The Drosophila homolog of the mammalian imprint regulator, CTCF, maintains the maternal genomic imprint in Drosophila melanogaster

Background CTCF is a versatile zinc finger DNA-binding protein that functions as a highly conserved epigenetic transcriptional regulator. CTCF is known to act as a chromosomal insulator, bind promoter regions, and facilitate long-range chromatin interactions. In mammals, CTCF is active in the regulatory regions of some genes that exhibit genomic imprinting, acting as insulator on only one parental allele to facilitate parent-specific expression. In Drosophila, CTCF acts as a chromatin insulator and is thought to be actively involved in the global organization of the genome. Results To determine whether CTCF regulates imprinting in Drosophila, we generated CTCF mutant alleles and assayed gene expression from the imprinted Dp(1;f)LJ9 mini-X chromosome in the presence of reduced CTCF expression. We observed disruption of the maternal imprint when CTCF levels were reduced, but no effect was observed on the paternal imprint. The effect was restricted to maintenance of the imprint and was specific for the Dp(1;f)LJ9 mini-X chromosome. Conclusions CTCF in Drosophila functions in maintaining parent-specific expression from an imprinted domain as it does in mammals. We propose that Drosophila CTCF maintains an insulator boundary on the maternal X chromosome, shielding genes from the imprint-induced silencing that occurs on the paternally inherited X chromosome. See commentary: http://www.biomedcentral.com/1741-7007/8/104


Background
The correct establishment and propagation of epigenetic states are essential for normal development, and disruption of these processes leads to disease. Genomic imprinting is a striking example of the effect of epigenetics on gene regulation. In genomic imprinting, a mark, the imprint, is imposed on the two parental genomes during gametogenesis. In the zygote, the imprint is maintained through each mitotic division and results in the parental alleles of a gene, or entire homologous chromosomes, adopting different epigenetic states. As a result of these different epigenetic states, one parental allele can be silenced while the allele from the other parent, although identical in DNA sequence, is active.
The CCCTC-binding factor, CTCF, is a key player in maintaining epigenetically distinct chromatin domains. CTCF is an evolutionarily conserved zinc finger-containing DNA-binding protein that can function both directly in gene regulation as a transcription factor and also indirectly by mediating long-range chromatin interactions. In this latter role, CTCF acts as a chromatin insulator by isolating enhancer and promoter regulatory units and as a barrier to the spread of heterochromatin [1]. CTCF binds at multiple sites throughout the genome [2][3][4], indicating a widespread role in generating chromatin domains. Epigenetic isolation is necessary for correct maintenance of genomic imprints as imprinted domains are often interspersed among nonimprinted domains [5,6], necessitating their isolation from flanking regulatory regions. Additionally, the two homologous alleles must be isolated as differential gene expression patterns, chromatin conformations, and replication timing have all been associated with imprinted alleles [7]. CTCF binding has been reported at multiple mammalian imprinted domains [5], illustrating the importance of insulator function in maintaining parent-specific expression. The role of CTCF in imprinting has been best characterized for the mammalian Igf2/H19 genes, in which only the maternal H19 allele and paternal Igf2 alleles are expressed [8][9][10]. On the maternal chromosome, CTCF binds to a differentially methylated domain (DMD) located between the Igf2 and H19 genes, preventing interaction of downstream enhancer sequences with the promoter of Igf2, effectively silencing the gene. Methylation of the paternal DMD effectively blocks CTCF binding, allowing activation of Igf2 expression while also initiating the silencing of the H19 gene. Binding of CTCF is necessary to maintain the epigenetic state of the imprinted alleles. Consequently, if the CTCF binding site in the Igf2/H19 DMD is mutated, the monoallelic expression arising from the imprint is lost [11,12]. An additional facet of CTCF binding appears to be the facilitation of higher-order chromatin structures through DNA looping, a property which fortifies the silencing of Igf2 and the activation of H19 on the maternal chromosome [13,14]. The details of CTCF binding and its consequences are less well studied at other imprinted loci; however, its insulator function and role in establishing higher-order chromatin function appear to be shared features of other mammalian imprinted loci which bind CTCF [5,[15][16][17]. The KvDMR1 imprinted domain, which contains two CTCF binding sites, regulates the tissue-specific expression of the gene Cdkn1c. It has been suggested that the tissue-specific imprinting of Cdkn1c is due to tissue-specific binding of CTCF to the KvDMR1 imprint domain [18,19]. The imprinted domain Wsb1/Nf1 also requires CTCFmediated interchromosomal association with the Igf2/ H19 imprinted domain for proper parent-specific expression [20].
Although CTCF appears to be the major insulator protein in vertebrates, the more compact Drosophila genome uses a variety of insulator proteins, among which is the Drosophila CTCF homolog dCTCF [21][22][23]. The insulator activity of dCTCF has been well characterized in the bithorax complex, where it demarcates the chromatin domains that define separate regulatory regions [22,24,25], and, as in mammals, dCTCF is widely used as an insulator throughout the Drosophila genome and also acts directly as a transcription factor [26][27][28]. Though the role of CTCF in the formation of distinct chromatin domains is conserved from Drosophila to mammals, the roles of CTCF in epigenetic processes such as genomic imprinting have been assumed to differ [1]. To assess the effect of dCTCF on Drosophila imprinting, we used a well-characterized imprinting assay system, the Dp(1;f)LJ9 mini-X chromosome, in which a readily visible eye color gene, garnet (g), is juxtaposed to an imprint control region and so becomes a marker for imprinting [29]. Regulation of the Dp(1;f)LJ9 imprint previously has been shown to share properties of mammalian imprinting, including transcriptional silencing of gene clusters and differential chromatin states between homologues [30][31][32]. Here we present the first demonstration that dCTCF has a role in the regulation of genomic imprinting in Drosophila. As is the case in mammalian imprinting, dCTCF in Drosophila is involved in the regulation of the maternal imprint by maintaining parent-specific expression from the maternally inherited X chromosome.

Characterization of CTCF alleles
The CTCF EY15833 allele (FBrf0132177) was produced by insertion of the P{EPgy2} element into the +26 position relative to the transcription start site of the dCTCF gene by the Berkeley Drosophila Genome Project (BDGP) Gene Disruption Project [33]. Homozygous CTCF EY15833 adults appear healthy and are reasonably fertile. CTCF 30 , created by a partial deletion of the P{EPgy2} element, is homozygous lethal. CTCF 30 lacks the entire 5′ end of the P{EPgy2} element but retains 4860 bp of the 3′ end. Flanking dCTCF sequences and the quemao (qm) gene remain intact in CTCF 30 . The reduced severity of CTCF EY15833 , with an intact P{EPgy2} element, suggests that a promoter in the 5′ end of P{EPgy2} may partially rescue dCTCF expression. To test this idea, we measured the dCTCF transcript levels by quantitative real-time PCR (qRT-PCR) of third instar larvae. Expression in homozygous CTCF EY15833 larvae is 20% ± 4% of that in wild-type (y w) controls. This is consistent with previous studies showing that CTCF EY15833 homozygotes producẽ 50% of wild-type dCTCF protein levels [22]. Homozygous CTCF 30 larvae cannot be recovered in sufficient numbers for qRT-PCR, but heterozygous CTCF 30 /+ larvae display 68 ± 9% of wild-type transcript levels, consistent with a severe reduction in expression by this mutation. Taken together, the phenotypic and expression analysis of dCTCF alleles indicates that both CTCF EY15833 and CTCF 30 alleles have reduced dCTCF expression and that the 5′ end of P{EPgy2} may drive sufficient expression to allow recovery of CTCF EY15833 adults.
Drosophila CTCF maintains the maternal imprint of the garnet gene on the Dp(1;f)LJ9 mini-X chromosome To test the effect of the dCTCF alleles on Drosophila imprinting, we used the Dp(1;f)LJ9 mini-X chromosome.
Maternal inheritance of the Dp(1;f)LJ9 mini-X chromosomes (Dp(1;f)LJ9 MAT ) generate full expression of the marker gene garnet, whereas paternal inheritance (Dp(1; f)LJ9 PAT ) generates variegated garnet expression ( Figures  1a and 1b, control). The variegated garnet gene phenotype arising from paternal transmission is mitotically stable and so results in distinct clonal regions exhibiting garnet expression in an eye devoid of garnet expression. Expression of garnet affects both red (pteridine) and brown (ommochrome) eye pigments, which makes this mini-X chromosome an easily assayed system in which to assess the effect of dCTCF alleles on imprinting in Drosophila.
Transmission of the Dp(1;f)LJ9 mini-X chromosome through the female results in y 1 z a g 53d /Dp(1;f)LJ9; +/+ mini-X chromosome-bearing male progeny with essentially wild-type expression of the garnet imprint marker gene. Eyes are phenotypically wild type, with 85.3 ± 1.4% wild-type red pigment levels and 85.7 ± 3.4% wildtype brown pigment levels (Figure 1a, control). To determine the effects of dCTCF on the maternal maintenance of imprinted garnet expression, the CTCF X alleles (X represents CTCF 30 or CTCF EY15833 ) were crossed to females with the Dp(1;f)LJ9 mini-X chromosome: y 1 z a g 53d /Y; CTCF X /TM3, Sb Ser males x X^X/Dp (1;f)LJ9 females. This cross-generated progeny with a mutant dCTCF allele and a maternally imprinted mini-X chromosome (y 1 z a g 53d /Dp(1;f)LJ9 MAT ; CTCF X /+), which were compared with progeny similarly carrying a maternally imprinted chromosome, but wild type for dCTCF.
For each dCTCF allele tested, the mutant allele substantially reduced expression of the maternally transmitted imprint marker gene (Figure 1a). Progeny with a maternally inherited mini-X chromosome (Dp(1;f) LJ9 MAT ) with CTCF EY15833 reduced pigment levels to 67.2 ± 3.1% (P < 0.001) and 68.4 ± 4.2% (P < 0.001) for red and brown pigments, respectively. Dp(1;f)LJ9 MAT progeny coupled with CTCF 30 resulted in an even greater reduction of pigment levels: 58.9 ± 3.1% (P < 0.001) and 56.9 ± 4.7% (P < 0.001) for red and brown pigments, respectively. No variegated garnet Figure 1 Effect of dCTCF alleles on the imprinted Dp(1;f)LJ9 garnet (g) gene expression. (a) Maternally transmitted Dp(1;f)LJ9 mini-X chromosome; the control (y 1 z a g 53d /Dp(1;f)LJ9) displays full garnet expression with no variegation observed. Both dCTCF mutant alleles tested (y 1 z a g 53d /Dp(1;f)LJ9; CTCF EY15833 /+ and y 1 z a g 53d /Dp(1;f)LJ9; CTCF 30 /+) disrupt maintenance of the maternal imprint, causing variegated garnet gene expression. Significant reduction in both red and brown pigment levels is observed in the presence of CTCF EY15833 or CTCF 30 alleles. (b) Paternally transmitted Dp(1;f)LJ9 mini-X chromosome; the control (y 1 z a g 53d /Dp(1;f)LJ9) exhibits variegated garnet gene expression, whereas the introduction of CTCF EY15833 and CTCF 30 alleles had no significant effect on garnet gene variegation. Pigment assay values are expressed as a percentage of wild-type pigment levels ± standard deviation. Values that are significantly different from the controls are marked with an asterisk signifying P < 0.001. expression was observed in flies with Dp(1;f)LJ9 MAT and wild type for dCTCF (Figures 2a and 2b). However, when Dp(1;f)LJ9 MAT was inherited along with mutant dCTCF alleles, variegated garnet expression was observed (Figures 2a and 2b). These results demonstrate that the maintenance of the maternal imprint is highly sensitive to dCTCF dosage.
This cross also produced sibling progeny that have a maternally inherited Dp(1;f)LJ9 MAT , but with the balancer chromosome, and so wild type for dCTCF (described in "Methods," genotype: y 1 z a g 53d /Dp(1;f) LJ9 MAT ; TM3, Sb Ser /+). These internal control flies are genotypically identical to the external controls but have the male parent mutant for CTCF X and so would allow detection of any paternal effect. None was detected (Figures 3a and 3b).

Figure 3
Absence of maternal or paternal effects from mutant dCTCF on Dp(1;f)LJ9 garnet expression. External control progeny (no modifier allele) have the same genotype as internal control progeny (y 1 z a g 53d /Dp(1;f)LJ9; TM3, Sb Ser/+) but are generated from a separate cross with parents that have never encountered a mutant CTCF X allele. (a) Maternally inherited Dp(1;f)LJ9 CTCF X internal control eye pigment levels; no significant difference in pigment levels was observed between external control progeny (no modifier allele) and internal control progeny from fathers carrying CTCF X (CTCF X internal), demonstrating that no paternal effect occurs. (b) Phenotypes of maternally inherited Dp(1;f)LJ9 ranging from 0% to 100% pigmentation; No modifier allele n = 300, CTCF EY15833 internal n = 300, CTCF 30 internal n = 300. No garnet variegation was detected from the internal controls. (c) Paternally inherited Dp(1;f)LJ9 CTCF internal control eye pigment levels; no significant difference in pigment levels was observed between the external control progeny (no modifier allele) and internal control progeny from mothers carrying CTCF X (CTCF X internal), demonstrating that no maternal effect occurs. (d) Phenotypes of paternally inherited Dp(1;f)LJ9 ranging from 0% to 100% pigmentation; no modifier allele n = 300, CTCF EY15833 internal n = 300, CTCF 30 internal n = 300. No significant change in garnet variegation was detected from the internal controls. Red eye pigment levels are measured by absorbance at 480 nm, and pigment quantification mean values for each group are based on n = 5 samples (40 heads total); error bars represent standard deviation.
contrast to the effects observed when the Dp(1;f)LJ9 is maternally imprinted, dCTCF mutants had no effect on the paternal expression of the garnet imprint marker gene. Paternal inheritance of the mini-X chromosome (X^Y/Dp(1;f)LJ9 males crossed to y 1 z a g 53d /y 1 z a g 53d ; TM3, Sb Ser/+ females) results in y 1 z a g 53d /Dp(1;f)LJ9 PAT ; +/+ progeny with variegated garnet expression and a marked reduction in eye pigment levels (35.5 ± 1.5% red and 52.5 ± 0.2% brown wild-type pigment levels; Figure 1b, control). The introduction of dCTCF mutant alleles (y 1 z a g 53d /y 1 z a g 53d ; CTCF X /TM3, Sb Ser females to X^Y/ Dp(1;f)LJ9 males) to generate progeny with either mutant CTCF EY15833 or CTCF 30 alleles and a paternally imprinted Dp(1;f)LJ9 PAT mini-X chromosome (y 1 z a g 53d / Dp(1;f)LJ9; CTCF X /+), yielded no significant change in either red or brown eye pigment levels or phenotype (Figures 1b and 2c).
Sibling progeny with a paternally inherited Dp(1;f) LJ9 PAT along with the balancer chromosome are wild type for dCTCF (described in "Methods," genotype: y 1 z a g 53d /Dp(1;f)LJ9; TM3, Sb Ser/+) and can be used to determine whether there is a maternal effect from mothers mutant for CTCF X . No maternal effect was detected; progeny wild type for dCTCF from mothers with either CTCF EY15833 or CTCF 30 showed no significant change in garnet expression levels (Figures 3c  and 3d).
Drosophila CTCF does not regulate the establishment of the maternal or paternal imprint of the Dp(1;f)LJ9 mini-X chromosome To determine whether the effect of dCTCF was on the somatic maintenance of the imprint or its establishment in the germline of the parents, we examined the phenotype of progeny from male or female parents with both the Dp(1;f)LJ9 mini-X chromosome and a mutant CTCF 30 allele. If dCTCF affects the establishment of the imprint, the imprint should be disrupted in the progeny of mutant CTCF 30 parents, but not wild-type (CTCF + ) parents. When we compared the phenotype of progeny wild type for dCTCF but differing in their parental genotype, no significant alternation in garnet expression levels resulted between Dp(1;f)LJ9 MAT progeny from mothers carrying CTCF 30 (Figure 4a; Mat-Est. CTCF 30 ) and either of the external or internal controls wild type for dCTCF (Figure 4a; Ex. Control and Mat-Est. CTCF + , respectively). This was reflected in the unchanged phenotype of progeny from mothers mutant or wild type for dCTCF (Figure 4b; Mat-Est. CTCF 30 and Mat-Est. CTCF + , respectively). Likewise, mutant dCTCF did not effect the establishment of the paternal imprint. Dp(1;f) LJ9 PAT progeny from fathers carrying Dp(1;f)LJ9 and CTCF 30 (Figure 4c; Pat-Est. CTCF 30 ) had no significant change in garnet expression compared with either the external or internal controls (Figure 4c; Ex. Control and Pat-Est. CTCF + , respectively). Again, these Dp(1;f)LJ9 PAT progeny also had no observable change in phenotype between fathers mutant or wild type for dCTCF ( Figure  4d; Pat-Est. CTCF 30 and Pat-Est. CTCF + , respectively). These findings distinguish the function of dCTCF in the maintenance versus the establishment of the imprint on the Dp(1;f)LJ9 mini-X chromosome; dCTCF is involved in the maintenance of the imprint in the soma of progeny as its reduction disrupts the maternal imprint. However, dCTCF is not involved in the establishment of the imprint as the presence of mutant CTCF 30 in either the maternal or paternal germline during establishment of the imprint does not affect regulation of the imprint.

Drosophila CTCF is not a general modifier of positioneffect variegation
To determine the effect of dCTCF mutant alleles on the Dp(1;f)LJ9 MAT imprint, we tested the effect of CTCF 30 on In(1)w m4 , a classical variegating rearrangement [34], and two fourth chromosome transgenic constructs [35] in which the white (w) gene is variegated. Like the Dp(1; f)LJ9 mini-X chromosome, the variegated silencing in In (1)w m4 is induced by the centric heterochromatin of the X chromosome [35] and the fourth chromosome has been proposed to be evolutionarily related to the X chromosome [36]. We found that the CTCF 30 allele decreased silencing of white in In(1)w m4 ;CTCF 30 /+ females while having no significant effect on white expression levels in In(1)w m4 ;CTCF 30 /+ males compared with sibling In(1)w m4 ;Tb/+ controls (Figure 5a). Similarly, the 6-M193 strain responded to CTCF 30 with a modest decrease in white reporter silencing in females only (Figure 5b), whereas the 39C-33 strain showed no significant change in white reporter silencing from CTCF 30 (Figure 5c). These results demonstrate that CTCF 30 is not a ubiquitous modifier of variegated heterochromatic silencing in Drosophila, consistent with the absence of an effect on silencing of the paternally inherited Dp(1;f)LJ9 mini-X chromosome. Furthermore, the decreased silencing of the nonimprinted variegators is opposite to the effect of CTCF 30 on Dp(1;f)LJ9 MAT silencing. Thus, the role of dCTCF in the maintenance of the maternal Dp(1;f)LJ9 imprint represents a distinct parent-specific function for dCTCF on the imprinted Dp (1;f)LJ9 mini-X chromosome.

Discussion
CTCF is essential for insulator function in vertebrates, where it plays an active role in regulating imprinted gene expression. In Drosophila, dCTCF has likewise been shown to be involved in the insulator function of boundary elements [28,37]. Our results show that parent-specific expression from an imprinted domain in Drosophila is dependent on dCTCF function. Maintenance of expression from maternally inherited Dp(1;f) LJ9 mini-X chromosome is highly sensitive to dCTCF; even a modest decrease in dCTCF mRNA alters the maternal imprint so that it resembles the paternal imprint.
The effect of dCTCF on maternal-specific expression is limited to the maintenance of imprint. The presence of mutant dCTCF in either the maternal or paternal parents, when the imprint is being established, does not affect the imprint in the progeny. These results are strikingly similar to the role of CTCF in mammalian imprinting, where CTCF assists in the postfertilization formation of an imprinted region, but is dispensable for the establishment of an imprint [11,12,38].
Furthermore, the requirement for dCTCF for maintenance of the maternal Dp(1;f)LJ9 imprint is specific and does not represent a ubiquitous role for dCTCF in regulating heterochromatic silencing. Not only is the paternal Dp(1;f)LJ9 imprint unaffected by mutant dCTCF, but other variegating Drosophila reporter genes respond differently to mutant dCTCF. Thus, the association of dCTCF expression with the maintenance of the maternal Dp(1;f)LJ9 imprint boundary demonstrates a distinct function for dCTCF in imprinted gene expression.
In mammals, maternally imprinted regions that bind CTCF rely critically on this binding to insulate the imprinted loci and establish distinct chromatin domains. Our results show that a reduction in dCTCF levels disrupts the maternal imprint boundary on the Drosophila Dp(1;f)LJ9 mini-X chromosome, and consequently the marker gene, garnet, is silenced. Variegated silencing of garnet from Dp(1;f)LJ9 PAT inheritance is a consequence of heterochromatin formation, nucleated from the paternal imprint control region, spreading in cis [29,32]. The Figure 4 dCTCF does not affect establishment of the Dp(1;f)LJ9 imprint. All genotypes tested are y 1 z a g 53d /Dp(1;f)LJ9; +/+ but differ in parental genotype. External control progeny (Ex. Control) are generated from parents carrying Dp(1;f)LJ9 that have never been exposed to a mutant dCTCF allele. Progeny generated from a parent carrying both the imprinted Dp(1;f)LJ9 mini-X chromosome and a mutant dCTCF allele test for effects on imprint establishment (Mat or Pat-Est. absence of an effect upon the introduction of dCTCF mutant alleles to Dp(1;f)LJ9 PAT suggests that dCTCF binding and boundary function occurs only on the maternal chromosome. Thus, it is conceivable that a reduction in dCTCF levels enables the spreading of heterochromatin on the maternal Dp(1;f)LJ9 MAT in a manner similar to that of the paternal Dp(1;f)LJ9 PAT . This would suggest that dCTCF defines the boundary of a distinct maternal-specific imprinted chromatin domain required to maintain maternal-specific gene expression on the X chromosome.
The model organism Encyclopedia of DNA Elements (modENCODE) project provides detailed mapping of regulatory elements throughout the Drosophila genome [39]. Large-scale profiling of dCTCF insulator sites from early embryo modENCODE data reveals several candidate dCTCF insulator sites present proximal to the predicted heterochromatic breakpoint of the Dp(1;f)LJ9 mini-X chromosome. These dCTCF insulator sites, located between the centric heterochromatic imprinting center and the imprint marker gene garnet, could account for the sensitivity of the maternal imprint to dCTCF expression. If dCTCF were bound only when the X chromosome was transmitted maternally, mutations to dCTCF would disrupt insulator function and lead to maternal silencing of the imprint marker gene. Although such binding remains to be tested, it is similar to the function of CTCF at mammalian imprinted regions.
That the structure of CTCF and its role as an insulator, barrier, and transcriptional regulator is conserved between mammals and insects have been well established [23,27,28]. However, the finding that CTCF maintains its function in regulating the imprinting of diverse genes in such phylogenetically distinct organisms is remarkable. CTCF is a versatile DNA binding factor; subsets of its zinc fingers are adept at binding diverse DNA sequences, and the rest of the protein is able to maintain common regulator interactions and insulator function [40]. This feature may explain how CTCF can regulate imprinting in organisms as diverse as insects and mammals, in which the imprinted target sequences are different.
Previously, the evolutionary origin of imprinting has been extrapolated from the conservation of imprinting among specific genes. Such studies have led to the proposal that mammalian imprinting is of relatively recent origin and restricted to eutherian mammals [41,42]. However, studies showing that the molecular mechanism of imprinting is highly conserved have suggested a much more ancient origin [7,30,43]. Mammalian imprint control elements inserted into transgenic Drosophila act as discrete silencing elements [44,45] and can retain posttranscriptional silencing mechanisms involving Figure 5 The effect of CTCF 30 on white variegation in In(1)w m4 and the fourth chromosome variegating strains 6-M193 and 39C-33. (a) Pigment levels were measured for In(1)w m4 /w 1118 ; CTCF 30 /+ and In(1)w m4 /Y;CTCF 30 /+ genotypes compared with the corresponding sibling In(1)w m4 ;Tb/+ controls. In(1)w m4 heterozygous for CTCF 30 results in an increase in white expression; however, this increase is only significant in female progeny (white bars). Red pigment quantification mean values for each group are based on n = 10 (50 heads total). Error bars represent standard deviation, and values significantly different from the controls are marked with an asterisk (P < 0.001). (b) Pigment levels were independently measured for both the maternally (6-M193 MAT ) and paternally (6-M193 PAT ) inherited fourth chromosome variegator 6-M193. 6-M193 heterozygous for CTCF 30 results in an increase in white expression when either maternally or paternally inherited; however, this increase is only significant in female progeny (white bars). Red pigment quantification mean values for each group are based on n = 9 (45 heads total) for 6-M193 MAT male and female progeny; n = 10 (50 heads total) for 6-M193 PAT male and female progeny; 6-M193; +/+ control male progeny; and n = 12 (60 heads total) for 6-M193; +/+ control female progeny. Error bars represent standard deviation, and values significantly different from the controls are marked with an asterisk (P < 0.001). (c) Pigment levels were independently measured for both the maternally (39C-33 MAT ) and paternally (39C-33 PAT ) inherited fourth chromosome variegator 39C-33. 39C-33 heterozygous for CTCF 30 results in no significant change in white expression when either maternally or paternally inherited. Red pigment quantification mean values for each group are based on n = 10 (50 heads total) for 39C-33 MAT and 39C-33 PAT male and female progeny, and n = 20 (100 heads total) for 39C-33;+/+ control male and female progeny. Error bars represent standard deviation.
noncoding RNA [46]. Whereas these transgenic imprinting elements lose their parent-specific functions, the retention of epigenetic silencing mechanisms suggests an ancient and conserved origin of imprinting mechanisms. Our finding that CTCF has a role in the maintenance of maternal imprints in insects, as it does in mammals, supports the possibility of evolutionary conservation for both CTCF function and the mechanisms of genomic imprinting.

Conclusions
CTCF is a multifunctional protein with a conserved role as a chromosomal insulator in both mammals and Drosophila. To determine whether dCTCF is involved in imprinted regulation in Drosophila as it is in mammals, we generated a dCTCF mutant allele with severe reduction in dCTCF expression and tested its effects on the expression of the imprint marker gene, garnet, on the Dp(1;f)LJ9 mini-X chromosome. Full garnet gene expression, which occurs when the Dp(1;f)LJ9 mini-X chromosome is maternally inherited, was disrupted when dCTCF expression levels were reduced. No effect of reduced dCTCF expression was observed on the Dp (1;f)LJ9 mini-X chromosome when it was inherited paternally. The effect of dCTCF mutations is on the maintenance rather than on the establishment of the imprint, and is specific to the Dp(1;f)LJ9 mini-X chromosome. These results demonstrate that dCTCF is involved in maintaining parent-specific expression from the maternally inherited X chromosome in Drosophila, a role paralleling its involvement in mammalian imprinting.

Drosophila culture
All crosses were maintained at 22°C and cultured on standard cornmeal-molasses Drosophila media with methyl benzoate (0.15%) as a mold inhibitor. Each set of crosses was performed in 55-ml shell vials and contained 10-15 virgin females and 10-15 males. Each of the crosses was subcultured three or four times at 3-day intervals before the parents were discarded. Each cross was replicated four to six times, and the progeny were pooled. All stocks were obtained from the Bloomington Drosophila stock center, with the exception of the CTCF 30 allele and the variegating fourth chromosome transgene strains. The CTCF EY15833 allele was created by insertion of P{EPgy2} 27 bp downstream of the dCTCF transcription start site. The homozygous lethal CTCF 30 allele was generated by imprecise excision of CTCF EY15833 . The CTCF 30 deletion was characterized by amplification across the break and sequencing of the PCR product. CTCF 30 is deleted for all 5′ transposon sequences but retains 4860 bp at the 3′ end. No genomic sequence was removed by the CTCF 30 deletion. The fourth chromosome variegating strains were generated using the transposable P element P [hsp26-pt, hsp70-w], which contains the hsp70-driven white gene that is susceptible to silencing caused by heterochromatin formation [47]. The 6-M193 strain has the construct inserted within a 1360 transposon and inside the Syt7 gene (fourth chromosome coordinate: 323400), whereas the 39C-33 strain is generated from the construct being inserted into gene of the RNA binding protein gawky (fourth chromosome coordinate: 680211), which is in close proximity to a 1360 transposon [47].
To determine the effect of mutant dCTCF on imprint maintenance, the CTCF 30 and CTCF EY15833 alleles were crossed into y 1 z a g 53d background to yield stable stocks of y 1 z a g 53d /y 1 z a g 53d ; CTCF X /TM3, Sb Ser (where CTCF X is the CTCF 30 or CTCF EY15833 allele). To test the effect of a dCTCF allele on the paternal imprinting of garnet, Dp(1;f)LJ9, y + g + /X^Y males were crossed to y 1 z a g 53d / y 1 z a g 53d ; CTCF X /TM3, Sb Ser females, and the reciprocal cross with Dp(1;f)LJ9, y + g + /X^X virgin females was performed to test the effect on the maternal imprinting of garnet ( Figure 6). y 1 z a g 53d /Dp(1;f)LJ9; CTCF X /+ male progeny were collected on the basis of wild-type yellow (y + ) body color, which independently confirms the presence of the Dp(1;f)LJ9 chromosome, whereas the zeste allele (z a ) reduces background eye color of the g allele (g 53d ). The y 1 z a g 53d /Dp(1;f)LJ9; TM3, Sb Ser/+ sibling males were used as internal controls ( Figure 6). The "no modifier" control test cross for paternal garnet imprinting consisted of Dp(1;f)LJ9, y + g + /X^Y males crossed to y 1 z a g 53d /y 1 z a g 53d ; TM3, Sb Ser/+ females, with the reciprocal cross serving as the maternal control: y 1 z a g 53d /Dp (1;f)LJ9; +/+ and y 1 z a g 53d /Dp(1;f)LJ9; TM3, Sb Ser/+ male progeny were collected as controls.
To test for the effects of CTCF on germline imprint establishment, mutant CTCF must be present in parents carrying the Dp(1;f)LJ9 mini-X chromosome. To detect the effect of CTCF 30 on the establishment of the imprint, Dp(1;f)LJ9; e/e flies were balanced over X^X; CTCF 30 /e for maternal establishment (Figure 7), or X^Y; CTCF 30 /e for paternal establishment (Figure 8). X^X/Dp (1;f)LJ9; CTCF 30 /e females were crossed to y 1 z a g 53d /Y males to test maternal imprint establishment (Mat-Est. CTCF 30 ), and the reciprocal cross tested for paternal imprint establishment (Pat-Est. CTCF 30 ). Maternal establishment controls (Mat-Est. CTCF + ) consisted of X^X/Dp(1;f)LJ9; e/e females crossed to y 1 z a g 53d /Y males, and paternal establishment controls (Pat-Est. CTCF + ) were X^Y/Dp(1;f)LJ9; e/e males crossed to y 1 z a g 53d / y 1 z a g 53d females. External controls were also produced by crossing F1 generation X^X/Dp(1;f)LJ9; e/e females to y 1 z a g 53d /Y males for maternal establishment, and the reciprocal cross for paternal establishment. Figure 6 Mating schematic for testing the effect of dCTCF on the maintenance of the Dp(1;f)LJ9 imprint. Two sets of progeny are generated from this cross: progeny that have independently inherited the Dp(1;f)LJ9 mini-X chromosome and a CTCF X mutant allele (modifier progeny) and progeny that have inherited Dp(1;f)LJ9 and the TM3, Sb Ser balancer, but had a parent carrying a CTCF X mutant allele (maternal and paternal effect test progeny).

Measurement of dCTCF expression
qRT-PCR was used to measure dCTCF expression. Total RNA was prepared from three groups of 50 larvae for each genotype. One microgram of total RNA was Figure 7 Mating schematic for testing dCTCF for an effect on maternal establishment of the Dp(1;f)LJ9 imprint. Two primary sets of progeny and an external control were generated from this cross: progeny with a maternally-inherited Dp(1;f)LJ9 mini-X chromosome from mothers with the CTCF 30 mutation (Mat-Est. CTCF 30 ) and progeny that also have a maternally imprinted Dp(1;f)LJ9 chromosome but from mothers wild type for CTCF (Mat-Est. CTCF + ). External control crosses were produced by crossing F1 generation X^X/Dp(1;f)LJ9; e/e females to y 1 z a g 53d /Y males (not depicted).
reverse transcribed using random hexamers and ImProm-II reverse transcriptase (Promega). Quantitative PCR was performed as previously described [48]. dCTCF primers (CTCF F2400, ACGAGGAGGT GTTGGTCAAG and CTCF R2485, ATCATCG TCGTCCTCGAAC) were used at 300 nM. Two technical replicates from each sample were amplified. Expression was normalized to Dmn, a gene that has proved reliable for this purpose [48].

Quantification of eye pigment levels
Expression of the imprint marker gene garnet was quantified both visually and through the use a spectrophotometric assay of extracted eye pigments. The visual assay assigns each eye a score in relation to its variegation class as described by Joanis and Lloyd [32]; 0-25% pigmentation, 25-50% pigmentation, 50-75% pigmentation, and 75-100% pigmentation. The prevalence of each variegation class is expressed as a percentage of all eyes assayed. As the variegated phenotype of maternally inherited Dp(1;f)LJ9 in the presence of mutant dCTCF is skewed toward fully pigmented eyes, a second assay with the following variegation classes was performed: >80% pigmentation, 80-90% pigmentation, 90-100% pigmentation, and 100% pigmentation.
The spectrophotometric assay was adapted from Real et al. [49]. For red (pteridine) pigment, flies of each test genotype were aged for 4 days and then placed in Figure 8 Mating schematic for testing dCTCF for an effect on the paternal establishment of the Dp(1;f)LJ9 imprint. Two primary sets of progeny and an external control were generated from this cross: progeny with a paternally-transmitted Dp(1;f)LJ9 mini-X chromosome from fathers with the CTCF 30 mutation (Pat-Est. CTCF 30 ) and progeny that also have a paternally imprinted Dp(1;f)LJ9 mini-X chromosome but from fathers wild type for CTCF (Pat-Est. CTCF + ). External control crosses were produced by crossing F1 generation X^Y/Dp(1;f)LJ9; e/e males to y 1 z a g 53d /y 1 z a g 53d females (not depicted).
1.5-ml Eppendorf microtubes and stored at -30°C. For each sample set, eight heads were placed into a 0.6-ml microtube containing 400 μl of acidified ethanol (30% EtOH, acidified to pH 2 with HCl). Pigment was extracted on an orbital shaker at 150 rpm in the dark for 48 hours. Absorbance of the extracted pigments was measured at 480 nm. Each 400-μl sample of extracted pigment was split into two 200-μl volumes, independently measured, and the values were averaged. Five tubes were run per sample set, with the values averaged and expressed as a percentage of wild-type pigment levels ± standard deviation. For brown (ommochrome) pigment, flies of each test genotype were aged for 4 days and then placed in 1.5-ml Eppendorf microtubes and stored at -80°C. Ten heads were placed in a 1.5-ml Eppendorf tube and homogenized with 150 μl of 2 M HCl and 0.66% sodium metabisulfite (wt/vol). A total of 200 μl of 1-butanol was added, and the mixture was placed on an orbital shaker at 150 rpm for 30 min before being centrifuged at 9000 g for 5 min. The organic layer was removed, washed with 150 μl of 0.66% sodium metabisulfite in dH 2 O and placed back on the orbital shaker for a further 30 min, and then this step was repeated for a second wash. The organic layer was removed and measured for absorbance at 492 nm. Five tubes were run per sample set, with the values averaged and expressed as a percentage of wild-type (O. R) pigment levels ± standard deviation. Absorbance was determined with a Pharmacia Biotech Ultrospec 2000 spectrophotometer. Representative eye pictures were photographed with a Zeiss AxioCam MRc5 mounted on a Zeiss Stemi 2000-C dissecting microscope.
Spectrophotometric assay for quantifying the expression of the white transgene on the fourth chromosome variegating strains and In(1)w m4 followed the same procedure for fly aging and head collection. Heads were split into groups of five and placed into 150 μl of 30% EtOH acidified to pH 2 with HCl. Pigment extraction consisted of sonicating samples for 5 seconds at 50 MHz (Sonic 300 Dismembrator Sonicator) prior to soaking samples at room temperature for 24 hours in the dark. For each sample, 90 μl of extracted pigment was loaded into 96well microtiter trays and quantified with a Microplate Reader (Benchmark Bio-Rad) at a wavelength of 480 nm. The results from each sample group were pooled for a final mean pigment value. Pigment levels for imprint establishment were quantified using the same spectrophotometric assay used for quantifying the white variegating strains, except one head was used per sample set.

Statistical Analysis
Kolmogorov-Smirnov two-sample tests were used to determine the statistical significance between the visual eye scores from control and dCTCF mutant data sets.
Statistical significance for both red and brown pigments was determined by using an ANOVA followed by Student's t-test with Bonferroni-corrected P values between the mean of the experimental dCTCF mutant data sets and the mean of the results from the appropriate control cross.

CTCF modENCODE data
The modENCODE data for dCTCF insulator sites from 0-to 12-hour embryos were obtained from the White Lab project on the modENCODE web site http://www. modencode.org [50].