Báo cáo khoa học: CXXC finger protein 1 restricts the Setd1A histone H3K4 methyltransferase complex to euchromatin doc

Murine embryonic stem ES cells lacking Cfp1 CXXC1 ⁄ are viable but show increased levels of global histone H3K4 methylation, suggesting that Cfp1 functions to inhibit or restrict the ac

CXXC finger protein 1 restricts the Setd1A histone H3K4 methyltransferase complex to euchromatin

Courtney M Tate, Jeong-Heon Lee and David G Skalnik

Herman B Wells Center for Pediatric Research, Section of Pediatric Hematology ⁄ Oncology, Departments of Pediatrics and Biochemistry and Molecular Biology, Indiana University School of Medicine, Indianapolis, IN, USA

Introduction

DNA in eukaryotic cells is complexed with histones and

other proteins in the form of chromatin The core histone

tails are subject to a variety of covalent modifications,

including acetylation, phosphorylation, methylation,

ubiquitination, sumoylation, and ADP-ribosylation

[1,2] Histone methylation plays critical roles in gene

expression, epigenetic regulation, and disease [3]

Histone methylation is catalyzed by a family of histone methyltransferase (HMT) enzymes, many of which are characterized by an evolutionarily conserved catalytic SET [Su(var)3–9, Enhancer of Zeste, Trithorax] domain [4] A major function of the SET domain-containing proteins is to modulate gene activity [5] Lys residues

of histones can be monomethylated, dimethylated,

Keywords

chromatin; epigenetics; histone methylation;

subnuclear targeting

Correspondence

D Skalnik, Cancer Research Building, 1044

West Walnut Street, Indianapolis, IN 46202,

USA

Fax: +1 317 278 9298

Tel: +1 317 274 8977

E-mail: dskalnik@iupui.edu

(Received 21 September 2009, revised 28

October 2009, accepted 4 November 2009)

doi:10.1111/j.1742-4658.2009.07475.x

CXXC finger protein 1 (Cfp1), encoded by the CXXC1 gene, is a compo-nent of the euchromatic Setd1A histone H3K4 methyltransferase complex, and is a critical regulator of histone methylation, cytosine methylation, cel-lular differentiation, and vertebrate development Murine embryonic stem (ES) cells lacking Cfp1 (CXXC1) ⁄ )) are viable but show increased levels of global histone H3K4 methylation, suggesting that Cfp1 functions to inhibit

or restrict the activity of the Setd1A histone H3K4 methyltransferase com-plex The studies reported here reveal that ES cells lacking Cfp1 contain decreased levels of Setd1A and show subnuclear mislocalization of both Setd1A and trimethylation of histone H3K4 with regions of heterochroma-tin Remarkably, structure–function studies reveal that expression of either the N-terminal fragment of Cfp1 (amino acids 1–367) or the C-terminal fragment of Cfp1 (amino acids 361–656) is sufficient to restore appropriate levels of Setd1A in CXXC1) ⁄ ) ES cells Furthermore, functional analysis

of various Cfp1 point mutations reveals that retention of either Cfp1 DNA-binding activity or association with the Setd1 histone H3K4 methyl-transferase complex is required to restore normal Setd1A levels In con-trast, expression of full-length Cfp1 in CXXC1) ⁄ ) ES cells is required to restrict Setd1A and histone H3K4 trimethylation to euchromatin, indicat-ing that both Cfp1 DNA-bindindicat-ing activity and interaction with the Setd1A complex are required for appropriate genomic targeting of the Setd1A complex These studies illustrate the complexity of Cfp1 function, and identify Cfp1 as a regulator of Setd1A genomic targeting

Abbreviations

CTD, C-terminal repeat domain; DAPI, 4¢,6-diaminidino-2-phenylindone; Dnmt1, DNA methyltransferase 1; ES, embryonic stem; FITC, fluorescein isothiocyanate; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; H3K4me3, trimethylated histone H3K4; HMT, histone methyltransferase; PHD, plant homeodomain; RNAP, RNA polymerase; Ser5-P CTD, C-terminal repeat domain phosphorylated at Ser5; SID, Set1 interaction domain.

or trimethylated, and the functional relevance of these

modifications depends on the position For example,

dimethylated and trimethylated histone H3K4 is found

associated with promoters and 5¢-regions of active genes

[6], whereas dimethylated and trimethylated

his-tone H3K9 is present at transcriptionally inactive

chro-matin sites [7–9] Yeast express a single H3K4 HMT,

Set1, which associates with a complex known as

COM-PASS (Complex Proteins Associated with Set1) [10] and

is required for telomeric and rDNA silencing [11,12] In

contrast, mammalian cells contain numerous HMTs that

show specificity for histone H3K4, including Setd1A,

Setd1B, Mll1, Mll2, Mll3⁄ Halr, Mll4 ⁄ Alr, Ash1L,

Smyd1, Smyd2, Smyd3, and Set7⁄ 9, which are present as

distinct multiprotein complexes and play critical roles in

gene expression and development [4,13–16]

The molecular mechanisms that control the targeting

and activity of HMT complexes are not well

under-stood Methylation at histone H3K4 correlates with

transcriptional activation and is directly coupled to the

transcription process [17] In yeast and mammals, Set1

and Setd1A localize to the 5¢-end of actively

tran-scribed genes and interact with the RNA polymerase

(RNAP) II C-terminal domain (CTD) phosphorylated

at Ser5 (Ser5-P CTD), a repeat marker associated with

transcription initiation [18–20] In yeast, Paf1C

interac-tion with RNAP II is required for recruitment of the

Set1–COMPASS H3K4 HMT complex to actively

transcribed genes [19] In mammals, Setd1A is tethered

to RNAP II by Wdr82, an integral component of the

Setd1A complex [18] Wdr82 associates with the RNA

recognition motif within Setd1A, and directly

recog-nizes Ser5-P CTD of RNAP II [18] In mammals, Mll1

interacts with RNAP II containing Ser5-P CTD and

mediates histone H3K4 methylation at a subset of

transcriptionally active genes [21] In addition, menin,

a component of the Mll2 H3K4 HMT complex,

associ-ates with RNAP II containing Ser5-P CTD [22] In

yeast and mammals, the Setd2 H3K36 HMT primarily

associates with the elongating hyperphosphorylated

form of RNAP II [23,24] Therefore, histone

methyla-tion mediated by HMTs is involved in regulating both

transcription initiation and elongation

Although generally widely expressed, mammalian

H3K4 HMTs have nonredundant functions For

exam-ple, Mll2 is important for expression of the HOXB

gene cluster, but not the HOXA cluster [13], whereas

HOXA9 and HOXC8 are exclusive Mll1 targets

[22,25] The HMTs Ash1L and Mll1 occupy the

5¢-regions of active genes, and their localization is

nearly indistinguishable, which suggests redundancy of

function [14] However, in vivo depletion of either

enzyme results in diminished methylation of histone

H3K4 at active HOXA genes [14] In addition, loss

of a single member of the H3K4 HMT family can lead

to disease or death [26,27] MLL1 is frequently the target of chromosomal translocations involved in acute lymphoid and myeloid leukemias [28–31] In addition, genetic disruption of murine MLL1 or MLL2leads to embryonic lethality [13,32] In addition, Smyd3 expression is upregulated in colorectal and hepatocellular carcinomas, and its H3K4 HMT activity activates oncogenes and other genes associated with the cell cycle, whereas depletion of Smyd3 by small interfering RNA treatment leads to suppression of cell growth [27]

With the exception of the enzymatic Setd1 compo-nent, the subunit composition of the mammalian Setd1A and Setd1B HMTase complexes are identical [16], each containing CXXC finger protein 1 (Cfp1), Rbbp5, Wdr5, Ash2, and Wdr82 [15,16] Setd1A and Setd1B mRNA are ubiquitously expressed in murine tissues, and Setd1A and Setd1B do not show differen-tial cell type expression [16] However, confocal immu-nofluorescence reveals that endogenous Setd1A and Setd1B show largely nonoverlapping subnuclear locali-zation [16] This suggests that Setd1A and Setd1B are targeted to unique sets of genomic sites, and that each has unique functions in the regulation of chromatin structure and gene expression Consequently, it is likely that the nonredundant function of each H3K4 HMT is a result of distinct target gene specificity [16] Cfp1 is a critical epigenetic regulator of both cytosine methylation and histone methylation, and interacts with both the maintenance DNA methyltrans-ferase [DNA methyltransmethyltrans-ferase 1 (Dnmt1)] [33] and with the Setd1A H3K4 HMT complex [15] Cfp1 localizes nearly exclusively to euchromatic nuclear speckles, and associates with the nuclear matrix [34] Cfp1 contains two Cys-rich plant homeodomains (PHDs); a PHD is a Cys-rich CXXC DNA-binding domain that shows specificity for unmethylated CpG dinucleotides, an acidic domain, a basic domain, a coiled-coil domain, and a Cys-rich Set1 interaction domain (SID), which is required for interaction with the Setd1A and Setd1B H3K4 HMT complexes [33,35,36]

Disruption of murine CXXC1 results in embryonic lethality shortly after implantation [37] Murine embry-onic stem (ES) cell lines lacking Cfp1 (CXXC1) ⁄ )) are viable but show a variety of defects, including an increased population doubling time due to increased apoptosis, a 70% decrease in global cytosine methyl-ation, decreased Dnmt1 protein expression and main-tenance DNA methyltransferase activity, and an inability to achieve in vitro differentiation [38] In

addition, CXXC1) ⁄ )ES cells express elevated levels of

histone H3K4 dimethylation and trimethylation, and

reduced levels of histone H3K9 dimethylation [15]

Consequently, Cfp1 plays an important role in the

reg-ulation of cytosine methylation, histone methylation,

and cellular differentiation

The purpose of this study was to obtain insights into

the molecular mechanisms regulating the activity and

targeting of the Setd1A H3K4 HMT complex The

results reported here reveal that CXXC1) ⁄ ) ES cells

contain reduced levels of Setd1A and show

mislocal-ization of both Setd1A protein and trimethylated

histone H3K4 (H3K4me3) to areas of

heterochro-matin Surprisingly, expression in CXXC1) ⁄ ) ES cells

of either the amino half of Cfp1 (amino acids 1–367)

or carboxyl half of Cfp1 (amino acids 361–656) is

sufficient to restore appropriate levels of Setd1A

However, full-length Cfp1 is required to restrict the

subnuclear localization of both Setd1A and H3K4me3

to euchromatin

Results

ES cells lacking Cfp1 contain decreased levels of

Setd1A

Exogenous expression of Setd1A fragments in HEK293

cells competes with endogenous Setd1A binding with

the Setd1A H3K4 HMT complex, resulting in decreased

stability of endogenous Setd1A [16] To examine

whether loss of Cfp1 has a similar effect, western blot

analysis was performed to determine protein levels of

Setd1A complex components in wild-type ES cells

(CXXC1+⁄ +), ES cells heterozygous for the disrupted

CXXC1 allele (CXXC1+⁄)), ES cells lacking Cfp1

(CXXC1) ⁄ )), CXXC1) ⁄ ) ES cells transfected with a

full-length Cfp1 expression vector (Rescue), and

CXXC1) ⁄ )ES cells cells carrying the empty expression

vector (Vector) A significant decrease ( 50%) in the

level of Setd1A was observed in CXXC1) ⁄ ) ES cells

(Fig 1A) Appropriate levels of Setd1A were restored

upon introduction of a Cfp1 expression vector (Rescue),

but not in ES cells carrying the empty expression vector

(Vector) CXXC1+⁄) ES cells express approximately

50% as much Cfp1 as CXXC1+⁄ + ES cells [38], and

show a slight decrease in Setd1A levels In contrast, no

difference in protein levels was observed for the other

Setd1A HMT complex components (Rbbp5, Wdr5,

Wdr82, and Ash2) in CXXC1) ⁄ )ES cells (Fig 1A)

Previous work demonstrated that Cfp1 functions as

a transcriptional activator in cotransfection assays

[34,36] Thus, further studies were performed to

exam-ine whether reduced Setd1A levels in ES cells lacking

Cfp1 are due to reduced transcription of the cognate gene Surprisingly, quantitative real-time PCR analysis demonstrated that Setd1A mRNA levels were elevated four-fold to five-fold in CXXC1) ⁄ ) ES cells as com-pared with CXXC1+⁄ +and CXXC1+⁄) ES cells, and are restored to wild-type levels in rescued ES cells but not in CXXC1) ⁄ )ES cells carrying the empty expres-sion vector (Fig 1B) Therefore, the decreased levels of Setd1A observed in CXXC1) ⁄ ) ES cells is not explained by reduced transcription of SETD1A Previous work by our laboratory demonstrated that disruption of the interaction between endogenous Setd1A and other components of the intact HMT complex led to a reduction of Setd1A levels as a conse-quence of a reduced Setd1A half-life [16] Additional studies were therefore performed to assess the role of protein stability in Setd1A levels in CXXC1) ⁄ ) ES cells These experiments revealed that treatment of CXXC1) ⁄ ) ES cells with the proteosome inhibitor MG132 led to an elevation of Setd1A levels to near wild-type levels (Fig 1C)

Cfp1 is required to restrict Setd1A and H3K4me3

to euchromatin The molecular mechanisms regulating HMT activity and genomic targeting remain largely unknown Previ-ous studies revealed the paradoxical finding that ES cells lacking the Cfp1 component of the Setd1A H3K4 HMT complex have increased levels of histone H3K4 methylation These findings suggest that Cfp1 may inhibit or restrict the activity of the Setd1A HMT complex To examine this issue further, subnuclear localization of Setd1A relative to 4¢,6-diaminidino-2-phenylindone (DAPI) staining was examined by confocal immunofluorescence DAPI is a fluorescent DNA stain that preferentially binds to the condensed structure of pericentromeric heterochromatin [39] Quantification of colocalization revealed that Setd1A showed only a slight ( 4%) overlap with DAPI-bright heterochromatin in wild-type ES cells However,

a significant (four-fold to five-fold) increase in colocal-ization of Setd1A with DAPI-bright heterochromatin was observed in CXXC1) ⁄ ) ES cells (Fig 2A) Rescue

of appropriate restriction of Setd1A to euchromatin was observed in CXXC1) ⁄ ) ES cells expressing full-length Cfp1 (1–656), but not in cells carrying the empty expression vector (Fig 2A)

The subnuclear localization of H3K4me3, a product

of Setd1A HMT activity, was similarly analyzed

by confocal immunofluorescence Consistent with the findings of Setd1A mislocalization in CXXC1) ⁄ ) ES cells, quantification of overlap between H3K4me3 and

DAPI-bright heterochromatin indicated that H3K4me3

showed only a slight overlap with DAPI-bright

heteo-chromatin in wild-type ES cells However, a significant

(five-fold to six-fold) increase in colocalization of

H3K4me3 with DAPI-bright heterochromatin regions

was observed in CXXC1) ⁄ ) ES cells (Fig 2B) Rescue

of appropriate subnuclear localization of H3K4me3

was observed in CXXC1) ⁄ ) ES cells expressing

full-length Cfp1 (1–656), but not in cells carrying the

empty expression vector (Fig 2B) These results

demonstrate that ES cells lacking Cfp1 show partial

mislocalization of both Setd1A and H3K4me3 to

DAPI-bright regions of heterochromatin, and reveal

that Cfp1 restricts the Setd1A H3K4 HMT complex to

euchromatin

Retention of either Cfp1 DNA-binding activity or association with the Setd1A HMT complex is required to restore appropriate levels of Setd1A The defects in Setd1A level and localization observed

in CXXC1) ⁄ ) ES cells were corrected upon introduc-tion of a full-length Cfp1 expression vector (Figs 1 and 2), thus providing a convenient method for assessment

of the structure–function relationships of Cfp1 Vari-ous cDNA expression constructs encoding FLAG-tagged Cfp1 truncations and mutations were stably expressed in CXXC1) ⁄ ) ES cells to identify the functional domains of Cfp1 that are necessary and suf-ficient to restore normal levels of Setd1A (Fig 3A) Isolated ES cell lines were screened for protein

Fig 1 ES cells lacking Cfp1 contain decreased levels of Setd1A (A) Whole cell protein extracts were isolated from the ES cell lines CXXC1+⁄ +, CXXC1+⁄), CXXC1) ⁄ ), and CXXC1) ⁄ ), expressing full-length Cfp1 (Rescue), and CXXC1) ⁄ ), carrying the empty expression vector (Vector) Extracts were subjected to western blot analysis, using antisera directed against the Setd1A HMT complex components Setd1A, Cfp1, Ash2, Rbbp5, Wdr5, and Wdr82 The graph presents the relative level of Setd1A normalized to b-actin expression from at least three independent experiments, and error bars indicate standard error Asterisks denote statistically significant (P < 0.05) differences as compared with CXXC1 +⁄ + ES cells (B) Quantitative RT-PCR was performed to assess Setd1A mRNA levels in the indicated ES cell lines The graph presents Setd1A transcript levels relative to those for GAPDH from three independent experiments, and error bars indicate standard error Asterisks denote statistically significant differences (P < 0.05) as compared with CXXC1+⁄ + ES cells (C) Western blot analysis was performed as described in (A) to assess Setd1A levels in CXXC1) ⁄ )ES cells following treatment with 5 l M MG132 for 6 h.

expression by western blot analysis, using an antibody

against Cfp1 CXXC1+⁄) ES cells express  50% as

much Cfp1 as CXXC1+⁄ + ES cells, but show normal

levels of cytosine methylation and histone methylation,

and are able to differentiate in vitro [38] Consequently,

clones were selected for analysis that have at least

50% of the level of Cfp1 observed in CXXC1+⁄ +ES

cells [44]

Expression of a C-terminal deletion fragment of

Cfp1 that lacks PHD2 (amino acids 1–481), or an

N-terminal deletion fragment that lacks PHD1, the

CXXC domain and the acidic domain (amino

acids 302–656), leads to restoration of normal levels of

Setd1A, indicating that none of these Cfp1 domains

are necessary for this rescue activity (Fig 3B)

Surpris-ingly, expression of either the amino half of Cfp1

(amino acids 1–367, containing PHD1, and the CXXC, acidic and basic domains) or the carboxyl half of Cfp1 (amino acids 361–656, containing the coiled-coil domain, SID, and PHD2) is sufficient to restore appropriate levels of Setd1A, indicating that Cfp1 contains redundant functional domains that support Setd1A levels, and that no single Cfp1 domain is essential for this function (Fig 3B)

The N-terminal fragment of Cfp1 (amino acids 1–367) contains the CXXC DNA-binding domain, and the C-terminal Cfp1 fragment (amino acids 361–656) contains the SID [33] Previous work determined that mutation of a conserved Cys residue (C169A) within the CXXC domain ablates Cfp1 DNA-binding activity [35], and mutation of a conserved Cys residue within the SID (C375A) ablates the interaction of Cfp1 with

Fig 2 Cfp1 is required to restrict Setd1A and H3K4me3 to euchromatin (A) The sub-nuclear distribution of endogenous Setd1A was determined in CXXC1 + ⁄ +

, CXXC1) ⁄ ) and CXXC1) ⁄ )ES cells expressing full-length Cfp1 (amino acids 1–656) or carrying the empty expression vector, using rabbit antibody against Setd1A and FITC-conju-gated bovine anti-rabbit IgG as secondary antibody Nuclei were counterstained with DAPI and observed by confocal microscopy Colocalization is indicated by yellow in the merged and colocalization image The num-bers inside the colocalization image indicate the percentage colocalized signal for the presented nucleus The numbers outside of the image summarize the average percent-age overlap of Setd1A with DAPI-bright het-erochromatin and standard error for at least

30 nuclei Asterisks denote a statistically significant difference (P < 0.05) as com-pared with CXXC1 + ⁄ + ES cells (B) Subnu-clear distribution of endogenous H3K4me3 was detected in the indicated ES cell lines, using rabbit antibody against H3K4me3 and FITC-conjugated bovine anti-rabbit IgG as secondary antibody, as described above The asterisks denote a statistically signifi-cant difference (P < 0.05) as compared with CXXC1 + ⁄ +

ES cells.

Fig 3 Cfp1 DNA-binding activity or

associa-tion with the Setd1A complex is required for

appropriate levels of Setd1A (A) Schematic

representation of full-length Cfp1 (amino

acids 1–656) and Cfp1 truncations and

mutations that were stably expressed in

CXXC1) ⁄ )ES cells The filled circle at the

N-terminus of Cfp1 represents the FLAG

epitope, and NLS represents a nuclear

locali-zation signal Mutations that ablate

DNA-binding activity (C169A) or interaction with

Setd1A (C375A) are indicated by ‘X’ (B)

Western blot analysis was performed

on whole cell extracts collected from

CXXC1+⁄ +, CXXC1) ⁄ )and CXXC1) ⁄ )ES

cells expressing full-length Cfp1 (amino

acids 1–656) or the indicated Cfp1

muta-tions (or carrying the empty expression

vec-tor), using antisera directed against Setd1A

[16] The level of b-actin serves as a loading

control The graph represents relative

Setd1A levels normalized to b-actin from at

least three independent experiments, and

error bars indicate standard error Asterisks

denote statistically significant (P < 0.05)

differences as compared with CXXC1) ⁄ )ES

cells expressing full-length Cfp1 (amino

acids 1–656).

the Setd1A HMT complex [33] Additional studies

were performed to assess the functional significance of

these Cfp1 properties for the ability to restore normal

levels of Setd1A CXXC1) ⁄ ) ES cells expressing

full-length Cfp1 that lacks DNA-binding activity (amino

acids 1–656, C169A) or interaction with the Setd1A

H3K4 HMT complex (amino acids 1–656, C375A)

contain normal levels of Setd1A This was expected,

given that expression of either half of Cfp1 is sufficient

to restore normal Setd1A levels However, ablation of

DNA-binding activity within the N-terminal fragment

of Cfp1 (amino acids 1–367, C169A), or disruption of

Setd1A interaction with the C-terminal Cfp1 fragment

(amino acids 361–656, C375A), results in the loss of

Setd1A rescue activity (Fig 3B) Finally, rescue

activ-ity was lost upon introduction of both point mutations

into full-length Cfp1 (amino acids 1–656, C169A⁄

C375A) These data indicate that retention of either

Cfp1 DNA-binding activity or interaction with the

Setd1A H3K4 HMT complex is required to restore

appropriate Setd1A levels in CXXC1) ⁄ )ES cells

Full-length Cfp1 is required to restrict Setd1A

and H3K4me3 to euchromatin

CXXC1) ⁄ ) ES cells expressing various Cfp1

trunca-tions and mutatrunca-tions were analyzed by confocal

immu-nofluorescence to determine the functional domains of

Cfp1 required to restrict the subnuclear localization

of Setd1A and H3K4me3 to euchromatin The vast

majority of Setd1A and H3K4me3 was localized to

DAPI-dim euchromatic regions in CXXC1) ⁄ ) ES cells

expressing full-length Cfp1 (amino acids 1–656)

(Figs 4 and 5) In contrast to the pattern of Cfp1

res-cue activity seen for Setd1A levels, however,

expres-sion of the N-terminal (amino acids 1–481 or 1–367)

or C-terminal (amino acids 302–656 or 361–656)

frag-ments of Cfp1 in CXXC1) ⁄ )ES cells is not sufficient

to exclude Setd1A and H3K4me3 from DAPI-bright

heterochromatin (Figs 4 and 5) In addition,

CXXC1) ⁄ ) ES cells expressing full-length Cfp1 that

lacks DNA-binding activity (amino acids 1–656,

C169A) or fails to interact with the Setd1A H3K4

HMT complex (amino acids 1–656, C375A) also fail

to restrict Setd1A and H3K4me3 to euchromatin

(Figs 4 and 5) As expected, ablation of the

DNA-binding activity within the N-terminal fragment of

Cfp1 (amino acids 1–367, C169A), disruption of the

Setd1A interaction with the C-terminal fragment of

Cfp1 (amino acids 361–656, C375A) or introduction

of both mutations within full-length Cfp1 (1-656

C169A, C375A) also results in a failure to exclude

Setd1A and H3K4me3 from DAPI-bright

heterochro-matin (Figs 4 and 5) Therefore, full-length Cfp1 is required to restrict Setd1A and H3K4me3 localization

to euchromatin, and Cfp1 DNA-binding activity and interaction with the Setd1A H3K4 HMT complex are both required for proper restriction of Setd1A and H3K4me3 to euchromatin

Discussion

The results of the studies reported here reveal that ES cells lacking the epigenetic regulator Cfp1 contain decreased levels of the histone H3K4 methyltransferase Setd1A Yeast cells lacking Spp1, the Cfp1 homolog, also express reduced amounts of Set1 [40], and Spp1 is thought to stabilize Set1 [40] Furthermore, expression

of Cfp1-interacting Setd1A fragments in human cells disrupts the association of endogenous Setd1A with the intact HMT complex, resulting in reduced Setd1A levels as a consequence of reduced Setd1A half-life [16] Thus, the reduced levels of Setd1A observed in

ES cells lacking Cfp1 may be due to decreased Setd1A stability The observed increase of Setd1A level in CXXC1) ⁄ )ES cells following treatment with the prote-osome inhibitor MG132 supports this hypothesis In contrast, the levels of the other components of the Setd1A complex (Ash2, Rbbp5, Wdr5, and Wdr82) are not altered in CXXC1) ⁄ ) ES cells, which may be due

to their association with additional H3K4 HMT com-plexes (Setd1B, Mll1, Mll2, and Mll3) [16,18,22,28, 41–43] Despite reduced Setd1A levels, CXXC1) ⁄ ) ES cells express an approximately five-fold increased level

of Setd1A mRNA, suggesting that these cells increase transcription of the SETD1A gene to compensate for reduced levels of Setd1A

Expression of either an N-terminal fragment (amino acids 1-367) or C-terminal fragment (amino acids 361–656) of Cfp1 is sufficient to restore normal levels

of Setd1A in CXXC1) ⁄ ) ES cells These results are consistent with previous findings that expression

in CXXC1) ⁄ ) ES cells of either Cfp1(1–367) or Cfp1(361–656) is sufficient to rescue defects in ES cell plating efficiency, cytosine methylation, and in vitro differentiation [44] Interestingly, Cfp1(1–367) fails to interact with the Setd1A complex [33], but still restores appropriate levels of Setd1A, indicating that a physical interaction of Cfp1 with the Setd1A complex is not required for appropriate levels of Setd1A In addi-tion, analysis of point mutations within the CXXC domain (C169A) or SID (C375A) reveals that reten-tion of either Cfp1 DNA-binding activity or interac-tion with the Setd1A H3K4 HMT complex is necessary to restore normal levels of Setd1A in CXXC1) ⁄ )ES cells

ES cells that lack Cfp1 show increased levels of

histone H3K4 dimethylation and trimethylation [15],

despite expressing decreased levels of Setd1A,

suggest-ing that Cfp1 restricts the activity of the Setd1A

H3K4 HMT complex Consistent with this model,

confocal immunofluorescence reveals that both

Setd1A and H3K4me3 are partially mislocalized to DAPI-bright regions of heterochromatin in CXXC1) ⁄ ) ES cells In contrast to the pattern of Cfp1 rescue activity observed for Setd1A levels, expression of full-length Cfp1 in CXXC1) ⁄ ) ES cells

is required to properly restrict subnuclear localization

Fig 4 Full-length Cfp1 is required to

restrict Setd1A to euchromatin The

subnu-clear distribution of endogenous Setd1A

was detected in CXXC1) ⁄ )ES cells

expressing full-length Cfp1 (amino

acids 1–656) or the indicated Cfp1

trunca-tions and mutatrunca-tions, using rabbit antibody

against Setd1A and FITC-conjugated bovine

anti-rabbit IgG as secondary antibody, as

described for Fig 2 Asterisks denote a

statistically significant difference (P < 0.05)

as compared with CXXC1) ⁄ )ES cells

expressing full-length Cfp1 (amino

acids 1–656).

of Setd1A and H3K4me3 to euchromatin These

studies further indicate that Cfp1 DNA-binding

acti-vity and interaction with the Setd1A H3K4 HMT

complex are both required for proper subnuclear

localization of Setd1A The requirement for an intact

Cfp1 CXXC domain for proper genomic localization

may indicate that Cfp1 DNA-binding activity restricts

the Setd1A H3K4 HMT complex to euchromatin by

binding to unmethylated CpG dinucleotides in euchromatin

Individual CXXC1) ⁄ ) ES cell nuclei show a range (5–30%) of colocalization between Setd1A and H3K4me3 with DAPI-bright heterochromatin, and 20–30% mislocalization of Setd1A and H3K4me3 is observed in 35–40% of CXXC1) ⁄ )ES cell nuclei It is possible that cell-to-cell variation in the degree of

Fig 5 Full-length Cfp1 is required to restrict H3K4me3 to euchromatin The subnuclear distribution of H3K4me3 was detected in CXXC1) ⁄ )ES cells expressing full-length Cfp1 (amino acids 1–656) or the indicated Cfp1 truncations and mutations, using rabbit antibody against H3K4me3 and FITC-conjugated bovine anti-rabbit IgG as secondary antibody, as described for Fig 2 Asterisks denote statistically significant differences (P < 0.05) as compared with CXXC1) ⁄ )ES cells expressing full-length Cfp1 (amino acids 1–656).

colocalization may be cell cycle-dependent However,

significant mislocalization of Setd1A and H3K4me3 is

never observed in wild-type ES cells or in rescued

CXXC1) ⁄ ) ES cells expressing full-length Cfp1 The

persistence of DAPI-bright staining colocalizing with

H3K4me3 indicates that deposition of this

euchroma-tin epigenetic mark is insufficient to induce general

chromatin remodeling in these heterochromatin

regions

Little is known regarding the relative contributions

of each mammalian histone H3K4 HMT complex

However, Cfp1 has been shown to be an integral

ponent of only the Setd1A and Setd1B HMT

com-plexes [15,16] The localization of Setd1B in the

absence of Cfp1 has not been determined, but the

find-ing that the extent of Setd1A mislocalization is similar

to that of H3K4me3 localization suggests that the

Setd1 HMT complexes are responsible for the bulk of

histone H3K4 trimethylation This conclusion is

con-sistent with a recent report that small interfering

RNA-mediated depletion of Setd1A and Setd1B leads

to a dramatic global reduction in histone H3K4

trime-thylation [45]

The full-length Cfp1 that is required to restrict

subnuclear localization of Setd1A and H3K4me3 to

euchromatin contains two PHDs PHDs are thought

to be involved in chromatin-mediated transcriptional

control [46], and can serve as binding modules for

unmodified and methylated histone H3K4 and

methy-lated histone H3K36 [17,47–51] For example, PHD1

of Spp1, the yeast homolog of Cfp1, binds

dimethylat-ed and trimethylatdimethylat-ed histone H3K4 [51] In addition,

the PHD finger of the tumor suppressor Ing2 directly

associates with H3K4me3, and this interaction is

criti-cal for proper occupancy of the Ing2–HDAC1 complex

at target promoters during the DNA damage response

and active transcriptional repression [48] Therefore,

the PHDs of Cfp1 may be important for binding

mod-ified histone H3K4 and targeting the Cfp1–Setd1A

complex to specific genomic sites

The mechanisms responsible for appropriate

subnu-clear localization of histone H3K4 HMTs are complex,

and involve gene-specific recruitment by DNA-binding

factors For example, the insulator DNA-binding

pro-tein Boris recruits Setd1A to the MYC and BRCA1

genes [52]; NF-E2 recruits Mll2 to the b-globin locus

[53]; the Ap2d transcription factor recruits Ash2L and

Mll2 to the HOXC8 locus [54]; and the paired-box

transcription factor Pax7 recruits Mll2 to the MYF5

gene [55]

In addition, several integral components of the

mammalian Set1-like histone H3K4 HMT complexes

have been implicated in genomic targeting Wdr5,

which is common to each member of the mammalian Set1-like HMT complex family, has been reported to bind directly to histone H3 [56–59] In addition, the Wdr82 component of the Setd1A and Setd1B HMT complexes binds to RNAP II containing Ser5-phos-phorylated CTD, thus recruiting these complexes to sites of transciption initiation [18] Furthermore, the compositions of the Setd1A and Setd1B HMT com-plexes are identical, except for the identity of the enzy-matic (Setd1) component [15,16], but confocal microscopy reveals that these complexes show a nearly nonoverlapping euchromatic subnuclear localization [16] This finding strongly suggests that these closely related complexes regulate distinct sets of target genes, and that this specificity is mediated by each Setd1 pro-tein, presumably through interactions with distinct tar-geting effector molecules The data reported here reveal that Cfp1 plays a novel role in restricting the subnuclear localization of Setd1A and H3K4me3 to euchromatin, thus identifying Cfp1 as another critical regulator of histone H3K4 HMT genomic targeting

Experimental procedures

Cell culture

Generation of murine CXXC1) ⁄ )ES cell lines was as previ-ously described [38] ES cells were cultured on 0.1% gela-tin-coated tissue culture dishes in high-glucose DMEM (Gibco BRL, Life Technologies, Grand Island, NY, USA) supplemented with 20% fetal bovine serum (Gibco BRL),

100 unitsÆmL)1 penicillin⁄ streptomycin (Invitrogen, Carls-bad, CA, USA), 2 mm l-glutamine (Invitrogen), 1% nones-sential amino acids (Invitrogen), 0.2% leukemia inhibitory factor-conditioned medium, 100 nm b-mercaptoethanol, 0.025% Hepes (pH 7.5) (Invitrogen), and 1% Hank’s balanced salt solution (Invitrogen)

Plasmid construction and transfection of ES cells

pcDNA3.1⁄ Zeo (Invitrogen) The Cfp1 expression vector or the empty expression vector was electroporated into CXXC1) ⁄ ) ES cells as previously described [38] Single amino acid substitutions within Cfp1 were performed using the QuikChange II site-directed mutagenesis kit (Strata-gene, La Jolla, CA, USA) according to the manufacturer’s protocol, as previously described [33,35] For structure– function studies, cDNA constructs encoding full-length FLAG epitope-tagged human Cfp1 (amino acids 1–656) and various Cfp1 truncations and⁄ or mutations were subcloned into the pcDNA3.1⁄ Hygro mammalian expres-sion vector (Invitrogen) The N-terminal bipartite nuclear localization signal of Cfp1 (amino acids 109–121) was