Supplemental Figure and Table Legends

Figure S1: (A) Illustration of how 7TGC-transduced cells were analyzed. First, viable and single cells were gated by FSC/SSC, then mCherrypositive cells identified to insure a constant background, and then GFPhigh cells were counted. (B) Serial enrichment for the 5% L3.6pl GFPhigh cells leads to a temporary augmentation of the GFPhigh cell fraction. Shown is the enrichment over 5 cycles. (C) Western blot analysis of highly enriched GFPhigh vs. GFPlow cells for phospho-β-Catenin (S552) and total β-Catenin. (D) Cytosolic protein lysate of the same cell fractions as in (C) reveals higher expression of Wnt signaling pathway components. (E) Total cell lysates show lower epithelial/differentiation markers and higher mesenchymal marker in GFPhigh PDAC cells. (F) Bulk tumor cells of PDAC cell lines as in Figure 1A were analyzed for EMT markers by Western blotting. (G) Immunofluorescent stains for GFP (green), DAPI (blue), and β-Catenin (red, left panel) or CK19 (red, right panel) in orthotopic pancreatic cancer xenografts.

Figure S2: (A) Sphere-formation ability is increased in GFPhigh cells (here in L3.6pl and Patx1). (B and C) Bright-field microscopy of several epithelial and mesenchymal PDAC cell lines in adherent or sphere conditions: epithelial cells reveal distinct clusters (adherent) and tightly packed spheres, whereas mesenchymal cells are spindly (adherent) and form grape-like, loosely structured spheres. (D) Sphere formation is reduced with canonical Wnt inhibitor IWR1-endo in total number as well as percentage of mesenchymal spheres [M]/epithelial spheres [E] (right graph and images). (E) Similar to Figure 2F, quantification of holoclones of a second L3.6pl clone according to morphology. (F) Bright-field and fluorescent microscopy of holoclones shows no GFP signal in epithelial clones, highest number in mesenchymal holoclones, and mixed in between. (G) Cell cycle analysis of GFPhigh and GFPlow L3.6pl: significant difference in cell cycle distribution with shift to G2 in the latter. (H) Cell cycle analysis of a panel of PDAC cell lines: CSC-enriched spheres reveal a high percentage in the G0/G1 phase, indicating quiescence.

Figure S3: (A) Micro-MRI of second-generation orthotopic pancreatic tumor was performed at two time points (after 4 [upper panel] and 9 [lower panel] weeks). The selected GFPlow tumor (#80 in Figure 3D) revealed no tumor first and modest tumor volumes at 9 weeks; the GFPhigh tumor showed tumor volumes after 4 weeks comparable to 9 weeks in GFPlow and a huge tumor mass at endpoint. (B) Metastatic spread of a GFPhigh subcutaneous tumor into the pancreas (arrowhead, upper image), peritoneum (black arrows, both images), and diaphragm (arrowhead, lower image). (C and D) Cells isolated from liver metastases [LM] (grey bar) have higher number of GFPhigh cells compared to cells from a primary tumor [P] (white bar) (D) and form more mesenchymal spheres (green arrowhead) than epithelial spheres (black arrowhead) (D). (E) In a second case, metastatic cells of ascites or omental metastasis possess high numbers of GFPhigh cells compared to the primary tumor. (F) Western blot analysis shows higher expression of LRP6, Dvl2 and p-ERK in LM compared to primary tumor or normal pancreas.

Figure S4: (A) Number of GFPhigh cells in tumors treated with vehicle (Co1, Co2) or gemcitabine (GEM1, GEM2) as evaluated by flow cytometry. (B) Flow cytometry: PDAC cell lines were quantitatively analyzed for GFPhigh cells in controls or after gemcitabine treatment in vitro. (C) Flow cytometry of Patx1 cells after 72 h of treatment with DMSO, IWR1-endo [10 uM], gemcitabine (GEM) [100 ng/mL], or a combination thereof. (D) qRT-PCR: Shown is the ratio of gene expression of cells treated with gemcitabine over control cells in Panc1 and Patx1 adherent cells or Patx1 grown as spheres. n=3 (E) qRT-PCR of different Wnt target genes in adherent and sphere cells of different generations. (F) qRT-PCR of Wnt target gene AXIN2 in adherent and sphere cells of different PDAC cell lines as well as after cell sorting in GFPhigh and GFPlow. (G) high low IC50 of gemcitabine for GFP and GFP (related to Figure 4 F). (H) Annexin V - APC/Sytox Blue analysis of cells derived from tumors 78 (upper panel) and 80 (lower panel) after TRAIL [10 ng/mL] or gemcitabine [100 ng/mL] treatment and its quantification (right bar graphs).

Figure S5: (A) Log2 copy number units of RSPO1-4 in normal pancreas vs. PDAC from the TCGA dataset (Oncomine). RSPO2 and 3 reveal significant differences, with RSPO2 higher in PDAC and RSPO3 lower in PDAC compared to normal pancreas. (B) From the same dataset, specification of Log2 copy number units of RSPO2 according to tumor stage (left) or nodal stage (right). (C) RSpo2 low (left) and high (right) protein expressions in tumor microarray stains of pancreatic cancer.

Figure S6: (A) GFP analysis by flow cytometry with and without stimulation in Capan2, Panc1, and MiaPaCa2 (analogue to Figure 5A). (B) Western blot analysis of high responders BxPC3, Capan2, and Panc1 for GFP, p-ERK, and ERK (similar to Figure 6A). (C) Western blot analysis of KPC-1050 cells after sorting into CD24/CD44low and CD24/CD44high subsets. It shows slightly higher nuclear expressions of β-Catenin and p- ERK in CD24/CD44high cell populations. (D) Western blot for GFP, (p)-LRP6, (p)-ERK and β-Actin in L3.6pl with or without RSpo2 and increasing doses of UO126 (left panel). (E) The same cells were analyzed for GFP expression by flow cytometry.

Figure S7: (A) Western blot for RSpo2, p-ERK, and ERK in bulk tumor cell lysates of PDAC cell lines used in Figure 1. Actin is part of Figure S1F. (B) RSpo2 IHC of a PDAC specimen from the public database The Human Protein Atlas: positive staining in connected epithelial tumor cell clusters (black arrow) and increased staining intensity in single detached cells (blue arrow). Of note, single cells seem EMT transformed due to smaller and spindle-like appearance. (C) TMA IHC staining for Zeb1, RSpo2, and p- ERK in subsequent sections show correlation of these three markers. (D) Sphere- formation ability is markedly inhibited by an anti-RSpo2 antibody [5 ug/mL] compared to untreated (white bar) and control antibody-treated cells (dark gray bar). (E) Cell viability is only marginally influenced by the same treatment.

Figure S8: Graphical illustration of the proposed model. PDAC bulk tumor cells seem to rest in a Wnt-inactive state; however, Wnt ligand production might stem from this source. Susceptible and plastic pancreatic CSCs intrinsically or extrinsically activate canonical Wnt signaling with critical enhancement by RSpo2. This leads to ERK phosphorylation, downstream activation of EMT, and a higher stemness level with display of critical CSC hallmarks such as chemoresistance and metastasis. Therapeutic intervention might be advantageous on multiple levels against Wnt and ERK, especially in CSCs, as well as standard therapeutic regimens in highly proliferative cells (options are shown in red).

Table S1: Complete RPPA datasets for L3.6pl and Patx1.

Table S2: List of murine primer sequences used in qRT-PCR analysis in this study.

Table S3: List of human primer sequences used in qRT-PCR analysis in this study.

Table S4: List of antibodies, assays, and dilutions utilized in this study.

Table S5: Evaluation of RSpo2 protein expression in tumor microarrays