PAWI-2 overcomes tumor stemness and drug resistance via cell cycle arrest in integrin β3-KRAS-dependent pancreatic cancer stem cells

Today, pancreatic cancer (PC) remains a major health problem in the US. The fact that cancer stem cells (CSCs) become enriched in humans following anti-cancer therapy implicates CSCs as key contributors to tumor dormancy, metastasis, and relapse in PC. A highly validated CSC model (FGβ3 cells) was used to test a novel compound (PAWI-2) to eradicate CSCs. Compared to parental bulk FG cells, PAWI-2 showed greater potency to inhibit cell viability and self-renewal capacity of FGβ3 cells. For FGβ3 cells, dysregulated integrin β3-KRAS signaling drives tumor progression. PAWI-2 inhibited β3-KRAS signaling independent of KRAS. This is clinically relevant. PAWI-2 targeted the downstream TBK1 phosphorylation cascade that was negatively regulated by optineurin phosphorylation via a feedback mechanism. This was confirmed by TBK1 genetic knockdown or co-treatment with TBK1-specific inhibitor (MRT67307). PAWI-2 also overcame erlotinib (an EGFR inhibitor) resistance in FGβ3 cells more potently than bortezomib. In the proposed working model, optineurin acts as a key regulator to link inhibition of KRAS signaling and cell cycle arrest (G2/M). The findings show PAWI-2 is a new approach to reverse tumor stemness that resensitizes CSC tumors to drug inhibition.

Compared to cells treated with PAWI-2 alone, co-treatment with TBK1 kinase inhibitor (MRT67307) 27 and PAWI-2 enhanced inhibition of cell viability (30% greater; Supplemental Table S2; Fig. 2F) and self-renewal capacity (25% greater; Fig. 2G) in FG and FGβ 3 cells. In FGβ 3 cells, enhancement of cell killing with co-treatment with PAWI-2 and MRT67307 was not associated with induction of cell apoptosis. For example, in the presence of PAWI-2 and MRT67307, caspase activation and PARP cleavage was comparable to treatment of PAWI-2 alone (Supplemental Fig. S3A,B). The enhanced inhibition of cell viability and self-renewal capacity (Fig. 2F,G) was associated with OPTN phosphorylation (2-4 fold activation; Supplemental Fig. S3C) similar to the result observed in the genetic knockdown of TBK1 (2-5 fold activation; Fig. 2E). Moreover, pharmaceutical inhibition of TBK1 by MRT67307 also down-regulated phosphorylation of p62 (pS403-p62) but not pS172-TBK1 or pS177-OPTN (Supplemental Fig. S3C). This result showed that phosphorylation of p62 induced by PAWI-2 was most likely related to TBK1 activity but phosphorylation of OPTN may not be solely associated with TBK1 activity.

Effect of PAWI-2 on OPTN phosphorylation in the presence of other inhibitors. Integrin
β 3 -mediated self-renewal capacity is associated with drug resistance in FGβ 3 cells 19 . Co-administration of erlotinib with proteasome inhibitor bortezomib was examined to determine effects on cell viability (Supplemental Table S2) and self-renewal capacity 19,21 . In FGβ 3 cells, co-administration of "PAWI-2 and erlotinib" enhanced inhibition of erlotinib alone on cell viability (30% greater) and self-renewal capacity (80% greater), compared to co-administration of "erlotinib and bortezomib" (Fig. 3A,B). Chou-Talalay analysis of synergism or antagonism was calculated based on a dose-dependent inhibition of cell viability for drug alone or drug-drug combinations. Synergism or antagonism between drugs was defined by combination index (CI values), showing PAWI-2 synergized erlotinib (but antagonized bortezomib, CI values > 1; Table 1) with greater synergism for FGβ 3 cells compared to FG cells (CI values < 1; Table 1). Synergism between "erlotinib and bortezomib" was observed in FG cells but was less apparent in FGβ 3 cells ( Table 1). Synergism of "PAWI-2 and erlotinib" paralleled induction of cell apoptosis (i.e., "PAWI-2 and erlotinib" enhanced activation of caspase and PARP cleavage compared to PAWI-2 or erlotinib alone; Supplemental Fig. S4A,B).  www.nature.com/scientificreports www.nature.com/scientificreports/ Bortezomib works on inhibition of late stage autophagy that promotes accumulation of p62 28 . However, in our hands, the effect of bortezomib on autophagy alone or in combination with erlotinib in FG and FGβ 3 cells was modest (p62 and LC3 accumulation was < 2-fold; Fig. 3C, Supplemental Fig. S4C). The distinct pattern of changes of LC3-I to LC3-II was not significantly affected by co-treatment of PAWI-2 with erlotinib. This showed synergism of PAWI-2 with erlotinib was not dominated by an autophagy-related effect. Synergism between PAWI-2 and erlotinib and antagonism between PAWI-2 with bortezomib were highly correlated with OPTN phosphorylation based on a plot of CI values versus pS177-OPTN fold-change (correlation coefficient r 2 > 0.8).
Co-administration of erlotinib and PAWI-2 increased pS177-OPTN 4-fold in FGβ 3 cells. In contrast, in the presence of PAWI-2 and bortezomib, OPTN phosphorylation was at control value. Similar results were observed for p62 and this can be explained because OPTN acts like a p62-like receptor 29 .
Combination chemotherapy of gemcitabine and nab-paclitaxel has been widely used in the treatment of advanced PC 30,31 . This drug combination showed comparable inhibition of FG cell viability with PAWI-2 alone (Supplemental Table S3; Fig. 3D). However, in FGβ 3 cells, co-administration of gemcitabine and paclitaxel did not show significant enhancement on inhibition of cell viability compared to gemcitabine or paclitaxel alone (Fig. 3D). Drug resistance of this combination in FGβ 3 cells was not associated with activated apoptosis because a comparable effect (activation on caspase activity and PARP cleavage for the combination compared to gemcitabine or paclitaxel alone; Supplemental Fig. S5A,B) was observed in both FG and FGβ 3 cells. Synergism between gemcitabine and paclitaxel was associated with OPTN phosphorylation (Fig. 3E). OPTN phosphorylation may be linked to microtubule (MT) disturbance because this effect was also observed in paclitaxel (MT stabilizer)-treated cells (Fig. 3E).
Double thymidine block arrests cells at the G1/S boundary and subsequent release to fresh media arrests cells at different boundaries 32 . Experiments were done to synchronize FGβ 3 cells at the G1/S boundary and release upon treatment with vehicle control, PAWI-2 or paclitaxel (Fig. 4D). Activation of pS177-OPTN was detected at later G2/M phase (8 hours after release). This was closely associated with onset of cyclin D3 degradation and also inhibition of TBK1 phosphorylation (Fig. 4D). Similar results were observed for phosphorylation of p62 on Ser403 (pS403-p62). Based on intracellular distribution studies in FGβ 3 cells, OPTN and pS177-OPTN (and also p62, NDP52) were mainly found in the cytoplasmic fraction under vehicle control conditions but accumulated in the nucleus with PAWI-2 (Fig. 4E). Accumulation of OPTN in the nuclear fraction was an indicator of G2/M arrest 33 . Cellular trafficking mediated by PAWI-2 was also associated with acetylated tubulin localization to nuclei (Fig. 4E). Similarly, nuclear cyclin D3 downregulation and accumulation of p21 (and its phosphorylated form) in cytoplasm were observed after administration of PAWI-2, paclitaxel or colchicine to FGβ 3 cells (Fig. 4F), providing strong evidence that these MT disturbing agents caused FGβ 3 cell G2/M arrest 34,35 . Together, these data show that PAWI-2 induced OPTN phosphorylation was highly associated with cell cycle arrest during mitosis.

Discussion
We have shown that PAWI-2 could reverse cancer stemness and overcome drug resistance in an integrin β 3 KRAS-dependent hPCSCs (i.e., FGβ 3 cells). A working model of PAWI-2 was proposed (Fig. 5). In this model, OPTN plays a central role in regulation of TBK1 functional activity to reverse tumor stemness and drug resistance in FGβ 3 cells. Phosphorylation of conserved OPTN residue (Ser177) by PAWI-2 promotes OPTN translocation www.nature.com/scientificreports www.nature.com/scientificreports/ into the nucleus and causes G2/M arrest. Concomitantly, OPTN phosphorylation induced by PAWI-2 has negative feedback control on TBK1 (dephosphorylation of TBK1 at S172) to inhibit dysregulation of KRAS-NF-κB signaling in FGβ 3 cells. This model links a role of OPTN to the functional interplay between G2/M cell cycle arrest and provides a mechanism to explain how PAWI-2 overcomes tumor stemness.
Previously, we showed PAWI-2 activated DNA-damage checkpoint and mitochondrial p53-dependent apoptotic signaling in other non-CSC cancer cells [22][23][24] . Data herein showed this was also observed for hPCSCs (FGβ 3 cells). For dysregulated KRAS-RalB-NF-κB signaling in FGβ 3 cells, galectin-3 plays a critical role in clustering www.nature.com/scientificreports www.nature.com/scientificreports/ integrin α v β 3 to induce KRAS and enable multiple processes in cancer cells 21 . In the study herein, PAWI-2 did not disrupt KRAS interactions with other effectors. This differentiates PAWI-2 from other drugs (e.g., GCS-100), that act as galectin-3 inhibitors and pharmacologically disrupt biochemical association between integrin α v β 3 and KRAS 21 . RalA/B serves as molecular regulators of integrin α v β 3 -KRAS-NF-κB signaling. PAWI-2 also did not measurably affect the inactive/active forms of RalA/B. These findings suggest that PAWI-2 inhibited KRAS-NF-κB signaling regardless of KRAS or Ral status. Given the fact that >90% of KRAS is activated by mutations in PC 36 and RAS or Ral inhibitors of these pathways have not proven effective clinically 19 , this suggests that PAWI-2 may possess advantages in clinical applications.
TBK1 is a serine/threonine kinase that is activated by autophosphorylation at Ser172 within the kinase activation loop 37 . Association of TBK1 with RalB of the major oncogene (RAS) in the integrin α v β 3 -KRAS-NF-κB signaling pathway promotes tumorigenesis 19,21 . TBK1 inhibitors (e.g., momelotinib) show limited utility in PC even in combination with other effective PC therapeutics 38 . As a key kinase in several signaling pathways, TBK1 also phosphorylates p62 or OPTN to enhance their binding capacity with poly-ubiquitin (poly-UB) chains 26,39 . TBK1 constitutively interacts with OPTN to act as a key modulator to initiate elimination of damaged mitochondria via selective mitophagy (PINK1/Parkin-dependent mitophagy), that is involved in tumor suppression pathways 40,41 . PAWI-2 was previously reported to affect mitochondrial function (i.e., membrane trafficking, mitochondrial membrane potential changes) 23,24 . However, in FGβ 3 cells neither PINK1 nor Parkin proteins were altered by PAWI-2. This data excludes mitophagy mechanisms initiated via OPTN by PAWI-2. PAWI-2 did not change autophagy biomarker LC3-I to lipidated form LC3-II (Supplemental Fig. S4C). Activation of OPTN phosphorylation by PAWI-2 may be related to other signaling cascades not solely dependent on TBK1. OPTN has also been shown to directly regulate TBK1 42 . A negative feedback control of TBK1 activation by OPTN helps explain the proposed working mechanism of PAWI-2. PAWI-2-induced OPTN phosphorylation negatively regulates TBK1 functional activity (i.e., autophosphorylation inhibited), and causes inhibition of KRAS-NF-κB signaling. This was further shown by exacerbated effects of PAWI-2 on the action of genetic knockdown of TBK1 and pharmacological inhibition (MRT67307) of TBK1 activation. Interestingly, MRT67307 does not affect accumulation of pS172-TBK1 (reversely activated). This shows that in contrast to previous reports 27,43 , TBK1 activation may not be the sole autocatalytic mechanism responsible operating for MRT67307.
In addition to being a downstream regulator of TBK1 function, OPTN is involved in a variety of other biological functions, including protection against apoptosis, Golgi organization, exocytosis, antiviral innate immune response, selective autophagy and other membrane trafficking mechanisms 29,41 . OPTN does not have any reported enzymatic activity but usually acts as an adaptor protein that links two different proteins (e.g., TBK1 and PINK1/Parkin) 29,41 . For tumorigenesis or tumor stemness, OPTN phosphorylation has been largely attributed to regulation of mitophagy 44,45 mediated by TBK1, but that was not observed herein for PAWI-2. Phosphorylation of OPTN at Ser177 also plays a pivotal role in mitotic progression and induces OPTN translocation into the nucleus 46 . OPTN-dependent G2/M cell cycle arrest induced by PAWI-2 in FGβ 3 cells parallels this process. Previously, G2/M arrest was independently observed in PAWI-2-treated colon cancer cells 22 . This regulatory mechanism is abolished at the end of the G2/M phase as a consequence of nuclear translocation of OPTN and leads to increased activity of TBK1 (Supplemental Fig. S6).
Synergism between PAWI-2 and other validated drugs (i.e., erlotinib) was controlled by phosphorylation of OPTN. In contrast, in FGβ 3 cells, if antagonism was observed (e.g., PAWI-2 with bortezomib), phosphorylation of OPTN was abolished. This observation helps explain drug resistance observed for FGβ 3 cells treated with well-documented PC chemotherapies (e.g., gemcitabine with paclitaxel, Fig. 3E) 30,31 . OPTN may work as an over-arching branch-point for PAWI-2 inhibition of cell viability to overcome self-renewal capacity in FGβ 3 cells and also to synergize other pathway inhibitors (i.e., erlotinib).
In PC cells, PAWI-2 binds to tubulin to stabilize/destabilize microtubules (MTs) and activate apoptotic signaling [22][23][24] . Phosphorylation of OPTN was closely associated with MT stabilization because this effect was also observed in cells treated with other MT stabilizers (e.g., paclitaxel or docetaxel; Fig. 4B). OPTN foci distribution www.nature.com/scientificreports www.nature.com/scientificreports/ is dependent on the integrity of MTs 46,47 , but no relationship between OPTN phosphorylation and MT disturbance has been reported thus far. Nothing describing synergism between clinically-validated cancer drugs through regulation of OPTN has been reported. Accumulation of pS177-OPTN in the presence of MT stabilizers may be due to the essential role of MTs in coordinating and organizing many crucial cellular steps 48 . Thus, OPTN phosphorylation induced by PAWI-2 or other MT stabilizers could modulate synergism effects to overcome drug resistance and combat more aggressive CSCs.
In conclusion, PAWI-2 synergized specific pathway inhibitors (e.g., TBK1 inhibitors, EGFR inhibitors) against CSCs. Selective pharmacological potency of PAWI-2 in CSCs (e.g., FGβ 3 cells versus FG cells) showed the utility of PAWI-2 to inhibit CSCs versus bulk cancer cells. This observation provides a basis for PAWI-2 as an efficient treatment of PC, especially in highly aggressive/metastatic cancer with stem-like properties and intrinsic or acquired drug resistance.

Methods
Cell lines. FG and FGβ 3 cells were provided by Dr. David Cheresh (UC San Diego and The Scripps Research Institute). FGβ 3 cells have been thoroughly documented as an aggressive cell line showing CSC-like properties and cancer drug resistance [19][20][21] . FG and FGβ 3 cells were grown in DMEM with 10% FBS. After thawing, cell lines were cultured at 37 °C in a humidified 5% CO 2 atmosphere and routinely screened for mycoplasma contamination.
Other drugs/inhibitors used in this study are listed in the Supplementary Materials and Methods.
Cell viability and apoptosis assays. FG and FGβ 3 cells were seeded onto plates and treated with test compounds (vehicle, 0.5% DMSO; PAWI-2 or other drugs; 2 to 5000 nM) for 3 days. Cell viability was determined using CellTiter-Glo (Promega). Data were expressed as percentage of survival compared to survival of vehicle-treated cells. A similar protocol was used to test synergy of PAWI-2 in the presence of erlotinib and/or bortezomib. Chou-Talalay analysis used commercial software (ComboSyn) 50 . Cell apoptosis was determined by quantifying caspase-3/7 activity using Caspase-Glo 3/7 (Promega). Tumor-sphere culture and self-renewal assay. FG and FGβ 3 cells were seeded on ultra-low attachment plates at single-cell suspensions (1,000 cells ml −1 ) in DMEM/F12 medium containing insulin-transferrin-selenium (Corning) supplemented with EGF and bFGF (Gibco). Primary tumor spheres were formed after 7 days. Cells were then treated with test compounds for 24 hours. Primary tumor spheres larger than 50 µm in diameter were counted for each condition in triplicate. Single-cell suspensions were dissociated from primary tumor spheres by filtration through a 40 µm cell strainer and seeded using the same conditions. Secondary tumor spheres were formed after 7 days and treated and counted similarly as that for primary tumor spheres.

Subcellular fractionation, immunoprecipitation and immunoblotting. Subcellular fractionation
and immunoblot experiments were carried out as before 24 . Whole-cell extracts were obtained after lysis with RIPA buffer (Supplementary Materials and Methods) and subcellular fractions were obtained after homogenization in isolation buffer and centrifugation. Immunoprecipitation experiments were carried out as before with specific antibodies 24 . Protein extracts were resolved by SDS-PAGE followed by immunoblotting using antibodies specific for target proteins (Supplementary Materials and Methods). Densities of immunoblot bands were quantified using ImageJ (NIH). Genetic knockdown. FG and FGβ 3 cells were transfected with TBK1 small hairpin RNA (shRNA; Dharmacon; Supplementary Table S4) using lipofectamine 3000 reagent (Invitrogen). Gene knockdown was confirmed by immunoblotting.
Ral activation assay. Affinity pulldown assays for RalA/B were carried out following manufacturer's instructions (Cell Biolabs). Cells were cultured in suspension and treated with vehicle or PAWI-2 (50 nM) for 8 hours. Lysate obtained was incubated with RalBP1 PBD agarose bead slurry at 4 °C for 1 hour with gentle agitation. Activated forms of Ral (GTP bound) bound to beads were collected, washed and resolved by SDS-PAGE followed by immunoblotting using RalA/B antibodies. Double thymidine block and release. FGβ 3 cells were first incubated with 2 mM thymidine (Sigma) for 18 hours and released into fresh medium for 8 hours. Thymidine treatment was repeated, and a second release was conducted for 0-8 hours by releasing cells for treatment with vehicle, PAWI-2 or paclitaxel. For G1/S boundary, cells were collected at 0 hour. For the G2/M boundary studies, cells were collected at 8 hours for analysis of protein by immunoblots.
Statistical analysis. IC 50 and EC 50 values were calculated using a nonlinear regression analysis (GraphPad Prism) of the mean and standard deviation (SD) or standard error of the mean (SEM) of at least triplicate samples for each biological assay. Student t tests were used to calculate statistical significance and a P-value < 0.05 was considered significant.