Gastrointestinal Hamartomatous Polyposis in Lkb1 Heterozygous Knockout Mice¹

PJS³ is an autosomal dominant disease characterized by hamartomatous polyps of the gastrointestinal tract as well as mucocutaneous pigmentation of the lips, buccal mucosa, and digits (1, 2). Patients with PJS show an increased risk of developing cancer at relatively young ages (3, 4). Recently, LKB1 (also termed STK11) on chromosome 19p13.3, which encodes a serine/threonine kinase of unknown function, was identified as a gene the mutations of which are responsible for PJS (5, 6). To date, various types of germ-line LKB1 mutations have been identified in PJS families (reviewed in Ref. 7). In addition, LOH on chromosome 19p near LKB1 or somatic mutations in the LKB1 gene have been reported in hamartomas and adenocarcinomas developed in a subset of PJS patients, suggesting that LKB1 is a potential tumor suppressor gene (8, 9, 10, 11, 12, 13). Recent cell culture studies have also suggested that LKB1 functions as a tumor suppressor. Forced expression of LKB1 in tumor cell lines results in suppression of cell growth by inducing cell cycle arrests at the G₁ phase (14). Phosphorylation of LKB1 by cyclic AMP-dependent kinase or by p90 ribosomal S6 kinase is essential for the suppression of cell growth (15, 16). LKB1 is also required for brahma-related gene 1-induced growth arrest (17). Moreover, LKB1 plays a key role in p53-dependent apoptosis (18). Taken together, mutations in LKB1 may contribute to tumorigenesis through suppression of either growth arrest or apoptosis in epithelial cells. Recently, an Lkb1 gene knockout mouse line has been reported (19). The homozygotes are embryonically lethal because of multiple developmental defects including aberrant vessel formation in the yolk sac and placenta. Expression of vascular endothelial growth factor is deregulated in the Lkb1 homozygous tissues. It is possible that lack of LKB1 supports tumor cell growth through angiogenesis by induction of vascular endothelial growth factor. However, these studies did not investigate whether heterozygous Lkb1 (+/−) mice develop hamartomatous tumors. Here, we describe construction of Lkb1 gene knockout mice and demonstrate that Lkb1 (+/−) mice develop gastrointestinal hamartomas without inactivation of the remaining wild-type Lkb1 allele.

A partial cDNA fragment of Lkb1 was amplified by RT-PCR using a primer set of E1F (5′-CTC GTA CGG CAA GGT GAA GGA G-3′) and E6R (5′-TAC AGG CCC GTG GTG ATG TTG T-3′). The amplified fragment was used as a probe for screening a 129/SvJ mouse genomic DNA library (Stratagene, La Jolla, CA). A BamHI fragment including exons 1 to 7 was used for vector construction (Fig. 1 A).

Mouse ES cells RW4 (Genome Systems, St. Louis, MO) were used. Homologous recombinants were screened by PCR using primers PGKR (5′-CTA AAG CGC ATG CTC CAG ACT-3′) and LKB1EX (5′-GGC GTC CCT AGA CAC ATT TCC-3′) and confirmed by genomic Southern analysis. Genotyping of the mice was performed by PCR using the primers E6R and PGKR. A BglII-BamHI fragment probe was used for genotyping by Southern hybridization (Fig. 1 A).

The methods have been described previously (20). The primary antibodies anti-αSMA monoclonal antibody (Progen Biotechnik, Heidelberg, Germany) and anti-β-catenin polyclonal antibody (Sigma Chemical Co. St. Louis, MO) were used at 50- and 500-fold dilutions, respectively.

Total RNA was prepared from the glandular stomach using ISOGEN solution (Nippon Gene, Toyama, Japan), and cDNA was synthesized with SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA). Two primer sets were used to amplify a 585-bp fragment from exons 1 to 6 (E1F and E6R) and an 809-bp fragment from exons 1 to 8 (E1F and E8R; 5′-TGT CTG GGC TTG GTG GGA TAG G-3′), respectively.

Genomic DNA was extracted as described (20). Two primer sets were used to detect the targeted allele (PGKR and 5′-GTC ATC CAC AGC GAA AGG GTG C-3′) and the wild-type allele (5′-TAC TTC CGC CAG CTG ATT GAC G-3′ and 5′-GAG GTC GGA GAT CTT GAG TGT G-3′), respectively.

Sense and antisense RNA probes were prepared from a BglII-PstI (1025-bp) fragment of the mouse Lkb1 cDNA. Deparaffinized sections were treated with 0.2 m hydrochloric acid, followed by 5 μg/ml of proteinase K. Subsequently, they were fixed in 4% paraformaldehyde and treated with triethanolamine buffer with stepwise additions of acetic anhydride. After prehybridization, the specimens were hybridized with digoxigenin-labeled RNA probes overnight. DIG nucleic acid detection kit (Roche) was used for signal detection.

Tissue samples were homogenized and sonicated in lysis buffer [10 mm HEPES (pH 7.4), 50 mm NaCl, 50 mm sodium PP_i, 50 mm NaF, 5 mm EDTA, 5 mm EGTA, 100 μm Na₃VO₄, 0.1% Triton X-100, and 500 μm phenylmethylsulfonyl fluoride]. After centrifugation at 2000 × g at 4°C for 10 min, 20 μg of the supernatant protein were separated in a 10% SDS-polyacrylamide gel. Sheep polyclonal anti-LKB1 antibody (Upstate Biotechnology, Lake Placid, NY) and ECL detection system (Amersham Pharmacia, Uppsala, Sweden) were used to detect the specific signals.

The strips of mucosa from the normal and polyp tissues were prepared by peeling off the muscle layer, rinsed with the washing medium (MEM supplemented with 1% BSA, 100 units/ml penicillin, and 100 μg/ml streptomycin), and minced into <1 mm² with scissors. The minced tissue was incubated in PBS containing 1% collagenase (Invitrogen) at 37°C for 1 h with shaking. Epithelial cells were collected by centrifugation at 1000 rpm for 5 min, suspended in culture medium as described (21), and seeded on collagen-coated culture dishes to avoid contamination of interstitial cells.

cDNA fragments that covered the Lkb1 full coding region were amplified with three sets of primers as follows: 5′-CGA AGG GGA CGA GGA CAA AGA-3′ and 5′-GAA CAA TGC CCT GGC TGT GTA G-3′ for the 5′ part; 5′-CTG CGG CAT CGG AAT GTG A-3′ and 5′-TGT CTG GGC TTG GTG GGA TAG G-3′ for the middle part; and 5′-TAT GAG CCG GCC AAG AGG TTC T-3′ and 5′-CTC CAA CGT CCC GAA GTG AGT G-3′ for the 3′ part. Amplified cDNA fragments were sequenced directly using the same primers.

The Lkb1 gene was inactivated in ES cells by homologous recombination in which exons 2, 3, and 4 were deleted and replaced with a neomycin resistance gene cassette (Fig. 1,A). The chimeric mice derived from a recombinant ES clone transmitted the targeted Lkb1 allele to the offspring. The progeny genotypes were examined by genomic Southern analysis (Fig. 1 B). No homozygous mutant pups were obtained by intercrossing the heterozygotes, indicating that Lkb1 (−/−) mice were embryonically lethal. This is consistent with a previous report analyzing another Lkb1 knockout mutant (19).

The transcript from the targeted Lkb1 allele was examined by RT-PCR analysis using primers to amplify cDNA fragments for exons 1–6 and for exons 1–8, respectively (Fig. 1,A). The wild-type Lkb1 cDNA fragment of the expected size was amplified with either primer set (Fig. 1,C, arrows). Smaller fragments were also amplified with both primer sets only when Lkb1 (+/−) cDNA was used as the template (Fig. 1 C, arrowheads). The length of such shorter cDNA fragments precisely matched the predicted size of the exon 2, 3, and 4-deleted Lkb1 cDNA. These results suggest that some alternative splicing events took place in the transcript from the targeted Lkb1 allele. However, such alternatively spliced mRNA, in which exon 1 is ligated to exon 5, should result in a frame-shift mutation after codon 97, eliminating most of the kinase domain (residues 50–337).

Lkb1 (+/−) mice developed normally and showed no overt phenotypes up to 20 weeks of age. However, gastric polyps developed in 93% of Lkb1 (+/−) mice >20 weeks of age (Fig. 2). After 40 weeks of age, the incidence of the gastric polyps reached 100%. Polyps developed in the glandular stomach, often in the pyloric region (Fig. 3,A). Polyps were also found in the small intestine of the aged Lkb1 (+/−) mice (Figs. 2 and 3,B). The incidence of the small intestinal polyps in the heterozygotes >50 weeks was 31% (Fig. 2).

Some Lkb1 (+/−) mice became moribund after 50 weeks of age, which was attributed to pyloric constriction and obstruction of the stomach. Others survived beyond 70 weeks of age (data not shown).

Although smaller gastric polyps were of sessile morphology, larger ones were pedunculated (Fig. 3,C). They consisted of glandular and cystic epithelial layers. A bromodeoxyuridine incorporation assay showed that tumor cells proliferated continuously in the polyps (Fig. 3,D). However, they did not show any dysplastic morphology, but the normal epithelial layers of different types were mixed in the polyp tissue (Fig. 3,E, inset). These histological characteristics indicate that the Lkb1 (+/−) polyps are hamartomas like those in human PJS. To further compare the pathological characteristics of the Lkb1 (+/−) with PJS polyps, we immunohisotochemically studied the mouse gastric polyp sections using an anti-αSMA antibody. The PJS hamartomas show a particular histology where benign glands are surrounded by fronds of lamina propria containing the muscularis mucosae. In the Lkb1 (+/−) mice, as expected, an arborizing network of smooth muscle bundles was stained with anti-αSMA antibody, extending into branching fronds in the polyp tissue (Fig. 3,F). Thus, the Lkb1 (+/−) polyps exhibit strikingly similar characteristics to those of PJS. The small intestinal polyps were also composed of the normal epithelial cell layers (Fig. 3 G). Few adenomatous changes were observed in any hamartomas examined, which is consistent with the PJS hamartomas that show low rates of neoplastic changes (22).

These histological characteristics of the Lkb1 (+/−) hamartomas are very different from those of the Apc knockout (Apc^Δ716) mouse intestinal polyps, where proliferating dysplastic adenoma cells are found without smooth muscle fibers of an arborizing pattern (20). To rule out the possibility that the Wnt signaling pathway is activated in the Lkb1 (+/−) hamartomas, we determined the subcellular localization of β-catenin by immunostaining. When the Wnt pathway is activated, β-catenin is stabilized and translocated to the nucleus, as we have demonstrated previously in the Apc^Δ716 polyps (23). In the Lkb1 (+/−) polyps, however, β-catenin was localized to the basolateral membrane and remained outside the nucleus in the polyp epithelial cells (Fig. 3, H and I). These results suggest that the Wnt pathway is not activated in the hamartomatous polyps, which is consistent with a previous report that neither β-catenin mutation nor APC LOH is detected in PJS hamartomas (9, 12).

To determine whether loss of the LKB1 function caused hamartomas in the gastrointestinal epithelium, LOH for the Lkb1 gene was analyzed in the polyp tissues. A genomic PCR analysis using allele-specific primers showed both the wild-type and targeted Lkb1 alleles in all hamartomas examined (Fig. 4,A). Then the nucleotide sequence was determined for the Lkb1 mRNA expressed in the hamartoma epithelial cells. As a result of direct sequencing analyses, no Lkb1 mutation was found in any of the four independent polyps examined (data not shown). Finally, LKB1 protein expression was examined by Western blotting. In Lkb1 (+/−) mice, the protein expression was decreased to 47% of that in the wild-type mice, reflecting the Lkb1 gene dosage (Fig. 4, B and C). Interestingly, the same level of LKB1 as in the normal stomach was expressed in the gastric hamartoma tissues of the Lkb1 (+/−) mice. No shorter protein bands were detected that could be derived from the targeted allele encoding the NH₂-terminal 97 residues. Taken together, these data exclude the possibility for biallelic Lkb1 inactivation as the cause of the hamartoma initiation.

To determine the cell types that express Lkb1 in the hamartomatous polyps, we performed an in situ hybridization analysis. The Lkb1 mRNA was predominantly detected in the hamartoma epithelium, as well as in the normal epithelium of the gastric mucosa (Fig. 4, D and F). This result is consistent with a previous report that shows LKB1 expression in the polyp epithelium of PJS patients (24). These data, taken together, suggest that another genetic event(s) in addition to the heterozygous inactivation of Lkb1 may be required for the hamartoma formation.

In extraintestinal organs, most Lkb1 (+/−) mice developed hepatocellular carcinoma after 30 weeks of age, suggesting that LKB1 plays a tumor suppressor role in the liver. Different from PJS, we could not find any other neoplastic lesions in Lkb1 (+/−) mice (data not shown).

LKB1 is the gene the mutations of which are associated with PJS (5, 6), and various types of germ-line LKB1 mutations have been reported in PJS (8). The present study using Lkb1 knockout mice provides the first direct evidence that a heterozygous Lkb1 mutation is responsible for gastrointestinal hamartomas that share the histological features with those in human PJS. However, the polyp localization in Lkb1 (+/−) mice is not necessarily identical with that in PJS, because polyps develop frequently in the small intestine as well as in the stomach and colon. It is conceivable that the additional genetic change that affects hamartoma formation (see below) differs between mice and human. According to some studies, LOH or somatic mutations of LKB1 are found in informative cases of hamartomas and adenocarcinomas in PJS patients, suggesting that LKB1 is a tumor suppressor gene (8, 9, 10, 11, 12, 13). It has been also proposed that a somatic mutation in another gene may be required for the neoplastic changes of the PJS polyps (12, 13). However, it has not been determined whether loss of the LKB1 function is essential for the hamartoma initiation. Here, we have excluded Lkb1 LOH in the gastric hamartomas of the Lkb1 heterozygous mice. Furthermore, we have demonstrated LKB1 expression at the haploid level in the hamartoma tissues. These data strongly suggest that biallelic inactivation of Lkb1 is not necessarily required for the hamartoma development. It is conceivable that additional genetic alterations in other genes trigger the hamartoma formation. Considering the LKB1 functions, it is possible that LOH in the Lkb1 gene contributes to progression of the hamartomas to malignant tumors. Because LKB1 is involved in the cell cycle arrest and p53-dependent apoptosis (14, 17, 18), reduction in its level may lead to suppression of growth arrest and apoptosis and help accumulate somatic mutations. Such cells might be predisposed to transformation. Gene chip analysis may help identify the gene mutation(s) that trigger the hamartoma formation. In this report, we have shown that the Wnt signaling cascade was not activated in the Lkb1 (+/−) hamartomas. The result is consistent with a previous report that showed β-catenin mutation in addition to LKB1 LOH only in adenomatous lesions of the hamartomatous polyps (9, 12). Thus, activation of the Wnt pathway may be an important step in hamartoma-carcinoma progression.

Previously, we demonstrated that heterozygous Smad4 knockout mice develop gastrointestinal hamartomas that have similar histopathology to those in JPS, such as stromal expansion or association with adenomatous lesions (23). SMAD4 plays a key role in the TGF-β-regulated signal transduction as a transcriptional factor (25), and its germ-line mutations account for a subset of familial JPS cases (26). Furthermore, heterozygous TGF-β gene knockout mice also develop benign gastric tumors, suggesting the involvement of altered TGF-β signaling in gastrointestinal hamartomas (27). Although histological characteristics of hamartomas in the Lkb1 (+/−) and Smad4 (+/−) mice are distinct as those in PJS and JPS, it is possible that similar molecular mechanisms underlie the formation of gastric hamartomas. Recently, Smith et al. (28) have shown that LKB1 and SMAD4 form a complex mediated by LIP1 (for LKB1 interacting protein 1). They suggest a novel mechanistic link between PJS and JPS. Therefore, it would be of great interest to investigate whether synergistic effects are observed in gastrointestinal hamartomas in Lkb1 (+/−) Smad4 (+/−) compound mutant mice. It is intriguing to propose that the Lkb1 heterozygosity leads to suppression of SMAD4 transcriptional activity, resulting in hamartoma formation. It should be also important to investigate whether target genes of TGF-β signaling is involved in hamartoma formation.

Although Knusdson’s two-hit model explains the majority of tumorigenesis by tumor suppressor gene inactivation, several cases have been reported recently that suggest haploinsufficiency responsible for tumor formation. For example, haploinsufficiency of CBFA2 causes familial thrombocytopenia with propensity to develop acute myelogenous leukemia in humans (29). In mice, tumors heterozygous for p53 mutation appear slightly later than those with biallelic inactivation (30). Likewise, mice hemizygous for p27^Kip1 develop tumors without inactivation of the remaining wild-type allele (31). It is worth noting that the homozygous mutant mice for these genes are viable and are more tumorigenic than the heterozygotes. On the other hand, homozygous mutation in Lkb1 causes embryonic lethality. Our results indicate that haploinsufficiency may by itself predispose to tumor development and provide the first step in oncogenesis.

We thank J. Inuzuka for histological and immunohistochemical sample preparations and T. Kurosawa for encouragement.