Abstract
Germline mutations of the LKB1 gene lead to Peutz–Jeghers syndrome (PJS), which is associated with a predisposition to gastrointestinal polyposis and cancer. In this study we tested for germline mutations of LKB1 in 11 patients with PJS from nine families and analyzed the expression patterns of the LKB1 and cyclo-oxgenase-2 (COX-2) proteins in 28 Peutz–Jeghers polyps (PJPs) and five carcinomas from these patients by immunohistochemical (IHC) analysis. In eight of those families we identified seven different mutations, which consisted of two splice site mutations, two nonsense mutations, one small inframe deletion, one frame-shift mutation, and one silent mutation. Immunostaining revealed nuclear and cytoplasmic expression of LKB1 protein in 23 PJPs and five carcinomas, nuclear expression alone in one PJP, and loss of LKB1 protein expression in four PJPs, indicating a heterogeneous LKB1 expression pattern in PJPs. Overexpression of COX-2 was detected in 23 (82%) of 28 PJPs and in all carcinomas. Despite heterogeneity in staining of LKB1 among individuals and even among samples from the same individual, we found statistically significant correlations in staining of LKB1 relative to COX-2. These results suggest that COX-2 plays a role in tumorigenesis in PJS and may therefore be considered as a potential target for PJS chemoprevention.
P
Cyclo-oxygenase (COX) is the rate-limiting enzyme in the formation of prostaglandins from arachidonic acid. COX exists in two isoforms, COX-1 and COX-2. COX-2 is constitutively absent in normal tissues but is rapidly induced by certain inflammatory cytokines, tumor promoters, growth factors, and oncogenes (Eberhart and DuBois 1995; Smith et al. 1996). Increased COX-2 levels have been identified in cancer of the colon, lung, breast, stomach, and prostate, and also in pancreatic adenocarcinomas (Prescott and Fitzpatrick 2000). Overexpression of COX-2 has been shown to result in resistance to apoptosis, modulation of adhesion of epithelial cells to the extracellular matrix, and promotion of angiogenesis.
Because both LKB1 and COX-2 are involved in tumorigenesis and suppression of apoptosis, we wanted to determine if expression patterns of these two genes are related in PJPs and in carcinomas from PJS patients. Here we developed an antibody that specifically recognizes the LKB1 protein and examined the expression of LKB1 protein and COX-2 protein in Peutz-Jeghers polyps by immunostaining. We found that COX-2 was overexpressed in most of the PJPs, whereas LKB1 exhibited heterogeneity in its expression patterns. There is a statistically significant correlation between immunostaining of COX-2 and LKB1.
Materials and Methods
Patients and Samples
Peripheral blood samples and formalin-fixed, paraffin-embedded sections of 28 PJPs and five carcinomas were obtained from 11 patients with PJS after each participant had signed release and informed consent documents. The histopathological diagnosis of PJPs and carcinomas was confirmed in all cases by a single pathologist. Consecutive 4–6-μm tissue sections were cut from each tissue block for IHC analysis.
DNA Extraction
Ten ml of blood was drawn in Vacutainer tubes containing EDTA (Becton Dickinson Vacutainer System; Rutherford, NJ) from each study subject. DNA was isolated with an Applied Biosystems 341 Nucleic Acid Purification System (Applied Biosystems; Foster City, CA) according to the manufacturer's instructions.
Mutation Analysis
Nine exons and the adjacent intronic regions of the LKB1 gene were amplified for analysis of germline mutation. The primers used in PCR were the same as those reported previously (Jenne et al. 1998), except for the primers used for exon 1. The primer pair of DJ698F (5′-GGTCCCCGAGGACGAAGTTGA-3′) and STK11 exon 1R (5′-ATCAGGTACTTGCCGATGAG-3′) was used for the first part of exon 1, and the primer pair of STK11 exon438 1F (5′-TCATCTACCAGCCGCGCCGCAA-3′) and DJ673R (5′-ACCATCAGCACCGTGACTGG-3′) was used for the second part of exon 1. The PCR products were treated using a PCR product Pre-Sequencing Kit (USB; Cleveland, OH) and were sequenced in both directions with the PCR primers. Positive results were confirmed in the CLIA-certified laboratory of GeneDx (GeneDx; Gaithersburg, MD;
LKB1 Antibody Preparation
Sigma-Genosys (The Woodlands, TX) generated an anti-peptide polyclonal antibody GN2733. The LKB1 peptide (KKKKLRRIPNGEAN) corresponding to the amino terminal region (AA81–AA94) of human and murine LKB1 was synthesized and conjugated to keyhole limpet hemocyanin before injection into rabbits. The sequence-specific antibodies were then affinity-purified from these antisera using peptide columns. We also studied an additional polyclonal antibody in the carboxyl terminus region of the LKB1 but could not reliably detect expression in normal colon mucosa or in PJPs by immunobiochemistry.
Western Blotting
Total protein (20 μg) was analyzed by 10% SDS-PAGE and blotted according to standard protocols (Sambrook et al. 1989). GN2733 was used as primary antibody to detect the LKB1 protein in DiFi, a colon cancer cell line (Dolf et al. 1991) and in mouse pancreas tissue lysates by an ECL Western blotting analysis system (Amersham Pharmacia Biotech; Fair Lawn, NJ)
Immunochemical Analysis
Slides were deparaffinized and endogenous peroxidase activity was blocked by incubation in 3% H2O2 for 10 min at room temperature. Sections were then placed in PBS and microwaved for 10 min for antigen retrieval. The anti-LKB1 antibody was applied at 1:500 dilution and incubated overnight at 4C. Immunodetection was performed with an LSAB 2 system (DAKO; Carpinteria, CA). Hematoxylin was used as a counterstain. A rabbit anti-LKB1 polyclonal antibody GN2733 generated against the 14 amino acid sequences at the N-terminus of LKB1 described above was used for the anti-LKB1 immunostaining at a dilution of 1:500. A mouse monoclonal antibody against human COX-2 (Cayman Chemical Catalog no. 160112; Ann Arbor, MI) was used for anti-COX-2 immunostaining at a dilution of 1:1000. For every sample, immunostaining was repeated at least twice to confirm the results. Staining was graded for intensity as negative (–), low (+), intermediate (+ +), or high (+ + +) by two independent observers (C. Wei and A. Rashid) and there were no major discrepancies between the two observers.
Statistical Analysis
We used Spearman correlation to evaluate joint staining of LKB1 and COX-2 in polyps and cancers (analyzed both jointly and separately). We also used analysis of variance to compare the levels of staining, contrasting cancers with polyps and polyps in the small bowel to other sites. We used SAS (1999) version 8.1 for all analyses.
Results
Germline Mutation of the LKB1 Gene in Patients with PJS
Mutation analysis was performed on DNA obtained from probands of nine unrelated PJS families (Table 1). Mutations were detected in all but one family. This later family was previously shown to display linkage to the LKB1 locus (Amos et al. 1997). Patients 1 and 7 had a mutation at a splice site, patients 3 and 4 had nonsense mutations, and patient 5 and 6 each had the same frameshift mutation. All of these mutations result in a truncated LKB1 protein. Patient 2 had a 12-bp in-frame deletion, and patient 8 had a single nucleotide mutation that did not affect the LKB1 protein sequence but which may create an alternate splice site in exon 8. In all, of the seven different mutations that we detected, five are novel.
Expression Patterns of LKB1 in PJPs and Carcinomas from Patients with PJS
To analyze the endogenous LKB1 protein in the PJPs and carcinomas, we generated a rabbit anti-peptide polyclonal antibody GN2733. After affinity- purification, this sequence-specific anti-LKB1 antibody specifically recognized the human and murine 53KDa LKB1 protein on Western blotting analysis (Figure 1). We performed anti-LKB1 immunostaining of normal human colon samples (Figure 2A), which were used as control specimens, as well as the 28 PJPs and five carcinomas from the above-mentioned 11 patients with PJS (summarized in Table 2). LKB1 was expressed in both the nucleus and cytoplasm of the epithelial cells and in stromal cells of the normal colon mucosa. In the epithelial cells of 23 polyps (Figure 2B) and five carcinomas (Figure 2C), LKB1 was expressed in both the cytoplasm and nucleus. This was indistinguishable from the normal expression pattern except that, in most of these cells, LKB1 expression levels were elevated compared with those in the normal epithelial cells. In contrast, one PJP expressed LKB1 in the nucleus, without cytoplasmic staining (Figure 2D), and four PJPs demonstrated loss of nuclear and cytoplasmic staining in epithelial cells, although there was detectable LKB1 protein expression in the stromal cells (Figure 2E). Of note was heterogeneity of the staining pattern in PJPs from the same patients. For example, two PJPs from patient 4 had elevated LKB1 expression in both nucleus and cytoplasm of the epithelial cells, whereas the other two PJPs had loss of LKB1 expression. These results suggest that the IHC expression also depends on the somatic alteration of the polyps, although the germline status is the same.
Germline mutation of the LKB1 gene in patients with Peutz–Jeghers syndrome
aIntervening sequence.
bSame family.
Endogenous LKB1 protein detected in a Western blot probed with anti-LKB1 antibody GN2733. Lane 1 and Lane 2 contain human DiFi cell and mouse pancreas tissue lysates respectively.
Expression Pattern of COX-2 Protein in PJPs and Carcinomas from Patients with PJS
We also performed COX-2 IHC on normal colon samples, PJPs, and carcinomas from patients with PJS, using a mouse monoclonal anti-COX-2 antibody to assess COX-2 expression (summarized in Table 2). In normal colon tissue specimens, COX-2 was undetectable in epithelial cells but there was some immunostaining in stromal cells (Figure 2A). In contrast, COX-2 was overexpressed in the cytoplasm of the epithelial cells in 23 PJPs (Figure 2B) and five carcinomas (Figure 2C). In the remaining five PJPs, COX-2 expression was absent (Figures 2D and 2E). We compared the results of anti-LKB1 and anti-COX-2 immunostaining and found that COX-2 staining corresponded to LKB1 staining. All these five PJPs without COX-2 staining showed loss of LKB1 expression in the cytoplasm and nucleus (Figure 2E) or in the cytoplasm only (Figure 2D). Statistical analysis of the expression of polyps and carcinomas listed in Table 2 showed a highly significant correlation in antibody expression. Over all the samples, the Spearman correlation between staining for cytoplasmic staining of LKB1 and COX-2 was 0.93. For nuclear staining of LKB1 vs COX-2 the correlation was 0.91, and for nuclear vs cytoplasmic staining of LKB1 the correlation was 0.92. All of these results are significant at p<0.0001. When the data were stratified by polyp vs carcinoma histology, the correlations remained over 90% for all comparisons and the correlation between cytoplasmic LKB1 staining and COX-2 staining in carcinomas was 100%. Analysis of variance showed no significant differences in antibody staining between carcinomas and polyps or between polyps of the small bowel and other locations.
Representative results of LKB1 and COX-2 immunostaining of normal human colon samples and tissue samples from patients with Peutz–Jeghers syndrome. The antibodies used were rabbit polyclonal anti-LKB1 antibody GN2733 and mouse monoclonal anti-COX-2 anti-body. LKB1 and COX-2 were detected with DAB (brown) and nuclei were counterstained with hematoxylin. In normal colon tissue sample (
Discussion
In this study we described the LKB1 expression patterns in PJPs and carcinomas from 11 patients with PJS. We found two predominant patterns of staining in the samples. In one there was both nuclear and cytoplasmic expression. In a second pattern there was loss of expression for LKB1. Most of the PJPs and all carcinomas showed expression in the epithelium. Both wild-type and in most cases truncated LKB1 protein contribute to the staining from the anti-LKB1 antibody GN2733 corresponding to the N-terminal of LKB1. It has also been reported that LKB1 was expressed in polyps from LKB1+/– mice (Miyoshi et al. 2002; Rossi et al. 2002) and in the epithelium of polyps from PJS patients using in situ hybridization analysis (Rowan et al. 2000). Interestingly, we observed heterogeneity of the staining patterns in different PJPs from the same patient. LKB1 expression was detected in some polyps but not in others from the same patient, showing genetic heterogeneity in PJS. This heterogeneity may be due to differences in somatic alteration in the PJPs in these patients. Previous studies suggest genetic heterogeneity in PJS (Mehenni et al. 1997; Olschwang et al. 1998; Westerman et al. 1999; Entius et al. 2001). Studies of hamartomatous polyps and carcinomas from patients with PJS have revealed somatic mutations or loss of heterozygosity (LOH) of the LKB1 gene, as well as other genetic alterations in the APC, K-ras, β-catenin, and p53 genes (Miyaki et al. 2000; Entius et al. 2001). All these genetic changes may alter LKB1 expression at the IHC level. Loss of expression in two polyps from patient 4 and one polyp from patient 11 may suggest biallelic inactivation in PJPs. Considering that LOH has been observed in some cases but is not a consistent feature of these tumors (Miyaki et al. 2000; Entius et al. 2001), it is possible that biallelic inactivation of LKB1 is not necessarily required for polyp formation but that haploin-sufficiency resulting from LKB1 heterozygosity is sufficient for polyp formation. Studies of mouse models have suggested that haploinsufficiency is sufficient for polyp formation, and our results in human tissues agree with this finding (Miyoshi et al. 2002; Rossi et al. 2002). Although we attempted to stain LKB1 with two different anti-LKB1 antibodies, only the N-terminal antibody GN2733 that we generated gave reliable staining. Of the patients studied, one patient had a splice-site mutation upstream of the epitope used by this antibody and still showed cytoplasmic and nuclear staining for LKB1. It remains possible that the mutated LKB1 is detected by antibody GN2733 in some of the patients studied, while the remaining wild-type protein has been deleted.
Summary of anti-LKB1 and anti-COX-2 immunostaining results of specimens from patients with Peutz–Jeghers syndrome
aPeutz–Jeghers polyps associated with colon carcinoma.
bMetastatic breast carcinoma in chest wall and bilateral intraductal breast carcinomas.
We have identified seven types of germline mutations in eight patients with PJS from eight different families. According to Nezu et al. (1999), mutations that cause truncation of the LKB1 protein at amino acid 311 or before result in loss of kinase function. On this basis, four types of mutations found in families 1, 3, 4, 5, 6, and 7 resulted in disruption of the kinase domain and kinase deficiency. Tiainen et al. (2002) demonstrated that kinase-deficient LKB1 mutants predominantly display nuclear immunostaining. This correlation between loss of kinase activity and nuclear accumulation of LKB1 suggests that the cytoplasmic localization is required for kinase activity. Here, in 10 (77%) of 13 PJPs and in carcinoma samples from these six families, the LKB1 protein displayed a localization pattern indistinguishable from that in the normal tissue samples. Our results indicate two possible explanations for these findings. One possibility is that the subcellular localization of LKB1 may not be solely related to kinase activity. If so, we would have detected more nuclear staining in those subjects with mutations before amino acid 311. The second possibility is that biallelic inactivation due to LOH is not a common event in PJPs, and so the other allele for these PJPs also contributes to the pattern of staining seen.
The nuclear localization pattern found for only specimen 9 from patient 7 is highly consistent with the inability of the mutant SL26 to retain cytoplasmic LKB1, as reported by Nezu et al. (1999). The SL26 mutation is a small in-frame deletion mutation that substitutes four amino acids (303IRQH), and was identified in a PJS patient reported by Hemminki et al. (1998). This small in-frame deletion disrupts the consensus sequence of the potential nuclear localization signal (NLS) (Robbins et al. 1991), which consists of the pattern of two basic residues followed by 10-residue spacer, and another basic region in which at least three of five residues were basic residues. The splice error detected in patient 7 will cause exon 8 deletion (starting from codon 308), which disrupts the same potential NLS in the SL26 mutation, presumably altering the subcellular distribution of the LKB1 protein so that it may accumulate in the nucleus only, although we know the status of only one of the LKB1 alleles. The other allele needs to be further investigated.
We demonstrated that COX-2 was overexpressed in the epithelium of 85% (28/33) of the PJPs and carcinomas. This is a somewhat higher rate than the 70% (16/23) and 64% (7/11) recently reported in two smaller studies by Rossi et al. (2002) and McGarrity et al. (2003), respectively. By comparing the LKB1 and COX-2 immunostaining results, we found that the polyps that expressed LKB1 overexpressed COX-2, and the polyps that showed loss of LKB1 expression also showed lack of COX-2 expression. Our results indicate that COX-2 expression correlates strongly with LKB1 expression. An exception was the polyps from patient 7, where LKB1 protein was detected in the nucleus yet we failed to detect COX-2 expression. The regulatory mechanism of COX-2 overexpression in PJPs is unknown. A mouse model study (Rossi et al. 2002) suggested that the Ras/Raf-1/MEK/ERK signal transduction pathway is most likely to mediate COX-2 induction in murine LKB1+/– polyposis. Further investigation to identify the physiological LKB1 substrates will facilitate a more comprehensive understanding of how LKB1 is involved in signaling pathways to mediate its role both in polyp formation and in COX-2 induction.
It has been reported that increased levels of COX-2 in intestinal epithelial cells increased their adhesion and simultaneously decreased their response to certain apoptotic stimuli (Tsujii and DuBois 1995). Another study showed that SC-58125, a selective COX-2 inhibitor, suppressed the growth of HCA-7 cells that had high levels of COX-2 expression in culture and in tumor implants in nude mice, but had no effect on HCT-116 cells that lack COX-2 expression in culture or when they were implanted in nude mice (Sheng et al. 1997). Therefore, it is quite possible that the over-expression of COX-2 in PJPs and carcinoma samples will result in resistance to apoptosis, thus increasing the tumorigenic potential. Another report indicates that LKB1 is involved in p53-dependent apoptosis and that loss of LKB1 function in PJS leads to a deficiency in apoptosis, which results in the formation of multiple polyps and the subsequent development of malignant tumors (Karuman et al. 2001). Our results of COX-2 overexpression in the majority of the PJ samples suggest that COX-2 may be involved in tumori-genesis in patients with PJS.
In summary, our results demonstrate that COX-2 is overexpressed in polyps and carcinomas from patients with PJS. Treatment with selective COX-2 inhibitors, such as celecoxib and NSAIDs, has been shown to induce apoptosis in a variety of cancer cells, including those of the colon, stomach, and prostate. Therefore, the finding of overexpression of COX-2 in PJPs and carcinomas provides insight into the development of chemopreventive strategies for this disease.
Footnotes
Acknowledgements
Supported in part by a NIH Grant R03 (CA 89843–1), a NIH Cancer Center Support Grant CA16672, American Cancer Society Grant RPG 99–030–01-CCE-3, and the Janis Davis Gordon Memorial Fellowship from the Division of Cancer Prevention at MD Anderson Cancer Center.
We thank the patients and their families for their cooperation. We also thank Mariann Crapanzano for her editorial comments.