To study the loss of function of Hltf in vivo, we mutated Hltf in mice by homologous recombination. A gene-targeting vector was generated to replace the exons 1-5 coding for residues 2-208 of Hltf with a nuclear localized LacZ (nls-LacZ) cDNA that was fused in-frame with the endogenous Hltf starting codon (Figure 1A). Correct homologous recombination in the targeted ES cells was confirmed by Southern blot analysis using both 5' and 3' probes located upstream and downstream of the targeted Hltf sequence (Figure 1B and 1C). Two independent ES clones were used to produce germline-transmitting chimeras that were further bred with a ubiquitous Cre transgenic line (EIIa-Cre) to delete the loxP flanked pGK-neo selection marker from the targeted locus (Figure 1A). The resultant Hltf ^+/- mouse, which was maintained on either a C57BL/6 or C57BL/6/129S1 background, was used to produce Hltf homozygous mutant mice for phenotypic characterization.

To determine whether Hltf expression is completely inactivated in Hltf ^-/- mice, we performed Northern blot analysis of total RNA prepared from control and mutant E10.5 embryos. Whereas Hltf transcripts were identified in control (wild-type and Hltf ^+/-) embryos using the 5' and 3' Hltf cDNA probes, transcripts were not detected in Hltf ^-/- samples (Figure 1D). To further demonstrate the absence of Hltf expression in Hltf ^-/- mice, we also applied a more sensitive real time RT-PCR assay to examine the expression of Hltf in derived Hltf ^-/- mouse embryonic stem (ES) cells or mouse embryonic fibroblast (MEF) cells as well as in Hltf ^-/- E10.5 embryos. With the primer set located on the 3' of Hltf cDNA, a strong amplification of Hltf was detected in the control group, but no signal was found in Hltf ^-/- samples (Figure 1E). These data clearly indicate that the targeted mutation that we introduced into the Hltf mouse genomic locus gave rise to a null allele.

With the established Hltf null allele, we then asked whether loss of Hltf function could affect mouse development and induce the formation of intestinal tumors. As a first approach to address this, we determined the expression pattern of Hltf during development. We took advantage of the inserted nls-LacZ reporter gene in Hltf knockout mice, whose expression is controlled by the endogenous Hltf regulatory elements. Both Hltf ^+/- and Hltf ^-/- mice showed the same LacZ expression pattern. By determining LacZ activity in these mice, we found that Hltf was specifically expressed in the heart at an early developmental stage (E8.5 to E9.5) (Figure 2A). Hltf exhibited a broader expression pattern at E10.5, with LacZ signals detected in somites, branchial arches, limb bud and brain (Figure 2B). At later embryonic developmental stages, such as E16.5, Hltf showed wide and strong expression in many tissues, including heart, lung, liver, kidney, spleen and pancreas (Figure 2C). This wide-spread expression of Hltf was also observed in adult mice (data not shown). In both adult intestine and colon, Hltf expression was mainly detected in the crypts and in the intestinal epithelial cells (Figure 2D).

This broad expression pattern of Hltf during mouse development might suggest that Hltf is important for development. However, Hltf ^-/- mice were found to have normal embryonic and postnatal development. The Hltf null mice were fertile, and exhibited similar a body weight and life span as their wild-type (wt) littermates. In addition, Hltf ^-/- mice did not display obvious abnormalities of intestinal proliferation or differentiation as determined by the BrdU incorporation assay as well as by the analyses with different intestinal cell lineage markers (Additional file 1 and Additional file 2). To determine whether loss of Hltf function could induce the formation of intestinal tumors, we monitored a cohort of 60 mice fed with a normal diet for a two-year period (30 Hltf ^-/- mice on C57BL/6 background and 30 wt mice matched for age and genetic background). Similar to the wt control group, 10% of Hltf ^-/- mice (3 out of 30) developed liver tumors or lymphomas between 16 and 24 months. However, none of Hltf ^-/- mice formed intestinal tumors or gastric tumors.

Taken together, our results demonstrate that Hltf is dispensable for normal development. Our work also indicates that loss of Hltf function by itself is insufficient to drive the oncogenic process in the gastrointestinal tract, which implicates that the epigenetic inactivation of HLTF, as commonly found in human colon cancers, could be involved in the late stage of intestinal tumor development.

To determine whether the loss of HLTF function could have a role in the progression of intestinal tumors, we introduced the Hltf null mutation into Apc^min/+ mice. The Apc^min/+ mice harbor a premature stop codon in one allele of the Apc tumor suppressor gene, and develop multiple intestinal adenomas that mimic human familial adenomatosis polyposis 27. The adenomas in Apc^min/+ mice rarely progress to invasive intestinal or colon cancers 28, making this mouse model an excellent genetic tool for studying the later events involved in intestinal carcinogenesis.

To avoid strain-specific modification of the Apc^min/+ intestinal tumor phenotype 2930, we generated Hltf ^-/-/Apc^min/+ mice on the C57BL/6 background. Hltf ^-/-/Apc^min/+ mice became moribund at a similar age as Apc^min/+ mice. Post-mortem examination revealed that Hltf ^-/-/Apc^min/+ and Apc^min/+ mice developed similar numbers of tumors in the small intestine (average of 65 macroscopic tumors/per mouse in both groups). However, upon histological characterization, most Hltf ^-/-/Apc^min/+ mice (24 out of 30) were found to form invasive intestinal adenocarcinomas characterized by deeper invasion of tumor cells into the muscularis propria (Figure 3B and Additional file 3). The invaded neoplastic glandular cells exhibited high β-catenin activity as reflected by the nuclear accumulation of β-catenin detected by immmuno-staining with anti-β-catenin antibody (Figure 3B). In contrast, most Apc^min/+ mice developed only well-defined tubular adenomas or adenomas within the lamina propria (Figure 3A).

Hltf ^-/-/Apc^min/+ mice were also found to develop a similar number of neoplastic lesions in the colon as Apc^min/+ mice (average of 1.5 macroscopic tumors/per mouse). Histologically, although both Hltf ^-/-/Apc^min/+ and Apc^min/+ colon tumors showed tumor growth with a pedunculated morphology protruding into the colonic lumen, the majority of Hltf ^-/-/Apc^min/+ colon tumors displayed a more dramatic glandular atypia that resulted in the formation of numerous mucin-filled cysts (Figure 4A, B), a characteristic pathological feature for human colon cancers 31. In addition, most Hltf ^-/-/Apc^min/+ colon tumors were also found to have a strong desmoplastic stromal reaction, suggesting invasion of the lamina propia (Figure 4B). Furthermore, Hltf ^-/-/Apc^min/+ colon tumors were also found to have stronger and more ubiquitous nuclear-stained β-catenin signals than the tumor cells in Apc^min/+ mice (Figure 4C). These data indicate that colon tumors from Hltf ^-/-/Apc^min/+ mice had progressed to much higher grades than those from Apc^min/+ mice. To further demonstrate the malignancy of colon tumors developed in Hltf ^-/-/Apc^min/+ mice, we derived cells from these tumors and then subcutaneously injected them into Rag1^-/-/IL2^-/- immunodeficient mice. Within 30 days after injection, these cells formed subcutaneous tumors that maintained the morphological and immunohistochemical features of the parental colon tumors (Figure 4D). This capability of propagating tumor cells into another host, which was not observed with cells derived Apc^min/+ colon tumors, further supports that many colon tumors developed in Hltf ^-/-/Apc^min/+ mice could be malignant cancers.

Collectively, our results demonstrate that loss of Hltf function is able to promote intestinal or colonic carcinogenesis in mice with a mutant Apc background. This finding, together with human epigenetic evidence that HLTF is frequently silenced in the advanced human colon adenomas or cancers but uncommonly in early adenomas 1719, indicates that epigenetic inactivation of HLTF could be an important event involved in the transition of benign adenomas to malignant colon cancer.

Chromosomal instability (CIN) is a genetic hallmark of most human colon cancers, and it has been demonstrated to be important for the progression of colon tumors by increasing the rate of genetic aberration 23233. Since HLTF has been shown to be required for the maintenance of genomic stability 1213141516, it is likely that intestinal or colonic tumors developed in Hltf ^-/-/Apc^min/+ mice would have increased genomic instability. To examine this, we analyzed metaphase spreads generated from tumor cells (passage 2) derived from colon tumors developed in Apc^min/+ and Hltf ^-/-/Apc^min/+ mice. These tumors all showed pronounced cribiform tumor growth without invasive neoplastic glands and were histologically characterized as late stage colonic adenomas.

Consistent with previous findings 3435, colon tumor cells from Apc^min/+ mice showed a near-diploid karyotype and did not harbor obvious chromosomal abnormalities (Figure 5A). In contrast, three colon tumor cell lines derived from Hltf ^-/-/Apc^min/+ mice displayed an aneuploid phenotype (Figure 5B and 5C). Spectral karytotype (SKY) analysis on these tumor cells further revealed a high incidence of genetic alterations in the forms of nonreciprocal translocations, chromosomal fusions (which include both dicentric chromosomes and centromeric fusions), and chromosomal fragments and breaks (Figure 5D-G and Table 1). In addition, many chromosomal fusions in Hltf ^-/-/Apc^min/+ colon tumor cells were found to lack telomere signals (Figure 5I), indicating the presence of telomere dysfunction in these tumor cells. Collectively, these results demonstrate that Hltf ^-/-/Apc^min/+ colon tumors had gross chromosomal abnormalities, similar to the CIN phenotype as described in most human colorectal cancers.

To determine whether loss of HLTF function would also lead to a similar tumor-promoting effect in human colon cancer as we demonstrated in the Hltf ^-/-/Apc^min/+ mouse model, we applied a lentiviral based shRNA knockdown approach to down-regulate HLTF expression in HCT116 human colon cancer cells. HCT116 cells are one of a few colon cancer cell lines that lack HLTF promoter methylation 17, and express high levels of HLTF (Figure 6A). In addition, these cells are near-diploid with a stable karyotype 36, which would also allow us to determine whether loss of HLTF function in human colon cancer cells could induce chromosomal abnormalities in a manner similar to the Hltf-deficient mouse colon tumor cells.

As shown in Figure 6B, HLTF expression was significantly reduced to almost undetectable protein level in the transduced HCT116 cells that were stably integrated with HLTF shRNAs (HCT116^HLTFshRNA1 and HCT116^HLTFshRNA2) as compared to the scramble shRNA control (HCT116^Scramble). To determine whether this down-regulation of HLTF expression was able to modulate HCT116 tumor growth, HCT116 cells with or without HLTF knockdown were injected subcutaneously into Rag1^-/-/IL2^-/- immunodeficient mice. All the mice with injected tumor cells (10⁵ cells/per mouse) developed subcutaneous tumors within 20 days. However, the tumors formed by HCT116^HLTFshRNA1 or HCT116^HLTFshRNA2 cells showed significantly increased tumor size as compared with that of the scramble control (Figure 6C). Moreover, both HCT116^HLTFshRNA1 and HCT116^HLTFshRNA2 subcutaneous tumors also frequently exhibited the invasion of tumor cells into the subcutaneous muscle layer that was uncommonly observed in the control tumors (Figure 6D). To confirm that HLTF expression was indeed down-regulated in these tumors, we performed immunohistochemistry with an anti-HLTF antibody. Both HCT116^HLTFshRNA1 and HCT116^HLTFshRNA2 subcutaneous tumors showed significantly lower expression of HLTF as compared to the scramble control tumors (Figure 6D). Taken together, our data indicate that down-regulation of HLTF in human colon cancer cells may promote tumor growth and malignancy, similar to what we found in the Hltf ^-/-/Apc^min/+ mouse model.

We then performed cytogenetic analysis to determine whether down-regulation of HLTF expression in human colon cancer cells induced chromosomal abnormalities. Detailed SKY analysis of metaphase spreads from both short term (10 days) cultured HCT116 parental cells and HCT116^scramble cells revealed the presence of near-diploid chromosomes with a few highly conserved chromosomal translocations (Figure 7A), consistent with the documented karyotype for HCT116 parental cells 37. This near-diploid karyotype was also observed in the second generation of HCT116 cells and HCT116^scramble cells that were derived from the subcutaneous tumors formed by these cells (Figure 7C), further indicating that HCT116 cells have a stable chromosome karyotype as demonstrated previously 3839. In contrast, both short-term cultured and the second generation of HCT116^HLTFshRNA1 and HCT116^HLTFshRNA2 tumor cells exhibited significantly increased chromosomal abnormalities in the forms of chromosomal loss and chromosomal fusions and breaks as compared to the control groups (p < 0.001) (Figure 7B-C). In addition, approximately 15% of metaphases from HCT116^HLTFshRNA1 and HCT116^HLTFshRNA2 tumor cells showed high numbers of trisomy (Figure 7D). These results clearly indicate that down-regulation of HLTF in human colon cancer cells is able to induce chromosomal abnormalities.

The Hltf gene-targeting vector was generated based on a PCR-based cloning strategy 47. Briefly, the mouse Hltf genomic fragments required for the 5' and 3' arms of homology were PCR-amplified from the genomic DNA of R1 ES cells (on 129S1 background) with a high-fidelity polymerase (TaKaRa). The primer sequences are listed in Additional file 5. After validation by DNA sequencing, the Hltf 5'arm DNA fragment was ligated upstream of a nlsLacZpA-loxP-pGKneo-loxP cassette. This vector was further inserted by a DNA fragment containing Hltf 3'arm and pGK-DTA cassette to generate the final gene-targeting vector that contains Hltf 5'arm-nlsLacZpA-loxP-pGKneo-loxP-Hltf 3'arm-pGK-DTA (Figure 1A).

The Hltf gene-targeting vector was linearized with SacII and then electroporated into R1 ES cells. The transfected ES cells were selected with G418 (250 μg/ml), and G418-resistant ES clones were screened by Southern blot analysis for the correctly targeted allele using BamHI (for the 5' external probe) and EcoRI (for the 3' external probe) digestions. Both the 5' and 3' external probes were PCR amplified with primers listed in Additional file 5. Two independently targeted ES cell clones were used to generate chimeric mice by ES cell⇔diploid embryo aggregation. Transmitting chimeric males were crossed with an EIIa-Cre line (Jackson laboratories, Bar Harbor, Maine) to delete the loxP flanked neo cassette from the targeted allele. The resultant mouse allele was further bred with 129S1 females to produce hemizygous transgenic offspring on 129S1/C57BL/6 background. Hltf ^+/- mice were also backcrossed to C57BL/6 for eight generations. All mouse experiments were performed in accordance with procedures approved by the University of Manitoba Animal Care and Use Committee.

A PCR based genotyping method was applied to genotype Hltf knockout mice. Primers to amplify the targeted allele were the sense primer (P1) (5'-GGAGCTTTATCAGGCTGTCTGGGA-3') and antisense primer (P2) specific for the LacZpA cassette (5'-AGGAAGATCGCACTCCAGCCAGCT-3'). To detect the wild-type HLTF allele, wt P1 primer (5'-GTCCATGTCCTAGCCATGAGTA-3') and an antisense primer (P3) locating in exon 1 (5'-GCTTGGTAAGGACTACAAAGCA-3') were used for PCR.

The Apc^Min/+ mice were purchased from The Jackson Laboratory and maintained on a C57/B6 genetic background. These mice were bred with Hltf ^+/- (C57/B6 background) to generate a cohort of Hltf ^-/-/Apc^min/+, Apc^min/+ and Hltf ^+/-/Apc^min/+ mice for intestinal tumor analysis. Mice were sacrificed by cervical dislocation, and the entire small intestine and colon were flushed with cold PBS and dissected longitudinally. The number, location and size of tumors in the intestine or colon were examined and recorded. Small intestinal and colonic tumors were further histologically analyzed by hemtoxylin/eosin staining and immunohistochemistry.

Tumor cells from Hltf ^-/-/Apc^min/+ and Apc^min/+ colon tumors were generated as described 35. The cells were grown in 6 cm culture dishes plated with mitomycin-treated mouse embryonic fibroblast (MEF) cells using the culture conditions as described 35. The early passage (p2) cells were collected either for tumorigenicity analysis by injecting 10⁵ cells subcutaneously to Rag1^-/-/IL2^-/- mice (purchased from Jackson Laboratory), or for cytogenetic characterization.

IHC was performed as described previously 48. Briefly, dissected intestinal and colonic tumors were fixed in 10% formalin in PBS overnight. The fixed tissues were embedded in paraffin, and 5 μm paraffin-embedded tissue sections were pretreated with Retrieval solution (Dako) and blocked with either mouse IgG blocking reagent (Vector Laboratory) or serum free blocking reagent (Dako), and then incubated with primary antibodies overnight at 4°. Primary antibodies used were mouse anti-BrdU (Roche, dilution 1:100), mouse anti-Ki67 (BD Pharmingen, dilution 1:100), mouse anti-β-catenin (BD Pharmingen, dilution 1:200), rabbit anti-HLTF (Sigma, dilution 1:100) rabbit anti-lysozyme (Dako, dilution 1:200), and rabbit anti-chromogranin A (ImmunoStar, dilution 1:4000). After washing with TBS containing 0.1% Tween20 buffer, sections were incubated with biotinylated anti-rabbit or anti-mouse antibody for 60 min at room temperature. The sections were washed 3 times with TBS-0.1% Tween20 buffer and incubated with avidin-biotin-peroxidase complex (Vector Laboratory), according to the manufacturer's protocol. Color was developed using metal 3,3'-diaminobenzidine tetrachloride.

Lentiviral vectors expressing shRNA against human HLTF (NM 003071, V2LHS 254799, _153142, _153143, _153144, _153146) or scrambled control were obtained from the GIPZ lentiviral shRNAmir library (Open Biosystems, Thermo Scientifc) acquired at the University of Manitoba. The pGIPZ lentiviral vector expresses a short hairpin RNA modeled by human microRNA-30 (miR30) primary transcript for increased Drosha and Dicer processing. TurboGFP reporter gene and puromycin drug resistance marker are expressed as part of a bicistronic transcript. Preparation and infection of lentiviral particles into HCT116 cells (from ATCC) were performed based on an established procedure as described previously 49. After infection, HCT116 cells were treated with 0.75 μg/ml puromycin for four days. The puromycin-resistant clones were pooled to determine HLTF expression by Western blot analysis with anti-HLTF antibody (Sigma). 10⁵ cells from these established stable cell lines were subcutaneously injected to Rag1^-/-/IL2^-/- mice to determine the effect of HLTF knockdown on tumor growth. These cells, as well as cells derived from subcutaneous tumors, were also utilized for cytogenetic analysis.

Metaphase spreads for SKY analysis were performed using the SKT kit (ASI) according to the manufacturer's instructions and previously published methods 50. In each cell line, at least 20 metaphase spreads were studied. Q-FISH analysis was performed as previously described 51.

Total RNA from freshly dissected mouse E10.5 embryos was extracted using TRIzol (Life Technologies, Inc.). 20 μg of total RNA was separated on a 1% agarose-formaldehyde gel and transferred to Hybond nylon membrane (Amersham). Hybridization was carried out in PerfectHyb (Sigma) with 1.5 × 10⁶ cpm/ml of 5' and 3' Hltf cDNA probes. Both 5' and 3' cDNA probes were PCR-amplified with the primers listed in Additional file 5.

Dissected mouse embryos or tissues were fixed for 30 min (for E8.5 to E9.5) or 60 min (for E10.5 embryos and tissues) with 4% paraformaldehyde (PFA), 0.2% glutaraldehyde and 0.02% NP-40 in PBS. Subsequently, fixed samples were washed three times with PBS containing 0.02% NP-40, and stained at 37°C overnight with a staining solution containing 4 mM K₄Fe(CN)₆, 4 mM K₃Fe(CN)₆, 2 mM MgCl₂, and 0.2% X-gal in PBS.

For Western blots, cells were lysed in RIPA buffer. 30 μg lysate was separated by 8% SDS-PAGE and transferred to nitrocellulose membrane, which was blocked with 5% non-fat milk and incubated with rabbit anti-HLTF antibody (Sigma, dilution 1:500) at 4° overnight. Protein was detected using an enhanced chemiluminescence system (Amersham Life Science).

A paired t-test (two tailed) was used to analyze the statistical significance of chromosomal abnormalities in tumor cells. P < 0.05 was considered indicative of statistical significance.