Haploinsufficiency of Krüppel-Like Factor 5 Rescues the Tumor-Initiating Effect of the Apc^Min Mutation in the Intestine

Mice

Founder C57BL/6J male mice heterozygous for the Apc^Min allele (Apc^Min/+) were purchased from the Jackson Laboratory. Founder C57BL/6J mice heterozygous for the Klf5 alleles (Klf5^+/−) were previously developed (15). Apc^Min/+ males were mated with Klf5^+/− females to obtain mice wild-type (WT) for both alleles, mice heterozygous for either the Klf5 allele (Klf5^+/−) or the Apc^Min allele (Apc^Min/+), or mice heterozygous for both the Apc^Min and Klf5 alleles (Apc^Min/+/Klf5^+/−).

Tumor assessment

Mice were sacrificed at age 16 to 18 wk or at age 28 wk by CO₂ asphyxiation. The entire small intestine was dissected longitudinally. Intestinal tissues were examined under a dissecting microscope in a blinded fashion for the presence of adenomas. Intestinal adenomas were counted and grouped by size as follows: <1, 1 to 2, 2 to 3, and >3 mm.

Cell culture

Cell culture medium and fetal bovine serum (FBS) were purchased from Invitrogen. COS-1 cells were acquired from American Type Culture Collection and maintained in DMEM and 10% FBS with 100 U/mL penicillin and 100 μg/mL streptomycin.

Plasmid constructs and antibodies

Information regarding plasmids and antibodies used can be found in Supplementary Materials and Methods.

Immunohistochemistry

Immunohistochemical staining using a chromogenic substrate was performed as described previously (19). For immunofluorescence staining, tissue sections were deparaffinized, rehydrated, subjected to antigen retrieval using Rodent Decloaker (Biocare Medical), and incubated with primary antibody overnight at 4°C. Slides were then incubated with Alexa Fluor secondary antibodies (Invitrogen) and counterstained with Hoechst 33258 to visualize nuclei. Fluorescent staining was viewed with a Zeiss Axioskop 2 plus microscope or a Zeiss LSM 510 confocal microscope (Carl Zeiss MicroImaging, Inc.). Images were analyzed for intensities of nuclear KLF5 staining using Metamorph Imaging software (Molecular Devices).

Immunocytochemistry

Cells were grown in 4-well chamber slides cells, fixed in 4% ultrapure formaldehyde (Polysciences), and blocked in 3% bovine serum. Cells were stained overnight at 4°C with primary antibodies followed by staining with Alexa Fluor secondary antibodies (Invitrogen).

Reverse transcription-quantitative PCR

Total RNA was isolated from mouse distal jejunum, which was inverted and scraped to isolate mucosal tissue. First-strand cDNA was prepared using the SuperScript III First-Strand Synthesis System for reverse transcription-PCR (RT-PCR; Invitrogen). Primers used are listed in Supplementary Materials and Methods. PCR reactions were amplified and detected with the iCycler iQ real-time PCR Detection System (Bio-Rad Laboratories), using iQ SYBR Green Supermix (Bio-Rad Laboratories). Relative fold changes were calculated based on the comparative C_T (−ΔΔC_T) method (21).

Luciferase reporter assays

COS-1 cells at 80% to 90% confluence were transfected with pCI-neo, pCI-neo-β-catenin-S33Y-FLAG, pCI-neo-KLF5, or in combination using Lipofectamine 2000 (Invitrogen). Cells were additionally transfected with firefly luciferase reporter constructs: pGL2, pTOPFLASH or pA3-cyclin D1, and a Renilla luciferase reporter to normalize for transfection efficiencies. Luciferase activities were determined at 24 h posttransfection using the Dual Luciferase Assay System kit (Promega).

For small interfering RNA (siRNA) experiments, validated Stealth Select siRNAs for mouse KLF5 (1126 or 1127) were obtained from Invitrogen. HCT 116 or DLD-1 cells were transfected using siRNAMax (Invitrogen) with either nonspecific siRNA or siRNAs specific for KLF5, 1126, or 1127 (Invitrogen) and the firefly luciferase reporter, pTOPFLASH. Cells were assayed for luciferase activity 48 h after transfection.

Western blot analysis

Equal amounts of protein lysates were subjected to standard Western blotting procedures. Lysates from COS-1 cells were prepared using 1 × Cell Extraction Buffer (Biosource/Invitrogen). Whole cell lysates from mouse jejunum were prepared from 1-cm pieces of small intestine isolated 10 cm proximal to the cecum using Cell Disruption Buffer (Applied Biosystems/Ambion). Before lysates preparation, tissues were screened for absence of polyps.

Northern blot analysis

Total RNA was prepared from cells using Trizol reagent (Invitrogen), per manufacturer's instructions. Northern blots were carried out as previously described (13).

Coimmunoprecipitation

In COS-1 cells, nuclear extracts were prepared using the Nuclear Complex Co-IP kit (Active Motif) and immunoprecipitations were performed according to the manufacturer's protocol. For immunoprecipitation from DLD-1 lysates, cells were pretreated with 2 mmol/L dithiobis[succinimidylpropionate] to cross-link interacting proteins (22) before lysate preparation.

Subcellular fractionation

Fractionation of crude cell lysates was carried out as described previously (23).

Statistical analysis

Numerical values in text and figures are expressed as the mean ± SE. Data were evaluated for normality and homogeneity of variance by Kolmogorov-Smirnov and Levene's tests, respectively. When assumptions of normality and equal variance were met, multigroup data were analyzed statistically by ANOVA with multiple pair-wise comparisons between means performed using Tukey's tests. Data failing to meet the assumptions required for analysis by parametric statistical tests were evaluated using distribution-free methods. Differences between means were considered to be statistically significant when P value are <0.05. Probability values of greater than P values of 0.001 are reported as exact P values.

Haploinsufficiency of Klf5 inhibits intestinal adenoma formation in Apc^Min/+ mice

To investigate the role of KLF5 in the well-established Apc^Min/+ mouse model of intestinal tumorigenesis, we crossed Apc^Min/+ mice with Klf5^+/− mice, and compared adenoma formation between Apc^Min/+ mice and Apc^Min/+/Klf5^+/− double heterozygous mice. Mice were screened at age 16 to 18 weeks when intestinal adenomas were clearly visible or at age 28 weeks, nearing the end of the approximate life span of the Apc^Min/+ mice. As seen in Fig. 1A, at age 16 to 18 weeks, only 41% of Apc^Min/+/Klf5^+/− mice developed adenomas in the small intestine (n = 22), compared with 100% of Apc^Min/+ mice (n = 22). Furthermore, the number of adenomas per mouse was reduced by 96% in Apc^Min/+/Klf5^+/− mice, which developed an average of 1.5 ± 0.8 adenomas per mouse, compared with 37.7 ± 4.5 in Apc^Min/+ mice (Fig. 1B). Average numbers of colonic tumors in the Apc^Min/+/Klf5^+/− mice were not significantly different compared with the Apc^Min/+ mice (0.36 ± 0.14 versus 0.50 ± 0.16 adenomas per mouse, respectively). The reduction in tumor incidence in the double heterozygous Apc^Min/+/Klf5^+/− mice compared with the Apc^Min/+ mice was maintained as the mice aged. At 28 weeks, Apc^Min/+/Klf5^+/− mice had an average of 0.73 ± 0.38 small intestinal adenomas per mouse (n = 11) compared with 37.3 ± 1.8 in the Apc^Min/+ mice (n = 11; Fig. 1B). These results indicate that tumor formation was not merely delayed in the Apc^Min/+/Klf5^+/− mice but that haploinsufficiency of Klf5 prevents tumor initiation.

In addition to tumor numbers, tumor sizes were compared between the two genotypes. As seen in Fig. 1C, at 16 to 18 weeks, the mode for adenoma size was <1 mm for both Apc^Min/+ and Apc^Min/+/Klf5^+/− mice. By 28 weeks, the mode for Apc^Min/+ mice was increased to 1 to 2 mm, indicating growth of the tumors, yet the mode for Apc^Min/+/Klf5^+/− mice remained at <1 mm. These data suggest that Klf5 haploinsufficiency may affect tumor growth as well as tumor initiation.

KLF5 and β-catenin levels are reduced in the Apc^Min/+/Klf5^+/− mice

To confirm that KLF5 protein was reduced in the small intestine of Apc^Min/+/Klf5^+/− mice, KLF5 levels were examined by immunofluorescence (Fig. 2A). KLF5 was restricted to the nuclei of epithelial cells in proliferating crypts of the small intestine. In Klf5^+/− mice, KLF5 levels were significantly reduced to approximately half the levels in WT mice (P = 0.015) as determined by quantification of immunofluorescence intensities, correlating with haploinsufficiency of Klf5 (Fig. 2B). Similarly, KLF5 levels in Apc^Min/+/Klf5^+/− mice were reduced to approximately half those in Apc^Min/+ mice (P = 0.016). In the Apc^Min/+ mice, KLF5 showed intense staining in the transit-amplifying cells in the crypts of normal-appearing tissue (Fig. 2A), but intensities were not significantly increased compared with WT levels.

Because β-catenin is the key effector of APC mutations in intestinal cancers, we examined whether β-catenin levels and/or localization were influenced by Klf5 haploinsufficiency. Immunohistochemical staining of small intestinal tissues showed strong nuclear and cytoplasmic staining of β-catenin at the base of the crypts in the Apc^Min/+ mice, whereas nuclear staining was less pronounced in the WT, Klf5^+/−, and Apc^Min/+/Klf5^+/− mice (Fig. 2C). By Western blot analysis, β-catenin levels were higher in intestinal tissues from Apc^Min/+ mice compared with WT or Klf5^+/− mice (Fig. 2D), confirming previous findings that mutations in the Apc gene increase the stability and accumulation of β-catenin (7). Importantly, this effect was abrogated in Apc^Min/+/Klf5^+/− mice, where β-catenin levels were comparable with those in the WT and Klf5^+/− mice (Fig. 2C and D). Similarly, when comparing levels of β-catenin in nuclear extracts from intestinal tissues by Western blot analysis, β-catenin was reduced in Apc^Min/+Klf5^+/− mice compared with Apc^Min/+ mice (Supplementary Fig. S1), agreeing with results in Fig. 2C indicating reduced nuclear localization of β-catenin in the crypts of Apc^Min/+/Klf5^+/− mice. Thus, haploinsufficiency of Klf5 may “offset” the effect of the Apc^Min mutation on the β-catenin protein.

KLF5 interacts with β-catenin

To investigate the mechanisms by which KLF5 affects β-catenin protein levels and localization, KLF5 and β-catenin were coexpressed in COS-1 cells. Protein levels of β-catenin were higher in cells expressing both ectopic β-catenin and KLF5 compared with cells expressing β-catenin alone (Supplementary Fig. S2A). This affect was not due to higher transcript levels, as both endogenous and exogenous β-catenin mRNA levels were unchanged in the presence of ectopic KLF5 (Supplementary Fig. S2B). Another potential mechanism by which KLF5 could affect β-catenin protein levels is through physical interaction and stabilization. Therefore, immunoprecipitation experiments were conducted in cells expressing KLF5-HA and β-catenin-FLAG to determine whether the two proteins interact. β-Catenin-FLAG coprecipitated with KLF5-HA, and likewise, KLF5-HA immunoprecipitated with β-catenin-FLAG (Fig. 3A). Neither protein was precipitated with the IgG antibody control. To verify interaction of endogenous β-catenin and KLF5 proteins, KLF5 was immunoprecipitated from DLD-1 human colon cancer cell lysates. β-catenin coprecipitated with KLF5 but was not precipitated in the preimmune serum negative control (Fig. 3B).

To further evaluate the interaction between β-catenin and KLF5, the two proteins were examined for colocalization by immunofluorescence staining. Expression of KLF5-HA alone showed intense, diffuse staining of KLF5 throughout the nucleus, whereas β-catenin-FLAG-alone transfected cells showed both cytoplasmic and nuclear staining of β-catenin (Fig. 3C). When KLF5 and β-catenin were coexpressed, KLF5 was redistributed within the nucleus in a pattern that corresponded to β-catenin localization. In addition, there was an increase in nuclear staining for β-catenin in the cotransfected cell (Fig. 3C). Redistribution of KLF5 was also seen for endogenous KLF5 in cells transfected with β-catenin-FLAG alone. Although KLF5 exhibited diffuse nuclear staining in cells lacking ectopic β-catenin (Fig. 3D, orange arrow), KLF5 was redirected to distinct intranuclear structures in cells expressing β-catenin–FLAG (Fig. 3D, white arrow). This altered localization of KLF5 correlated with the expression pattern of ectopic β-catenin.

KLF5 facilitates nuclear localization of β-catenin

To investigate in-depth the ability of KLF5 to promote nuclear localization of β-catenin, cells transfected with either β-catenin alone or β-catenin and KLF5 were analyzed for subcellular distribution of β-catenin by immunofluorescence and protein fractionation. In examining a field of cells stained for β-catenin and KLF5, β-catenin exhibited more intense nuclear localization in the presence of ectopic KLF5 than when expressed alone (Fig. 4A). β-catenin localization was also analyzed by subcellular fractionation, with protein lysates separated into nuclear, cytoplasmic, and membrane fractions (Fig. 4B). Amounts of β-catenin localized to the membrane fraction were approximately equal in the two sets of transfections. However, a significant proportion of β-catenin was redistributed from the cytoplasm to the nucleus upon coexpression with KLF5. Thus, cells with ectopic expression of both β-catenin and KLF5 had higher levels of nuclear β-catenin.

To determine whether loss of KLF5 would have a reciprocal effect on β-catenin localization, HCT116 human colon cancer cells were transfected with KLF5-specific siRNA and examined for sub-cellular localization of β-catenin by immunofluorescence (Fig. 4C). In cells transfected with nonspecific siRNA, β-catenin exhibited visible nuclear staining. However, this localization was in large part disrupted in cells transfected with two different KLF5-specific siRNAs, where β-catenin was localized predominantly to the cell membranes and cytoplasm. A similar effect was seen in another colon cancer cell line, DLD-1 (Fig. 4D), Taken together, these data indicate that KLF5 facilitates localization of β-catenin to the nucleus.

KLF5 modulates transcriptional activity of β-catenin

Because increased expression of KLF5 enhances nuclear localization of β-catenin, KLF5 may also promote the transcriptional activity of β-catenin. To test this possibility, KLF5 and β-catenin were coexpressed in COS-1 cells with pTOPFLASH, a luciferase reporter construct that measures TCF/β-catenin activity (8). Compared with empty vector, cells transfected with β-catenin alone showed a 3.6-fold increase in TOPFLASH luciferase activity (P = 0.003), whereas ectopic expression of KLF5 alone had no effect (Fig. 5A). Cells transfected with both β-catenin and KLF5 showed a 6.2-fold increase in luciferase activity, an effect that was significantly greater than with β-catenin alone (P = 0.006). As a transcriptional target for both KLF5 and β-catenin (24, 25), Ccnd1 promoter activity was also measured (Fig. 5A). Transfection with β-catenin alone did not significantly increase Ccnd1 luciferase activity compared with cells transfected with empty vector. In cells transfected with KLF5 alone, luciferase activity was increased by 3.2-fold over vector control (P < 0.001). Coexpression of β-catenin and KLF5 resulted in a 7.6-fold increase in activity, which was significantly greater than with β-catenin (P < 0.001) or KLF5 alone (P < 0.001). Moreover, when β-catenin and KLF5 were coexpressed, they interacted to produce a synergistic effect on Ccnd1 promoter luciferase activity (P < 0.001). Thus, coexpression of KLF5 with β-catenin enhances the transcription of β-catenin targets.

To determine if loss of KLF5 would have a corresponding negative effect on TCF/β-catenin transcriptional activity, HCT 116 and DLD-1 cells were transfected with a nonspecific siRNA or two different siRNAs specific for KLF5 and the TOPFLASH luciferase reporter (Fig. 5B). Transfection with either KLF5-specific siRNA dramatically reduced TOPFLASH activity in both cell lines compared with transfection with nonspecific siRNA (P < 0.001), correlating with loss of β-catenin from the nucleus in Fig. 4C and D.

Expression of β-catenin targets is reduced in Apc^Min/+/Klf5^+/− mice

To address whether reduced KLF5 expression affects transcriptional activity of β-catenin in vivo, RNA was isolated from the jejunum of WT, Klf5^+/−, Apc^Min/+, and Apc^Min/+/Klf5^+/− mice, and expression of two transcriptional targets of β-catenin that promote proliferation were examined by reverse transcriptase-quantitative PCR. Transcript levels of Ccnd1, the gene encoding cyclin D1, were increased 1.9-fold in the Apc^Min/+ mice compared with WT (P = 0.009; Fig. 5C). In contrast, levels of Ccnd1 in the Apc^Min/+/Klf5^+/− mice were similar to WT controls (Fig. 5C). As a transcriptional target of KLF5 (24, 25), Ccnd1 transcript levels were significantly reduced in the Klf5^+/− mice compared with WT (0.41-fold; P = 0.03), suggesting that this target gene is strongly influenced by KLF5 levels. In examining c-Myc transcript levels, c-Myc was increased 1.4-fold (P = 0.002) in Apc^Min/+ mice compared with WT mice. This change was offset in the Apc^Min/+/Klf5^+/− mice, with levels comparable with WT (P < 0.001). No significant change in c-Myc levels was observed in the Klf5^+/− mice. Target gene expression was further confirmed by immunohistochemical staining of the protein products (Fig. 5D). In concordance with mRNA levels, cyclin D1 staining was more intense in small intestinal tissues of Apc^Min/+ mice compared with WT mice (Fig. 5D), whereas no induction of cyclin D1 was observed in the Apc^Min/+/Klf5^+/− mice. Likewise, c-Myc expression was increased in the Apc^Min/+ mice, but this effect was abrogated in the Apc^Min/+/Klf5^+/− animals, which exhibited c-Myc levels similar to WT (Fig. 5D). Therefore, KLF5 is necessary for the downstream activation of β-catenin target genes in the Apc^Min/+ mouse.

KLF5 and β-catenin colocalize in intestinal adenomas

Because KLF5 seems to contribute to the tumorigenic effects of β-catenin in the Apc^Min/+ setting, immunohistochemical analyses were performed to determine the distribution of KLF5 and β-catenin in intestinal tumors derived from Apc^Min/+ mice. As seen in Fig. 6A, there was intense nuclear staining of KLF5 in epithelial cells in the dysplastic epithelium adjacent to the adenomas as well as cells throughout the adenomas. β-catenin exhibited strong cytoplasmic staining throughout the adenomas, and showed a high degree of nuclear staining, particularly in cells at the leading edges of the adenomas. To examine colocalization of KLF5 and β-catenin in Apc^Min/+ adenomas, tissues were stained by immunofluorescence (Fig. 6B). Both KLF5 and β-catenin showed nuclear staining in a majority of the cells in the adenomas, with considerable colocalization of the two proteins in individual cells. Thus, KLF5 and β-catenin are coexpressed in Apc^Min/+ adenomas, and can work together to drive growth-promoting transcriptional activity in those tissues.

When comparing β-catenin levels in the adenomas from Apc^Min/+ and Apc^Min/+/Klf5^+/− mice by immunofluorescence, no significant differences were observed in β-catenin intensities in small-size adenomas (<1 mm) from the two genotypes (Supplementary Fig. S3). Both sets of adenomas exhibited nuclear localization of β-catenin. The only subjective difference seemed to be a more diffuse cytoplasmic localization of β-catenin in the Apc^Min/+ mice compared with a more intense localization to cell junctions in the Apc^Min/+/Klf5^+/− mice. These results reinforce the notion that haploinsufficiency of Klf5 affects adenoma formation at the earliest stages; and that these effects can be bypassed at low frequency to result in high β-catenin levels and subsequent establishment of adenomas.