Abstract
The mammalian epigenetic phenomena of X inactivation and genomic imprinting are incompletely understood. X inactivation equalizes X-linked expression between males and females by silencing genes on one X chromosome during female embryogenesis. Genomic imprinting functionally distinguishes the parental genomes, resulting in parent-specific monoallelic expression of particular genes. N-ethyl-N-nitrosourea (ENU) mutagenesis was used in the mouse to screen for mutations in novel factors involved in X inactivation. Previously, we reported mutant pedigrees identified through this screen that segregate aberrant X-inactivation phenotypes and we mapped the mutation in one pedigree to chromosome 15. We now have mapped two additional mutations to the distal chromosome 5 and the proximal chromosome 10 in a second pedigree and show that each of the mutations is sufficient to induce the mutant phenotype. We further show that the roles of these factors are specific to embryonic X inactivation as neither genomic imprinting of multiple genes nor imprinted X inactivation is perturbed. Finally, we used mice bearing selected X-linked alleles that regulate X chromosome choice to demonstrate that the phenotypes of all three mutations are consistent with models in which the mutations have affected molecules involved specifically in the choice or the initiation of X inactivation.
EPIGENETIC gene regulation mechanisms result in chromatin modification and genome reprogramming during mammalian development. The two best-studied examples of epigenetic regulation, X inactivation and genomic imprinting, are thought to be evolutionarily related (Pannuti and Luchessi 2000). X inactivation results in the chromosome-wide silencing of one of the two X chromosomes in each cell of a female mammal (Avner and Heard 2001). Genomic imprinting results in the differential expression of the paternal and maternal alleles of specific genes (Ferguson-Smith and Surani 2001). Although both processes produce monoallelic expression of a subset of genes in the genome, imprinted genes reside in clusters throughout the autosomes and are regulated in a parent-of-origin-specific manner, while the effect of X inactivation is limited to genes on one X chromosome and the choice of which X is to be inactivated in any given embryonic cell is random. The relationship between the pathways of X inactivation and genomic imprinting is best exemplified by the extraembryonic tissues of certain rodents, including mice. In these tissues, X inactivation is imprinted with preferential silencing of the paternal X chromosome (Takagi and Sasaki 1975). Classic genetic analyses have revealed multiple molecules involved in these epigenetic pathways and have suggested that the interplay of such factors via large protein complexes is likely to be crucial to all epigenetic gene regulation mechanisms (Marinet al. 2000; Pannuti and Luchessi 2000). Nevertheless, details of the underlying molecular mechanisms, including which aspects are shared by or are unique to the imprinting or X-inactivation pathways, remain poorly understood. Further characterization of the components of these regulatory complexes is necessary for understanding mammalian epigenetic reprogramming.
X chromosome inactivation equalizes X-linked expression between males and females through the transcriptional silencing of a majority of genes on one X chromosome during early female development (Lyon 1961). To achieve stable silencing of one X chromosome, a cell must complete distinct stages of the pathway: selection or choice of one of the two X chromosomes, initiation and propagation of silencing, and maintenance of the inactive state during subsequent cell divisions (Avner and Heard 2001; Boumil and Lee 2001). The establishment of inactivation is under the control of elements within the genetically defined X-inactivation center (Xic) on the X chromosome. The inactive X maintains its silent state by acquiring distinct epigenetic modifications, including histone 3 lysine methylation, replication timing variation, CpG methylation, histone hypoacetylation, and enrichment with the histone variant macroH2A (Costanzi and Pehrson 1998; Chadwick and Willard 2001; Csankovszkiet al. 2001; Heardet al. 2002).
At least three elements within the Xic are involved in the establishment of X chromosome silencing. A key component is the Xist gene that is required in cis during the early stages of inactivation for both initiation and propagation of X chromosome silencing and that represents the major silencing element early in the pathway (Brockdorffet al. 1991; Brownet al. 1991; Pennyet al. 1996). Tsix, a gene antisense to Xist, suppresses Xist’s silencing activity prior to inactivation, locks in the states of the active and inactive X chromosomes, and plays an important role in imprinted X inactivation in extraembryonic tissues (Leeet al. 1999; Lee and Lu 1999; Miseet al. 1999; Lee 2000; Wutz and Jaenisch 2000; Boumil and Lee 2001). The X-controlling element (Xce) within the Xic regulates X chromosome choice in the mouse via different alleles (Cattanach and Isaacson 1965). The four Xce alleles that have been described, Xcea, Xceb, Xcec, and Xced, function in cis to affect the probability that an X chromosome remains active (Cattanachet al. 1969; West and Chapman 1978; Cattanach and Papworth 1981; Johnston and Cattanach 1981; Cattanach and Rasberry 1991, 1994). The Xce effect results in nonrandom X-inactivation choice in Xce heterozygous females, but random inactivation in Xce homozygous females (Plengeet al. 2000). Mutation studies have suggested that trans-acting autosomal factors interact with cis-acting elements within the Xic to determine X chromosome choice (Lyon 1971; Brown and Chandra 1973; Russell and Cacheiro 1978; Rastan 1983). Recently, one trans-acting factor, the chromatin insulator and transcription regulator CTCF (CCC TC-binding factor), has been suggested to bind sites within the Xic to direct X chromosome choice (Chaoet al. 2002). It is likely that additional trans-acting molecules are involved in the pathway (Percec and Bartolomei 2002).
While X chromosome inactivation specifically affects genes on the X chromosome, genomic imprinting modifies the activity of a subset of autosomal genes in a parent-of-origin-specific manner (Ferguson-Smith and Surani 2001). Imprinted genes, clustered throughout the genome in “imprinted domains,” are subject to coordinate regulation. Multiple factors regulate genomic imprinting by invoking allele-specific differences in DNA methylation, chromatin structure, and gene expression similar to those documented for genes undergoing X inactivation. These modifications enable the organism to erase, set, and maintain the appropriate epigenetic state of a given allele. While some of the molecules that govern these modifications have been identified, many, including those possibly common to X inactivation and genomic imprinting, remain unknown (Bourc’hiset al. 2001; Howellet al. 2001; Reiket al. 2001).
N-ethyl-N-nitrosourea (ENU) mutagenesis has been applied effectively to identify novel developmental and behavioral factors in the mouse (Habrede Angeliset al. 2000; Nadeau 2000; Nolanet al. 2000). We describe here how ENU mutagenesis screens may be applied successfully to dissect quantitative trait phenotypes of epigenetic pathways. We argue that the use of a sensitive, specific, high-throughput, and noninvasive assay with genetically matched phenotyping controls is crucial to identifying mutations with subtle effects on complex mechanisms. We previously described autosomal mutations with dominant effects on X chromosome inactivation in two pedigrees identified through a genome-wide, phenotype-driven mouse ENU mutagenesis screen for mutations with primary effects on X inactivation and mapped the mutation segregating in one pedigree to mouse chromosome 15 (Percecet al. 2002). We now have mapped two additional mutations on mouse chromosomes 5 and 10 in the second pedigree, fulfilling the prediction that multiple autosomal factors are involved in the X-inactivation pathway. Our results indicate that the mutations target the X chromosome inactivation mechanism occurring in the embryo proper and do not affect the imprinted form of X-inactivation or genomic imprinting mechanisms. Further, we demonstrate through genetic analysis of mutant animals that are homozygous at the Xce locus that the effects of the X-inactivation mutations in both pedigrees are specific to animals that are heterozygous at the Xce locus. Together, the studies presented here are consistent with a model in which the mutations target factors in the X-inactivation pathway specifically involved in the choice or the initiation of X inactivation.
MATERIALS AND METHODS
Mice: C57BL/6J, BALB/cByJ, B6CBAF1/J, and CAF1/J mice were purchased from The Jackson Laboratory and CD-1 mice were purchased from the Charles River Laboratory (Wilmington, MA). The tester stock of mice, containing M. m. castaneus (CAST/Ei) alleles along the length of the X chromosome on a 129S1/SvJ background, was generated as described (Plengeet al. 2000). The tester stock was maintained subsequently by brother-sister mating.
ENU mutagenesis: ENU was purchased (Sigma, St. Louis). Ten milliliters of 95% ethanol was injected into the Isopac container and 5 ml of the solution was discarded. To the remaining solution, 95 ml phosphate-citrate buffer (0.1 m Na2HPO4, 0.05 m NaCitrate, pH 5.0) was added. The ENU concentration was determined by spectrophotometry at 398λ and by applying the following formula: [ENU] = (OD398/0.72)(5). The injection volume was calculated as follows: ml ENU = (dose/[ENU])(grams body weight). ENU was injected intraperitoneally and doses requiring multiple injections were repeated weekly. The ENU doses, which ranged from a minimum of 200 mg to a maximum of 400 mg, were administered as single treatments or as multiple (fractionated) treatments (Table 1). Animals treated with single large doses (200 or 250 mg) and animals treated with fractionated high doses (four doses of 100 mg each) fared the most poorly. In general, F1 hybrid animals had a higher survival rate and recovered fertility faster than inbred BALB/cByJ animals treated with the same dose, consistent with previous studies (Russellet al. 1982; Favor 1988; Nolanet al. 1997; Justiceet al. 1999; Weberet al. 2000). The time required to restore fertility to a particular male was measured from the date of the last injection. Female progeny of ENU-treated males, designated as G1, were weaned at approximately 3 weeks of age. Because the primary screening phenotype—X inactivation—is female specific, male progeny were sacrificed. Matings were continued until ∼50 female pups were generated from each male.
Embryo dissections: Embryos were dissected according to specifications described elsewhere (Hoganet al. 1994). First day after mating was designated 0.5 dpc (days post-coitum). Timed embryos were aged more specifically by morphology (Rugh 1994).
RNA isolation: Toe and ear samples were collected from each animal at weaning (∼3 weeks of age). RNA from G1 animals was isolated using the Purescript RNA isolation kit (Gentra Systems, Minneapolis) for 5- to 10-mg tissue samples according to the manufacturer’s recommendations. Toe/ear and embryo RNA from animals of the G2 and all subsequent generations was isolated using the HighPure RNA tissue kit (Roche Molecular Biochemicals) with minor modifications to the manufacturer’s recommendations. Complete liver, kidney, heart, and tongue samples were collected. RNA was isolated from kidney, heart, and tongue using the NucleoSpin RNA II purification kit (CLONTECH, Palo Alto, CA) according to the manufacturer’s recommendations. RNA was isolated from whole liver with the LiCl2 isolation procedure as previously described (Auffray and Rougeon 1980). RNA samples from liver, kidney, heart, and tongue tissues were purified further for use in LightCycler assays by processing through the High-Pure RNA tissue kit.
cDNA synthesis: Approximately 0.75 μg of RNA was reverse transcribed in a 20-μl reaction containing 10 mm dithiothreitol (GIBCO BRL, Gaithersburg, MD), 500 μm of each dNTP (GIBCO BRL), 1× first strand buffer (GIBCO BRL), 25 ng random primers (GIBCO BRL), 200 units of M-MLV RT (GIBCO BRL) and 20 units of RNaseOUT ribonuclease inhibitor (GIBCO BRL). Samples were incubated for 10 min at room temperature, 60 min at 37°, and 10 min at 95°. Specific-primed reverse transcription was carried out for embryo samples to be characterized with the Xist assay by substituting 10 pmol of the Xist-R 5′-CCGATGGGCTAAGGAGAAG-3′ primer for the random primer.
Pctk1 and Xist allele-specific expression assay: The Pctk1 and Xist expression assays were conducted as previously described (Percecet al. 2002).
H19 allele-specific expression assay: The H19 allele-specific expression assay was carried out as previously described (Thorvaldsenet al. 2002).
Snrpn allele-specific expression assay: The Snrpn expression assay was conducted on cDNA using the LightCycler real time PCR system (Roche Molecular Biochemicals). Snrpn primers, Sn1 (5′-CTCCACCAGGAATTAGAGGC-3′) and Sn3 (5′-TAT AGTTAATGCAGTAAGAGG-3′), were used to amplify a 155-bp region of the Snrpn gene (Szabo and Mann 1995). Fluorescence resonance energy transfer hybridization probes were designed to the C57BL/6 amplicon. The Snrpn sensor probe (5′-GAAGCATTGTAGGGGAAGAGAA-FL-3′) spans a single nucleotide polymorphism at nucleotide 915 between C57BL/6 (C) and CAST/Ei (T) and was labeled with fluorescein at the 3′ end. The Snrpn anchor probe (5′RED640-GGCTGA GATTTATCAACTGTATCTTAGGGTC-P3′; Idaho Technologies, Salt Lake City) was labeled with LC-Red640 at the 5′ end and was phosphorylated at the 3′ end. To a Ready-To-Go PCR bead (Amersham Biosciences), 5.12 μl H2O, 0.38 μl TaqStart antibody (CLONTECH), and 1.5 μl 25 mm MgCl2 (final concentration 3.0 mm) were added, and the reaction was incubated at room temperature for 5 min. After incubation, a final concentration of 12% DMSO, 0.5 μm of each primer, and 0.3 μm of each probe were added to the mix and the volume brought to 12.5 μl. From this reaction mix 10 μl was removed and added to a LightCycler glass capillary (Roche), and 10 μl cDNA and H2O were added for a final reaction volume of 20 μl. After an initial denaturation step at 95° for 2 min, amplification was performed for 65 cycles at 95° for 1 sec, 50° for 15 sec, and 72° for 6 sec. A single fluorescence acquisition occurred at the end of each annealing step. After amplification, a final denaturation and annealing step was conducted (95° for 3 min, 35° for 2 min), followed by a melting curve analysis with fluorescence acquisition occurring continuously as the temperature was increased from 35° to 85° in 0.2° increments. After background subtraction, the contribution of each allele was calculated as the peak height of the melting curve generated at the allele-specific temperature, ∼56.5° for C57BL/6 and 51° for CAST/Ei (LightCycler Data Analysis Software).
Cdkn1c allele-specific expression assay: To a Ready-To-Go PCR bead, 0.3 μm of each primer and 1 μl of cDNA were added. Primers p576 (5′-CGGACGATGGAAGAACTCTGG-3′) and p574 (5′-TACACCTTGGGACCAGCGTACTCC-3′) amplified a 167-bp fragment (94° for 2 min followed by 35 cycles at 94° for 15 sec, 50° for 10 sec, and 72° for 20 sec) containing a polymorphism between C57BL/6 (T) and CAST/Ei (C; position 1256, MMU20553). The amplified product was digested with TaqI to reveal the CAST/Ei polymorphism (cleaved products are 118 and 49 bp) and resolved on a 7% polyacrylamide gel.
Makorin3 allele-specific expression assay: To a Ready-To-Go PCR bead, 0.3 μm of each primer and 1 μl of cDNA were added. Primers -Zf1 (5′-GACAGCCTTACCGAGGTCGC-3′) and Zf2 (5′-CATGGGGGTATGCACACCTG-3′) amplified a 215-bp C57BL/6 fragment and a 225-bp CAST/Ei fragment (94° for 2 min followed by 35 cycles at 94° for 15 sec, 55° for 10 sec, and 72° for 20 sec). Restriction digestion with MspA1I produces 152- and 67-bp bands for C57BL6/J cDNA while CAST/Ei cDNA remains uncleaved. The products were resolved on a 7% polyacrylamide gel.
Genotyping: All genotyping was carried out by PCR amplification of MIT microsatellite markers. For PCR amplification, ∼100 ng of genomic DNA (or 1 μl of supernatant from the DNA isolation procedure) and 0.5 μm of each primer were added to a Ready-To-Go PCR bead (Amersham Pharmacia Biotech) according to the manufacturer’s recommendations. After an initial denaturation step at 95° for 2 min, amplification was performed for 35 cycles at 95° for 15 sec, 55° for 10 sec, and 72° for 20 sec. Genotyping of the X chromosome was carried out routinely with DXMit53 (∼5 cM from Pctk1) and DXMit18 (<1 cM from Xist and Xce) to determine genotypes at Pctk1 and Xist/Xce, respectively. These markers are polymorphic between CAST/Ei and other inbred mouse strains.
RESULTS
X-inactivation pattern measured by allele-specific expression assays: We previously reported two mutant pedigrees segregating aberrant X-inactivation phenotypes (Percecet al. 2002) and here provide the details of the original mutagenesis screen. To detect mutations specifically affecting the early steps of X chromosome inactivation, we screened for abnormalities in the X-inactivation pattern, a quantitative measure of cells that have chosen to keep one or the other X chromosome active. This phenotype reflects the X chromosome choice made by the population of progenitor cells undergoing inactivation during early embryogenesis. We hypothesized that screening for aberrations of the X-inactivation pattern would encourage the recovery of subtle mutations with primary effects on the X-inactivation pathway and possible secondary effects on genomic imprinting. To measure the X-inactivation pattern, we used an efficient quantitative allele-specific RT-PCR assay that measures transcript levels from the X-linked gene, Pctaire-1 (Pctk1) in an allele-specific manner (Carrelet al. 1996; Plengeet al. 2000).
ENU treatment regimen
Nonrandom X chromosome choice occurs in the mouse when a “strong” Xce allele (i.e., one that is more likely to be on the active X, such as the Xcec allele derived from CAST/Ei mice) is inherited in combination with a “weak” allele (i.e., one that is less likely to be on the active X, such as Xcea or Xceb derived from M. m. domesticus mice). The effect on the X-inactivation pattern exerted by the Xcea and Xceb alleles of various inbred mouse strains relative to the Xcec allele derived from the CAST/Ei strain has been documented (Plengeet al. 2000). In these Xce heterozygous animals, ∼25-30% of cells choose the X bearing the weaker Xce allele as the active X, instead of the 50% expected for a random process (Cattanach and Williams 1972; Cattanach and Johnston 1981; Johnston and Cattanach 1981; Rastan 1982; Cattanach and Rasberry 1994). The screen took advantage of this quantitative model of choice in which the effect of different Xce alleles is used as a baseline measure of the X-inactivation pattern. We hypothesized that the use of Xcea/c and Xceb/c heterozygotes (in contrast to Xce homozygotes) would facilitate the recovery of mutations disrupting general aspects of X chromosome choice as well as those specifically involved in Xce allelic discrimination.
ENU mutagenesis screen: Because we hypothesized that many unidentified factors must be involved in the pathway of X inactivation (Plengeet al. 2000), we screened for dominant mutations affecting the X-inactivation pattern. A total of 100 male mice were treated with ENU (Table 1). These consisted of 50 BALB/cByJ, 40 B6CBAF1/J, and 10 CAF1/J males. A total of 62% of the F1 hybrid males recovered fertility after treatment with an average of 13 weeks of sterility. In contrast, only 20% of BALB/cByJ males recovered, after an average of 18 weeks of sterility. ENU-treated males that recovered fertility after treatment (G0) were mated to females from a tester stock of mice selected to have a CAST/Ei X chromosome. BALB/cByJ and CAF1 males possess an X chromosome bearing a Xcea allele while B6CBAF1/J males possess an X chromosome bearing a Xceb allele (Plengeet al. 2000). All female progeny (G1) thus inherited a maternal CAST/Ei X chromosome and a paternal non-CAST/Ei X chromosome.
In total, we screened 84 Xcea/c and 252 Xceb/c G1 female progeny of ENU-treated male mice (Percecet al. 2002). The former population represented the progeny of 10 BALB/cByJ and 5 CAF1 males and the latter represented the progeny of 26 B6CBAF1/J males that had recovered fertility after ENU treatment (Table 1). A maximum of 50 offspring were tested from each ENU-treated male to prevent the recovery of identical mutations from the limited population of spermatogonial stem cells (Justiceet al. 1999; Weberet al. 2000). As described previously (Percecet al. 2002), the phenotypic distribution of the X-inactivation pattern in wild-type control females provided a statistical basis for screening the offspring of ENU-treated males and for discriminating between normal and aberrant phenotypes in the female progeny of ENU-treated males. We considered any G1 animals with X-inactivation patterns ≥2 standard deviations (SD) from the mean of the wild-type population candidates for X-inactivation mutants. To avoid recovery of X-linked mutations unrelated to primary effects on X-inactivation choice, we focused on animals that preferentially maintained the mutagenized X chromosome as the active X.
—Progeny test of Xceb/c candidate mutant females. Four affected Xceb/c G1 females, 14.2, 18.6, 28.6, and 1.19, and one unaffected Xceb/c G1 female, 21.9, serving as a negative control, (○) were progeny tested and their respective G2 female progeny (•) are shown. The X-inactivation pattern was calculated with the Pctk1 assay as the proportion of transcription from the X chromosome bearing the weaker Xce allele to the total transcription from both X chromosomes. The mean for the respective wild-type Xce heterozygous distribution is depicted by a solid horizontal line and the two standard deviations from the mean of the wild-type population are depicted by dashed horizontal lines.
Two Xcea/c and four Xceb/c G1 females fulfilling the candidate mutant criteria were progeny tested (data not shown and Figure 1) to exclude statistical variants from further analysis. It is recognized that some true mutants may have been excluded, such as those with mutations of low penetrance, with maternally imprinted effects, with transmission ratio distortion, or those missed simply because of inadequate numbers of informative offspring (less than half of the female progeny were heterozygous at both the Pctk1 and Xce loci). As expected, the phenotyping control female (21.9) transmitted only wild-type X-inactivation patterns to her progeny. Three of the six candidate females tested (23.14, 14.2, and 18.6) were classified into the statistical variant category (data not shown and Figure 1). Significantly, however, three G1 females (24.21, 1.19, and 28.6) transmitted aberrant X-inactivation patterns to a subset of their female offspring, consistent with the segregation of a dominant mutation (data not shown and Figure 1). (Given the nature of the quantitative cutoff used in this screen, each offspring with an X-inactivation pattern >2 SD above the mean has only a 2.5% probability of having that pattern by chance.) Candidate mutations segregating in the three animals were induced independently of one another, as these animals were the progeny of different ENU-treated founder males, with female 24.21 as the daughter of an ENU-treated BALB/cByJ male, and females 1.19 and 28.6 as the daughters of two ENU-treated B6CBAF1/J males. We investigated further one female from each Xce genotype and genetic class and thus examined females 24.21 and 1.19 as candidate mutants of X chromosome inactivation.
Establishment of the mutant phenotypes: To confirm and characterize further the phenotypes of the two candidate mutants, the Xist gene was assayed as a marker of X chromosome activity instead of the Pctk1 gene used for the primary mutagenesis screen. Because phenotyping required heterozygosity at the locus assayed (originally Pctk1) and at Xce, a greater number of informative animals were generated when Xist, a marker tightly linked to Xce, was used for the allele-specific expression analysis. The Xist assay was applied to control unmutagenized Xcea/c and Xceb/c females to establish wild-type phenotypic parameters. The X-inactivation pattern in 55 wild-type Xcea/c females was normally distributed with a mean of 0.29 and an SD of 0.07 (Figure 2A). The X-inactivation pattern in 44 wild-type Xceb/c females also was normally distributed with a mean of 0.32 and an SD of 0.06 (Figure 2B). As expected, the phenotypic distributions were similar to those observed with the Pctk1 assay (Percecet al. 2002). Phenotyping of several generations of control animals outcrossed to the tester stock confirmed that the X-inactivation pattern does not vary with segregation of the genetic background (Figure 2, A and B), consistent with prior studies (Plengeet al. 2000). The molecular phenotyping criteria for the Xist assay were designated in the same manner as they were for the Pctk1 assay: affected females were defined by phenotypes that were ≥2 SD above the mean of the respective wild-type Xce genotypes. Carrier animals (males or females) are defined as those that have transmitted the mutant phenotypes to two or more female offspring. (The likelihood of such transmission occurring by chance is <0.001.)
To confirm heritability and to characterize further the inheritance pattern of the mutant phenotypes from candidate mutant G1 females 24.21 and 1.19, we phenotyped all informative female progeny generated throughout their lifetime (Figure 2, C and D). Affected females (defined as above) examined from both families transmitted the mutant phenotypes to the subsequent generation in a dominant manner. Both families stably segregated the mutant X-inactivation patterns through at least the seventh generation, G7 (Figure 2, C and D, and data not shown). Carrier females from both pedigrees displayed X-inactivation patterns significantly different from those predicted by their respective Xce genotypes, as the X chromosome bearing the “weaker” Xce allele was the active X in a significantly higher than expected proportion of cells (compare Figure 2A to 2C and Figure 2B to 2D). The mean X-inactivation pattern of transmission-confirmed carrier females segregating in family 24.21 is 0.48 with a standard deviation of 0.07 while in family 1.19 these values are 0.51 and 0.04, respectively (Figure 2, C and D).
—Multigeneration progeny tests of wild-type and mutant animals. For the progeny test, females from each generation were bred to tester males and all informative (heterozygous at Xist and Xce) female progeny were phenotyped. Xist phenotyping of four generations of Xcea/c (A) and Xceb/c (B) unmutagenized control animals demonstrated that the X-inactivation phenotype remains constant despite outcrossing to the tester stock. For each control generation, an open circle and vertical bar depict the X-inactivation pattern mean and 2 SD demarcation, respectively. Transmission of the aberrant phenotype is demonstrated here in a dominant manner through the sixth generation in family 24.21 (C) and family 1.19 (D). In C and D, circled animals were progeny tested and gave rise to the subsequent generation. An arrowhead depicts the mean for the respective wild-type Xce heterozygous distribution and the dashed horizontal lines demarcate values within 2 SD from the mean of the wild-type populations. The X-inactivation pattern was calculated with the Xist assay as the proportion of transcription from the X chromosome bearing the stronger Xce allele to the total transcription from both X chromosomes.
Mutant phenotype from family 1.19 segregates with two independent autosomal loci: Previous genetic segregation and mapping analysis demonstrated that the mutations segregating in both pedigrees map to autosomal loci and, more specifically, that mutation from family 24.21 maps to chromosome 15 (Percecet al. 2002). To localize the mutation in pedigree 1.19, four affected and transmission-tested females and five unaffected distantly related females from the fifth (G5) and sixth (G6) generations were genotyped with 81 informative microsatellite markers evenly spaced across the genome with a maximum swept radius of 30 cM (Figure 3 and data not shown). Informative markers were selected that distinguish ENU-treated alleles C57BL/6J and CBA/J from wild-type 129S1/SvJ and CAST/Ei alleles. Alleles from the mutagenized B6CBAF1/J founder male were observed at the expected frequency, ∼2.8 and 1.2% in unaffected G5 and G6 animals, respectively (as compared to the expected 1/32 or 3% and 1/64 or 1.5% of alleles in wild-type G5 and G6 animals, respectively), serving as genotyping controls (data not shown). No regions of the genome exhibited nonrandom segregation of any markers in the unaffected females tested. In affected females, however, two regions of the genome, distal 23 cM of chromosome 5 and ∼30 cM of the proximal chromosome 10, demonstrated significant linkage (P < 0.03) to the mutant phenotype, calculated as the probability of random segregation of the mutant C57BL/6J and CBA/J alleles to the animals tested (Figure 3B). The results of the genome scan indicate that two regions of the genome are segregating with the mutant phenotype in pedigree 1.19.
—Genome scan of family 1.19 localizes mutant phenotype to chromosomes 5 and 10 (A). Four affected G5 (4) and G6 (1-3) females were genotyped with 81 microsatellite markers spaced evenly throughout the genome with a maximum spacing of 30 cM. Arrowheads on the left of each chromosome designate the location of the markers. Genotypes for the four individual females are represented as vertical bars 1-4 for each chromosome. Open regions of the bars represent wild-type genotypes (i.e., from the nonmutagenized parent), while solid regions represent C57BL/6J or CBA/J alleles inherited from the ENU-treated founder male. The probability that the ENU-treated alleles were inherited by chance in the regions of chromosome 5 and chromosome 10 is P < 0.03. Mapping of chromosomes 5 and 10 (B) suggests that the mutations are located within the distal 23 cM of chromosome 5 and within a 30-cM region of the proximal chromosome 10. The MIT microsatellite markers used for genotyping are designated at the left of each chromosome with the approximate Mouse Genome Database (MGD) genetic map location shown on the right. The map locations of markers employed for this genotyping were consistent between the MGD, Ensemble, and Celera databases.
To determine definitively whether one or both candidate regions are responsible for the mutant phenotype, we progeny-tested animals carrying both candidate regions, as well as animals carrying one or the other candidate regions (Table 2, Figure 4, and data not shown). Animals carrying both regions clearly transmitted the mutant phenotype to offspring (Figure 4, A and B). Significantly, animals carrying solely the candidate region on chromosome 10 transmitted the mutant phenotype to their offspring in a dominant manner, indicating that there is a mutation in this region that is sufficient to induce the aberrant X-inactivation pattern (Figure 4C). Furthermore, animals carrying solely the candidate region on chromosome 5 also transmitted the mutant phenotype to their offspring in a dominant manner, indicating that there is a mutation in this region as well that is sufficient to induce the aberrant X-inactivation pattern (Figure 4D). These data suggest that two independent mutations are segregating in this pedigree and that each of the two mutations is sufficient to cause the mutant X-inactivation phenotype. The changes in the X-inactivation pattern induced by each mutation were quantitatively similar to one another as well as to the mutant phenotype exhibited by double heterozygotes, indicating that any additive effect on the phenotype is likely to be subtle. In addition, the phenotypes of each of the two mutations were similar when transmitted independently through male and female lines, suggesting that the mutant factors are not subject to an imprinting effect (Table 2, Figure 2D, Figure 4, and data not shown).
Candidate region segregation in transmission-proven carrier animals from family 1.19
Mutant phenotypes do not perturb imprinted X inactivation in family 24.21: The analysis of the mutant X-inactivation phenotypes in families 24.21 and 1.19 has thus far been limited to somatic tissues including adult tissues and tissues of the embryo proper. X inactivation in extraembryonic tissues of the mouse differs, however, from its somatic counterpart by undergoing paternally imprinted X inactivation (Takagi and Sasaki 1975; Huynh and Lee 2001). To determine whether the mutations in our families also affect the imprinted form of inactivation, we examined the X-inactivation pattern in extraembryonic tissues of early embryos with the Xist assay (Table 3). Because the Xce locus does not affect X chromosome choice in extraembryonic tissues, we phenotyped Xce heterozygous embryos (Takagi and Sasaki 1975; Rastan and Cattanach 1983). In extraembryonic tissues of wild-type 8.5- and 10.5-dpc embryos, we observed exclusive expression of Xist from the paternal allele, consistent with paternal-specific silencing of the X chromosome. In extraembryonic tissues of 8.5- and 9.5-dpc embryos from family 24.21, we also observed exclusive expression of Xist from the paternal allele, indicating that paternal imprinting is not perturbed by the mutation segregating in family 24.21.
—Partial pedigrees of family 1.19 depict phenotype-genotype associations in animals carrying candidate regions on chromosomes 5 and 10 (A and B), animals carrying solely the candidate region on chromosome 10 (C), and animals carrying solely the candidate region on chromosome 5 (D). Circles represent females and squares represent males. Half-solid symbols represent animals carrying both chromosome 5 and 10 candidate regions, half-shaded symbols represent animals carrying only the chromosome 5 candidate region, half-striped symbols represent animals carrying only the chromosome 10 candidate region, and open symbols represent animals carrying neither candidate region. The XCI pattern is noted below each animal. These data suggest that the two mutations segregate independently in this pedigree and that each mutation is sufficient to cause the mutant X-inactivation phenotype.
Imprinted X inactivation in extraembryonic tissues of embryos of family 24.21
Mutant phenotypes do not perturb genomic imprinting: To maintain and perhaps initiate X chromosome silencing, the X-inactivation pathway employs factors that are common to other epigenetic processes, including genomic imprinting. Therefore, it is possible that the factors targeted by our candidate mutations may be common to multiple epigenetic silencing pathways. To address this directly, we inquired whether the candidate mutations affect genomic imprinting of the well-characterized H19, Snrpn, Cdkn1c (p57kip2), and Makorin3 (Zfp127) genes (Bartolomeiet al. 1991; Leffet al. 1992; Hatada and Mukai 1995; Jonget al. 1999). H19 and Cdkn1c normally are expressed only from the maternal allele while Snrpn and Makorin3 are expressed only from the paternal allele. Affected and unaffected females from both pedigrees were examined with allele-specific expression assays and all exhibited appropriate imprinting (Table 4). For the analysis of family 1.19, we examined double heterozygous animals that carried both chromosome 5 and chromosome 10 mutations. These observations indicate that genomic imprinting is not perturbed by any of the three mutations segregating in the two pedigrees.
Mutant phenotypes in Xce homozygous females: The mutant phenotypes in families 24.21 and 1.19 have been defined by their effect on the X-inactivation pattern in Xce heterozygous animals. Because the mutant phenotypes shift the X-inactivation pattern from the wild-type nonrandom inactivation choice characteristic of Xcea/c and Xceb/c genotypes to a phenotype closer to the random X-inactivation pattern characteristic of Xce homozygotes (Plengeet al. 2000), we asked whether the mutant effects can be detected in the X-inactivation pattern in Xce homozygotes from the two pedigrees by measuring the X-inactivation pattern in Xce homozygous female progeny of Xce heterozygous carriers.
For this analysis, phenotyping of the two families was conducted by assaying the Pctk1 gene as a marker of X chromosome activity. The Pctk1 assay was employed since Xist and Xce are tightly linked and both loci would be homozygous in these animals. The Pctk1 assay was applied first to wild-type control Xcec/c females to establish wild-type phenotypic parameters for the assay, including mean and standard deviations (Figure 5). All categories of Xce homozygous females (i.e., Xcec/c, Xcea/a, and Xceb/b) are expected to display random X-inactivation patterns (Plengeet al. 2000). As predicted, the X-inactivation pattern in nine wild-type Xcec/c females tested was normally distributed with a mean of 0.5 and an SD of 0.07, consistent with previous studies (Plengeet al. 2000). This phenotypic distribution therefore may be applied to the comparison of all classes of Xce homozygotes.
The X-inactivation pattern was characterized in the Xce homozygous female progeny of carrier females from both families. Again, the animals examined from family 1.19 were double heterozygous animals that carried both chromosome 5 and chromosome 10 mutations. The mean X-inactivation pattern in Xcec/c females tested from families 24.21 and 1.19 did not differ significantly from the wild-type Xcec/c population (Figure 5). Further, the X-inactivation pattern in Xcea/a female progeny of carrier females from family 24.21 did not differ from the wild-type Xcec/c population (Figure 5). In contrast, the variance differed significantly between the two 1.19 and control groups (P = 0.02). This difference is not observed between the 24.21 and control groups, which suggests that the mutations in the two pedigrees may act via different mechanisms to induce the aberrant X-inactivation phenotypes. As additional confirmation of the Xce homozygous phenotypes in adult tissues, the X-inactivation pattern was characterized in Xce homozygous wild-type control and mutant embryos from pedigree 24.21 (data not shown). As expected, both wild-type and mutant embryos displayed random X-inactivation patterns that were similar to one another as well as to those observed in mutant Xce heterozygous embryos of the same family. These data suggest that the candidate mutations from both families may have targeted factors specifically involved in X chromosome choice and in Xce allelic discrimination.
DISCUSSION
ENU mutagenesis for subtle molecular phenotypes: To advance the understanding of epigenetic regulatory mechanisms, we conducted a dominant, genome-wide, phenotype-driven ENU mutagenesis screen to detect novel regions of the mouse genome involved in X chromosome inactivation and hypothesized that this approach would result in a high rate of novel mutation detection, given the paucity of genes currently known in the pathway and the likelihood that many more genes are involved. Two recent large-scale studies demonstrated dominant mutation frequencies as high as 2% when a spectrum of phenotypes was screened for mutations, with >50% of the abnormal phenotypes detected in the screen resulting from heritable mutations (Habre de Angelis et al. 2000; Nolanet al. 2000). Small-scale ENU mutagenesis screens focusing on morphological phenotypes have proven to be as successful as the larger studies, if not more so (Kasarskiset al. 1998; Hentgeset al. 1999). The estimated minimum mutation frequency for our screen, 0.9% (3/346), is within the range of these latest parameters. Because not all females with abnormal X-inactivation phenotypes were characterized genetically, it is not known whether additional females carried heritable mutations; thus, our estimated mutation frequency is likely an underestimate.
Imprinting in affected and unaffected females from families 24.21 and 1.19
Our results clearly demonstrate that small-scale ENU mutagenesis screens can be applied successfully to the recovery of multiple mutations resulting in subtle molecular phenotypes. The use of high fractionated ENU doses and a sensitive and robust assay, and the likelihood that many factors are involved in the epigenetic control of the X-inactivation pathway, contributed to the successful recovery of mutants in our screen. These observations indicate that many other molecular pathways would be amenable to genetic dissection through the use of small-scale ENU mutagenesis screens if screening parameters were carefully optimized. While other ENU mutagenesis screens have attempted to identify mutations involved in epigenetic factors (Tsaiet al. 2001), fewer mutations have been detected from this type of screen than from screens for other developmental phenotypes. Presumably, this is due to the lethal nature of severely abnormal epigenetic phenotypes. Thus, the recovery of epigenetic mutations may be enhanced by the use of highly quantitative screening assays such as the one applied in our screen.
Mutant phenotypes are specific to X chromosome inactivation: The mutant phenotypes segregating in our pedigrees could result from mutations in global factors, such as housekeeping genes, or from mutations in factors specific to the X-inactivation pathway. In the former scenario, the mutant phenotypes would be manifested in both sexes and affect multiple developmental and genetic pathways. Such mutations may have targeted factors involved in epigenetic silencing mechanisms common to both X inactivation and genomic imprinting. Our data demonstrate intact genomic imprinting of four different imprinted genes in carrier females from both families, indicating that none of the mutations has targeted a factor essential to the epigenetic regulation of genomic imprinting. Consequently, our studies present no evidence for any effect of the candidate mutations beyond the X-inactivation pattern, suggesting that the mutations have targeted factors specific to the pathway of X inactivation.
Mutant phenotype of family 24.21 does not affect imprinted X inactivation: Previous studies of early postimplantation embryos from both families established that the mutant X-inactivation phenotypes are manifested early in development around the time that X inactivation occurs (Percecet al. 2002). The mutant phenotypes characterizing the early postimplantation embryos were similar to those observed in adult siblings, indicating that the phenotypes are well established prior to the developmental stages studied. While the X-inactivation pattern in somatic tissues of the embryo and adults is clearly affected by the mutations in both pedigrees, imprinted inactivation of the paternal X chromosome in extraembryonic tissues is not affected in family 24.21. These results indicate that the mutation in family 24.21 specifically affects the random (embryonic) form of X inactivation.
—Mutant X-inactivation phenotype in Xce homozygous females from both families. The X-inactivation phenotype at the Pctk1 locus in adult Xcec/c wild-type control females, in Xcec/c homozygous female progeny of family 1.19 carrier females, in adult Xcec/c (•), and in adult Xcea/a (○) female progeny of family 24.21 carrier females. An arrowhead depicts the mean X-inactivation pattern for each group. The Xcea/a and Xcec/c homozygous females from family 24.21 have the same mean. The X-inactivation pattern was measured with the Pctk1 assay.
Genetic composition of the candidate mutations in family 1.19: The identification of two regions of the genome that segregate independently with the mutant phenotype in family 1.19 raised the possibility that genetic modifiers may contribute to the mutant phenotype in this family. The prospect that a modifier is involved is especially noteworthy, as alleles from four different mouse strains [composed of two ENU-treated genomes, C57BL/6J and CBA/J (from the B6CBAF1/J founder), and the wild-type 129S1/SvJ and CAST/Ei genomes] segregate in this pedigree. C57BL/6J alleles segregate in the candidate region of chromosome 10 while CBA/J alleles segregate in the candidate region of chromosome 5. The multiple genetic strains segregating in this family may have encouraged the interaction of genetic modifiers between two strains or between an ENU-induced mutation and a modifier in the ENU-treated strain. While we cannot rule out that any of the candidate intervals in both pedigrees contain strain-specific modifiers rather than ENU-induced mutations, the identification of novel autosomal genes in the X-inactivation pathway is equally significant, whether resulting from the activity of modifiers or from ENU-induced mutations.
Candidate genes that are likely to be involved in the pathway of X chromosome inactivation include those that affect epigenetic silencing mechanisms through the modification of chromatin structure and function. Among such candidate genes on the distal interval of chromosome 5 are zinc-finger protein 70 (Zfp70), high mobility group box 1 (Hmgb1), breast cancer 2 (Brca2), heat-shock protein 25 kD (Hsp25), and zinc-finger proliferation 1 (Zipro1). Candidate genes on chromosome 10 include BRCA2-associated factor 35 (Braf35), histone deacetylase 2 (Hdac2), methyl CpG-binding domain protein 3 (Mbd3), DNA methyltransferase 3l (Dnmt3l), and zinc-finger protein, autosomal (Zfa). Candidate genes in the previously mapped pedigree 24.21 include RAD21 homolog (Rad21), zinc fingers and homeoboxes protein 1 (Zhx1), and KH-domain-containing, RNA-binding, signal-transduction-associated 3 (Khdrbs3) on chromosome 15. Significantly, the candidate regions do not contain known genes for factors previously shown to be involved in any aspect of X inactivation, indicating that the mutations segregating in pedigrees 1.19 and 24.21 affect novel factors in the X-inactivation pathway.
Models of mutation activity: The results of the mutant phenotyping data presented here clearly delineate the effect of the mutations and are consistent with models in which the candidate mutations target specific factors in the X-inactivation pathway. Furthermore, our genetic analysis of mutant animals bearing selected Xce alleles establishes that the effect of the mutant phenotypes is exclusive to the X-inactivation mechanism in Xce heterozygous animals that undergo nonrandom X chromosome choice.
A variety of models may be invoked to explain the effect of the mutations segregating on mouse chromosomes 15, 5, and 10, designated X-inactivation autosomal factor 1, 2, and 3 (Xiaf1, Xiaf2, and Xiaf3), respectively. In both pedigrees, the X-inactivation pattern in carrier Xce heterozygous females is shifted away from the nonrandom X-inactivation pattern of wild-type Xce heterozygous females toward a more random X-inactivation pattern. The simplest explanation for this change is a defect in Xce allelic discrimination. Consequently, one or all candidate mutations may affect (different) factors involved in distinguishing Xce alleles during X chromosome selection (Figure 6). It has long been hypothesized that during initiation of X chromosome silencing an autosomally encoded dosage-sensitive blocking complex protects one Xic from inactivation (perhaps through action at the Xce locus), thereby choosing that chromosome to remain active (Russell 1963; Brown and Chandra 1973; Russell and Cacheiro 1978; Avner and Heard 2001). It is not known how the distinction of Xce alleles in an Xce heterozygous animal occurs, although differential methylation, CTCF binding, and chromatin structure may play a role (Simmleret al. 1993; Avneret al. 1998; Prissetteet al. 2001; Chaoet al. 2002). In wild-type Xce heterozygotes, preferential interaction of the blocking complex with the stronger Xce allele (e.g., the Xcec allele) would result in nonrandom X-inactivation patterns (Figure 6). It is possible that the loci targeted by our mutations encode or affect members of the putative blocking complex and/or other autosomal factors that interact with Xce. The mutations described here, for example, may disrupt the ability of the blocking factor to differentiate between Xce alleles, resulting in a reversion to random (i.e., ∼50:50) X chromosome choice in mutant animals despite heterozygosity at Xce.
—Model of mutation activity. During X chromosome selection in wild-type Xce heterozygotes, a dosage-sensitive, autosomally encoded Xic blocking complex preferentially recognizes the stronger Xce allele, maintaining its activity and resulting in a nonrandom X-inactivation pattern. In pedigrees 24.21 and 1.19, autosomal-dominant mutations may lead to conformational change or to haplo-insufficiency of a factor within the mutant blocking complex. According to this model, the mutations abrogate preferential activity of the blocking complex and result in partial reversion to random X chromosome choice.
In an alternative model, the shift in the X-inactivation pattern in carrier Xce heterozygous females may not represent a change toward random X chromosome choice, but rather an increase in the variance of the X-inactivation pattern (Nesbitt 1971). In this scenario, the candidate mutations may cause initiation and/or stabilization of X inactivation at an earlier stage when fewer cells are present in the embryo. In a parallel scenario, it is possible that the mutations may disrupt the stability of the inactivation choice such that a cell may no longer be able to “lock-in” X chromosome choice and thereby manifest a more random X-inactivation pattern. According to this model, the initiation step of the X-inactivation pathway may be affected by the mutation(s) with the other steps of the pathway proceeding normally, implicating as candidates any factors involved in the initiation of X inactivation. This includes RNA-binding molecules that may interact directly with the Xist or Tsix transcripts or molecules involved in chromatin structure, such as those that stabilize or lock in the inactive state during the shift from reversible to irreversible inactivation (Wutz and Jaenisch 2000). Finally, it remains formally possible that the mutations lead to haplo-insufficiency of a dosage-sensitive factor involved early in the inactivation pathway, thereby resulting in altered X-inactivation patterns. Positional cloning and characterization of the three Xiaf genes identified here will be required to discriminate among these or other models.
Conclusion: Our study of ENU-induced mutations of X chromosome inactivation has identified three mutations that act early in development in an autosomal-dominant manner to induce X-inactivation patterns in Xce heterozygous animals that are significantly different from those predicted by their Xce genotypes. The mutations do not affect the X chromosome inactivation pattern in Xce homozygous animals, indicating that they may have targeted factors involved in Xce allelic discrimination during the X chromosome choice process. Significantly, neither imprinted X inactivation nor genomic imprinting is perturbed in mutant animals. Consequently, the phenotypic analyses presented here are consistent with models in which the mutations have affected molecules involved specifically in the choice or initiation steps of the X-inactivation pathway. Through our genomic mapping analysis of the two pedigrees we have shown that at least three novel autosomal loci are involved in the pathway of X inactivation, as none of the candidate regions harbor genes known to be involved in the process of inactivation. The identification of factors targeted by these mutations and other epigenetic mutations detected through such ENU mutagenesis screens should play a significant role in our understanding of epigenetic regulation of the mammalian genome.
Acknowledgments
We thank Monica Justice and Jim Amos-Landgraf for helpful assistance with mutagenesis; Warren Ewens, David Beier, and Richard Spielman for statistical advice; Mellissa Mann for assistance with the Snrpn assay and the Makorin3 and Cdkn1c polymorphisms; and Keith Olszens for animal husbandry and dissections. We also appreciate helpful discussions and critical reading of the manuscript by Maja Bucan. This work was supported by National Institutes of Health (NIH) grant GM-45441 (H.F.W.) and by the Howard Hughes Medical Institute (M.S.B.). I.P. was supported by an NIH Radiation Biology predoctoral training grant.
Footnotes
-
Communicating editor: N. A. Jenkins
- Received January 27, 2003.
- Accepted April 4, 2003.
- Copyright © 2003 by the Genetics Society of America