Edinburgh Research Explorer LKB1 haploinsufficiency cooperates with Kras to promote pancreatic cancer through suppression of p21-dependent growth arrest

mechanism We investigated the of Lkb1 in a model of pancreatic cancer, both in terms of disease progression and the molecular level. To test the relevance of our to investigated levels of LKB1 and its potential targets in RESULTS: We deﬁnitively show that Lkb1 haploinsufﬁciency can cooperate with oncogenic to cause pancreatic ductal adenocarcinoma (PDAC) in the mouse. Mechanistically, this was associated with decreased p53/p21-dependent growth arrest. Haploinsufﬁciency for p21 ( Cdkn1a) also synergizes with Kras G12D to drive PDAC in the mouse. We also found that levels of LKB1 expression were decreased in around 20% of human PDAC and signiﬁcantly correlated with low levels of p21 and a poor prognosis. Remarkably, all tumors that had low levels of LKB1 had low levels of p21, and these tumors did not express mutant p53. CONCLUSIONS: We have iden-tiﬁed a novel LKB1-p21 axis that suppresses PDAC follow-ing Kras mutation in vivo. Down-regulation of LKB1 may therefore serve as an alternative to p53 mutation to drive pancreatic cancer in vivo.

P ancreatic cancer is the fourth most common cause of cancer deaths worldwide, with an estimated 5-year overall survival of Ͻ5%. 1 The highly aggressive nature of this disease, combined with the anatomical location of tumors, results in 90% of patients having surgically unresectable disease at the time of diagnosis. 2 The pancreas consists of 3 main cell types-islet cells, acinar cells, and duct cells. Tumors can arise from any of these cell types, but approximately 90% of cases are pancreatic ductal adenocarcinoma (PDAC). PDAC arises from precursor lesions called pancreatic intraepithelial neoplasms (PanINs). 3 The formation of PanIN lesions and the progression to invasive adenocarcinoma are driven by activation of the KRAS oncogene in about 90% of cases, 4 accompanied by loss of function of tumor suppressors, most commonly the Ink4a, p53, and Smad4 tumor suppressors. 3 Certain inherited genetic lesions have also been shown to confer a predisposition to pancreatic cancer. Mutations in the LKB/STK11 tumor suppressor gene result in the Peutz-Jeghers syndrome, 5,6 an autosomal-dominant condition characterized by hamartomatous polyps of the gastrointestinal tract and a dramatically increased risk of epithelial malignancies at other sites, including a Ͼ100fold increased risk of pancreatic cancer. [7][8][9] Restoration of silenced LKB1 in human pancreatic carcinoma cells induces apoptosis in vitro. 10 Furthermore, LKB1 gene inactivation has been observed in intraductal papillary mucinous neoplasms of the pancreas. 11 Lkb1 knockout mice are not viable, and embryos survive only until embryonic day E9.5 because of neural tube defects and vascular abnormalities. 12 However, Lkb1 ϩ/Ϫ mice are viable and mirror human Peutz-Jeghers syndrome in that they develop benign intestinal polyps (hamartomas) and have an increased risk of a range of cancers later in life. [13][14][15][16] However, the consequences of Lkb1 deficiency in the pancreas have not been well-studied thus far, and the mechanisms by which its loss may contribute to pancreatic cancer are unknown.
Lkb1 encodes a serine/threonine kinase that activates a number of downstream kinases, including the adenosine monophosphateϪactivated protein kinase (AMPK), which responds to energy stress by negatively regulating the mammalian target of rapamycin kinase. 17 Lkb1 is also able to regulate cell growth and apoptosis, potentially through interaction with the tumor suppressor p53. 18 Ectopic expression of Lkb1 in cells lacking the endogenous protein induces p21 expression and cellcycle arrest in a p53-dependent manner, and Chromatin Immunoprecipitation analysis has revealed that Lkb1 is recruited to the p21 promoter by p53. [19][20][21] Lkb1 deficiency has also been shown to prevent culture-induced senescence, although paradoxically it renders cells resistant to subsequent transformation by Ha-Ras. 13 Using cre-lox technology to target endogenous expression of Kras G12D to the mouse pancreas through the Pdx1 pancreatic progenitor cell gene promoter results in formation of PanINs. 22 However, these lesions fail to rapidly progress and only develop into invasive pancreatic adenocarcinoma at low frequency unless additional genetic lesions are introduced. In this study, we have assessed whether Lkb1 loss can promote tumorigenesis in this model and found a dramatic acceleration of tumorigenesis in mice carrying a single conditional knockout allele of Lkb1. We have also demonstrated that this is associated with decreased p21-dependent growth arrest.

Immunohistochemistry
Immunohistochemical analysis was performed on formalin-fixed paraffin-embedded sections according to standard protocols. For detailed protocols, see Supplementary Materials.

Senescence-Associated ␤-Galactosidase Staining
We stained cryosections of mouse pancreas or tumor for senescence-associated ␤-galactosidase activity according to manufacturer's protocol (Cell Signaling Technology, Danvers, MA) and counterstained them with nuclear fast red solution.

Laser Capture Microdissection and RNA Isolation
Frozen tissue was sectioned (at 15-20 m) onto PALM-PEN membrane slides and lightly stained with hematoxylin. Laser capture microdissection was performed using the P.A.L.M. MicroLaser System. RNA was isolated with the RNA easy extraction kit (Qiagen, Hilden, Germany).

Reverse-Transcription Polymerase Chain Reaction
Total RNA was reverse transcribed to complementary DNA using the Superscript III kit (Invitrogen, Carlsbad, CA) according to manufacturer's instructions. For further information and primers, see Supplementary Materials.

Tissue Microarray Analysis
The human pancreatico-biliary tissue microarray was created within the West of Scotland Pancreatic Unit, University Department of Surgery, Glasgow Royal Infirmary. For further information, see Supplementary Materials.
This increased pancreatic cancer predisposition was not limited to invasive tumors; the number of PanINs observed in 6-week-old Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) mice was significantly increased when compared with Pdx1-Cre, Kras G12D/ϩ (KC) mice ( Figure 1B, P ϭ .007). In addition, we also observed an increase in PanIN 2 and PanIN 3 lesions compared with KC mice ( Figure 1C). Histological sections of tumors arising in Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) animals were analyzed to ascertain the phenotype of these PanINs and tumors. PanIN lesions exhibited characteristic histologic changes of the normal duct, including expansion of the cytoplasm with associated mucin accumulation, which was confirmed by Alcian blue staining ( Figure 1D, right), formation of papillary architecture, loss of polarity, appearance of atypical nuclei, and luminal budding ( Figure 1D). A majority of Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) tumors were PDAC ( Figure 1E); however, some tumors exhibited a more cystic morphology ( Figure 1E, middle panel), and en- hanced immune cell infiltration was apparent in some tumors ( Figure 1E, right panel), compared with the small number of tumors observed in older Pdx1-Cre, Kras G12D/ϩ (KC) mice. Our results show that Lkb1 deficiency can synergize with activated Kras to induce pancreatic tumor formation.

Homozygous Loss of Lkb1 Is Sufficient to Initiate Pancreatic Tumorigenesis
We also investigated whether loss of Lkb1 as a sole initiating genetic event was sufficient to induce pancreatic tumor formation in the mouse. We crossed Lkb1 flox/ϩ mice to Pdx1-Cre mice and interbred the offspring to generate cohorts of Pdx1-Cre Lkb1 flox/ϩ (LC) and Pdx1-Cre Lkb1 flox/flox (LLC) mice. We found that Pdx1-Cre Lkb1 flox/flox (LLC) mice develop pancreatic tumors with an incidence of 100% and a median survival of 68 days, while Pdx1-Cre, Lkb1 flox/ϩ (LC) mice remained disease-free for 500 days (Figure 2A). Pdx1-Cre Lkb1 flox/flox (LLC) mice presented with abdominal distention, and tumors arising in these mice were mucinous cystadenomas characterized by the presence of multiple large cysts, in some cases at the expense of most of the normal pancreas tissue ( Figure 2B-E).
Tumors also exhibited excessive mucin production, as confirmed by Alcian blue staining ( Figure 2E). We conclude that Lkb1 loss as a sole event is sufficient to initiate pancreatic tumor growth; however, those tumors are benign mucinous cystadenomas and Lkb1 loss alone is not sufficient to drive formation of PDAC. These results agree with a previous analysis of mice lacking Lkb1 specifically within the pancreas, in which mice developed pancreatic serous cystadenomas. 27
Laser capture microdissection was also performed to isolate tissue from preneoplastic PanIN lesions and tumors arising in Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) mice, and reverse-transcriptase polymerase chain reaction analysis showed transcription of wild-type Lkb1 in the resulting tumors ( Figure 3D). Further, immunoblot analysis revealed only a decrease in Lkb1 levels, and a reduction in  Figure  1). These results demonstrate that Lkb1 is a haploinsufficient pancreatic tumor suppressor, and that lack of only 1 allele is sufficient, when combined with Kras mutation, to cause PDAC.

Lkb1 Deficiency Limits Expression of the Tumor Suppressors p53 and p21 in PanIN Lesions
We sought to further delineate the mechanism by which Lkb1 haploinsufficiency synergizes with activated Kras to promote pancreatic tumorigenesis. Consistent with its in vivo tumor suppressor function, Lkb1 deficiency has been shown to prevent culture-induced senescence. 13 Re-expression of Lkb1 in cancer cell lines deficient for Lkb1 has also been shown to result in p53-dependent cell-cycle arrest and enhanced expression of p21. 19,20 On the basis of these results, we wondered whether Lkb1 might act to suppress pancreatic tumorigenesis by promoting growth arrest in vivo through transcriptional activation of p21, because preneoplastic pancreatic lesions in Elas-tTA/tetO-Cre, Kras G12V mice have previously been reported to undergo oncogene-induced senescence, as indicated by positive staining for a number of senescence markers. 29 We hypothesized that preneoplastic lesions in our Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) mice would exhibit diminished p21 and p53 expression compared with those lesions found in Pdx1-Cre, Kras G12D/ϩ (KC) mice. We performed immunohistochemical analysis for both p21 and p53 in PanIN lesions in these mice. High levels of both p21 and p53 were observed in PanINs arising in Pdx1-Cre, Kras G12D/ϩ (KC) mice ( Figure 4A, Supplementary Figure 2), compared with normal ducts in these mice, as expected (data not shown). Significantly, however, in PanIN lesions arising in Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) mice, we observed a considerable reduction in levels of p21 and of p53 ( Figure 4B, Supplementary Figure 2). We quantified the proportion of cells staining positive for p21 and p53 expression in PanINs from both Pdx1-Cre, Kras G12D/ϩ (KC) and Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) mice and confirmed that expression of both was significantly reduced in Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) PanINs with a median of 14.3% p21-positive cells and 12.0% p53positive cells, compared with 32.2% and 36.1%, respectively, in Pdx1-Cre, Kras G12D/ϩ (KC) PanINs ( Figure 4C, P Ͻ .002; Figure 4D, P Ͻ .004). Quantitative real-time polymerase chain reaction analysis performed on microdissected tissue demonstrated that transcription of p21 is also decreased in Pdx1-Cre, Kras G12D/ϩ Lkb1 flox/ϩ (KLC) mice, to 0.64% of the levels observed in Pdx1-Cre, Kras G12D/ϩ (KC) PanINs (data not shown).

Decreased Lkb1 Expression in Human PDAC Correlates With Low p21 Expression and Reduced Survival
We next sought to investigate whether this Lkb1/ p21 pathway was relevant to human PDAC development. Lkb1 and p21 immunohistochemistry was performed in a tissue microarray containing 114 cases of primary human PDAC. As expected, we observed Lkb1 staining primarily in the cytoplasm of epithelial cells ( Figure 7A), while p21 staining was evident in the nuclear compartment (Supplementary Figure 5). Lkb1 staining was present in 98% of stained normal ductal tissue. In PDAC, 19% of cases expressed Lkb1 at a low level (histoscore Ͻ100). Expression levels of Lkb1 did not differ in terms of lymph node status or tumor size; however, high tumor grade and stage were significantly associated with lower median Lkb1 expression level ( Figure 7B; P ϭ .01 and P ϭ .02, respectively). In univariate analysis, low Lkb1 expression (n ϭ 20) was associated with significantly decreased survival compared with high expression (n ϭ 86) after resection ( Figure 7D, left panel 13.5-20.6]; P ϭ .006). Most importantly, in a multivariate Cox proportional-hazards regression analysis, low Lkb1 expression remained an independent predictor of poor survival, with a hazard ratio of 1.87 (95% CI: 1.09 -3.22; P ϭ .022).
Given our preclinical data suggesting that low Lkb1 levels caused low levels of p21, we next investigated the expression of p21 on the same human PDAC tissue microarray. Expression levels of p21 were not significantly altered in relation to any clinicopathological parameter; however, low expression of p21 (n ϭ 78) was associated with decreased cumulative survival after surgical resection, compared with high expression (n ϭ 28) ( Figure 7D, right Figure  7C, Spearman's correlation coefficient 0.34; P Ͻ .001). Significantly, high expression of both Lkb1 and p21 identified a group of patients with a more favorable outcome and a median survival of 25.7 months ( Figure 7E, 95% CI: 12.9 -40.3). Other predictors of poor survival were higher tumor stage, high histologic grade, larger tumor size, and positive resection margin; however, p21 status did not independently influence outcomes (Supplementary Table 1).
Because the TP53 tumor suppressor gene is frequently mutated in human pancreatic cancer (40%-70%) 30 and LKB1 is down-regulated in around 20% of PDAC, we hypothesized that loss of Lkb1-mediated p53/p21 induc- tion might be able to circumvent the need for p53 mutation in human PDAC and thus should not be downregulated in those tumors with p53 mutation. We therefore investigated levels of p53 accumulation, indic-ative of p53 mutation, by immunohistochemical staining of the human PDAC tissue microarray. Strikingly, in those tumors that had low levels of LKB1, and, hence, low levels of p21, we did not observe accumulation of mutant p53 (median histoscore ϭ 4.08, n ϭ 20). In contrast, in the subset of tumors that had low p21 with high LKB1 expression, we found significantly higher levels of p53, indicative of accumulation of mutant p53 (median histoscore ϭ 71.3, n ϭ 58, P ϭ .05) ( Figure 7F). In human pancreatic cancer, we have shown that Lkb1 deficiency correlates with loss of p21 expression and with poorer prognosis, and that Lkb1 deficiency may act as an alternative to p53 mutation in human pancreatic tumorigenesis. These results support the hypothesis that Lkb1 acts as a tumor suppressor in the pancreas, and that it functions, at least in part, by inducing p21 expression. Loss of Lkb1 can thus facilitate escape from Ras-induced p21-mediated growth arrest, and promote Ras-induced tumorigenesis in the pancreas.

Discussion
These data show that Lkb1 haploinsufficiency synergizes with activated Kras in pancreatic tumorigenesis. Mechanistically, we believe this is because of reduced growth arrest/senescence through low levels of p21 in PDAC from these mice. Importantly, our study of human PDAC strongly supports this finding, as low levels of p21 and LKB1 are correlated in human PDACs. Our data are consistent with the previous findings that Lkb1 loss prevents culture-induced cellular senescence, 13 allows BRAF mutant melanoma cells to proliferate, 31 and cooperates with activating Kras mutations in a mouse model of lung cancer. 32 Indeed, our studies in both the pancreas and intestine suggest strong synergy with Kras signaling, with heterozygosity for Lkb1 sufficient to drive signaling downstream of Kras. 33 Overall, these data indicate that levels of Lkb1 are critical in determining the cellular response to Kras activation.
One important question that has been raised through our work and that of others 13 is whether biallelic mutations in LKB1 are required for tumorigenesis or whether they may in fact be limiting for tumor progression. Peutz-Jeghers syndrome patients develop benign hamartomas of the gastrointestinal tract and develop intraductal papillary mucinous neoplasm and cystadenomas.
Here we have confirmed the previous study of Hezel and colleagues, 27 who showed that complete loss of Lkb1 in the pancreas leads to formation of benign cystadenomas. Taken together, these data argue that complete loss of Lkb1 leads to formation of benign tumors, that a cooperating oncogenic event is required to drive carcinoma formation, and that the timing of the cooperating oncogenic event may be critical-if it occurs too late the tumor may not progress from a benign state. From the data presented here, we suggest that in sporadic cancer, a single LKB1 mutation or down-regulation of protein expression would be sufficient to synergize with KRAS mutation to drive tumor progression. Analysis of human pancreatic cancers is consistent with this hypothesis; 20 of 106 tumors show a down-regulation of LKB1 compared to normal ductal epithelium and, remarkably, low levels of LKB1 can act as an independent prognostic indicator of poor outcomes of resected pancreatic cancer. In agreement with our findings in the pancreas, when the LKB1 gene sequence was determined in primary lung adenocarcinomas, only 8 of 27 tumors (of 80 cancers total) that had a mutation or deletion of LKB1 exhibited biallelic loss, 32 suggesting that a monoallelic mutation in LKB1 is sufficient to drive cancer progression. The lack of LKB1 mutations so far observed in human RAS-driven pancreatic tumors may instead be explained by downregulation at the protein level, or inactivation of the gene by epigenetic means, because hypermethylation of Lkb1 in hamartomatous polyps and in tumors commonly associated with Peutz-Jeghers syndrome has been demonstrated in the absence of mutation of the gene. 34 We propose that the mechanism for the synergy between Lkb1 heterozygosity and Kras activation is an escape from Kras G12D -induced growth arrest by loss of p53 mediated p21 up-regulation. The reasons for this are multiple, including increased numbers of PanINs in Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) mice, increased proliferation of PanINs, concomitant reduced expression levels of p53 and p21, reduced expression of senescenceassociated ␤-galactosidase, rapid tumorigenesis in Pdx1-Cre, Kras G12D/ϩ , p21 ϩ/Ϫ mice, and the human data showing the correlation between LKB1 and p21 levels. Remarkably, no human tumors that had low Lkb1 expression had high p21 expression. There was a subset of human tumors that had low p21 with high LKB1 expression, presumably because of the fact that multiple different events can cause p21 down-regulation, for example, p53 mutation or TBX2 overexpression. 35 Indeed, this group of tumors exhibited high levels of p53, indicative of mutant p53 accumulation, suggesting that Lkb1 deficiency can substitute for p53 mutation in human pancreatic tumorigenesis.
Given the plethora of pathways that LKB1 impinges on, it is likely that other pathways may also contribute to the phenotype we see here. However, we failed to see clear up-regulation of phosphoϪmammalian target of rapamycin within the Pdx1-Cre, Kras G12D/ϩ , Lkb1 flox/ϩ (KLC) PanINs and tumors when compared with the Pdx1-Cre, Kras G12D/ϩ (KC) PanINs and tumors, although we clearly see reduced levels of the target phospho-AMPK (data not shown). It is possible that within the pancreas, reduced AMPK activation is not sufficient to exert a clear phenotype and indeed heterozygous AMPK knockout mice have no reported phenotype. 36 Other potential phenotypes of LKB1 deficiency, such as a loss of polarity and differentiation to mucus secretory lineages, could accelerate tumorigenesis in this system. 27,37,38 However, one of the characteristics of Kras-driven PanINs is an increase in mucin secretion and loss of polarity and, because heterozygosity for Lkb1 has never been sufficient to drive either of these 2 events, we believe that these are not major contributors to our phenotype, although they may act in synergy with Kras activation.
In conclusion, we have shown that Lkb1 heterozygosity can accelerate Kras G12D -induced PDAC formation. We have observed a marked reduction of p53 and p21 expression in PanIN lesions in these mice compared with mice bearing intact Lkb1. This correlation is borne out in human PDAC. We therefore propose that Lkb1 acts as a tumor suppressor in the pancreas through its ability to limit the p53/p21 pathway, thus allowing precursor lesions to more easily overcome the Ras-induced growtharrest barrier to tumor formation.

Genetically Modified Mice and Animal Care
Animals were kept in conventional animal facilities and monitored daily. Experiments were carried out in compliance with UK Home Office guidelines. Mice were genotyped by polymerase chain reaction analysis. Tumor and metastatic burden was assessed by gross pathology and histology. Animals were sacrificed by cervical dislocation as per institutional guidelines. Organs/tumors were removed and either fixed in 10% buffered formalin overnight at room temperature or snap frozen in liquid nitrogen. Fixed tissues were paraffin-embedded, and 5-m sections were placed on sialynated/poly-L-lysine slides for immunohistochemical analysis.

Immunohistochemistry
Formalin-fixed paraffin-embedded sections were deparaffinized and rehydrated by passage through Xylene and a graded alcohol series. Endogenous peroxidase activity was inactivated by treatment with 3% hydrogen peroxide, after which antigen retrieval was performed using microwave-heated antigen unmasking solution (Vector Labs, Burlingame, CA) or by incubation in citrate buffer in a pressure cooker. Sections were blocked in 5% serum for an hour, and then incubated with primary antibody for an hour at room temperature or overnight at 4°C. Primary antibodies used were anti-Lkb1 (Abcam, Cambridge, UK) 1:200, anti-pAMPK (Cell Signaling Technology) 1:50, anti mouse p53 (Vector) 1:100, anti-p21 (Santa Cruz Biotechnologies, Santa Cruz, CA) 1:500, anti-Ki67 (Vector) 1:200, and anti-human p53 (Dako, Carpinteria, CA). Sections were incubated in secondary antibody for an hour (Dako Envision ϩ Kit, or Vectastain ABC system) and the staining was visualized with 3,3=-diaminobenzidine tetrahydrochloride. Alcian blue staining was carried out by incubation in Alcian blue solution (pH 2.5) for 30 minutes, followed by counterstaining in nuclear fast red solution for 5 minutes.

Reverse-Transcriptase Polymerase Chain Reaction
Polymerase chain reactions were performed on a PTC-200 DNA Engine (Bio-Rad Laboratories, Hercules, CA), using the GoTaq polymerase kit (Promega, Madison, WI) according to manufacturer's instructions. Polymerase chain reaction products were run on a 2% agarose gel, stained with ethidium bromide, and visualized using the GelDoc-It 300 imaging system (UVP, Cambridge, UK). Primers sequences used were: Lkb1 F: GGTCACACTTTA-CAACATCAC, and R: CTCATACTCCAACATCCCTC.

Tissue Microarray Analysis
All patients gave written, informed consent for the collection of tissue samples, and the local Research Ethics Committee approved collection. All cases had undergone a standardized pancreaticoduodenectomy. A total of 1500 cores from a total of 224 cases with pancreaticobiliary cancer (including 114 pancreatic ductal adenocarcinomas) with a full spectrum of clinical and pathological features were arrayed in slides. At least 6 tissue cores (0.6 mm diameter) from tumor and 2 from adjacent normal tissue were sampled. Complete follow-up data were available for all cases within the tissue microarray analysis. Lkb1, p21, and p53 expression levels were scored based on staining intensity and area of tumor cells using a weighted histoscore calculated from the sum of (1 ϫ % weak staining) ϩ (2 ϫ % moderate staining) ϩ (3 ϫ % strong staining), providing a semi-quantitative classification of staining intensity. The cutoff for high and low expression of Lkb1 and p21 was a histoscore of 100 and 40, respectively. Statistical correlation between Lkb1 expression and p21 expression in human PDAC was determined by the Spearman Correlation Coefficient analysis. Kaplan-Meier survival analysis was used to analyze the overall survival from the time of surgery. Patients alive at the time of follow-up point were censored. To compare length of survival between curves, a log-rank test was performed. A Cox proportional hazards model was used for univariate analysis to adjust for competing risk factors, and the hazard ratio with 95% CIs was reported as an estimate of the risk of disease-specific death. Variables that were found to be significant on univariate analysis at P Ͻ .10 were included in multivariate analysis in a backward stepwise fashion. Statistical significance was set at a P value of Ͻ.05. All statistical analyses were performed using SPSS version 15.0 (SPSS Inc, Chicago, IL).