The Krüppel-associated Box Repressor Domain Can Induce Reversible Heterochromatization of a Mouse Locus in Vivo

Background: The KRAB module mediates ectopic and drug-controllable transcriptional repression.

Results: Targeting of KRAB to a mouse gene body results in reversible heterochromatization and gene silencing in adult and embryonic cells.

Conclusion: KRAB binding to gene bodies does not induce stable DNA promoter methylation as previously thought.

Significance: These proof-of-principle experiments provide the basis for the development of novel KRAB-based tools in vivo.

Abstract

The study of chromatin and its regulators is key to understanding and manipulating transcription. We previously exploited the Krüppel-associated box (KRAB) transcriptional repressor domain, present in hundreds of vertebrate-specific zinc finger proteins, to assess the effect of its binding to gene bodies. These experiments revealed that the ectopic and doxycycline (dox)-controlled tet repressor KRAB fusion protein (tTRKRAB) can induce reversible and long-range silencing of cellular promoters. Here, we extend this system to in vivo applications and use tTRKRAB to achieve externally controllable repression of an endogenous mouse locus. We employed lentiviral-mediated transgenesis with promoterless TetO-containing gene traps to engineer a mouse line where the endogenous kinesin family member 2A (Kif2A) promoter drives a YFP reporter gene. When these mice were crossed to animals expressing the TetO-binding tTRKRAB repressor, this regulator was recruited to the Kif2A locus, and YFP expression was reduced. This effect was reversed when dox was given to embryos or adult mice, demonstrating that the cellular Kif2A promoter was only silenced upon repressor binding. Molecular analyses confirmed that tTRKRAB induced transcriptional repression through the spread of H3K9me3-containing heterochromatin, without DNA methylation of the trapped Kif2A promoter. Therefore, we demonstrate that targeting of tTRKRAB to a gene body in vivo results in reversible transcriptional repression through the spreading of facultative heterochromatin. This finding not only sheds light on KRAB-mediated transcriptional processes, but also suggests approaches for the externally controllable and reversible modulation of chromatin and transcription in vivo.

Introduction

The sequencing and annotation of vertebrate genomes has revealed that Krüppel-associated box (KRAB)⁵-containing zinc finger proteins (KRAB-ZFPs) constitute the largest class of transcriptional repressors encoded by higher species (1–4). The hundreds of mouse and human KRAB-ZFPs all encompass an N-terminal KRAB domain and C-terminal C₂H₂ zinc fingers that are predicted to interact with DNA in a sequence-specific manner (5, 6). At least some if not all KRAB-ZFPs can mediate transcriptional repression by recruiting the co-regulator KAP1 (also known as TRIM28, TIF1b, or KRIP-1) via their KRAB domain (7–9). Upon KRAB-ZFP-mediated tethering to specific genomic loci, KAP1 indeed acts as a scaffold for the recruitment of different heterochromatin-inducing factors and complexes, such as heterochromatin protein 1 (HP1), the H3K9me3-histone methyltransferase SETDB1 and the nucleosome remodeling and deacetylase complex NuRD (10–14). This formation of facultative heterochromatin is further characterized through the loss of histone acetylation and the increase of histone 3 lysine 9 trimethylation (H3K9me3) (15, 16).

Targeting of an ectopic KRAB domain to promoters of interest results in transcriptional repression (17, 18). This finding was exploited to develop systems for conditional gene expression and knockdown. Specifically, the externally regulated binding of KRAB to promoters can be achieved through the use of the ectopic repressor tTRKRAB, where the KRAB domain of human KOX1 is linked to the DNA binding tet repressor (tTR) protein found in Escherichia coli. The resulting tTRKRAB repressor binds to bacterial-specific tet operator (TetO) repeats with a high affinity that is reversed by the addition of doxycycline (dox). Finally, a drug-dependent system for conditional promoter activity is created by placing heterologous TetO sequences near sites of interest and by adding the tTRKRAB repressor. Previous results demonstrated that different components of this system can be delivered by retro- or lentiviral (LV)-based vectors (19–21), resulting in stable and dox-controllable gene or shRNA expression in cell lines (20) and primary tissues (22, 23).

An early study indicated that transgenic mice generated by the perivitelline injection of LV vectors exhibited irreversible silencing of TetO-containing constructs if tTRKRAB was left to bind to these sequences during early embryogenesis (22). Subsequent work revealed that, in these circumstances, the tethering of KRAB to the vicinity of a promoter led to its de novo DNA methylation, and hence to its stable silencing even in the absence of repressor binding (24). Importantly, in these cases KRAB was targeted to the vicinity of a CpG-island containing promoter in the context of a lentiviral vector. Therefore, we set out to test the idea whether targeting of KRAB to the body of an endogenous gene may have a different effect. For this, we made use of a system, which we previously developed to investigate the effect of KRAB/KAP1 recruitment on euchromatic gene bodies in cell culture. Here, the ectopic tTRKRAB repressor is targeted to TetO-carrying gene traps in a dox-dependent way (25, 26). This so-called “Trapping/Silencing” or “TrapSil” approach uses retroviral and LV vector-based gene traps, which carry a promoterless reporter gene. Hence, the reporter is only expressed if the activity of a transcribing endogenous promoter is “trapped” after proviral integration. TetO-repeats inserted within the gene trap further allow us to monitor the effect of tTRKRAB-induced “silencing” on these genomic sites. Using this system at hundreds of trapped loci in cultured cells, we found that efficient KRAB/KAP1-mediated silencing preferentially takes place in gene-rich and active genomic environments (26). In addition, molecular analyses of cellular “TrapSil” clones revealed that, in contrast to former belief, KRAB/KAP1 can mediate reversible long-range transcriptional repression through the spread of H3K9me3-containing heterochromatin (25). The model emerging from this data, suggests that KRAB/KAP1 initially recruit heterochromatic factors, such as the H3K9me3-inducing SETDB1, and the H3K9me3-binding HP1 to induce the spreading of heterochromatin in a self-propagating loop. Finally, affected promoters are repressed through these increased heterochromatic levels, which impair proper RNA Pol II recruitment and transcriptional initiation (25).

In the present study, we wanted to characterize the effect of KRAB/KAP1 recruitment to euchromatic gene bodies in vivo. For this, we generated a “TrapSil” mouse line in which the kinesin family member 2A (Kif2A) gene is trapped and drives the expression of a YFP reporter (Venus). When a tTRKRAB transgene was left to bind to the Kif2A locus, its transcriptional activity decreased. Moreover, dox treatment not only reversed silencing in early embryos but also in adult mice, suggesting the absence of stable tTRKRAB-induced repression. Correspondingly, analyses of embryonic fibroblasts demonstrated the absence of KRAB-mediated DNA methylation of the trapped Kif2A promoter, but instead revealed that binding of the trans-repressor led to the spread of facultative H3K9me3-containing heterochromatin at the Kif2A locus. Our results, therefore, indicate that KRAB can reversibly induce both heterochromatin spreading and transcriptional repression in vivo when it is tethered to the body of genes.

EXPERIMENTAL PROCEDURES

LV Vectors

The LV based gene trap vector M (kindly provided by Jens-Erik Dietrich and Takashi Hiiragi) was modified through addition of a unique XbaI site by site-directed mutagenesis (Stratagene XL QuickChange mutagenesis kit) upstream of the 3′-LTR. The 3′-LTR was then excised by restriction digest with XbaI and SnaBI and replaced with the tetO containing LTR of pLVTH (20). Importantly, the H1 promoter of the pLVTH 3′ SIN-LTR was excised with EcoRI and ClaI beforehand, while the remaining vector was religated. The modified 3′-LTR was then sequentially digested with SalI, filled in with Klenow and then digested with XbaI before being ligated into the M gene trap. The resulting vector was called M-LVT. The procedure of LV particle production by transient transfection in 293T cells is described elsewhere (27).

Mice and LV Vector-mediated Transgenesis

We generated transgenic mice by perivitelline injection of fertilized oocytes, as described in Refs. 22, 28, 29. A highly concentrated LV gene trapping vector was used. More specifically, the M-LVT virus had a titer of 2 × 10⁸ viral genomes per ml in HCT116 cells and an infectious particle to physical particle ratio of 1:1400. Genotyping of the offspring was done with the TissueDirect Multiplex PCR system (Genscript) on DNA extracted from ear biopsies or blood samples. PCR condition were: 5 min at 95 °C, 35 cycles including 45 s at 95 °C, 45 s at 58 °C and 45 s at 72 °C, followed by a final extension of 5 min at 72 °C. The primers used can be found in supplemental Table S2. All mice were maintained in the EPFL animal facility and housed in individually ventilated cages. Animal procedures were performed according to protocols approved by the Schweizer Bundesamt für Veterinärmedizin.

Dox Treatment and Analysis

Mouse embryonic fibroblasts were isolated as described (24) and where applicable were grown in the presence of 1 μg/ml dox (Sigma). Dox was administered to mice in their drinking water, which was also supplemented with 4% glucose at a concentration of either 2 g/liter or 0.2 g/liter. Vehicle-treated mice were given water with 4% glucose only. Peripheral mouse blood samples were isolated at the indicated time points by tail vain bleeding. Leukocytes were isolated through a Histopaque 1083 (Sigma) gradient according to manufacturer's instructions. FACS analysis was performed on a Beckton Dickinson FACScan and results were analyzed using the FlowJo 8.1.1 software.

Ligation-mediated PCR (LM-PCR)

Integration sites were mapped by LM-PCR using a previously described protocol (26).

Southern Blot Analysis

Genomic DNA extraction and Southern blotting from mouse-tail biopsies were performed using standard procedures. Briefly, 20 μg of genomic DNA were digested with 40 units of NcoI and run on an agarose gel. After that the DNA on the gel was denatured, neutralized and transferred on to a nylon membrane (Hybond-N, Amersham Biosciences). The DNA was immobilized on the membrane with 200 mm NaP_i buffer (1 m Na₂HPO₄, pH 7.2) and by baking for 2 h at 80 °C. Pre-hybridization was done in Church buffer (0.5 m Na₂HPO₄, pH7.2, 1 mm EDTA) at 68 °C before an IRES-specific probe was added, and the incubation was extended overnight. The probe was produced by random DIG-containing nucleotide labeling of a 600 bp DNA fragment generated by restriction digest of the M-LVT plasmid with NcoI and AscI. The membrane was washed in 40 mm Na₂HPO₄, pH 7.2, 1% SDS twice for 5 min at 68 °C before it was washed at room temperature in buffer 1 (1 m maleic acid, 1.5 m NaCl, pH 7.5, 0.3% Tween). The membrane was then blocked in 1× blocking reagent (Roche, 1096176) dissolved in 1 m maleic acid, 1.5 m NaCl, pH7.5 for 30 min. After that anti-DIG antibody (DIG-High Prime, Roche) was added, and the incubation was continued. The membrane was washed twice in buffer 1 before the CDP Star substrate (DIG-High Prime, Roche) was added. The membrane was exposed to a film for a couple of hours and then developed using standard procedures.

DNA Methylation Analysis

Genomic DNA from MEFs was extracted with the DNeasy kit (Qiagen). 200 ng was used for bisulfite conversion with the EpiTect bisulfite kit (Qiagen) according to supplier's recommendations. Biotinylated amplicons were then produced by touchdown PCR: 5 min at 94 °C, 10 cycles including 1 min at 94 °C, 1 min at 60–50 °C, 1 min at 72 °C, 30 cycles including 1 min at 94 °C, 1 min at 50 °C, 1 min at 72 °C, followed by a final extension of 7 min at 72 °C. Primers can be found in the supplemental Table S2. 1 μl of bisulfite converted DNA was used as template in a final volume of 100 μl. The PCR products were purified using standard methods. The amount of methylated CpG present in the purified PCR products was determined by pyrosequencing reactions performed with the Pyrosequencing PSQ 96MA system (Biotage) at the Center for Integrative Genomics in Lausanne, Switzerland.

Immunohistochemistry

After dissections olfactory bulbs were fixed in 4% paraformaldehyde (PFA) for 4 h at 4 °C and then put in increasing concentrations of sucrose (15–30%) for 24 h. The next day they were rinsed in PBS and embedded in OCT. 20 μm cryostat slices were mounted on Superfrost+ slides (Menzel-Glaser), dried for 40 min at room temperature and then pre-incubated in 0.5% Triton, 10% FCS, 1× PBS for 1 h at room temperature. Slides were incubated overnight at 4 °C with the chicken anti-GFP antibody (1:1000, Abcam). Then slides were washed three times in 0.5% Triton, 1× PBS for 10 min. The primary antibody was revealed with goat anti-chicken alexa 488 (1:1000, Invitrogen) for 90 min. After three washes with 0.5% Triton, 1× PBS for 10 min, slides were mounted in DAPI-containing (1 μg/ml) antifade reagent (DABCO). Pictures were acquired using a Zeiss 510 META confocal microscope.

Whole Mount in Situ Hybridization (WISH)

Embryos were dissected in PBS and fixed in 4% PFA. WISH was performed according to standard protocols with a YFP-specific probe (30).

RNA Procedures

Total RNA from cultured cells was isolated using an RNeasy plus kit (Qiagen). RNA extracted from olfactory bulbs came from samples that were immediately snap frozen after dissection. More specifically, the bulbs were lysed in QIAzol lysis reagent (Qiagen) by brief sonication before being subjected to RNA isolation with the RNeasy lipid kit (Qiagen). Before being used for downstream applications, the RNA was subjected to rigorous DNase treatment with TURBO DNA-free DNase (Ambion) following the provider's recommendations.

For quantitative Real-time PCR (qPCR) RNA conversion to cDNA was primed with random hexamers and catalyzed using the Super Script II Reverse Transcriptase (RT) kit (Invitrogen). Diluted cDNA was then used to assay gene expression with 400 nm of gene-specific primers in 1× Power Sybr mastermix (Applied Biosystems). Specific amplification was verified by adding RT negative reactions and by adding melting curve analysis after the PCR runs. The fluorescence increase was analyzed with the SDS software (Applied Biosystems) using automatic settings. A mean quantity of RNA was calculated from technical duplicates or triplicates for each sample and normalized to the geometric mean of the selected normalization gene (31).

Chromatin Immunoprecipitation (ChIP)

ChIP was performed as previously described (25). The precipitating antibody was H3K9me3 (Abcam ab 8898). The relative enrichment was calculated based on the relative quantities of IP and total input samples after qPCR analysis. Primer-specific standard curves were used to estimate the relative amounts. ChIP efficiencies were comparable between samples as measured at the specific control locus Dhfr (32).

RESULTS

Generation of Mice Containing a tTRKRAB-regulated Gene Trap Allele

We used LV-mediated transgenesis (28) to generate mice harboring the “TrapSil” system (Fig. 1, A and B). We administered promoterless LV-based gene traps to fertilized zygotes through perivitelline injection (Fig. 1, A and B). The resulting adult mice were then tested for YFP reporter activity in peripheral blood cells (Fig. 1B). This system was chosen because it allows rapid and non-invasive screening of YFP levels indicating the presence of a cellular promoter-trapping event. Since the gene traps within these mice also contained TetO repeats, they were bred to mice stably expressing a tTRKRAB transgene (23, 33) to assess the effects of targeting KRAB/KAP1 to a gene body in vivo.

FIGURE 1.

Generation of mice containing a tTRKRAB-regulated gene trap allele. A, to explore if targeting of the ectopic repressor tTRKRAB results in inducible transcriptional repression of endogenous promoters in mice, we made use of the gene trap-based “TrapSil” system. Here, lentiviral-based constructs carrying the YFP reporter gene in the absence of promoter elements are used to “trap” the activity of endogenous promoters. By adding TetO repeats to these traps, which permit doxycycline (dox)-dependent binding of tTRKRAB, transcriptional “silencing” is induced. Thus, the activity of YFP is a direct read-out of the transcriptional effects mediated by the KRAB/KAP1 silencing complex. B, mice carrying trapped genes were generated through lentiviral (LV)-mediated transgenesis, where high-titer virus is used to infect fertilized zygotes. Animals with integrated proviruses were then screened for “trapping events” by analyzing levels of YFP expression in blood leukocytes. Finally, YFP-positive founder animals (F₀) were crossed with mice harboring the tTRKRAB transgene (hPGK::tTRKRAB) to give rise to F₁ offspring. C, Southern blot analysis was used to estimate the number of proviruses within the “trapped” founder (F₀) and offspring (F₁) animals. A unique band present only in YFP-positive F₁ animals was identified (Kif2A^tr, arrow). D, ligation-mediated PCR was used to map this “trapping” integration site to the kinesin family member 2A (Kif2A) locus, where the proviral:genomic junction is represented by a red box. The first Kif2A promoter drives expression of many splice variants, including Kif2A-001, Kif2A-002, and Kif2A-201, and is located ∼5 kb upstream of the gene trap, making this transcriptional start site the “trapped” promoter driving YFP expression in the Kif2A^tr allele. An alternative Kif2A promoter is located further downstream of the provirus driving the expression of the Kif2A-005 variant. E and F, Kif2A promoter activity in the adult brain was analyzed by fluorescence microscopy. E, this revealed YFP reporter expression within olfactory bulbs (ob), cortex (co), and cerebellum (ce). F, strong YFP signals were further detected within neurons of the ob, consistent with the essential role of Kif2A in neuronal microtubule dynamics. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI).

Trapped promoters were identified through a combination of Southern blot analysis (Fig. 1C) and ligation-mediated PCR (LM-PCR) (Fig. 1D). For this, genomic DNA from a YFP positive founder animal (F₀) and its offspring (F₁) was digested and blotted with an LV-specific IRES-binding probe. The different-sized bands on the Southern blot, therefore, reveal different proviral integration sites (Fig. 1C, asterisks). To determine the identity of an integration site trapping an active promoter, we correlated YFP expression in F₁ animals with the presence of a unique band (Fig. 1C, arrow). Two reporter-expressing F₁ animals (333, 338) harboring this proviral band only were subjected to LM-PCR to map the proviral:genomic junction and to identify thereby the candidate trapped promoter. In both cases, the trapping integrant was located in an intron of the kinesin family member 2A (Kif2A) gene locus (Fig. 1D, red box). The presence of the Kif2A-trapped allele (Kif2A^tr) was subsequently verified in all YFP expressing animals of this line using a trap-specific PCR-based genotyping approach (supplemental Table S1). The proviral gene trap at the Kif2A locus is situated between two alternative Kif2A promoters (Fig. 1D). Since the architecture of the trap only allows for reporter expression from an upstream promoter, the first Kif2A promoter is very likely to be the driver of the Kif2A^tr allele. The gene trap carries a strong polyadenylation signal, which abrogates the transcription of the different full-length Kif2A splice variants originating from the trapped promoter (Fig. 1D). In contrast, the gene trap does not disrupt the architecture of the downstream Kif2A promoter. Therefore, this constellation is consistent with a hypomorphic Kif2A gene trap allele, explaining the absence of post-natal lethality in our Kif2A^tr homozygous animals when compared with conventional Kif2A knock-out mice (34). The expression pattern of the trapped Kif2A promoter was further characterized in the brain (Fig. 1E), where strong YFP expression was detected in neurons of the olfactory bulb (Fig. 1F). This is consistent with the essential role of Kif2A in neuronal microtubule dynamics (34).

tTRKRAB-mediated Reversible Repression of the Kif2A Promoter in Vivo

To study the effect of tTRKRAB recruitment on Kif2A expression in vivo, we crossed mice heterozygous for our Kif2A^tr allele (Kif2A^tr/+) to a tTRKRAB transgenic line (PGK::tTRKRAB) (23, 33). These crosses were conducted while administering either dox or vehicle, in order to prevent or permit tTRKRAB repressor binding throughout embryonic and post-natal development (Fig. 2A). We then used FACS analysis to measure overall YFP fluorescence levels in blood leukocytes of the offspring. We found that tTRKRAB binding resulted in decreasing the mean fluorescence intensity (MFI) of YFP by around 30% (Fig. 2A). This tTRKRAB-mediated repression of the Kif2A^tr promoter was inhibited in the presence of dox, which allowed for reporter gene expression indistinguishable from levels present in mice without the tTRKRAB transgene (Fig. 2A). Taken together, these results imply that tTRKRAB mediates the dox-controllable repression of the cellular Kif2A promoter in vivo.

FIGURE 2.

tTRKRAB mediates the repression of the Kif2A promoter in a reversible way. A, effects of tTRKRAB and dox on Kif2A^tr expression were assessed by flow cytometry analyses of the blood. Peripheral nucleated lymphocytes were taken from 3–5-week old mice to determine the mean fluorescence intensity (MFI) of YFP values. Mice were generated by mating heterozygous mice containing a trapped and a wild-type Kif2A allele (Kif2A^tr/+) with tTRKRAB expressing mice (hPGK::tTRKRAB). One set of these crosses was carried out under vehicle conditions (4% glucose), while another set of crosses was conducted in the presence of doxycycline (dox, 2 g/liter in 4% glucose). The statistical analysis included the Mann-Whitney non-parametric test. The p value legend is ***: p < 0.001. B, induction of repressed Kif2A promoter activity was monitored through administration of dox, which relieves tTRKRAB-mediated silencing. Dox (0.2 g/liter in 4% glucose) was given to two groups of adult mice, which both contained a trapped and a wild-type Kif2A allele (Kif2A^tr/+), but differed in the presence of the tTRKRAB transgene. The induction of YFP expression in the tTRKRAB-containing animals (Kif2A^tr/+; tTRKRAB, n = 5) was monitored over 4 weeks by measuring the levels of MFI YFP in blood lymphocytes. These values were then compared with levels of YFP expression in the group without tTRKRAB (Kif2A^tr/+, n = 3), which were set at 100% after every measurement. The plotted values are therefore expressed as % of Kif2A ^tr/+ = % (MFI YFP Kif2A ^tr/+;tTRKRAB/MFI YFP Kif2A ^tr/+).

Next, we examined if targeting of tTRKRAB to a gene body may lead to reversible gene silencing, as opposed to its docking to the vicinity of a promoter. We did this by administering dox to individual adult mice, previously raised in the absence of dox, and by assessing the effect on tTRKRAB-mediated repression. More specifically, we exposed two groups of adult mice to dox and measured their expression of YFP in peripheral blood leukocytes over a 4-week period. Animals of both groups were heterozygous for the Kif2A-trapped allele (Kif2A^tr/+), but only one group additionally expressed tTRKRAB. We assessed MFI YFP levels on a weekly basis. Plotting the ratio of the two groups as a percentage of Kif2A^tr/+ (% of Kif2A^tr/+ = % (MFI YFP Kif2A^tr/+; tTRKRAB/MFI YFP Kif2A^tr/+)) over time, we found that the reporter expression in Kif2A^tr/+; tTRKRAB animals increased substantially after dox administration. Moreover, after the 4-week sampling period the MFI YFP values from the two groups were nearly identical. This indicates that dox administration can induce the complete reversal of KRAB-mediated repression of an endogenous promoter in adult mice (Fig. 2B). When we discontinued dox treatment, YFP expression was not re-silenced in Kif2A^tr/+; tTRKRAB animals after 1 or 2 months (supplemental Table S1). This is probably due to the deposition of this antibiotic in the bone from where it is released even long after administration has been discontinued (23).

KRAB-induced Reversible Long-range Heterochromatin Spreading at the Kif2A Locus

Previous studies indicated that tTRKRAB binding during the first few days of embryogenesis can induce CpG DNA promoter methylation, which results in irreversible transcriptional silencing even after the repressor is removed (22, 24). In contrast to these results, we find here that tTRKRAB-mediated repression is fully reversible in differentiated leukocytes, despite tTRKRAB binding at the Kif2A locus during development. To assess the molecular events underlying this phenomenon, we examined the corresponding mouse embryos. We first established the effect of tTRKRAB on embryonic Kif2A^tr expression by monitoring YFP RNA levels using whole mount in situ hybridizations in E11.5 embryos. We found that the presence of the tTRKRAB transgene decreased the levels of YFP transcripts, when compared with an embryo that did not express the ectopic repressor (Fig. 3A). This established that tTRKRAB mediates transcriptional repression of the Kif2A^tr locus in embryos. Therefore, we went on to isolate mouse embryonic fibroblast (MEF) cell lines from homozygous Kif2A trap allele (Kif2A^tr/tr) mice, which did or did not harbor the tTRKRAB transgene. The resulting MEF lines were therefore all Kif2A^tr/tr, with some of them additionally producing the tTRKRAB trans-repressor. FACS analysis indicated that, as predicted, the presence of tTRKRAB correlated with a reduction of YFP levels of around 30% (Fig. 3B). Furthermore, similar to the results obtained in peripheral blood cells, this repression was fully reversed when dox was added to the Kif2A^tr/tr;tTRKRAB cells, as verified both at the protein (Fig. 3B) and RNA (Fig. 3C) levels. Using Kif2A-specific primers, we found that levels of Kif2A:YFP fusion transcripts were reduced upon tTRKRAB binding (Fig. 3C). The Kif2A region far downstream of the gene trap cassette also displayed significantly decreased expression upon tTRKRAB recruitment (Fig. 3C). Notably, the degree of silencing could be increased when tTRKRAB expression in these MEF cell lines was augmented through additional transduction with a tTRKRAB expressing LV vector (supplemental Fig. S1), indicating that the lack of complete repression in the transgenic mice was at least partly due to suboptimal levels of trans-repressor. Finally, comparable dox-dependent transcriptional effects were recorded in the olfactory bulbs of Kif2A^tr/+;tTRKRAB mice (supplemental Fig. S2), further corroborating the general nature of these findings.

FIGURE 3.

Reversible long-range regulation of the Kif2A locus. A, E11. 5 embryos from a cross between Kif2A^tr/+ and hPGK::tTRKRAB were subjected to whole mount in situ hybridization using a YFP-specific probe. The different genotypes are indicated. B–E, MEFs were generated from individual embryos and were used for different analyses. MEFs were homozygous for Kif2A ^tr/tr, but differed in the presence of tTRKRAB. B, effect of tTRKRAB binding on Kif2A-driven YFP protein level was assayed through FACS analysis. For this, we determined the MFI YFP values of the different MEF lines. The statistical analysis performed consisted of the Mann-Whitney non-parametric test. The p value legend is *: p < 0.05, **: p < 0.01. C, effect of tTRKRAB recruitment on Kif2A locus gene expression was assessed in Kif2A^tr/tr; tTRKRAB lines cultured in the presence or absence of dox (1 μg/ml). We used qPCR to assay levels of the fusion transcript between the Kif2A exon upstream of the gene trap (Kif2A trapped) and YFP present in the trap. In addition, transcript levels including sequences 3′ to the gene trap, partly originating from the alternative Kif2A promoter located ∼20 kb downstream of the gene trap (Kif2A downstream), were measured. The GAPDH gene was used as a control gene, while the relative amounts of the different mRNAs were measured and normalized to actin. Similar results were found when GAPDH was used as a normalization gene. The statistical analysis included the Mann-Whitney non-parametric test. The p value legend is *: p < 0.05, n = 4. D, mean CpG DNA methylation of the YFP reporter gene body and the trapped Kif2A promoter was assessed through pyrosequencing of Kif2A^tr/tr and Kif2A^tr/tr;tTRKRAB MEF cell lines (n = 4). E, levels of H3K9me3 at the Kif2A locus were measured by chromatin IP followed by qPCR at the indicated sites (A–G). Kif2A^tr/tr; tTRKRAB MEF cell lines and Kif2A^+/+ wild-type cell lines were assayed (n = 3).

The embryos giving rise to the Kif2A^tr/tr;tTRKRAB MEF cell lines had the KRAB repressor bound to the Kif2A locus during early development. The finding that tTRKRAB-mediated Kif2A repression is fully reversible after the addition of dox (Fig. 3, B and C) suggested that tTRKRAB had not induced promoter CpG methylation, as this modification would have resulted in repression even in the absence of repressor binding. Consistently, when we assessed the status of the CpG island-containing Kif2A promoter, we found low overall DNA methylation levels and no difference between mice with or without the tTRKRAB transgene (Fig. 3D). Furthermore, while CpG methylation levels at the YFP-encoding part of the trap were much higher, they still did not exhibit any difference whether or not tTRKRAB was present (Fig. 3D).

The downstream Kif2A promoter is located 20kb away from the gene trap. Decreased levels of transcripts produced from this promoter (Fig. 3C) suggested that KRAB-induced repression was effective over this distance. Consistently, we recently demonstrated that tTRKRAB can induce long-range heterochromatin spreading and reversible promoter silencing in cultured cells (25). To test whether this mechanism could explain our observations, we used our MEF system to assess H3K9me3 levels at the Kif2A locus. We found an increase of this heterochromatic mark over the whole region in the presence of tTRKRAB (Fig. 3E). Therefore, we conclude that in our “TrapSil” animals, tTRKRAB can mediate the reversible long-range spread of facultative heterochromatin, thereby inducing conditional transcriptional repression.

DISCUSSION

The present work provides the proof-of-principle that endogenous genomic loci can be modified to serve as a target for externally controllable epigenetic regulators in vivo. Our study indeed demonstrates that introducing the target sequence of the dox-controlled KRAB-containing tTRKRAB effector at the mouse Kif2A gene locus allows for the drug-regulated and reversible down-regulation of the corresponding transcriptional unit. Silencing was incomplete at the trapped promoter. While this may reflect the persistence of trace amounts of dox in the animals under study (35), repression could be boosted in MEFs harvested from mice carrying the gene trap by LV-mediated over-expression of tTRKRAB (supplemental Fig. S1). This indicates that the lack of complete repression in the transgenic mice was at least partly due to suboptimal levels of the trans-repressor, although we cannot exclude locus-specific cis-acting influences, which can sometimes interfere with KRAB/KAP1-mediated effects as we previously observed (25, 26). Nevertheless, the experiments described here provide a robust demonstration of the attractiveness of the TrapSil system for controlling gene expression and chromatin states in vivo. In addition, the use of KRAB/KAP1-mediated silencing to control transcription in the hematopoietic (23) as well as in the visual system (36) demonstrates the broad applicability of this ectopic repressor system.

Repression of the trapped promoter was dependent on binding of tTRKRAB to the introduced intragenic docking site, and could be fully reversed through the addition of dox both in embryo and adult mice, consistent with our finding that this promoter did not become DNA methylated. Because we allowed for the binding of tTRKRAB to its target during early embryogenesis of our TrapSil mice, this result may appear to conflict with our previous reports stating that the recruitment of KRAB/KAP1 at imprinting control regions and in the immediate vicinity of promoters during this period leads to irreversible DNA methylation (24, 37). However, our recent analysis of DNA methylation in embryonic stem cells indicates that, while this modification is generally found at KAP1-binding genomic sites adorned with repressive histone marks, it usually spreads over only very short distances from these sites, at most 2 kb on average.⁶ This is also consistent with the previous finding that DNA methylation occurred over the 5′-LTR and genomic flanking sequences of a KRAB/KAP1-repressed retroviral vector, but not over its 3′ regions (38). It is thus likely that the Kif2A promoter trapped in our experiments escaped KRAB/KAP1-induced DNA methylation because the trans-repressor was targeted to its gene body, around 5 kb away from sensitive CpG dinucleotides in its promoter.

The gene targeted in our demonstration, Kif2A, encodes a microtubule motor protein with depolymerizing activity. In addition to being essential for spindle assembly and chromosome movement in mitotic cells (39), this motor protein is also required for normal brain development and axonal growth in post-mitotic neurons (34). In our mouse model, we find that the Kif2A promoter is active in all tissues examined, consistent with an essential role of its product in mitotic and post-mitotic cellular tubulin dynamics. Of note, because the hereby-described trap mouse line reports on Kif2a promoter activity, it may be of value to generate mouse models of diseases such as squamous cell carcinoma of the tongue, where Kif2A is overexpressed and contributes to cancer progression (40). More generally, the drug-controllable repression of a TetO-containing locus of interest seems particularly well suited for developing preclinical mouse models of human diseases. In such a scenario candidate disease genes could be silenced, and then re-activated to test the potential benefit of their expression, mimicking therapeutic interventions. In conclusion, our results reveal that the KRAB module can induce reversible and dox-controllable heterochromatization and repression of a mouse locus in vivo when targeted to a gene body. This finding provides novel leads for the development of genetic tools, where the reversible modulation of both chromatin and transcription status could be induced at specific loci of interest.