STK38L kinase ablation promotes loss of cell viability in a subset of KRAS-dependent pancreatic cancer cell lines

Pancreatic ductal adenocarcinomas (PDACs) are highly aggressive malignancies, associated with poor clinical prognosis and limited therapeutic options. Oncogenic KRAS mutations are found in over 90% of PDACs, playing a central role in tumor progression. Global gene expression profiling of PDAC reveals 3-4 major molecular subtypes with distinct phenotypic traits and pharmacological vulnerabilities, including variations in oncogenic KRAS pathway dependencies. PDAC cell lines of the aberrantly differentiated endocrine exocrine (ADEX) subtype are robustly KRAS-dependent for survival. The KRAS gene is located on chromosome 12p11-12p12, a region amplified in 5-10% of primary PDACs. Within this amplicon, we identified co-amplification of KRAS with the STK38L gene in a subset of primary human PDACs and PDAC cell lines. Therefore, we determined whether PDAC cell lines are dependent on STK38L expression for proliferation and viability. STK38L encodes a serine/threonine kinase, which shares homology with Hippo pathway kinases LATS1/2. We show that STK38L expression is elevated in a subset of primary PDACs and PDAC cell lines displaying ADEX subtype characteristics, including overexpression of mutant KRAS. RNAi-mediated depletion of STK38L in a subset of ADEX subtype cell lines inhibits cellular proliferation and induces apoptosis. Concomitant with these effects, STK38L depletion causes increased expression of the LATS2 kinase and the cell cycle regulator p21. LATS2 depletion partially rescues the cytostatic and cytotoxic effects of STK38L depletion. Lastly, high STK38L mRNA expression is associated with decreased overall patient survival in PDACs. Collectively, our findings implicate STK38L as a candidate targetable vulnerability in a subset of molecularly-defined PDACs.


INTRODUCTION
Pancreatic ductal adenocarcinoma (PDAC) is typically associated with poor clinical prognosis and is the fourth leading cause of cancer-related deaths in the United States [1]. At present, conventional chemotherapies used to treat PDAC, such as 5-fluorouracil and gemcitabine, fail to improve patient survival beyond several weeks, underscoring the need to develop more effective therapeutic agents [2]. A critical barrier to the development of selective and efficacious therapeutics for the treatment of PDAC is the heterogeneous nature of the disease. Oncogenic KRAS mutations are found in over 90% of PDACs. Co-occurring alterations in CDKN2A/p16, TP53, and SMAD4/DPC4 tumor suppressor genes are also prevalent at high frequencies [3,4]. Despite the presence of these common genetic alterations, PDACs exhibit a high degree of inter-and intratumoral molecular and Research Paper histological heterogeneity. Whole genome transcriptional profiling reveals 3-4 PDAC molecular subtypes that are associated with distinct phenotypic traits and pharmacological vulnerabilities [4,5]. In order to develop more selective therapeutic modalities to treat PDAC, it is critical to understand the differences in activated signaling networks when comparing these contrasting subtypes.
The "aberrantly differentiated endocrine exocrinelike" (ADEX) PDAC subtype is characterized by high expression of epithelial differentiation genes and overexpression of oncogenic KRAS. These features are found in a subset of KRAS-mutant PDAC cell lines that exhibit robust dependence on KRAS for survival [6]. KRAS-dependent cells frequently harbor KRAS gene amplification and undergo apoptosis following KRAS depletion, indicating a state of KRAS oncogene "addiction" or dependency. Indeed, KRAS plays a critical role in PDAC initiation and maintenance, making it an attractive therapeutic target [7]. However, attempts to develop KRAS-directed therapies for clinical use have proven challenging [8][9][10]. This obstacle has prompted the search for alternative therapeutic targets to treat PDAC by identifying synthetic lethal KRAS interacting genes that confer a state of non-oncogene dependency [11]. We previously identified non-oncogene dependency for the nuclear Dbf2 and LATS1/2-related kinase STK38L (also known as NDR2) in the KRAS-mutant SW620 colon cancer cell line [12]. Subsequently, we noted that the STK38L and KRAS genes are located in close proximity on chromosome 12p11-12, a region frequently amplified in solid tumors, including those of the colon and pancreas [13][14][15][16]. Oncogene amplification is often related to oncogene "addiction/dependency," as is the case for the MYC oncogene [17]. This provided rationale to investigate whether STK38L gene copy number gain correlated with STK38L dependency on a larger scale in PDAC cell lines and to ultimately determine the potential of STK38L as a candidate therapeutic target. STK38L can play context-dependent, oncogenic or tumor suppressive roles. STK38L-mediated tumor suppression can occur via promotion of YAP phosphorylation and regulation of the Hippo signaling pathway [18]. Oncogenic functions of STK38L include regulation of MYC protein stability [19]. Additionally, STK38L regulates the stability of the CDK inhibitor p21 via direct phosphorylation [20]. Many of these functions of STK38L have been attributed to the closely related isoform STK38, also known as NDR1, although distinct STK38L-specific non-redundant functions are likely to exist [21]. Therefore, it is important to determine context-dependent STK38L-specific signaling mechanisms in PDAC cell lines.
Here, we assessed STK38L gene amplification as well as mRNA and protein expression in a panel of human PDAC cell lines and primary tumors. We determined correlative relationships comparing STK38L with KRAS gene amplification and expression. We investigated the role of STK38L in promoting proliferation and viability in PDAC cell lines. To identify potential mechanisms of STK38L-dependent survival signaling, we investigated downstream consequences of STK38L depletion in PDAC cell lines on LATS1/2 and p21 protein expression and function. Finally, we analyzed STK38L mRNA expression patterns in primary tumors from PDAC patients and correlated STK38L expression with overall survival.

STK38L gene copy number and protein expression levels are elevated in subsets of human PDAC
To assess the prevalence of STK38L gene copy number alterations in human PDAC, we analyzed the pancreatic adenocarcinoma (PAAD) dataset from The Cancer Genome Atlas (TCGA) ( Figure 1A). Consistent with previous studies, KRAS mutations are found in over 90% of cases in the TCGA PDAC cohort. STK38L gene amplification was observed in 2.7% percent of tumors. STK38L copy number gains were concordant with KRAS gene amplification. However, two tumors exhibited KRAS amplification alone. Using TCGA PDAC SNP and RNAseq-based gene expression data, we performed a linear regression analysis and observed a positive correlation (r = 0.9616; p < 0.0001) between KRAS and STK38L gene copy number values ( Figure 1B). We also observed a positive correlation between STK38L and KRAS mRNA expression (r = 0.5232; p < 0.0001) ( Figure 1C). In contrast, we failed to observe a significant correlation between expression levels of the STK38 isoform and KRAS (r = −0.0693; p = 0.4011) ( Figure 1D). Next, we performed correlation analyses of expression levels for other genes located within the chromosome 12p11-12 amplicon (RASSF8, SSPN, FGFR1OP2, MED21, BHLHE41 and ITPR2). With the exception of MED21, KRAS and STK38L mRNA expression levels were the most correlated of all the genes tested (Supplementary Figure 1). These findings indicate that the correlation of KRAS and STK38L expression levels is not simply due to proximal genomic context and could be due to transcriptional co-regulation.
We subsequently assessed STK38L protein levels in a panel of twelve established and well characterized human PDAC-derived cell lines (Supplementary Table 1) by Western blotting ( Figure 1E). STK38L protein expression varied considerably across the panel of cell lines. We also assessed KRAS protein levels in the cell line panel and observed a similar degree of variation, with some cell lines expressing high levels of KRAS protein, as previously reported [6]. Notably, we observed a significant correlation between STK38L and KRAS protein levels as determined by pixel densitometric analysis (r = 0.8122, p = 0.0014) ( Figure 1F). We conclude that STK38L and www.impactjournals.com/oncotarget KRAS gene copy number, mRNA, and protein levels are tightly correlated in human PDAC and tumor-derived cell lines.
To assess whether STK38L expression is elevated in primary PDACs, we analyzed a series of human PDAC tissue microarray (TMA) slides by immunohistochemistry ( Figure 1G). STK38L protein levels varied significantly across the primary tumor cohort. A subset of tumors was found to express high levels of STK38L protein.
In contrast, STK38L expression was weak or barely detectable in another subset of tumors. In summary, STK38L protein expression levels follow a variable spectrum both in cell lines and in primary PDACs.

STK38L depletion promotes cell death in a subset of human-derived PDAC cell lines
To determine the role of STK38L in promoting cell proliferation and survival in PDAC cell lines, we performed a series of functional experiments using STK38L-selective synthetic RNAi oligonucleotides or lentiviral-based shRNAs from various sources. First, we used pooled endoribonuclease-prepared small interfering RNAs (esiRNAs), which have been used effectively in genome-wide RNAi screens [22]. Transfection of STK38Ldirected esiRNAs was performed to determine effects on relative cell proliferation and viability (Figure 2A). STK38L depletion caused a spectrum of growth inhibitory effects in a PDAC cell line panel. Of note, a subset of cell lines was highly sensitive to the growth inhibitory effects of STK38L depletion, including the DAN-G and YAPC cell lines. In contrast, MIA PaCa-2 and KP-1N were insensitive to STK38L depletion. STK38L mRNA knockdown was validated by qPCR in representative cell lines from opposite ends of the sensitivity spectrum, DAN-G and KP-1N ( Figure 2B). We validated significant reduction in STK38L mRNA levels in both cell lines following esiRNA transfection. Importantly, the STK38Ldirected esiRNAs did not display off-target effects on related STK38, LATS1, and LATS2 mRNA expression (Supplementary Figure 2A). On the contrary, LATS1 and LATS2 levels were increased following STK38L depletion. To determine if esiRNA-mediated STK38L depletion caused apoptotic cell death, we employed a luminescence-based assay to measure levels of activated effector caspase-3 in cells following STK38L esiRNA transfection. In the YAPC PDAC cell line, we observed an esiRNA dose-dependent increase in activated caspase-3, indicating strong induction of apoptotic cell death following STK38L knockdown ( Figure 2C). In contrast, STK38L knockdown failed to induce apoptosis in a KRAS mutant lung cancer cell line H358. Like YAPC cells, H358 cells are KRAS dependent. However, unlike YAPC cells, H358 cells do not exhibit increased STK38L gene copy number. Therefore, esiRNAs directed against STK38L selectively induce caspase-3 activation and apoptotic cell death in a PDAC cell line that harbors combined KRAS and STK38L gene amplification.
Next, we determined the effect of STK38L depletion on levels of cleaved poly(ADP-ribose) polymerase (PARP), which is a marker of caspase-3-mediated apoptosis ( Figure 2D). For these assays, we used two STK38L-selective siRNA oligonucleotides. These siRNAs did not significantly affect protein levels of the STK38 isoform ( Figure 2D). The three cell lines that were most sensitive to the growth inhibitory effects of STK38L-directed esiRNAs, DAN-G, YAPC, and HuP-T4, displayed robust induction of cleaved PARP in response to STK38L depletion with both siRNAs. Although Panc 04.03 cells were strongly growth inhibited following esiSTK38L introduction, we did not observe any PARP cleavage in these cells using single siRNAs. The remaining cell lines were relatively unresponsive to the effects of STK38L depletion on cleaved PARP levels. We validated these contrasting effects on PARP cleavage using STK38L-targeted esiRNAs in DAN-G and KP-1N cells (Supplementary Figure 2B). To rule out the possibility that KP-1N cells are generally resistant to apoptosis, we treated DAN-G and KP-1N cells with the apoptosis-inducing agent anisomycin, and showed comparable effects in both cell lines (Supplementary Figure 2C). Finally, we used image-based cytometry of Hoechst dye-stained cellular nuclei in DAN-G cells to quantitate relative effects of siSTK38L transfection on cell cycle phases, including non-viable sub-G1 phase cells ( Figure 2E). Ablation of STK38L caused a robust increase in the proportion of sub-G1 cells (28% of the cell population) ( Figure 2F). Furthermore, STK38L ablation caused reduced numbers of cells in G1/S phases, and an accumulation of cells in G2/M phases perhaps indicative of a block in mitosis ( Figure 2F). Taken together, these findings suggest that selective knockdown of STK38L elicits a spectrum of antiproliferative effects in human PDAC cell lines, with a subset of cell lines undergoing PARP-cleavage associated apoptotic cell death.
Since oncogene dependency is often associated with copy number gain, we assessed STK38L gene copy number via genomic qPCR in PDAC cell lines (Supplementary Figure 3A). We observed varying STK38L gene copy number gains across the cell line panel. However, gene copy number failed to correlate (r = 0.4737; p = 0.1198) with the antiproliferative effects of STK38L depletion (Supplementary Figure 3B). We subsequently assessed STK38L mRNA transcript levels in the same cell line panel by qPCR (Supplementary Figure 3C). We observed a positive correlation (r = 0.8339; p = 0.0007) between STK38L transcript levels and STK38L dependency (Supplementary Figure 3D). Furthermore, the STK38Ldependent cell lines, DAN-G and YAPC, had significantly elevated STK38L protein levels compared to cell lines that were not STK38L-dependent ( Figure 1E). Collectively, these findings suggest that high STK38L mRNA and protein expression correlates well with STK38L dependency in PDAC cell lines. STK38L gene copy number gains can be found in a subset of cell lines, but do not correlate significantly with STK38L dependency.

The kinase activity of STK38L is necessary to promote cell survival
To validate the on-target specificity of STK38Ldirected siRNA oligonucleotides, we performed a series of rescue experiments using a constitutively-activated variant of STK38L (STK38L-PIF), which contains a PRK2 hydrophobic motif at the C-terminus [23,24]. We established DAN-G cells stably expressing GFP, HA-STK38L-PIF WT, or HA-STK38L-PIF K119R (kinase-dead variant) (Supplementary Figure 4). We did not observe any gross morphological or phenotypic abnormalities in cells expressing these STK38L variants. HA-STK38L-PIF K119R may be expected to have a dominant negative growth inhibitory effect. However, we did not observe any proliferative effects in DAN-G cells expressing the STK38L K119R mutant. This could be due to low expression levels of the mutant protein. We depleted endogenous STK38L in the stable DAN-G cell lines by transfection with an siRNA directed against the 3′ UTR of STK38L mRNA (siSTK38L #1), which does not target the exogenous STK38L mRNA. We quantified the total number of cells remaining following STK38L depletion by automated DAPI-stained nuclei counting ( Figure 3A). Cells expressing GFP or HA-STK38L-PIF K119R showed reduced cell numbers following STK38L depletion. In contrast, cells expressing HA-STK38L-PIF WT were more proliferative following STK38L depletion ( Figure 3A). Under similar experimental conditions, we analyzed changes in cleaved PARP by Western blotting  Figure 3B). DAN-G cells expressing GFP or HA-STK38L-PIF K119R exhibited nearly a four-fold increase in cleaved PARP when compared to the control treatment. However, cells expressing HA-STK38L-PIF WT exhibited no significant change in PARP cleavage. As an additional control, we used a siRNA directed against the coding sequence of STK38L (siSTK38L #2), which targets both endogenous and exogenous STK38L. This siRNA had similar antiproliferative effects in cell lines stably expressing GFP, HA-STK38L-PIF WT and HA-STK38L-PIF K119R ( Figure 3C). Taken together, these findings demonstrate that the effects of STK38L-directed siRNAs on cell viability are on-target and that the kinase activity of STK38L is necessary for cell survival in DAN-G cells.

STK38L depletion causes reduced clonogenic growth
To further validate the role of STK38L in promoting cell proliferation and survival, we performed a series of clonogenic assays, initially using 2D monolayer cell cultures. First, we stably depleted STK38L in DAN-G and PANC1 cells by lentiviral shRNA transduction and verified selective STK38L protein knockdown ( Figure 4A). Then, we performed a clonogenic assay using the same STK38Lselective shRNA. STK38L-depleted DAN-G cells formed significantly fewer and smaller colonies compared to shLuciferase (shLUC)-expressing cells ( Figure 4B). STK38L-depleted PANC1 cells also formed smaller colonies compared to control shLUC-expressing cells. However, the effects of STK38L depletion were relatively modest in PANC1 cells compared to DAN-G cells. Next, we analyzed clonogenic growth of DAN-G cells in 3D Matrigel TM assays, which provide a more physiologically relevant microenvironment that incorporates tumor cellextracellular matrix (ECM) interactions. When grown in Matrigel TM , STK38L-depleted DAN-G cells formed smaller colonies compared to shLuciferase (shLUC) expressing cells, as demonstrated by colony nuclear DNA and actin staining ( Figure 4C). STK38L-depleted PANC1 cells also formed smaller colonies compared to control shLUC expressing cells. However, the effect of STK38L depletion was relatively modest in PANC1 cells when compared to DAN-G cells. Automated computer-assisted quantitation of colony size validated these results for shLUC and shSTK38L expressing DAN-G and PANC1 cells ( Figure 4D). STK38L depletion caused a 68% reduction in the size of colonies formed by DAN-G cells but only a 28% reduction in the size of colonies formed by PANC1 cells. In summary, STK38L depletion causes substantially reduced clonogenic growth in DAN-G cells but only has a modest effect in PANC1 cells.

STK38L depletion causes induction of LATS2 expression concomitant with cell death
To elucidate mechanisms underlying STK38L dependency in PDAC cell lines, we characterized the effects of STK38L depletion on expression levels of key proteins in KRAS and general tumor-associated signaling pathways including p21, MYC, MAPK, PI3K, and autophagy (Supplementary Figure 5A). STK38L depletion caused increased ERK and ribosomal S6 protein phosphorylation in some cell lines, including DAN-G. We also noted that LC3-II levels were moderately increased in a few cell lines following STK38L depletion, indicating a possible block in autophagy flux. This is concordant with previous reports of NDR kinase regulation of autophagosome formation [25]. However, this effect of STK38L depletion on LC3-II levels was not generalizable across all PDAC cell lines tested and did not correlate with PARP cleavage-associated cell death. In over half of the cell lines tested, particularly STK38L-dependent cell lines, STK38L depletion led to increased protein levels of p21, consistent with previous reports [20,26,27]. The context-dependent role of p21 in promoting or blocking apoptosis is unclear [28,29]. Therefore, we tested a role for p21 in STK38L-regulated cell survival. To that end, we concomitantly depleted p21 and STK38L in DAN-G cells using combinations of siRNA oligonucleotides (Supplementary Figure 5B). Depletion of p21 alone had no effect on apoptotic cell death induction, as assessed by PARP cleavage. On the contrary, concomitant depletion of STK38L and p21 caused increased PARP cleavage compared to STK38L depletion alone. We also depleted p21 prior to STK38L depletion and, again, observed no significant effect on cell proliferation (Supplementary Figure 5C). Therefore, we conclude that increased p21 protein levels do not contribute to cytotoxicity following STK38L depletion.
To further validate the on-target effects of siRNAmediated STK38L depletion, we determined mRNA expression levels of related STK38, LATS1 and LATS2 kinase genes. STK38L depletion caused increased mRNA expression of LATS2 but had no effects on STK38 or LATS1 mRNA transcript levels ( Figure 5A). These effects of STK38L depletion on LATS1/2 and STK38/STK38L mRNA expression were consistent with changes in protein expression levels as assessed by Western blotting ( Figure  5B). Depletion of STK38L caused reduced expression of LATS1 in DAN-G cells but had no effect in KP-1N cells. In contrast, STK38L depletion caused increased protein levels of LATS2 in both cell lines. The effect of STK38L depletion on LATS2 mRNA and protein expression was more robust in DAN-G cells compared to KP-1N cells. Reciprocal control of LATS1/2 and STK38/STK38L expression levels has been documented in previous reports [30][31][32]. However, our studies represent the first report of STK38L-mediated control of LATS2 expression. Upon activation of the Hippo signaling pathway, LATS2 controls the phosphorylation, stability, and subcellular localization of the transcriptional co-activators YAP and TAZ [33,34]. Increased LATS2 protein expression in DAN-G cells, following STK38L depletion, was associated with decreased levels of YAP and TAZ ( Figure 5B). In contrast, STK38L depletion did not cause decreased YAP and TAZ protein expression in KP-1N cells.
Increased LATS2 protein expression can promote apoptosis in some contexts [35][36][37]. Thus, we hypothesized that induction of LATS2 expression contributes to cytotoxicity following STK38L depletion. To test this, we depleted STK38L alone or in combination with either LATS1 or LATS2 depletion in DAN-G cells. Depletion of either LATS1 or LATS2 alone had no significant effect on cell proliferation and viability ( Figure 5C). STK38L depletion alone caused reduced cell proliferation and viability in DAN-G cells, as seen previously. This effect did not change significantly with co-depletion of LATS1. In contrast, co-depletion of LATS2 with STK38L caused a 13% increase in cell proliferation ( Figure 5C). We confirmed on-target effects of siRNA cotransfections on LATS1/2 and STK38/STK38L protein expression levels by Western blotting ( Figure 5D). Taken together, we conclude that the cytotoxicity associated with STK38L depletion in DAN-G cells occurs, in part, via induction of LATS2 expression.
YAP and TAZ interact with the TEAD family of transcription factors to drive the expression of genes that promote cell survival and proliferation [38]. Thus, we hypothesized that reduced YAP/TAZ expression could confer antiproliferative effects in DAN-G cells, which are sensitive to STK38L depletion. To test this hypothesis, we depleted YAP/TAZ by lentiviral shRNA transduction in DAN-G cells. We then quantitated the total number of remaining cells by automated DAPI-stained nuclei counting ( Figure 6A). Reductions in total cell number were observed following depletion of YAP or TAZ relative to the luciferase control. However, this effect was more robust following TAZ depletion. To determine whether the reductions in total cell number were due to apoptotic cell death, we analyzed changes in cleaved PARP under the same experimental conditions ( Figure 6B). Cleaved PARP was observed in TAZ-depleted cells but not YAP-depleted cells. We conclude that loss of TAZ induces apoptosis in DAN-G cells whereas loss of YAP causes a reduction in cell proliferation.
To determine whether the loss of viability in DAN-G cells observed following STK38L depletion was due to a decrease in TAZ protein levels, we performed a rescue experiment using a mutant form of TAZ (4SA: S66A, S89A, S117A, and S311A) that cannot be phosphorylated by LATS2 [39]. We established DAN-G cells stably expressing an empty vector Control, 3xFLAG epitopetagged TAZ WT (3xFLAG-TAZ WT), or 3xFLAG-TAZ 4SA ( Figure 6C). Exogenous levels of TAZ WT were notably lower than TAZ 4SA, likely due to the increased protein stability of the 4SA mutant. We depleted STK38L in stable DAN-G cell lines and quantified the total number of cells remaining by automated DAPI-stained nuclei counting ( Figure 6D). Cells expressing the empty vector Control or TAZ WT showed significantly reduced cell numbers following STK38L depletion (73% and 67% respectively). Comparatively, cells expressing TAZ 4SA showed a significantly weaker reduction in cell number following STK38L depletion (39%). Collectively, these findings suggest the cytotoxic effect associated with STK38L depletion is due, in part, to LATS2-mediated degradation of TAZ ( Figure 6E).

High STK38L mRNA expression is associated poor prognosis in PDAC patients
To determine whether STK38L expression correlates with overall survival of PDAC patients, we performed Kaplan-Meier analyses using publically-available RNAseq-derived gene expression data from The Cancer Genome Atlas (TCGA-PAAD dataset) (Figure 7) [40]. We found that high STK38L mRNA expression associates with a significantly lower overall patient survival when compared to low STK38L expression (hazard ratio: 1.652, 95% confidence interval: 1.096-2.490, p = 0.0165). Of note, expression of LATS1/2 and STK38 kinase genes did not correlate with survival. As STK38L and KRAS are co-amplified in a subset of PDACs, we examined whether there is a relationship between KRAS expression and survival. We found that high KRAS expression was indeed associated with poor survival (hazard ratio: 1.766, 95% confidence interval: 1.174-2.656, p = 0.0063) (Supplementary Figure 6). To rule out the possibility that the lower overall survival associated with STK38L expression is a passenger effect, we analyzed the expression of other genes located in the chromosome 12p11-12p12 amplicon. Of the seven genes analyzed (RASSF8, BHLHE41, SSPN, ITPR2, FGFR1OP2,  TM7SF3 and MED21), none associated significantly with patient outcome (p > 0.05) (Supplementary Figure 6). Therefore, we conclude that elevated STK38L and KRAS mRNA expression levels are specifically associated with poor patient outcome in PDAC cases.

DISCUSSION
The inter-and intratumoral molecular heterogeneity of PDAC has confounded efforts to identify generally efficacious therapeutic agents. It is plausible that molecular subtype-selective targeted therapeutic agents could be effective in managing disease outcome by allowing for a more tailored and precise regimen. In this study, we identified STK38L gene amplification in a subset of human PDACs. Importantly, high STK38L gene expression associates with poor patient survival. Genetic STK38L depletion causes the selective killing of a subset of human PDAC cell lines, many of which have molecular hallmarks of the ADEX PDAC subtype [4]. Therefore, we implicate STK38L as a candidate, therapeutic target for a subset of ADEX-like PDACs. Analysis of a large PDAC cell line cohort demonstrated that STK38L gene copy number does not predict STK38L dependency. Our analysis of STK38L gene copy number was performed using a qPCR-based approach, which has inherent limitations. Future studies of STK38L amplification will utilize fluorescence in situ hybridization (FISH) to definitively quantitate STK38L gene copy number gain in cell lines, as well as primary tumors, and determine its utility as a biomarker. STK38L mRNA or protein expression levels may provide a more accurate biomarker for prediction of STK38L dependency in tumors. To further develop STK38L kinase inhibition as a therapeutic strategy, it will be necessary to identify additional biomarkers of STK38L dependency, such as cooccurring genetic alterations or gene expression signatures. Co-occurring, non-driver mutations may cooperate with oncogenic KRAS and STK38L to promote PDAC cell survival, lending complexity to the clinical potential of KRAS and STK38L-directed therapeutics. However, the identification of non-driver mutations and non-oncogene dependencies could allow for the development of potential combination-based therapies to effectively treat PDAC.
Previous studies demonstrate that STK38 and STK38L have redundant tumor suppressor function [18,30]. Consistent with this finding, STK38 and STK38L single gene deletions show no discernable phenotypes in mice, suggesting that selective STK38L inhibition will be well tolerated [32]. This provides support for STK38L as an attractive therapeutic target. Our findings highlight a context-dependent pro-tumorigenic role for STK38L. In concordance with our findings, a recent study demonstrates that STK38 promotes survival in some RAS-transformed cell lines by promoting clearance of damaged mitochondria [27]. However, STK38L does not compensate for loss of STK38 function in this context, indicating non-redundant function for these closely related kinases. Indeed, we found that STK38 expression levels do not affect cytotoxicity that is induced following STK38L depletion. The pro-survival function of STK38L is closely related to KRAS oncogene dependency in PDAC cell lines. We observed highly significant correlations between STK38L and KRAS expression levels in PDAC cell lines as well as primary tumors. However, the mechanistic relationship between STK38L and KRAS remains unclear. STK38L depletion failed to induce consistent effects on downstream activation of KRAS effector pathways. Furthermore, expression of mutant KRAS 4A and 4B isoforms does not affect STK38L expression levels (Supplementary Figure 7). Therefore, we conclude that STK38L and KRAS function in parallel pathways to promote tumor cell survival. Further characterization of the mechanistic relationship between STK38L and KRAS will be critical to fully understand context-dependent tumor cell survival signaling networks in KRAS mutant PDACs.
Mechanistically, we demonstrate for the first time that STK38L depletion causes cell death, in part, via induction of LATS2 protein expression. LATS2 is a well-established tumor suppressor that, under normal physiological conditions, can restrict cell proliferation and promote apoptotic cell death. The mechanisms underlying LATS2-mediated cytotoxicity in PDAC cell lines have yet to be fully determined, but could involve perturbations in the phosphorylation and function of NDR/ LATS substrates p21 and YAP [18,20,26,37,38]. YAP can be constitutively-activated in PDAC, and genetic ablation of YAP limits KRAS-driven PDAC progression in mouse models [41]. Consistent with this finding, we observed robust inhibition of cell proliferation following depletion of the transcriptional co-activator YAP or the closely related paralog TAZ. Further elucidation of the STK38L-LATS2-YAP/TAZ signaling axis will be interesting to explore in future studies. In summary, our findings highlight an important context-dependent role for STK38L kinase activity in promoting the survival of a distinct subset of KRAS-dependent PDAC cell lines via suppression of LATS2 apoptotic function. STK38L-dependent survival signaling represents an important vulnerability in subsets of PDAC cases that can potentially be exploited for therapeutic benefit. Thus, it will be important to develop ATP-competitive inhibitors to validate the kinase-dependent functions of STK38L in normal cells and in the pathophysiological context of STK38L-dependent tumor cell survival.

Cell culture
Human-derived PDAC cell lines were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 5% fetal bovine serum (FBS), 1mM sodium pyruvate, 100 Units/ml penicillin, and 100 µg/ ml streptomycin. 293T cells were cultured in DMEM (Dulbecco's Modified Eagle Medium) supplemented with 10% FBS. All cell lines were maintained in a humidified incubator at 37°C with 5% CO 2 .

Analysis of human PDAC patient data
The Pancreatic Adenocarcinoma (PAAD) dataset provided by The Cancer Genome Atlas was analyzed using cBioPortal for Cancer Genomics maintained by the Memorial Sloan Kettering Cancer Center [42,43]. This dataset was also analyzed using the UCSC Xena Functional Genomics Browser (http://genome.ucsc.edu/) to derive Kaplan-Meier plots of associations between selected gene expression levels and patient overall survival [44].

Western blotting
Cells were washed in 1X PBS followed by lysis in 1X Laemlli Buffer supplemented with Halt Protease and Phosphatase Inhibitor Cocktail (ThermoFisher Inc.). Protein lysates were sonicated and normalized for total protein with a Bicinchoninic Acid (BCA) Assay Kit (Pierce). Proteins were resolved by SDS-PAGE and then transferred to a polyvinylidene fluoride (PVDF) membrane for Western blotting. Detection of proteins was carried out by enhanced chemiluminescence (ECL) using SuperSignal West-Pico or West-Dura (Thermo Fisher Scientific) reagents. Imaging was carried out with a Syngene G-Box XT4 system and the GeneSys imaging software. Detailed information regarding primary and secondary antibodies used for Western blotting is provided in Supplementary  Table 2. The primary antibody for STK38L is previously described [45]. All antibodies were diluted in 2% w/v BSA in TBS-T.

Cellular viability and quantitation assays
Cell viability assays were carried out in 96-well format with Alamar Blue reagent. Cells were subjected to esiRNA-mediated gene depletion for 72 h and subsequently incubated with 50 μg/ml alamarBlue for 1 h at 37°C/5% CO 2 . Alamar Blue fluorescence was measured with a FLUOstar OPTIMA Microplate Reader (BMG Labtech) at λ Ex /λ Em 544/590 nm. The relative fraction of viable cells was determined by normalizing fluorescence intensity to parallel cultures transfected with control esiRNAs. Cell quantitation assays were carried out in 96-well format. Cells were subjected to siRNAmediated gene depletion for 48 h or treatment with drug diluted in 0.1% DMSO for 24 h. Cells were fixed with 4% paraformaldehyde in 1X PBS and subsequently treated with 5 µg/ml Hoechst 33258 nuclear stain in 1X PBS for 20 min. Plates were scanned with a Cytation 3 Imaging Multi-Mode Reader (BioTek) at λ Ex /λ Em 377/477 nm. Gen5 Microplate Reader and Imager Software was used to identify and count individual nuclei and determine total cell number. For image cytometry analyses of cellular subpopulations, DNA fluorescence intensity cut-off values were set in the Gen5 software to count cells in specific phases of the cell cycle, including sub-G1 (non-viable or apoptotic) cells.

Recombinant lentivirus generation
Recombinant, replication-defective lentiviral particles were generated using a three-plasmid system previously described [46,47]. Briefly, 293T cells were transfected with the desired lentiviral expression vector, the lentiviral packaging vector psPAX2 (gift from Didier Trono; Addgene plasmid #12260), and the lentiviral envelope vector pMD2.D-G VSV-G (gift from Didier Trono; Addgene plasmid #12259). Lentivirus production was induced the next day by treatment with DMEM supplemented with 30% FBS. The viral supernatant was collected 24 post-induction and syringe filtered to remove any potential cell contaminants. Target cells were subsequently spin-infected with lentiviral particles in the presence of 8µg/ml polybrene at 1200xg for 1 h. Selection of infected cells was carried out by treatment with 1 µg/ ml puromycin.

Clonogenic assays
For two-dimensional clonogenic assays, PDAC cells transduced with shRNA expression plasmids were seeded on a 6-well plate at a density of 1000 cells/well and maintained for 10 days. Colonies were fixed with 4% paraformaldehyde in 1X PBS and subsequently stained overnight with Geimsa dye (Ricca Chemical Company) diluted 1:50 in 1X PBS. Images of each well were captured with a Bio-Rad ChemiDoc XRS+ using the Image Lab TM Software. Three-dimensional clonogenic assays were performed using a method previously described [48]. PDAC cell lines transduced with shRNA expression plasmids were seeded on an 8-well chamber slide (Lab-Tek) coated with 200 µl Matrigel TM (Corning) diluted 1:1 in serum-free DMEM and maintained for 6 days. Representative images of each well were captured by bright-field microscopy and subsequently analyzed using CellProfiler software to determine average colony size [49]. For immunofluorescence imaging, colonies were fixed with 4% paraformaldehyde in 1X PBS and permeabilized with 0.1% Triton X-100. Colonies were sequentially incubated with 132 nM Alexa594-conjugated phalloidin in 1× PBS to stain actin filaments (Thermo Fisher Scientific) and then 2 µg/ml Hoechst 33256 dye in 1X PBS to stain nuclear DNA. Image acquisition was performed with a Cytation 3 Imaging Multi-Mode Reader (BioTek) at 10× magnification.

Quantitative PCR-based gene expression analyses
RNA was isolated from cells with an RNeasy Mini Kit (Qiagen Inc.) and subjected to reverse transcription using a High Capacity Reverse Transcription Kit (Applied Biosystems Inc.). Diluted cDNA was used in TaqMan or SYBR Green-based qPCR assays. Relative quantitation determinations were obtained with a StepOne Real-Time PCR System (Applied Biosystems). Primer sequences used for SYBR Green-based assays are listed in Supplementary  Table 3. Relative mRNA transcript levels were normalized to the reference gene GAPDH to allow for comparisons between cell lines.

Gene copy number assay
Genomic DNA was isolated from cells with a QIAamp DNA Mini Kit (Qiagen Inc). STK38L gene copy number was assessed by genomic qPCR using a TaqMan Copy Number Assay Kit (Applied Biosystems). Recommended/validated probes for STK38L and the reference gene RPPH1/RNaseP were used in this assay. Relative quantitation determinations were obtained with a StepOne Real-Time PCR System (Applied Biosystems). STK38L gene copy number was normalized to the reference gene RPPH1/RNaseP to allow for comparisons between cell lines.

Derivation of STK38L dependency
STK38L dependency was determined using a method similar to the one previously described for KRAS dependency [6]. The relative fraction of surviving cells following esiRNA-mediated depletion of STK38L was determined by Alamar Blue assay. The inverse of this fraction was then calculated to provide a quantitative value for STK38L dependency.

Immunohistochemistry (IHC)
STK38L expression was assessed in human PDAC samples by immunohistochemistry. Paraffin-embedded tumor microarrays (TMAs) were developed using samples that were obtained from consenting patients under IRB approved protocols (Massachusetts General Hospital/Dana Farber/Harvard Cancer Center, Department of Pathology). TMAs were stained using a Leica Bond III auto-stainer (Leica Biosystems) according to the manufacturer's protocol. Antigen retrieval was carried out with Sodium Citrate (pH 6.0), and a rabbit polyclonal antibody directed against STK38L was applied to the TMAs at a dilution of 1:250. Tissue sections were counter-stained with hematoxylin. Slides were imaged using an Aperio ScanScope CS system (Leica Biosystems).

Author contributions
AS supervised the study. TG, AKM, NG, and AS conceived and designed experiments. TG, AKM, AG, and