Crosstalk between PIM and LKB1 kinases regulates AMPK phosphorylation and tumorigenic growth

Kwan Long Mung University of Turku Faculty of Mathematics and Natural Sciences: Turun yliopiston luonnontieteiden ja tekniikan tiedekunta William Eccleshall University of Turku Faculty of Mathematics and Natural Sciences: Turun yliopiston luonnontieteiden ja tekniikan tiedekunta Niina M Santio University of Turku Faculty of Mathematics and Natural Sciences: Turun yliopiston luonnontieteiden ja tekniikan tiedekunta Adolfo Rivero-Müller Medical University of Lublin: Uniwersytet Medyczny w Lublinie Päivi J. Koskinen (  paivi.koskinen@utu. ) Turun Yliopisto https://orcid.org/0000-0002-6864-4457

In this study, we explored the interaction between PIM and LKB1 kinases (enzymes that catalyse the transfer of phosphate groups to their target proteins). While PIM can support the development of certain cancers, such as breast and prostate cancer, LKB1 is a known tumor suppressor. Through our research we have shown that PIM reduces the activity of LKB1, providing a novel mechanism by which PIM promotes tumor growth. By using CRISPR/Cas9-based genomic editing, we generated breast and prostate cancer cells lacking PIM, LKB1 or both (PIM/LKB1 knock-out cells), and monitored the behaviour of these cells both under standard cell culture conditions and in a chick embryo-based tumor xenograft model.
Prostate cancer cells lacking LKB1 formed signi cantly larger tumors than LKB1-expressing cells, but this effect disappeared when PIM was knocked out together with LKB1. By contrast, overexpression of PIM in LKB1-de cient prostate cancer cells further exaggerated the size of the tumors. Taken together, these ndings provide insight into the relationship between two enzymes that are important in the context of cancer, and suggest that PIM kinases may be a rational therapeutic target for the treatment of LKB1de cient tumors.

Background
The three serine/threonine-speci c PIM family members (PIM1, PIM2, and PIM3) are highly homologous and in part functionally redundant [1][2][3]. As PIM kinases are constitutively active in cells [4], their catalytic activities correlate well with their protein expression levels. These oncogenic kinases are often overexpressed in solid tumors or haematological malignancies, in which they promote cell proliferation, survival, motility and metabolism via phosphorylation-dependent activation or inactivation of a wide variety of substrates, such as the NFATC1 and NOTCH1 transcriptional regulators, the CDKN1A and CDKN1B cell cycle inhibitors, the BAD pro-apoptotic protein, the CXCR4 chemokine receptor, the CAPZ actin capping proteins, and the AKT1S1 and EIF4EBP1 translational inhibitors [1][2][3]5]. Accordingly, PIM kinases have emerged as attractive targets for cancer therapy, especially as they possess a structurally unique ATP-binding pocket [4], and as PIM triple knock-out (TKO) mice are viable and fertile with only a mild reduction in body size [6]. This phenotype may at least partially be due to reduced cytokine responses [7] and a diminished glycolytic phenotype [8].
The serine/threonine-speci c liver kinase B1 (LKB1), encoded by the STK11 gene, is a tumor suppressor, which is mutated in patients with the hereditary Peutz-Jeghers syndrome [9,10]. Somatic inactivating mutations have also been found in sporadic tumors: 5-17% of non-small cell lung carcinomas [11][12][13], 5% of pancreatic cancers and melanomas [14][15][16] and around 20% of cervical cancers [17,18]. Furthermore, STK11 has been identi ed as the third most frequently mutated gene in human lung adenocarcinoma, following TP53 and KRAS [19]. By contrast, LKB1 mutations have rarely been reported from breast, colorectal or gastric cancer [9]. The tumor suppressor function of LKB1 is largely attributed to its ability to phosphorylate the AMP-activated protein kinase (AMPK) [20][21][22] and 12 other closely related kinases [23]. AMPK in turn is a heterotrimeric protein comprising of a catalytic α subunit and regulatory β and γ subunits [24]. In response to changes in the AMP/ATP ratio resulting e.g. from energy deprivation, LKB1 phosphorylates the α subunit of AMPK at a conserved threonine site (commonly stated as Thr172 because of its pivotal nding in rats [25], while the corresponding site in the human protein is Thr183).
Phosphorylation of AMPK increases its catalytic activity more than 100-fold in vitro [26], and in cells this allows it to inhibit anabolic biosynthetic pathways and to promote catabolic processes to restore the energy balance in favour of ATP production [24,27]. Remarkably, failure to activate AMPK in response to energy stress has been proposed as an explanation for the massive cell death that occurs in LKB1de cient tumors after treatment with metabolic inhibitors, such as metformin or phenformin [28,29]. Interestingly, inhibition of PIM expression or activity has been shown to increase AMPK phosphorylation, possibly via LKB1 [30], but the exact mechanism behind this phenomenon has remained unclear.
As cancer cell growth and metabolism are regulated by the balance between oncogenic (e.g. PIM) and tumor-suppressive (e.g. LKB1) kinases, both overexpression of PIM kinases and loss of LKB1 expression are expected to promote tumorigenesis. In the present study with prostate and breast cancer cell lines expressing PIM and LKB1 kinases, we demonstrate that PIM kinases act as upstream kinases of LKB1 and that Ser334 in LKB1 is their major phosphorylation target site. Both pharmacological and CRISPR/Cas9-based approaches reveal that inhibition of expression or activity of all three PIM family members upregulates AMPK activity in an LKB1-dependent manner. Finally, double knock-out of both LKB1 and PIM kinases led to a striking reduction in cell proliferation and tumor growth, raising possibilities for PIM-targeted pharmaceutical interventions in suppressing the growth of LKB1-de cient tumors.

Methods
Cell Culture, reagents and DNA constructs MCF7 breast cancer and HeLa cervical cancer cells were cultured in Dulbecco's modi ed Eagle's medium (DMEM). PC3 prostate cancer cells and FDCP1 myeloid cell lines were cultured in RPMI-1640 medium. Construction and maintenance of FDCP1 cell lines have been described previously [31]. Both media were supplemented with L-glutamine, 10% fetal bovine serum and antibiotics. MEM Non-Essential Amino Acids (Gibco, #11140050) and sodium pyruvate (Gibco, #11360070) were further added to RPMI-1640 medium to facilitate cell growth. FuGENE® HD Transfection Reagent (Promega) was used for plasmid transfection according to the manufacturer's protocol. PIM-selective small molecule inhibitors DHPCC9 [32,33] and AZD1208(AstraZeneca, Cambridge, UK) were diluted in DMSO. Expression vectors pcDNA™3.1/V5-His-C, pGEX-6P-1 and pTagRFP-N for wild-type (WT) human PIM kinases have been described previously [34].  Figure S1 and S2), as are also the sgRNA sequences and sequencing primers (Additional File 1: Table S2 and S3). After knocking out individual PIM family members, triple PIM kinase knock-out cell lines were generated by sequentially knocking out additional genes.
Overnight bacterial cultures were grown at 30 °C until OD 600 of 0.6. Isopropyl-β-d-galactosidase (250 µM) was added to induce protein expression, and the cells were cultured for another 4 h (GST-PIMs) or 24 h (GST-LKB1, His-LKB1). The follow-up puri cation steps of GST-tagged and His-tagged protein have been described previously [5].

Western Blot
Cells were lysed for 10 min in ice-cold 50 mM Tris-HCl, pH 8.0 buffer containing 150 mM NaCl, 2 mM EDTA, 1% NP-40, 5 mM NaF, 1 mM Na 3  Nuclear/Cytoplasmic fractionation Nearly con uent cells (~ 80% con uence) were collected from 10 cm plates by scraping them into 1 ml aliquots of PBS. After 10 s centrifugation at 21,000 x g, supernatants were discarded and the pellets were lysed for 15 min in 500 µl of lysis buffer: 10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 3 mM MgCl 2 , 0.5% Nonidet P-40, 5 mM NaF ,1 mM PMSF and mini EDTA-free protease inhibitor tablet. After centrifugation at 500 x g for 5 min at + 4 °C, the supernatants contained the cytoplasmic compartments, while the nuclei were in the pellets. The pellets were washed three times with 500 µl lysis buffer and centrifuged each time at 500 x g for 5 min at + 4 °C, after which they were suspended in 200 µl of lysis buffer and sonicated for 30 s. After an additional centrifugation at 500 x g for 1 min, the supernatants were collected which contained nuclear fractions. The cytoplasm-containing solutions were centrifuged at 12000 x g for 15 min at + 4 °C, after which the supernatant was collected. Laminin A/C and beta-tubulin were used as nuclear and cytosolic markers, respectively, to evaluate fractionation e ciency.
In vitro kinase assays The procedure for performing radioactive in vitro kinase assays has been described previously [36]. Brie y, 0.5-2.0 µg of PIM kinase and its substrate were used in each reaction. Samples were separated by SDS-PAGE and stained by Page Blue™ protein staining solution (#24620, Thermo Fisher Scienti c). Additional in vitro kinase assays were performed similarly, but without radioactivity. Band intensities were quantitated by the Image Lab software Version 5.2.1 (Bio-Rad Laboratories, Inc., Hercules, CA, USA).

Fluorescence-lifetime imaging microscopy (FLIM)
FLIM was carried out as previously described [34] to probe for intracellular protein-protein interactions. Brie y, cells were plated on coverslips and transiently transfected with RFP-or GFP-tagged expression vectors.

IncuCyte analysis
Cells were seeded in a 96-well plate at a density of 3500 cells per well. After an overnight incubation, they were treated with DMSO or DHPCC-9 and imaged every 2 h using the IncuCyte S3 Live-Cell Analysis System (Essen BioScience, Ltd., Newark, United Kingdom). Phase images were acquired and the percentage of con uence of the cell layers was analysed using the IncuCyte® Software (v2019B) Basic Analyzer module.
In silico analysis The PhosphoSitePlus® database (phosphosite.org, Cell Signaling Technology, Inc., Danvers, MA, USA) was used to search for potential phosphorylation sites. IST Online™ (ist.medisapiens.com) was used to generate gene expression data derived from patient samples.

Statistical analysis and gure preparation
Bar graphs or scatter plots were produced by GraphPad Prism 6.0 and results were analysed by Student's t-test. Signi cant differences (p < 0.05 and p < 0.01) were marked by * and **, respectively. Error bars represent standard deviations. Inkscape was used for gure preparation.

PIM inhibition increases LKB1-dependent phosphorylation of AMPK
To investigate in more detail whether PIM kinases negatively regulate AMPK phosphorylation and activation, we used a pharmacological approach to inhibit PIM activity in PC3 prostate cancer, HeLa cervical cancer and MCF7 breast cancer cell lines. Cells were treated with either DMSO or 10 µM concentrations of two structurally distinct small molecule pan-PIM inhibitors, DHPCC9 or AZD1208, which inhibit the catalytic activity of all three PIM family members [33,38]. The relative phosphorylation level of AMPK was determined 24 h later by Western blotting with antibodies against AMPK or its phosphorylated Thr172 residue. As shown in (Fig. 1A), AMPK was expressed at a similar level in all three cell lines, but it was more prominently phosphorylated in PC3 cells than in the others. When PIM activity was inhibited by either DHPCC9 or AZD1208, AMPK phosphorylation was signi cantly enhanced in both PC3 and MCF7 cells, but not in HeLa cells.
These results could be explained by the observed expression of the AMPK upstream kinase LKB1 in PC3 and MCF7 cells, but not in HeLa cells. To con rm this conclusion, we used the CRISPR/Cas9-based genomic editing technique to knock out LKB1 from both PC3 and MCF7 cells (Additional File 2, Figure  S1A). As demonstrated by DNA gel electrophoresis (Additional File 2, Figure S2) and Western blotting (Fig. 1B), there was no LKB1 expression in the knock-out cells. When AMPK phosphorylation levels were analysed, no signi cant differences were observed between wild-type and LKB1-de cient cells that had been treated with DMSO (Fig. 1B). By contrast, treatment with the PIM inhibitor DHPCC9 induced a profound increase in AMPK phosphorylation in wild-type, but not knock-out cells. Furthermore, transient expression of FLAG-tagged LKB1 in LKB1-de cient PC3 or MCF7 cells restored the response to DHPCC9 (Fig. 1C). Altogether, these data indicate that LKB1 is necessary for the PIM inhibition-induced increase in AMPK phosphorylation.

AMPK phosphorylation levels are inversely correlated with PIM expression levels
In order to analyse the respective contribution of the different PIM family members in regulating AMPK phosphorylation and activity, we used the CRISPR/Cas9 technique to generate both individual and combined PIM knock-out cells (Additional File 2, Figure S1B-D). As con rmed by DNA gel electrophoresis (Additional File 2, Figure S2) and Western blotting ( Fig. 2A), single (KO) as well as triple (TKO) knock-out lines were successfully produced from both PC3 and MCF7 cells. Lack of any single PIM protein did not result in notable changes in AMPK phosphorylation in either cell line (Fig. 2B). By contrast, signi cantly elevated levels of phosphorylation were observed in the two independent PC3 and MCF7 TKO cell clones (Fig. 2C), while transient expression of His-tagged PIM1 in these cells restored AMPK phosphorylation back to its basal level (Fig. 2D). Furthermore, FDCP1 myeloid cells stably overexpressing PIM1 (FD/PIM1) exhibited signi cantly lower levels of AMPK phosphorylation than the corresponding control cells (FD/Neo) (Fig. 2E). Taken together, our data indicate that either pharmacological inactivation or knockout of all three PIM family members results in increased AMPK phosphorylation.

LKB1 is a novel substrate for PIM kinases
As LKB1 was indispensable for AMPK phosphorylation in both PC3 and MCF7 cells, this raised the question of whether PIM kinases downregulate LKB1 activity by directly phosphorylating it. To address this question, we subjected GST-tagged PIM family members, LKB1 or their combinations to radioactive in vitro kinase assays. Visualisation of 32 P-labeled phosphoproteins by autoradiography revealed that all three PIM kinases phosphorylate LKB1 in vitro, and that LKB1 does not undergo autophosphorylation (Fig. 3A). Thus, our data indicate that LKB1 indeed is a novel substrate targeted by all three PIM kinases.
These data were further con rmed by a non-radioactive in vitro kinase assay (Fig. 3B), where phosphoproteins were visualised by Western blotting with the phospho-Akt substrate (PAS) antibody. This antibody recognises not only the AKT-targeted sequence RXXS/T, but also the PIM-targeted consensus sequence RXRHXS/T [39] (Fig. 3C).
For LKB1, multiple phosphorylation sites have been identi ed [40,41]. However, only a few of them, including Ser334 and Ser428, resemble PIM-target sites that can be recognised by the PAS antibody. To determine whether one or both of them are PIM target sites, we mutated them separately to alanine residues and subjected the mutant proteins to in vitro kinase assays. When His-tagged wild-type (WT) or mutant proteins were incubated in the presence of GST-PIM1, there was a 40% decrease in the intensity of the 32 P-labeled signal for the S334A mutant as compared to the WT protein, while no signi cant changes were observed for the S428A mutant (Fig. 3D), indicating that Ser334 is a prominent PIM target site. This was con rmed by non-radioactive in vitro kinase assays followed by Western blotting with the PAS antibody (Fig. 3E). However, as the S334A mutation did not completely remove the residual signals in either assay, it remains possible that PIM kinases also target other sites in LKB1.
Having established PIM proteins as upstream kinases of LKB1, we examined their intracellular interactions. In co-immunoprecipitation assays, His-tagged PIM1 could be captured by FLAG-tagged LKB1 from both cell lines (Fig. 3F). In FLIM ( uorescence-lifetime imaging microscopy) analysis, signi cantly reduced GFP lifetimes were observed when GFP-tagged LKB1 and RFP-tagged PIM1 were coexpressed in either MCF7 or PC3 cells (Fig. 3G), suggesting that they physically interacted with each other.
PIM kinases target Ser334 in LKB1 to regulate AMPK phosphorylation To verify that PIM kinases phosphorylate LKB1 in cells, we transiently expressed FLAG-tagged LKB1 in both PC3 and MCF7 cells. At 24 h after transfection, cells were treated with DMSO or 10 µM DHPCC9 for another 24 h, after which cells were lysed, FLAG-LKB1 proteins were pulled down with the FLAG antibody and their phosphorylation levels were analysed by Western blotting with the PAS antibody. As shown in Fig. 4A, the relative phosphorylation levels of LKB1 were signi cantly reduced by PIM inhibition in both types of cells. In addition to the pharmacological approach, we analysed LKB1 phosphorylation in WT and TKO MCF7 cells transiently expressing either FLAG-tagged LKB1 or the corresponding S334A mutant.
In line with our data on PIM inhibition, LKB1 phosphorylation was dramatically decreased in TKO cells lacking all three PIM kinases (Fig. 4B). Notably, there was no signi cant difference between the phosphorylation level of the S334A mutant in WT and TKO cells, suggesting that Ser334 is a prominent PIM target site in LKB1.
To explore the impact of Ser334 phosphorylation of LKB1 on AMPK phosphorylation, FLAG-tagged LKB1 or the corresponding S334A mutant were transiently expressed in LKB1 KO derivatives of PC3 and MCF7 cells, and the cells were treated with either DMSO or 10 µM DHPCC9. As expected, DHPCC9 treatment did not trigger any considerable increase in AMPK phosphorylation in either type of FLAG-transfected cells lacking LKB1 (Fig. 4C). By contrast, reintroduction of WT LKB1, but not the phosphorylation-de cient S334A mutant restored the response of cells to DHPCC-9, resulting in a signi cant increase in AMPK phosphorylation. These data suggest that phosphorylation of LKB1 at Ser334 is involved in the regulation of AMPK phosphorylation by PIM kinases and their inhibitors.
AKT has been reported to phosphorylate LKB1 at Ser334, resulting in its nuclear sequestration by the 14-3-3 protein [42]. To determine whether PIM-dependent phosphorylation of this site has similar consequences in PC3 or MCF7 cells, we fractionated LKB1KO cells transiently expressing FLAG-tagged WT LKB1 or the phosphode cient S334A mutant. According to our analyses, both WT and mutant proteins were mainly localised in the nuclear fractions of both PC3 and MCF7 derivatives (Additional File 2: Figure S3A). By contrast, the endogenous LKB1 in parental cells was mostly localised in the cytosolic fractions, and this was not in uenced by pharmacological PIM inhibition (Additional File 2: Figure S3B) or by knocking out of all three PIM kinase members (Additional File 2: Figure S3C). However, the level of AMPK phosphorylation in the cytoplasmic fraction was increased in both cases. To con rm that there is no compensatory activation of AKT in the PIM TKO cells, we analysed AKT Ser473 phosphorylation levels from them, but did not observe any major changes as compared to WT cells (Additional File 2: Figure  S3D).

Combined knock-out of LKB1 and PIM kinases impairs cell proliferation and tumor growth
We next performed an in silico analysis of mRNA expression levels in patient-derived samples and observed that in prostate carcinomas and certain breast carcinomas, PIM expression was elevated and LKB1/STK11 expression was reduced (Additional File 2: Figure S4). However, in the breast medullary carcinoma dataset, both PIM and LKB1 expression levels were highly upregulated. As LKB1 expression and LKB1-dependent AMPK activation are often associated with cell growth suppression [43], this prompted us to evaluate the proliferation rates for WT PC3 or MCF7 cells or their LKB1-de cient derivatives in response to treatment with DMSO or 10 µM DHPCC9. Proliferation was followed for 5 days by measuring cell con uence with the IncuCyte live cell imaging system, where the two independent LKB1 KO clones behaved similarly to the WT cells (Fig. 5A). All the DMSO-treated cells proliferated well with sigmoidal growth curves, while the DHPCC9 treatment retarded cell growth.
The proliferation rates of TKO cells lacking all PIM kinases were reduced (Fig. 5B), which was in line with what we observed following the DHPCC9 treatment of WT cells. Surprisingly, cells lacking both LKB1 and PIM kinases (TKOLKB1 KO) grew even slower than TKO clones. As shown in (Fig. 5C), increased AMPK phosphorylation levels in TKO clones correlated well with their slower proliferation rates as compared to their WT counterparts. On the other hand, due to the absence of LKB1, the phosphorylation levels of AMPK in TKOLKB1 KO clones were lower than in TKO clones, yet the proliferation rates of TKOLKB1 KO clones were slower. These data suggest that under the conditions of PIM inhibition, changes in AMPK phosphorylation levels are not directly connected to the proliferation properties of LKB1 KO cells.
We next employed the chick embryo chorioallantoic membrane (CAM) xenograft model [37] to investigate the in vivo behaviour of the knock-out cells lacking both PIM and LKB1. WT MCF7 and PC3 cells or their KO and/or TKO derivatives were implanted onto the CAM of eggs on day 7 of incubation to allow the development of tumors. On day 14, tumors were excised from the CAM and weighed. Interestingly, there was a signi cant increase in the mass of tumors derived from LKB1 KO clones in PC3 but not MCF7 cells (Fig. 5D), but there was no signi cant difference between WT and TKO clones in either cell type. However, the mass of TKOLKB1KO MCF7 tumor cells was signi cantly lower than that of WT or LKB1KO samples. Intriguingly, while loss of LKB1 alone in PC3 cells increased tumor load, this effect was abolished, when combined with the loss of PIM kinases. Conversely, transient over-expression of PIM1 triggered further increases in tumor mass in two independent PC3 LKB1KO clone xenografts but not in WT samples (Fig. 5E). These data indicate that LKB1 and PIM kinases cooperate in the regulation of tumorigenic growth.

Discussion
To our knowledge, this is the rst report combining both pharmacological and CRISPR/Cas9-based genomic editing approaches to show that inhibition of the expression or activity of all three PIM kinases activates AMPK in cancer cells via LKB1-dependent phosphorylation at Thr172. Notably, knocking out of any particular PIM family member is not su cient to trigger AMPK activation, re ecting the previously observed functional redundancy of PIM kinases and the fact that all three PIM kinases are capable of phosphorylating LKB1 and thereby inhibiting its ability to phosphorylate AMPK. Besides demonstrating that PIM kinases are upstream regulators of LKB1, we have also identi ed Ser334 as the major, although possibly not the only PIM target site in LKB1.
In MDA-MB-231 breast cancer cells, phosphorylation of LKB1 at Ser334 by AKT has been reported to block the tumor suppressor activity of overexpressed LKB1 via nuclear sequestration by the 14-3-3 protein [42]. However, there may be cell type-speci c differences in the subcellular localisation of LKB1. While endogenously expressed LKB1 protein is exclusively localised in the nucleus of non-transformed IMR90 broblasts, it is predominantly located in the plasma membrane of polarised epithelial MDCK cells [43]. According to our fractionation data, overexpressed LKB1 and its S334A phosphode cient mutant derivative are both mostly found in the nuclear fractions of PC3 and MCF7 LKB1KO cells, while the endogenously expressed LKB1 protein of the parental cells resides in the cytoplasm, irrespective of whether PIM expression or activity is inhibited. These discrepancies in the cellular compartmentalisation between endogenous and ectopically expressed proteins warrants the usage of knock-in mutant cell lines to properly examine the physiological consequences of LKB1 phosphorylation.
It is not surprising that both PIM and AKT kinases target LKB1, as they also share several other substrates [2,3]. For example, both PIM and AKT protect cells from apoptosis by phosphorylating the proapoptotic BAD protein, albeit at different but proximate sites [44,45]. In addition, both PIM and AKT promote mTOR-and Cap-dependent protein synthesis by phosphorylating the AKT1S1 and EIF4EBP1 translational inhibitors [2,3]. However, these kinases also have cell type-speci c non-redundant roles, as we did not detect any compensatory increase in AKT activity in PC3 or MCF7 PIM TKO cells.
In terms of cell proliferation and tumor growth, the tumor-suppressive effects of LKB1 could be readily seen in the chick embryo CAM xenograft experiments, but not in the two-dimensional cell proliferation assays. Similar discrepancies with respect to the effects of LKB1 between in vitro and in vivo models have also previously been demonstrated [46]. Co-deletion of PTEN and LKB1 from prostate cancer cells results in aggressive tumors and lung metastases, while deletion of LKB1 alone has no such effect. This nding was in line with our CAM data, in which knocking out LKB1 elicited a robust increase in tumor mass in PTEN-de cient PC3 cells, but not in PTEN-expressing MCF7 cells. In PC3 cells, the resulting oncogenic insult could be either suppressed by knocking out all three PIM kinase members or exacerbated by upregulating PIM1 expression, highlighting the integral role of PIM kinases in supporting tumor growth in this setting. Notably, the combined PIM and LKB1 knock-out slowed the rate of cell proliferation and tumor growth as compared to the LKB1 knock-out alone, but without considerable changes in AMPK phosphorylation levels. Even though a decrease in AMPK activity is conventionally associated with the enhanced growth of tumors lacking LKB1, this idea has recently been challenged by ndings in K-Ras-driven models of non-small-cell lung carcinoma which indicated that loss of LKB1 and AMPK suppressed tumorigenesis [47]. In addition, emerging data have revealed that loss of salt-inducible kinases (SIKs), which are less-well studied LKB1 downstream targets, accounts for a signi cant proportion of the transcriptional changes and histological features of LKB1-de cient tumors [48,49]. Further studies are therefore needed to determine whether PIM kinases share signaling pathways with SIKs in affecting the growth of LKB1-de cient tumors as well as whether PIM inhibition can suppress the aggressive metastatic behaviour observed in tumors lacking both PTEN and LKB1.

Conclusions
Catabolic events invoked by the LKB1/AMPK signalling pathway are expected to antagonise the oncogenicity of PIM kinases. Our novel nding that PIM kinases act as upstream regulators of LKB1 uncovers a molecular pathway that allows the tumor-suppressive function of LKB1 and the oncogenic functions of PIM kinases to be tightly and precisely controlled. Inactivation of both PIM kinases and LKB1 results in a signi cant decrease in cell proliferation in vitro and tumor growth in vivo, suggesting that PIM-targeted pharmaceutical interventions could be exploited to suppress the growth of LKB1de cient tumors.