Deficiency of the Chromatin Regulator Brpf1 Causes Abnormal Brain Development*

Background: The chromatin regulator BRPF1 recognizes different epigenetic marks and activates multiple histone acetyltransferases, but little is known about its biological functions in mammals. Results: Forebrain-specific inactivation of the mouse gene causes neocortical abnormalities and partial agenesis of the corpus callosum. Conclusion: Mammalian BRPF1 is important for forebrain development. Significance: This study links a unique chromatin regulator to mammalian brain development. Epigenetic mechanisms are important in different neurological disorders, and one such mechanism is histone acetylation. The multivalent chromatin regulator BRPF1 (bromodomain- and plant homeodomain-linked (PHD) zinc finger-containing protein 1) recognizes different epigenetic marks and activates three histone acetyltransferases, so it is both a reader and a co-writer of the epigenetic language. The three histone acetyltransferases are MOZ, MORF, and HBO1, which are also known as lysine acetyltransferase 6A (KAT6A), KAT6B, and KAT7, respectively. The MORF gene is mutated in four neurodevelopmental disorders sharing the characteristic of intellectual disability and frequently displaying callosal agenesis. Here, we report that forebrain-specific inactivation of the mouse Brpf1 gene caused early postnatal lethality, neocortical abnormalities, and partial callosal agenesis. With respect to the control, the mutant forebrain contained fewer Tbr2-positive intermediate neuronal progenitors and displayed aberrant neurogenesis. Molecularly, Brpf1 loss led to decreased transcription of multiple genes, such as Robo3 and Otx1, important for neocortical development. Surprisingly, elevated expression of different Hox genes and various other transcription factors, such as Lhx4, Foxa1, Tbx5, and Twist1, was also observed. These results thus identify an important role of Brpf1 in regulating forebrain development and suggest that it acts as both an activator and a silencer of gene expression in vivo.

In eukaryotic cells, nuclear DNA is elegantly packaged with histones and other proteins into chromatin, which in turn controls gene activities according to spatiotemporal signaling cues (1)(2)(3). Such control is altered in human diseases and provides pathological epigenetic mechanisms. These mechanisms were initially recognized in leukemia and other cancers (4 -6), but it is now becoming clear that such mechanisms are also crucial in neurological disorders (7)(8)(9). For epigenetic regulation, chromatin is subject to various modifications, one of which is histone acetylation. The acetyltransferase CREB-binding protein and its paralog p300 are altered in Rubinstein-Taybi syndrome, a genetic disease with the characteristic of intellectual disability and predisposition to brain and other tumors (10,11). Histone deacetylases also regulate neurological processes. HDAC2 displays increased expression in patients with Alzheimer disease (12), and haploinsufficiency of HDAC4 causes brachydactyly mental retardation syndrome (13). The HDAC10 gene is mutated in amyotrophic lateral sclerosis (14), and the moodstabilizing drug valproic acid is a histone deacetylase inhibitor (15). These findings support the importance of histone acetylation modifiers in neurological diseases.
MOZ (monocytic leukemia zinc finger protein) is an acetyltransferase paralogous to MORF (MOZ-related factor) (16). MOZ was identified during analysis of acute myeloid leukemia (17), whereas MORF was uncovered based on sequence similarity to MOZ (18). Abnormal expression of mouse Moz contributes to metastasis of medulloblastoma (19), and inactivation of the mouse Moz gene partially phenocopies DiGeorge syndrome (20). Psychiatric illness is a late-occurring feature of this syndrome, and some patients are predisposed to schizophrenia. The MORF gene is mutated in four neurodevelopmental disorders as follows: Noonan syndrome-like disorder (21); Ohdo syndrome (22); genitopatellar syndrome (23,24); and blepharophimosis-ptosis-epicanthus inversus syndrome (25). Intellectual disability is common to these disorders, and the patients often display callosal aplasia or hypoplasia (26). Related to the intellectual disability, mice with low Morf expression exhibit an abnormal cerebral cortex (27). No callosal abnormalities, however, have been reported for the mutant mice.
BRPF1 interacts with MOZ and MORF to stimulate their acetyltransferase activity, restrict the substrate specificity, and regulate related transcription (28,29), suggesting that BRPF1 may be an important modifier of neurological disorders resulting from aberrant MOZ and MORF. Moreover, BRPF1 interacts with and activates HBO1 (histone acetyltransferase bound to ORC1), which possesses a catalytic domain highly homologous to those of MOZ and MORF (30). Interestingly, genetic analysis of Chameau, the fly ortholog of HBO1, revealed an unexpected role in silencing Hox gene expression (31). In addition to association with histone acetyltransferases, BRPF1 possesses double plant homeodomain-linked (PHD) zinc fingers for interaction with histone H3 (30), a bromodomain with acetyl-lysine-recognizing ability (32) and a PWWP domain possessing specific affinity for methylated histone H3 (33,34). Thus, BRPF1 is a multivalent chromatin regulator.
Biologically, mouse Brpf1 is essential for embryogenesis (35). Pathologically, the human BRPF1 gene is recurrently mutated in sonic hedgehog-driven adult medulloblastoma (36). Moreover, related to the aforementioned possibility that BRPF1 functions as a modifier of neurological disorders resulting from aberrant MOZ and MORF, mouse Brpf1 is highly expressed in the forebrain and cerebellum (35), suggesting a role in brain development. To investigate this, we have specifically inactivated the mouse Brpf1 gene in the forebrain. Here, we describe the resulting early postnatal lethality and characterize the developmental defects in the neocortex and corpus callosum. Strikingly, microarray-based gene expression analyses revealed dramatically elevated expression of multiple Hox genes and other transcription factors, thereby unveiling an unexpected role of Brpf1 in negative regulation of gene expression in vivo.

Maintenance of Mouse Colonies and Use of Mice-Mouse
strains were maintained in an established facility at McGill University, and all procedures about the use of mice were performed according to guidelines and protocols approved by the McGill Animal Use Committee. Heterozygous Brpf1 f/ϩ mice have been described recently (35). This conditional allele contains two LoxP sites for Cre-mediated excision to remove coding exons 4 -6 of the Brpf1 gene (35). Intercrosses generated Brpf1 f/f homozygotes, which were then crossed with Emx1-Cre mice (Jax). Further intercross yielded forebrain-specific Brpf1 f/f ;Emx1-Cre knock-out (or bKO) mice.
Behavioral Tests-Nest-building tests were performed as described (37). Between 4:00 and 5:00 pm, mice were transferred to individual testing cages with bedding but no environmental enrichment items. One nestlet cotton (3.0 g) was placed in each cage. The nests were assessed in the next morning on a 5-point scale as described (38). Digital cage photos were taken before and after the tests.
Hindlimb clasping tests were performed as reported (39). Mice were suspended by the base of the tail and videotaped for 14 s. Animals were assessed for hindlimb clasping score and duration. For the clasping score, the video was played back in half-time. The test period was divided into 2-s slots. An animal would receive a score of 1 point if it displayed any abnormal movement during the given time slot, thereby allowing for a maximum score of 7 points. An abnormal movement was defined as any dystonic movement of the hindlimbs, or a combination of hind-and forelimbs and trunk, during which the limbs were pulled into the body in a manner not observed in wild-type mice.
Nissl Stain-Tissue preparation was performed as described (40). Cryosections were prepared and air-dried before staining. Paraffin sections were dewaxed and rehydrated through a gradient series of ethanol. Cryosections or paraffin sections were stained in 0.1% cresyl violet solution (prepared with 0.3% glacial acetic acid and filtered) for 6 -8 min, rinsed in distilled H 2 O, dehydrated in a gradient of ethanol, cleared in xylene, and covered with glass coverslips for examination under a light microscope. Slides were also digitized with a Scanscope XT (Aperio) for further analysis.
Timm's Stain-This method stains vesicular zinc in the brain and was performed according to an on-line protocol (Nikki Sunnen, BCM 4/09). Wild-type and bKO mice were anesthetized at P8 with isoflurane and perfused transcardially with cold PBS and then with a buffered sulfide solution containing 1.2% Na 2 S. Brains were removed and immersed in the perfusion solution for 1 h. After rinsing with distilled H 2 O, the brains were fixed in neutral buffered formalin (Sigma, HT501128) for 24 h and post-fixed in 3% glutaraldehyde in 30% dextrose solution for 1.5 h. Samples were embedded in paraffin and serially sectioned sagittally at 5 m. Dewaxed and rehydrated slides were stained in a preheated (26°C) Timm's stain solution (30% gum arabic, 1.7% hydroquinone and 0.085% silver nitrate in the citrate buffer) for 45 min in the dark at room temperature and then for at least 20 min at 60°C. Slides were counterstained in 0.1% cresyl violet.
Myelin Staining-The Black Gold II myelin staining kit was used according to the manufacturer's instructions (Millipore, AG105). Wild-type and bKO mice at P13 were anesthetized with isoflurane and perfused transcardially with cold PBS and then with 4% paraformaldehyde in PBS. The brains were removed and immersed in 4% paraformaldehyde overnight at 4°C. 40 m of coronal cryosections were used.
Golgi-Cox Staining-After anesthesia, mice were perfused transcardially with cold PBS. Golgi-Cox impregnation was performed with the FD Rapid Golgi-Stain kit (FD NeuroTechnologies, PK401A) according to the manufacturer's instructions. For this, 180-m coronal slices were prepared on a Vibrotome (Bio-Rad/EM Corp., Micro-Cut H1200). Black and white photographs were taken under a bright field on an Axio Observer Z1 microscope controlled by the Zen software package (Zeiss).
Measurement of Cortical Layer Thickness-After DAPI staining, fluorescent images were taken, and cortical layer thickness was measured with the Zen software package (Zeiss). The neocortex above the hippocampus was selected for quantification. For E15.5 neocortex, the total thickness was divided into ventricular zone/subventricular zone, intermediate zone, and cortical plate. For quantification, three brains per genotype and three sections per brain were measured. Neonate cortex was considered to include the subventricular zone, intermediate zone, and layers I-VI; for quantification, three brains per genotype and 4 -5 sections per brain were measured.
After double immunolabeling with anti-Ctip2 and -Cux1 antibodies, images were taken with the Zen software package (Zeiss) to determine thickness of the Cux1 ϩ or Ctip2 ϩ layer, signal intensity of the Cux1 ϩ layer, and cell density in the Ctip2 ϩ layer. Three brains per genotype and three sections per brain were used for quantification.
Cortical Neuronal Migration Analysis-After intraperitoneal injection with 50 g of BrdU/g body weight at E12.5, E14.5, and E16.5, pregnant mice were sacrificed at P0 to retrieve neonate brains for fixation and subsequent preparation of 5-m paraffin sections. For immunostaining, the sections were dewaxed, rehydrated, and exposed to 2 N HCl at 37°C for 30 min before antigen retrieval by boiling in 10 mM sodium citrate at pH 6.0 for 20 min. Sections were then incubated in the blocking solution (2% BSA and 0.2% Triton X-100 in PBS) at room temperature for 1 h before incubation with a rat anti-BrdU antibody (Abcam, AB6326, 1:100) in the blocking solution at 4°C overnight. Biotin-SP-conjugated AffiniPure donkey anti-rat IgG(HϩL) (Jackson ImmunoResearch, 712-065-153, 1:200) was used as a secondary antibody. The signal was developed with the Vector ABC and DAB substrate kits (Vector Laboratories, PK-4001 and SK-4100). Stained sections were archived on a Scanscope XT (Aperio) with a 20ϫ objective. The developing neocortex above the hippocampus was used for quantification. For each cortex, the total thickness extending from molecular layer to the intermediate zone was divided into 10 bins with Photoshop (Adobe), and the cell number of darkly stained and lightly stained BrdU ϩ cells was counted manually and assigned to each bin to generate histograms of the number of labeled cells against bin. The genotypes were intentionally ignored to avoid bias. For each BrdU labeling time point, two pairs of littermates and four matched sections per pair were used for quantification.
Cell Cycle Analysis-Pregnant mice were intraperitoneally injected at E15.5 with 50 g of BrdU/g body weight, and fetal brains were recovered 1 h later. Paraffin sections were prepared for immunostaining with anti-Ki67 and -BrdU antibodies. For antigen retrieval, the sections were boiled in 10 mM sodium citrate at pH 6.0 for 20 min, blocked, and incubated simultaneously with rat anti-BrdU antibody (Abcam, AB6326, 1:100) and mouse anti-Ki67 antibody (Pharmingen, 556003, 1:400) at 4°C overnight. Alexa Fluor 488-conjugated goat anti-rat IgG (Invitrogen, A11006, 1:500) and Alexa Fluor 568-conjugated goat anti-mouse IgG (Invitrogen, A11031, 1:500) were used as the secondary antibodies. After counterstaining with DAPI, fluorescent images were taken on an Axio Observer Z1 fluorescence microscope (Zeiss) controlled by the Zen software package (Zeiss). The subventricular zone (SVZ) 2 above the hippocampus was boxed using the Zen software package (Zeiss) for manual counting of BrdU ϩ and/or Ki67 ϩ cells; during counting, the genotypes were intentionally ignored to avoid bias. The SVZ region was chosen for quantification based on published reports (43,44), and the SVZ precursor cells were identified as those set apart from the dense layer of BrdU ϩ cells in the ventricular zone as described (45). The quantification was performed on two pairs of littermates and 4 -5 matched sections per animal.
RT-PCR-This was performed as described (40), with primers listed in Table 1.
Microarray Analysis-Total RNA was isolated from three pairs of wild-type and mutant dorsal caudal cortices at P4 using the miRNeasy mini kit (Qiagen). The yield was measured on a Nanodrop 2000 spectophotometer (Thermo Scientific), and the integrity was assessed with an RNA Pico chip on an Agilent 2100 Bioanalyzer (Agilent Technologies). 500 ng of total RNA was subjected to a single round of amplification with the Amino Allyl MessageAmp TM II amplified antisense RNA kit (Ambion/ Invitrogen, AM1753) according to the manufacturer's instructions. The amplified product was labeled with a CyDye Post-Labeling Reactive Dye Pack (GE Healthcare). Cy3-labeled and Cy5-labeled aRNA probes (825 ng each) were co-hybridized to Agilent SurePrint G3 Mouse GE 8 ϫ 60K microarray (Agilent Technologies, G4852). Microarray slides were washed according to the manufacturer's instructions and scanned on an Agilent dual-laser scanner. Feature extraction was performed with Agilent Feature Extraction software (version 9.5.3.1).
The Limma package was used to normalize and determine the modulated genes from the original microarray data (46). Signal background correction was carried out by use of the normexp method and offset ϭ 50 as suggested by the Limma package. To determine the modulated genes between a mutant and its wild type, we applied two-channel normalization. Within-array normalization was carried out with the Loess method, although between-array normalization was performed by a quantile method (46). Empirical Bayes statistics were applied to determine the differential expression of the genes. Original microarray data and normalized datasets have been deposited in the GEO database under the accession number GSE63909. The resulting gene list is presented in supplemental Table S1.
RT-qPCR-Total RNA was isolated from three pairs of wild type and mutant dorsal caudal cortices at P12 with the miRNeasy mini kit (Qiagen). cDNA was synthesized using the Quantitect reverse transcription kit (Qiagen). Real time qPCR was conducted with the Green-2-Go qPCR Mastermix (Bio-Basic) on Realplex2 (Eppendorf). Primers were taken either from PrimerBank or designed on the web-based qPCR primer designer (IDT Biotechnology). Primers (Table 2) were synthesized by IDT Biotechnology.
Statistical Analysis-Log-rank tests (for survival curve) and Student's tests were carried out with Prism 5 (GraphPad).

RESULTS
Forebrain-specific Loss of Brpf1 Causes Early Lethality and Behavioral Abnormalities-To determine the function of Brpf1 in the mouse forebrain, we mated Brpf1 f/f mice (35) with the Emx1-Cre strain, in which the Cre recombinase is specifically expressed in the neocortex and hippocampus starting as early as E10.5 (47). The resulting Brpf1 f/ϩ ;Emx1-Cre mice appeared normal, and further crosses yielded Brpf1 f/f ;Emx1-Cre knock-out (or bKO) mice. The pups were born at a ratio slightly lower than the expected Mendelian ratio (data not shown), suggesting partial prenatal lethality. After birth, a majority of the viable bKO pups died prior to weaning around postnatal day (P) 21, with a minority surviving afterward (Fig. 1A). The mutant pups did not display obvious abnormalities at birth, and the growth appeared normal at P5. However, at P10, they were significantly smaller than their control littermates (Fig. 1B). Those mutant pups that survived after weaning were easily identifiable by their smaller size (Fig. 1C). They also consumed much less food than their control littermates (data not shown). Thus, forebrain-specific deletion of the Brpf1 gene resulted in early postnatal lethality and growth retardation.
We then conducted behavioral tests with those mutant mice that survived after weaning. Hindlimb-clasping tests are often used as an indicator for motor and coordination skills (48). The bKO but not control mice displayed a characteristic full-body clasping phenotype (Fig. 1D and supplemental video S1). Semiquantitative analysis of the clasping behavior was also carried out. For this, four control and four bKO mice were observed in a 14-s test procedure to monitor the clasping phenotype. Mice were scored for clasping severity and duration. The results revealed that bKO mice displayed a much higher tendency of hindlimb clasping (Fig. 1, E and F), indicating problematic motor coordination and neurological defects. We also assessed nest-building skills, which are learned by pups watching the mother and other cage mates and thus indicate learning ability and motor usage (37). Control mice consistently tore cotton pads provided in the cage and built perfect nests within the given time window (Fig. 1G, left panel). By contrast, mutant mice were unable to do so and left the cotton pad untouched (Fig. 1G, right panel). Semi-quantitative scoring revealed defective nest-building skills in the mutant (Fig. 1H). One rare mutant mouse survived to 45 days after birth and was old enough for marble-burying tests, which are often used to indicate repetitive, compulsive behaviors (49). Although the control littermate completely or partially buried 10 of the 12 marbles provided, the mutant left all 12 untouched (data not shown). Together, these results indicate that Brpf1 inactivation in the forebrain caused behavioral and neurological defects.
Brpf1 Deficiency Alters Neocortical Development-We then investigated abnormalities at the tissue level. Compared with that of the control littermates, the bKO brain was slightly smaller ( Fig. 2A, left panel), although the size remained proportional to the body size ( Fig. 2A, right panel). The rostral-caudal length of the neocortex was shorter in the bKO brain (Fig. 2B, marked by green arrowheads). Acquired late during evolution, the neocortex is a mammal-specific forebrain structure comprising a uniform laminar structure that has been traditionally divided into layers I-VI (50). We thus systematically examined tissue histology by performing Nissl staining of serial sagittal and coronal sections of the control and mutant brains. Shown in Fig. 2C are two representative images from these sections. Compared with the control, the mutant neocortex was thinner and showed differences in lamination (Fig. 2C). By contrast, the cerebellum appeared normal (Fig. 2D). These results are in agreement with the fact that the Emx1-Cre strain expresses the Cre recombinase from the endogenous Emx1 locus to drive excision in the forebrain but not the mid-or hindbrain (47,51).
Starting before birth, the cortical plate of the mouse brain gradually develops to form the laminated neocortex (50,52). Immediately after birth, the cortical plate of the mutant brain was also thinner than that in the wild type (Fig. 2E), as confirmed by DAPI staining and quantification (Fig. 2, F and G). Moreover, the defect remained evident at E15.5 (Fig. 2, H and I). E15.5 is a mid-corticogenesis stage when lower layer neurons have differentiated and are migrating to the cortical plate (53). The thinner cortical plate at E15.5 (Fig. 2, H and I) is consistent with the thinner layers V-VI at P0 (Fig. 2, F and G).
To characterize the lamination defects further, we performed Timm's stain on serial sagittal brain sections. As reported (54), the wild-type neocortex contained two darkly stained bands (Fig. 3A). In the mutant neocortex, the superficial band was present, but the inner band was missing (Fig. 3A). These results provide further support for defective lamination in the mutant neocortex.
Golgi-Cox staining has been widely used to reveal axonal and dendritic structures of neurons (55), so we next utilized this technique to examine how Brpf1 inactivation may affect cortical neuron distribution and structures. As shown in Fig. 3B, neurons in the bKO neocortex were less organized. Moreover, the neuron density was lower, and dendritic trees were less sophisticated in the mutant neocortex (Fig. 3C).
Taken together, results from the above morphological and histological analyses (Figs. 2 and 3) indicate that specific inactivation of the Brpf1 gene in the forebrain led to formation of abnormal neocortical organization. Brpf1 is thus required for neocortical development.
Brpf1 Regulates Cortical Neurogenesis-To gain molecular insights into the lamination defects, we analyzed expression of markers for different layers of the neocortex. For this, indirect immunofluorescence microscopy was performed with antibodies against Cux1 and Ctip2, well known markers of layers II/III/IV and V/VI of the neocortex, respectively (56,57). At P0, Cux1 expression was slightly altered in the bKO neocortex (Fig. 4A). Compared with the wild type, a narrower Ctip2 ϩ layer was observed in the mutant brain (Fig. 4, B and C). These results provide support for the lamination defects at the molecular level.
Through an inside-out mechanism, cortical neural precursors produce neurons of layers VI-V from E11.5 to E15.5 and then generate those in layers II-IV from E14.5 to E17.0 (50,52). Thus, we used neuronal birthdating assays, as described previously (58,59), to investigate whether neurons are produced at the correct time and then migrate properly to different cortical layers. For this, BrdU was injected into pregnant dams at E12.5, E14.5, and E16.5, and the dams were sacrificed at P0 to recover neonate brains for immunohistochemical analyses with anti-BrdU monoclonal antibody. The staining patterns observed in control neonate brain sections (Fig. 5) were very similar to what was reported by others (58,59). As in the control, E12.5 labeling revealed a peak of BrdU ϩ neurons in cortical layers V-VI, corresponding to bins 5-8, in the mutant (Fig. 5A). There was only a slight decrease in the number of densely labeled neurons in bin 8 of the mutant (Fig. 5B), indicating that a majority of neurons born at E12.5 migrate correctly to different neocortical layers. E14.5 labeling uncovered a significant decline of mutant BrdU ϩ neurons in bins 2-4, corresponding to cortical layers II-IV (Fig. 5, C and D). Moreover, an accumulation of mutant BrdU ϩ progenitors was detected in bins 9 and 10 corresponding to the intermediate zone (Fig. 5D). When labeled at E16.5, there was a decrease in bins 3-8 and 10 (Fig. 5, E and F). These results indicate that the lamination defects are evident as early as E14.5 and become more severe as the development progresses to E16.5. MARCH 13, 2015 • VOLUME 290 • NUMBER 11

Brpf1 in the Neocortex and Corpus Callosum
Neuronal progenitors are important for neocortical lamination (50,52). Because we observed a thinner cortical plate in the mutant as early as E15.5 (Fig. 2I), we analyzed proliferation of ventricular and subventricular cortical progenitors by immunostaining for Ser-10-phosphorylated histone H3, a widely used marker of mitosis. The number of mitotic cells at the ventricular surface was significantly lower in the mutant (Fig. 6, A  and B). To investigate how other phases of the cell cycle are affected, we carried out double labeling for BrdU and Ki67. For this, E15.5 pregnant mice were sacrificed after 1 h of BrdU labeling to retrieve fetal brains for immunofluorescence microscopy. Consistent with reduced mitotic cells in the ventricular surface (Fig. 6, A and B), total proliferating (Ki67 ϩ ) cells and the S phase (BrdU ϩ Ki67 ϩ ) cells declined significantly in the mutant subventricular zone (Fig. 6, C-G). However, the labeling index (i.e. the ratio of BrdU ϩ Ki67 ϩ versus Ki67 ϩ cells) remained unchanged in the mutant (Fig. 6H), indicating that neuronal progenitors cycle at similar rates of lamination (43,60). This is very similar to what was reported for mice lacking the T-box transcription factor Tbr2 (43,60). To assess whether the reduced cortical thickness is due to apoptosis, we performed immunostaining for activated caspase-3. No apoptotic signals were detected in the control or bKO fetal forebrain (data not shown). These results suggest that Brpf1 loss reduces the neuronal progenitor pool.
To substantiate this, we examined Tbr2 ϩ intermediate neuronal progenitors (or basal progenitors), which are known to be important for cortical lamination (43, 60). These progenitors Abnormal neocortex in the mutant brain. A, decreased brain weight was observed in the bKO mice at P12, but the small brain size was proportional to decreased body weight (n ϭ 3). *, p Ͻ 0.05; ns, not statistically significant. B, dorsal view of wild-type, heterozygous, and bKO brains at P14. Compared with the wild-type and heterozygote, the mutant brain displayed shorter rostral-caudal span (flanked with green arrowheads). C-E, Nissl staining of sagittal brain sections. At P10, the bKO mice had thinner cortex, especially layer V (C), although the mutant cerebellum was normal (D). The cortical difference was apparent at P0 (E). F and G, DAPI staining of brain sections at P0. Representative images are shown, and quantification of cortical layer thickness is presented. The entire cortex and layers V-VI were significantly thinner in the mutant. Data were based on images from three mice per genotype and 3-4 sections per mouse. H and I, DAPI staining of fetal brain sections at E15.5. The developing cortex in the mutant, notably in the cortical plate (CP), was thinner than that in the wild type. Data were from three fetuses per genotype and 4 -5 sections per fetus. decreased significantly in the subventricular zone of the mutant neonate cortex (Fig. 6, I and J). However, Sox2 ϩ neural stem cells were not affected (Fig. 6, K and L). These results support the idea that Brpf1 deficiency compromises neuronal progenitor production. Brpf1 Loss Leads to Callosal Hypoplasia-In addition to the abnormal neocortex in the mutant brain, we noticed the absence of the corpus callosum in the ventral and dorsal but not the middle transverse sections at P14 (Fig. 7A), The defect was also present at later time points such as P24 (Fig. 8A) in the caudal part of serial coronal sections. To investigate whether the observed defects originate prior to birth, we analyzed wildtype and mutant brain sections at earlier developmental time points. As shown in Fig. 8B, the wild-type corpus callosum was similar to the mutant one at E17.5. However, the defect became evident caudally at P0 (Fig. 8C), indicating that the abnormality occurs shortly before birth. The corpus callosum contains axon tracks and is heavily myelinated, so we performed myelin staining. This staining confirmed partial callosal agenesis and also revealed defective myelination (Fig. 7C). Genitopatellar syndrome patients carrying MORF mutations display callosal agenesis (23,24). Thus, the hypoplasia observed in Brpf1-deficient mice provide some insights into callosal agenesis in these patients.
Several groups of cells are crucial for callosal development, including the supracallosal gyrus (or indusium griseuem), which is located above the corpus callosum (61). Examination of Nissl-stained sections at high magnification revealed that the supracallosal gyrus was largely missing or ill-formed in the mutant (Fig. 7B). This provides a cellular mechanism whereby Brpf1 causes callosal agenesis. Molecularly, Tbr2 is important for callosal development, and its loss causes agenesis of the corpus callosum in human patients (62) and mice (60,63). Related to this, Tbr2 expression decreased in the bKO brain (Fig. 6, I and J). Another transcription factor, Nr2f1, is also required for the formation of corpus callosum by regulating axonal growth in mice (64). Moreover, the axon guidance regulator roundabout 3 (Robo3) is important for midline axon crossing (65). As described in the section below, Nr2f1 and Robo3 transcript levels decreased in the mutant. These results indicate that Brpf1 loss reduces Tbr2, Nr2f1, and Robo3 expression, thereby leading to callosal hypoplasia.
Brpf1 Regulates Transcriptional Activation and Gene Silencing-To understand the underlying molecular mechanisms, we first analyzed transcripts of different layer markers, including Reelin, Ror␤, Fezf2, and Tbr1, which are known to govern the identity of layers I, IV, V, and VI, respectively (57). As shown in Fig. 9A, the knock-out efficiency was expected. Importantly, mRNA levels of layer IV-VI markers were reduced in bKO brains at both P10 and P24 (Fig. 9B). In addition to lamination, arealization is important for neocortical patterning (66,67). Besides the layer V marker Fezf2, several other transcription factors, including Emx1, Pax6, and Nr2f1, are important for neocortical arealization. RT-PCR analysis revealed that their transcript levels were also altered at both P10 and P24 (Fig. 9B).
To investigate the impact on transcription systematically, we prepared total RNA from three pairs of control and mutant dorsal brain cortices at P4 for microarray-based gene expression analyses, leading to identification of ϳ200 genes with transcript levels reduced in the mutant by at least 3-fold. Among the top 50 are Robo3, osteocrin (also known as muslin), and the transcription factor Otx1 ( Fig. 9C and supplemental Table S1). Robo3 controls neurite outgrowth (65), so its decreased expression is consistent with the defects observed in cortical neurogenesis (Figs. 4 -6). Osteocrin is a novel peptide only known to be important for bone and muscle development (68), so its expression in the brain of young pups suggests a novel role in cortical development. Consistent with the observed cortical defects (Figs. 2-4), Otx1 is a marker of layer V pyramidal neurons, known to be composed of decreased Morf expression (27). RT-qPCR confirmed down-regulated expression of these three genes ( Fig. 9C and supplemental Table S1). Gene Set Enrichment Analysis revealed ribosomal structural proteins as the top group whose expression was down-regulated in the mutant. RT-qPCR detected minimal changes in the levels of related transcripts (supplemental Table S1), indicating that expression of ribosomal structural proteins is not significantly affected.
Microarray analyses also uncovered ϳ100 genes whose transcript levels were elevated in the mutant by at least 3-fold. The top 40 exhibited elevation by Ͼ4-fold. Unexpectedly, among them are multiple Hox genes, including Hoxa7, Hoxb5, Hoxb13,  Hoxc4, Hoxc6, Hoxc10, and Hoxd8 (Fig. 9C and supplemental Table S1). Moreover, the elevation was dramatic as some of them (e.g . Hoxb13, Hoxc4, Hoxc6, and HoxC10) increased by  Ͼ100-fold (Fig. 9C). Hox genes are important for body patterning and hindbrain development, and they are normally silenced in the neocortex (69). Also among the top 40 are two other homeobox proteins, Hmx1 and En2 (Fig. 9C and supplemental Table S1). Hmx1 is important for development of craniofacial structures, and mutations in its gene cause oculoauricular syndrome with characteristics of defective eye and ear development (70). En2 is critical for cerebellar development, and its expression is a marker of prostate cancer (71). Other transcrip-  MARCH 13, 2015 • VOLUME 290 • NUMBER 11 cal to that of the cytoskeletal protein Spag6, so we have referred to this new protein as Spag6l. RT-qPCR confirmed transcriptional up-regulation of the above transcription factors and Spag6l in the mutant (Fig. 9C). For example, by RT-qPCR, Hoxc4 mRNA increased in the mutant by Ͼ1000-fold (Fig. 9C).

Brpf1 in the Neocortex and Corpus Callosum
In agreement with this, the protein was not expressed in the heterozygotes (Fig. 9D, lanes 1-3) but became easily detectable in the mutant homozygotes (lanes 4 -6). Normally, a majority of these genes are not expressed in the cerebral cortex, so the results from microarray analysis (supplemental Table S1), RT- qPCR (Fig. 9C), and immunoblotting ( Fig. 9D) indicate that Brpf1 acts as a silencer to inhibit expression of these genes in the forebrain.

DISCUSSION
The human genome encodes hundreds of chromatin regulators. With many of them molecularly characterized, an emerging issue is how they control biological processes in vivo. This study identifies an important role of Brpf1, a unique chromatin regulator, in development of the neocortex and corpus callosum and regulating expression of Hox and other genes.
Brpf1 Regulates Forebrain Development-Forebrain-specific inactivation of the Brpf1 gene caused early postnatal lethality (Fig. 1A), severe growth retardation (Fig. 1, B and C), and aberrant behaviors (Fig. 1, D-H). Histological analyses revealed disorganization of the neocortex (Figs. 2-4), abnormal neurogenesis (Figs. 5 and 6), and partial agenesis of the corpus callosum ( Figs. 7 and 8). Although not characterized here, Brpf1 loss also led to hypoplasia of the dentate gyrus in the hippocampus (Fig.  8A). 3 Consistent with these results, Brpf1 is highly expressed in the neocortex and hippocampus (35). These novel findings support its important role in development of the mouse forebrain.
In addition to the forebrain, Brpf1 is highly expressed in the cerebellum (35), so it will be interesting to investigate whether Brpf1 plays a role in this structure. The Brpf1 f/ϩ line used herein should be valuable for this. Because the human BRPF1 gene is mutated in sonic hedgehog-driven medulloblastoma (36), the role of mouse Brpf1 in the cerebellum is pathologically important. Its loss in the forebrain led to elevated expression of En2 and Twist1 (Fig. 9C), known to be important for tumorigenesis and metastasis. Whether such regulation also occurs in the cerebellum is an interesting question worth further investigation.
Molecularly, Brpf1 loss reduced the transcription of multiple genes important for lamination and arealization of the neocortex (Fig. 9, A and B). This role in transcriptional regulation is in agreement with published molecular and cell-based studies that Brpf1 and its partners, Moz and Morf, function as transcriptional co-activators (29). At the cellular level, Brpf1 loss altered distribution of cortical neurons (Figs. 4 and 5) and Tbr2 ϩ intermediate neuronal progenitors (Fig. 6, I and J). These progenitors are important not only for neocortical but also for callosal development (60,72,73). Moreover, TBR2 loss causes agenesis of the corpus callosum in humans and mice (60,62,63). Thus, deregulation of these progenitors provides a poten- . Developmental origin of callosal agenesis. A, Nissl staining was performed on serial coronal sections from control and bKO brains at P24. Two sets of representative images are shown here. The agenesis was evident at the caudal section of the mutant brain. Note hyperplasia in the subiculum (marked with red arrowheads) and hypoplasia in the dentate gyrus of the mutant hippocampus. B and C, Nissl staining was performed on serial coronal sections from control and bKO brains at E17.5 and P0. Two sets of representative images are shown. At E17.5, both control and bKO mice formed the rostral corpus callosum (marked by green double asterisks). At P0, the mutant corpus callosum failed to cross the midline caudally (red double asterisks). The black asterisks in C mark two regions accidentally torn during section processing. Scale bars, 0.5 mm. tial mechanism whereby Brpf1 inactivation led to cortical abnormalities (Figs. 2-6) and partial callosal agenesis (Figs. 7 and 8).
Brpf1 is paralogous to Brpf2 and Brpf3 (16). Loss of mouse Brpf2 leads to multiple defects in neural tube closure, eye development, and erythropoiesis (74), but it remains unclear what roles Brpf2 may play in the brain. Moreover, little is known about the biological functions of Brpf3. The defective development observed here indicates that Brpf1 has functions that neither Brpf2 nor Brpf3 can substitute. Thus, this study reiterates the notion that the three mammalian proteins have distinct roles in vivo.
Interaction of Brpf1 with Moz, Morf, and Hbo1 in the Brain-According to molecular and cell-based assays, Brpf1 interacts with the three acetyltransferases Moz, Morf, and Hbo1 to stimulate their enzymatic activities and control their substrate specificity (28 -30), so an interesting question is how these enzymes may contribute to the defects caused by Brpf1 inactivation (Fig.  9E). Mice with residual Morf expression display abnormalities in the neocortex (27). Importantly, mutations in the MORF   gene causes Noonan syndrome-like disorder (21), Ohdo syndrome (22), genitopatellar syndrome (23,24), and blepharophimosis-ptosis-epicanthus inversus syndrome (25). Patients with these diseases display intellectual disability and ill-formed corpus callosum (26). Thus, the defective noecortex and corpus callosum in the Brpf1 mutant mice suggest that Morf may be an immediate target of Brpf1 in the mouse forebrain (Fig. 9E, top  panel). However, it should be noted that both mouse Moz and Hbo1 are also expressed in the cerebral cortex and hippocampus (35), so it remains to be investigated whether and how these two acetyltransferases mediate Brpf1-dependent actions in the forebrain. Further studies are needed to dissect the relative contribution of these three acetyltransferases to the role of Brpf1 in regulating forebrain development (Fig. 9E). As discussed below, rather than the linear relationship that cell-based studies have suggested (28 -30), Brpf1 may differentially interact with the acetyltransferases according to various developmental contexts. Brpf1 Activates or Represses Transcription in a Context-dependent Manner-Interestingly, microarray-based gene expression analysis revealed that in addition to down-regulation of gene expression, Brpf1 loss dramatically up-regulated transcription of different Hox genes and the genes of multiple transcription factors important in various developmental processes (supplemental Table S1 and Fig. 9C). RT-qPCR and immunoblotting confirmed this (Fig. 9, C and D). Notably, no such up-regulation was detected in Brpf1-deficient embryos or mouse embryonic fibroblasts (data not shown), supporting that the repressive role of Brpf1 is specific to the brain. Normally, a majority of these genes is not expressed in the forebrain, so the results indicate that Brpf1 is required for silencing gene expression in brain development. This is striking and unexpected.
In stark contrast, fish Brpf1 is important for maintaining Hox gene expression during skeletal development (75,76). Consistent with this, fish Moz is essential for Hox gene expression during skeletogenesis (77,78). In agreement with these fish genetic studies, mouse Moz regulates segmental identity and Hox gene expression (79). Thus, both Brpf1 and Moz activate Hox gene expression during skeletal development in fish and mice. It is noteworthy, however, that little is known about the function of Moz in the brain. One study indicates that mouse Moz regulates neural stem cell proliferation by repressing p16 expression (80), but it remains unclear how this repression occurs.
Interestingly, related to the repressive role of Brpf1 in forebrain Hox gene expression (supplemental Table S1 and Fig. 9, C and D), the fly HBO1 ortholog Chameau is essential for Hox gene silencing by polycomb group proteins (31). Moreover, Chameau functionally replaces yeast Sas2 (something about silencing) in vivo (31). Sas2 was initially identified as a gene silencer and is a founding member of the MYST family of histone acetyltransferases (81). In light of their similar roles in silencing Hox gene expression, it is tempting to propose that Brpf1 interacts with Hbo1 and acts as a gene silencer in the forebrain (Fig. 9E, bottom panel). Further studies are needed to investigate this hypothesis and to determine the function of Hbo1 in the mouse brain.
In conclusion, Brpf1 plays an important role in development of the neocortex and corpus callosum by regulating related neurogenesis and transcriptional programs. Our results also suggest that depending on cellular and developmental contexts, Brpf1 functions either as an activator or a silencer of gene expression in vivo.