightfield microscopy was used to monitor cell growth and ImageJ was used to measure PDAC cluster area ( Supplementary Fig. 1). DOX administration induced an approximate fourfold increase in average cluster area ( Fig. 1f,g, Supplementary Fig. 2), indicating that oncogenic Kras* expression was sufficient to facilitate tumor cell growth and proliferation in this model. These results are consistent with previous in vivo and in vitro findings 13 .
In 2D, oncogenic Kras had a limited effect on cell viability ( Supplementary Fig. 3a,b). Conversely, 3D growth was dependent on the expression of oncogenic Kras ( Supplementary Fig. 2), mimicking the requirement for oncogenic Kras in vivo. We used the 3D system to study the effect of inhibiting Kras* downstream effector pathways, specifically MAPK and PI3K, in the tumor cells. First, we performed immunostaining on iKras*p53* PDAC clusters grown in the presence of DOX for 6 days. Similar to human PDAC and tissue from the primary iKras*p53* tumor (Fig. 2a,b), iKras*p53* PDAC cells in 3D culture upregulated phosphorylated ERK (pERK) (Fig. 2a) and phosphorylated S6 (pS6) (Fig. 2b), indicating activation of the MAPK and PI3K pathway, respectively. Furthermore, we observed expression of membrane proteins E-Cadherin and Claudin-18 ( Fig. 2c-j), both epithelial cell markers. Taken together, these data suggest two implications for our system. First, iKras*p53* PDAC cells in our 3D culture system recapitulate common biological characteristics of human PDAC and other murine in vivo models. Second, it can be used to study the effect of signaling pathways in a system that mimics the spatial relationships of tumor cells, while allowing us to dissect signaling, in absence of the complexity of the microenvironment.

MEK inhibition inhibits PDAC growth in 3D and induces apoptotic lumen formation. To inhibit
MAPK signaling, clusters were grown for 6 days and then administered PD325901 (MEK inhibitor, abbreviated MEKi-P) for 4 days; DOX was present in the media at all points, so that oncogenic Kras* was constitutively expressed. Administration of the inhibitor was sufficient to abrogate Kras*-mediated cluster growth ( Fig. 3a-e, Supplementary Figs. 3c,d, 7a-c). Upon histologic analysis we found that MEK inhibition decreased the expression of pERK, suggesting successful inactivation of the MAPK signaling pathway (Fig. 3f-i). However, cultures survived, indicating resistance to MEK inhibition. Similar results were obtained by others using pancreatic xenograft models to test MEK inhibition 15 . The MEK-resistant cells formed single-layer clusters, while the lumen (occupying > 75% of cross-section area) contained apoptotic debris. This finding was dose-dependent ( Fig. 3j), as increasing concentration of MEKi led to increased prevalence of single-layered clusters with large lumens. These results suggested that cells at the periphery of the cluster, and thus in contact with the basement membrane, were uniquely resistant to MEK inhibition. Intriguingly, this finding recapitulated what we and others observed in vivo upon inactivation of oncogenic Kras 13,16 .
To confirm this initial finding with a different MEK inhibitor, PDAC cells were grown for 1 week and administered a clinically available MEK inhibitor, trametinib (MEKi-T) for 4 days. To visualize any morphologic changes induced by MEK inhibition, PDAC clusters were fixed, sectioned, and stained for hematoxylin and eosin. Similar to clusters treated with MEKi-P, MEKi-T treated clusters were found to have an apoptotic lumen-evidenced by unorganized hyaline aggregation as well hyperchromatic debris, which indicates nuclear fragmentation, features consistent with cells that undergo programmed cell death (Fig. 4c). The prevalence of clusters with apoptotic debris that occupied > 75% of a single lumen was found to be significantly increased approximately sevenfold following MEK-T administration (Fig. 4d). To establish that the cells had indeed undergone apoptosis, we stained sections of PDAC clusters for cleaved caspase-3 (CC3) (Fig. 4a,b). Thus, using two different MEK inhibitors, we showed that cancer cells vary in their sensitivity to MEK inhibition.
PDAC cells adjacent to Matrigel display a survival advantage. The prevalence of single-layer cluster morphology suggests that cells in contact with the Matrigel basement membrane mimetic have a survival advantage. In this 3D system, Matrigel coats the entire outside of the cluster in vitro, so following fixation and sectioning, the outermost layer of cells in cluster cross-sections are considered adjacent to the Matrigel (Sup-Scientific RepoRtS | (2020) 10:11133 | https://doi.org/10.1038/s41598-020-67814-9 www.nature.com/scientificreports/ Figure 1. In a 3D culture system, iKras*;p53* cells recapitulate morphologic characteristics of the primary tumor. (a) Schematic describing the genetic model of the iKras*;p53* mouse, wherein administration of doxycycline (DOX) leads to pancreatic-epithelial-cell-specific expression of oncogenic Kras G12D (dominantnegative p53 R172H is also constitutively expressed in the pancreatic epithelium). PDA were isolated from endogenous tumors arising. (b) Brief description of endogenous primary tumor formation; in adult mice, DOX was administered through the drinking water. Three days following DOX administration, pancreatitis was induced through two series of intraperitoneal injections of caerulein. Following endogenous tumor formation, tissue was harvested from the primary tumor and the cells were isolated and placed in medium containing DOX. (c) Hematoxylin/eosin stain of primary iKras*p53* PDAC tumors. (d) Brightfield images of PDAC cell lines in 2D culture, maintained in doxycycline (1 µg/mL) (Kras* on). (e) Hematoxylin/eosin stain of iKras*p53* PDAC cell cross sections, 6 days following plating in the "on-top" 3D system (cells were also maintained in doxycycline (1 µg/mL). (f) Brightfield images of iKras*p53* cells plated in 3D in the absence or presence of doxycycline (1 µg/mL) (Kras* on or off, respectively), 6 days following plating of cells. (g) Quantification of cluster area size, 6 days following plating thin the absence (black bars) or presence (yellow bars) of doxycycline (1 µg/mL). In quantification, at least 100 clusters were traced and quantified in combined duplicate treatment wells. Bars represent average cluster area ± SD. *p < 0.01 in Student's t test analysis. Scale bars 50 µm; 500 µm (low magnification).
Scientific RepoRtS | (2020) 10:11133 | https://doi.org/10.1038/s41598-020-67814-9 www.nature.com/scientificreports/ plementary Fig. 1). This can be visualized by staining for Type IV collagen (Col4), a primary component of Matrigel and the basement membrane in vivo. Staining reveals Col4 enrichment on the outer edge of clusters treated with either vehicle (DMSO) or MEKi-T, suggesting that MEK inhibition does not affect organization of basement membrane (Fig. 5a). Upon quantification of cells adjacent or nonadjacent to Matrigel, the MEK-T treated clusters showed an increased prevalence of cells adjacent to the Matrigel, suggesting that these cells have a survival advantage over cells that lack ECM attachment (Fig. 5b). The phenomenon of anoikis, a form of programmed cell death in response to loss of ECM signaling, is well described in other systems 17 . We hypothesized that Matrigel-adjacent cells were able to resist anoikis due to their ability to interact and exchange signals with ECM components. Human and murine PDAC cells have been shown to engage in complex signaling with the surrounding microenvironment, yet the implications of this signaling and their effects on disease pathogenesis are currently www.nature.com/scientificreports/ www.nature.com/scientificreports/ poorly understood. In multiple systems, the transmembrane, bidirectional signaling molecule β1 integrin has been implicated in coordinating cell-to-cell and cell-to-ECM interactions 18 . And in the normal pancreas, β1 integrin expression is necessary for acinar maintenance 19 . In our 3D system, we found that iKras*p53* PDAC cells expressed β1 integrin ( Supplementary Fig. 4), and this expression was not affected by administration of MEKi-T ( Supplementary Fig. 4). Given that Matrigel-adjacent PDAC cells show a survival advantage and express β1 integrin, we hypothesized that β1 integrin signaling mediated survival in this population.

Beta-1 integrin inhibition disrupted cell:cell organization and decreased Kras effector
signaling. The function of β1 integrin is highly dependent on cell type as well as the cell's immediate microenvironment 20,21 . To determine the functional importance of β1 integrin signaling in our 3D culture model, PDAC clusters were grown for 7 days and treated with either solubilized rat immunoglobulin G (IgG) or a β1 integrin neutralizing antibody [Supplementary Table 1] for 4 days. Following administration of either IgG or β1 integrin neutralizing antibody, cells were fixed, sectioned histologically, and analyzed. After 4 days of treatment with control IgG, brightfield microscopy of PDAC clusters showed distinct colonies with smooth and ordered interactions with the surrounding Matrigel (Fig. 6a). In contrast, β1 integrin blockade induced structural disorder of PDAC clusters, disrupting their interaction with surrounding Matrigel as well as induced some disintegration of clusters, as evidenced by an increase in scattered, single cells (Fig. 6b). Upon histological analysis, PDAC clusters treated with IgG formed organized colonies with 1 or multiple lumen ( Fig. 6c-f); conversely, β1 integrin blockade abrogated lumen formation (Fig. 6 g-j), suggesting that β1 integrin outside-in signaling is necessary for PDAC cell:cell adhesion and formation of higher ordered 3D structure. Moreover, β1 integrin blockade affected E-cadherin (ECAD) expression at the cell membrane. PDAC clusters in the treated group showed punctate, noncontiguous staining, indicating disruption of ECAD localization to cell membranes (Fig. 6d-f, h-j). www.nature.com/scientificreports/ We then evaluated the effects of β1 integrin blockade on Kras* effector signaling. Following administration of control IgG, PDAC clusters demonstrated activation of MAPK and PI3K signaling in Matrigel-adjacent and non-Matrigel-adjacent cells indicated by positive staining of pERK and pS6 (Fig. 7a,e,c,g, Supplementary  Fig. 4). Strikingly, following β1 integrin blockade, Kras* effector signaling was restricted to Matrigel-adjacent cells (Fig. 7b,f,d,h). In samples treated with the blocking antibody, immunostaining detects the antibody bound to the extracellular domain of β1-integrin, with minor disruption in localization (Fig. 7i-l), colocalization of β1 integrin and pERK or pS6 signals was rare (Fig. 7m-t). These results suggest that β1 integrin signaling is necessary for PDAC upregulation of Kras* downstream signaling in the absence of ECM signaling in our system. Furthermore, β1 integrin signaling is dispensable for Kras* effector signaling, if cells are physically adjacent to the ECM (Supplementary Fig. 5).

Dual blockade of MEK and β1 integrin synergize to induce cancer cell death. Prior studies
have demonstrated that epithelial cell lumen formation is an active process that requires both cell:cell adhesion and cell:ECM signaling. Since lumen formation appears to be protective for PDAC cells in the context of MEK inhibition (Figs. 4d, 5b) and β1 integrin was necessary for lumen formation in our model (Fig. 6g-j), we hypothesized β1 integrin signaling blockade would prevent lumen formation and potentiate the ability of MEK inhibition to induce apoptosis. To test this, PDAC cells were grown in 3D for 7 days and treated for 4 days with vehicle (DMSO + IgG), MEK-T, an anti-β1 integrin neutralizing antibody, or a combination of the compounds. Subsequently, PDAC clusters were fixed, sectioned and prepared for immunofluorescence to examine CC3 or TUNEL staining. Singular administration of anti-β1 integrin neutralizing antibody, but not MEK-T, increased apoptosis, as indicated by increased CC3 staining, compared to the vehicle group (Fig. 8a-c, e-g, i-k, m-o, Supplementary Fig. 6). Furthermore, dual inhibition led to significantly increased apoptosis when compared to singular blockade of either pathway (Fig. 8d,h,l,p,q). TUNEL staining demonstrated similar findings; However, singular blockade of MEK significantly increased cell death according to the TUNEL staining. Still, clusters in the dual blockade group showed significantly increased death when compared to singular blockade of either pathway (Fig. 8r,s). These results, taken together, suggest that β1 integrin signaling mediates PDAC resistance to MEK inhibition. www.nature.com/scientificreports/

Discussion
Our results show that MEK inhibition induced lumen formation in iKras*p53*-mouse-derived PDAC cells in 3D culture. Furthermore, the presence of Cleaved Caspase 3 and TUNEL positive cells in the lumen suggested that MEK inhibition sensitized PDAC cells to anoikis, programmed cell death in response to detachment from the ECM 25 . Cell:cell adhesion and ECM:cell are key signaling networks that may regulate PDAC progression [22][23][24] . Elucidation of these pathways may provide insight into mechanisms that regulate disease persistence, which would ideally lead to practical, effective molecular targeting and reduction of PDAC burden. Given that ECMadjacent PDAC cells displayed a survival advantage in the context of MAPK inhibition, we sought to characterize the role of ECM:cell signaling in our system using Matrigel. Matrigel is a basement membrane mimetic and the primary protein components include, but are not limited to: collagen IV, laminin, and enactin 26 . Similarly, the PDAC extracellular matrix, including the basement membrane, is mainly composed of type IV collagen and laminin 26,27 . We hypothesized that β1 integrin played an important role in the survival of ECM-adjacent PDAC cells because the B1 integrin receptor is a binding site for both collagen IV and laminin 28 and knockdown of B1 integrin in PDAC results in decreased cell adhesion to collagen IV and laminin 29 .
The integrin family of cell adhesion receptors mediates a multitude of cellular functions crucial to the initiation, progression and metastasis of solid tumors 30 . In human breast cancer, overexpression of β1 integrin is www.nature.com/scientificreports/ correlated with poor prognosis 31,32 , and in vitro β1 blockade induced apoptosis of malignant cells 33 . In pancreatic cancer, overexpression of β1 is correlated with poorer disease-free survival 34 . In a murine model of pancreatic β cell cancer, genetic ablation of β1 integrin led to reduced tumor cell proliferation and senescence 35 . Pancreatic cancer cell lines overexpress integrin ɑ subunits 1-6 and β subunits; moreover, the β1 subunit is found to be constitutively active, mediating adhesion and invasiveness in some PDAC lines 36 . We found this to be true, as the β1 integrin neutralizing antibody caused scattering of the cells in the Matrigel matrix. This decreased adhesion could be responsible for the cell death we saw. Further, we found that lumen formation as well as PDAC cell www.nature.com/scientificreports/ www.nature.com/scientificreports/ survival in the context of MEK inhibition were significantly decreased following β1 integrin neutralizing antibody administration, suggesting that PDAC resistance to MEK inhibition is mediated in part by β1 integrin signaling. Another possibility for the cause of the lumen formation is cell migration. β1 integrin plays an important role in cell migration and could be the reason for lumen forming during MEK inhibition but not singular β1 integrin blockade or dual treatment 37 . To test this theory, applying low concentrations of microtubule inhibitors that inhibit cell migration or observing the cells under live imaging could confirm that the cells are migrating away from the center of the cluster 38 .
The last facet of β1 integrin signaling to consider is inside-out signaling. While it is beyond the scope of our study, the role of β1 integrin-fibronectin inside-out signaling may play a role in this system. It is known that fibronectin-β1 integrin interactions play a role in invasion and metastasis and blocking these interactions with antibodies resulted in inhibition of anchorage-independent growth 39 . Additionally, β1 integrin binding to fibronectin supports ROCK mediated actomyosin contractility 40 , which can increase tissue tension and potentially plays an important role as PDAC tumor rigidity is correlated with epithelial-mesenchymal transition (EMT) and chemoresistance 41 . Further, loss of inside-out signaling can induce apoptosis through several different pathways 42 .
Strikingly, singular blockade of β1 integrin decreased Kras* effector signaling in cells lacking ECM attachment but failed to decrease MAPK or PI3K signaling pathway activation in Matrigel-adjacent PDAC cells. These data suggest two novel findings in relation to iKras*p53* PDAC cell biology in vitro. First, cells lacking ECM interaction require β1 integrin to upregulate downstream effectors of Kras*. This is consistent with previous findings that β1 integrin is involved in ERK signaling in PDAC 43 and inhibition promotes central necrosis 44 . The combination therapy likely exacerbated apoptosis due to β1 integrin's ability to activate PI3K signaling in PDAC 45 . Furthermore, β1 integrin has be shown to be necessary to activate PI3K when MEK is inhibited 46 and PI3K inhibition produces central necrosis in a pattern similar to our data 47 . The second conclusion drawn from our data suggest that ECM-adjacent cells upregulate Kras*-effector signaling in a β1 integrin-signaling-independent manner. One likely mechanism to explain this is β5 integrin signaling (Fig. 7u). β5 integrin is present in PDAC 44 and can activate MEK signaling in breast cancer 48 . Upon β5 integrin inhibition in breast cancer, proliferation is decreased 48 . Lastly, suppression of a specific heterodimer containing β5 integrin (α5β5) in colon cancer increases the function of a heterodimer containing β1 44 . This crosstalk between β1 and β5 integrins suggest that β5 integrin could upregulate Kras-effector signaling in ECM-adjacent cells when β1 integrin is inhibited. However, future tests will need to be done to examine the role of β5 integrin in PDAC as well as characterizing how subunit-specific signaling may modulate different characteristics of PDAC interaction with the microenvironment, subsequently regulating PDAC cell survival.
Ethical approval for use of human cell lines. All procedures performed in studies involving human participants and use of human cell lines were approved by the University of Michigan. All methods were performed in accordance with the ethical standards of the 1964 Helsinki declaration and its later amendments and all human patients provided informed consent for the study. For human cell lines, medical chart review was used to screen for potential study patients with pancreatic disease at the University of Michigan. Surgical specimens of either tumor tissue or adjacent normal pancreas were obtained from patients referred for Whipple procedure or distal pancreatectomy according to IRB HUM00025339.
Cell culture. iKras*p53* PDAC cells or human PDAC cells were maintained and passaged in 2D culture in IMDM media + 10%FBS + DOX 1 μg/mL 53 . The 3D assay used followed the "3D On-Top" cell culture protocol 14,53 . Four-well chamber slides (Corning 354,104) were coated with 50-100 μL Growth-Factor-Reduced (GFR) Matrigel. Afterward, the slides were incubated for 10 min at 37 °C to induce solidification of GFR Matrigel. PDAC cells were trypsinized and centrifuged. Pellets were resuspended in single-cell-suspension in Waymouth's media (Gibco 11,220-035) + 10% Fetal Bovine Serum (Gibco 10438-036) + 1% penicillin/streptomycin (Gibco 15140-122) and the solution was added atop the GFR Matrigel surface. Every 2-3 days, media was changed. In experiments using small molecule inhibitors, inhibitors were solubilized in DMSO or 1 × PBS, and vehicle or drug treatment groups were administered in either technical duplicate or triplicate. Representative experiments of at least 3 biological replicates has been shown unless otherwise noted.
Fixation and staining. The staining protocol was adapted from a protocol established for studying breast cancer cell subpopulations in 3D Matrigel culture 54 . PDAC clusters were fixed in formalin at 22 °C for at least 2 h. After fixation, chambers from chamber slides were removed, and PDAC clusters in GFR Matrigel were collected Scientific RepoRtS | (2020) 10:11133 | https://doi.org/10.1038/s41598-020-67814-9 www.nature.com/scientificreports/ and placed in a cryomold (Sakura 4,557) with Histogel (Thermo Scientific HG-4000-012) 55 . These samples were then processed, embedded in paraffin, and cut into 5 μm-thick sections. The primary antibodies were added and then antigen retrieval was accomplished using the respective antibodies and concentrations seen in Supplementary Table 1. Histology and immunofluorescence analysis were performed as described below.
Brightfield and cleaved-caspase 3 imaging and quantification. Brightfield and fluorescent images were acquired with an Olympus BX-51 microscope DP71 digital camera/software as well as Pannoramic SCAN II slide scanner and software. Brightfield, low-magnification images of clusters growing in chamber slides were used to quantify PDAC cluster area. Pictures of at least 5 non-overlapping areas were taken and a blinded observer used ImageJ to trace and measure cluster area. Results indicate the average area of at least 100 unique clusters per treatment group. To quantify cleaved-caspase 3 positivity, a blinded observer analyzed cross sections of fixed, sectioned, and stained PDAC clusters. The number of nuclei as well as positive and negative staining was recorded in at least 50 cross sections of unique PDAC cell clusters from 3 biological replicate experiments.
TUNEL staining and quantification. The sections from two biological replicate experiments were stained using a Terminal deoxynucleotidyl transferase (TdT) dUTP Nick-End Labeling (TUNEL) assay (Millipore S7165), according to the manufacturer's instructions. The slides were deparaffinized, pretreated, and antigen retrieval was done by first using a TdT enzyme and then a rhodamine antibody solution. DAPI was used as a nuclear counterstain. Images of the fluorescently labelled sections were obtained using a Leica SP5X Upright Two-Photon Confocal Microscope. To quantify TUNEL positivity, the number of TUNEL positive nuclei were obtained as well as total number of nuclei per cell cluster.