Identification of potential target genes for RFX4_v3, a transcription factor critical for brain development

Abstract

Regulatory factor X4 variant transcript 3 (Rfx4_v3) gene disruption in mice demonstrated that interruption of a single allele (heterozygous, +/–) prevented formation of the subcommissural organ, resulting in congenital hydrocephalus, while interruption of two alleles (homozygous, –/–) caused fatal failure of dorsal midline brain structure formation. To identify potential target genes for RFX4_v3, we used microarray analysis to identify differentially expressed genes in Rfx4_v3-deficient mouse brains at embryonic day 10.5, before gross structural changes were apparent. Of 109 differentially expressed transcripts, 24 were chosen for validation and 22 were confirmed by real-time PCR. Many validated genes encoded critical proteins involved in brain morphogenesis, such as the signaling components in the Wnt, bone morphogenetic protein (BMP) and retinoic acid (RA) pathways. Cx3cl1, a CX3C-type chemokine gene that is highly expressed in brain, was down-regulated in the Rfx4_v3-null mice. Both human and mouse Cx3cl1 proximal promoters contained highly conserved X-boxes, known cis-acting elements for RFX protein binding. Using the Cx3cl1 promoter as an example of a target gene, we demonstrated direct binding of RFX4_v3 to the Cx3cl1 promoter, and trans-acting activity of RFX4_v3 protein to stimulate gene expression. These data suggest that RFX4_v3 may act upstream of critical signaling pathways in the process of brain development.

Abbreviations used

BMP

bone morphogenetic protein

ChIP

chromatin immunoprecipitation

CX3CL1

CX3C ligand 1

DBD

DNA binding domain

E10.5

embryonic day 10.5

EMSA

electrophoretic mobility shift assays

Fzd10

Frizzled-10

IL-5Rα

interleukin-5 receptor α chain

MLC

megalencephalic leukoencephalopathy with subcortical cysts

RA

retinoic acid

RFX4_v3

regulatory factor X4 variant transcript 3

TSS

transcription start site

The RFX proteins belong to the winged-helix subfamily of helix-turn-helix transcription factors, and are defined by a highly conserved 76-amino acid DNA binding domain (DBD) that is distinct from any other DBD. These proteins recognize ‘X-box’ consensus sequences, 5′-GTNRCC(0–3 N)RGYAAC-3′ (where N is any nucleotide, R is a purine, Y is a pyrimidine, and the two half sites GTNRCC and RGYAAC are separated by 0–3 base pairs), in target DNA, and regulate gene expression (Gajiwala et al. 2000). RFX family members are present in a broad range of eukaryotic organisms. For example, in Schizosaccharomyces pombe, the RFX-like transcription factor SAK1 is thought to function downstream of the cAMP-dependent protein kinase, and to regulate the exit from the mitotic cell cycle (Wu and McLeod 1995). In Saccharomyces cerevisiae, CRT1, which contains a RFX-type DBD, functions as an effector in the DNA damage and replication block checkpoint pathway (Huang et al. 1998). DAF-19, the RFX family member identified in the Caenorhabditis elegans, is expressed specifically in all ciliated sensory neurons, and loss of its function results in severe sensory defects (Swoboda et al. 2000; Efimenko et al. 2005). DmRFX, the Drosophila RFX protein, is an embryonic type I sensory neuron marker, and is necessary for ciliated sensory neuron differentiation (Vandaele et al. 2001; Dubruille et al. 2002).

In mammals, five RFX proteins have been identified, named RFX1–RFX5. RFX5 has been the most intensively studied family member; it plays a critical role in the regulation of major histocompatibility complex class II gene expression, and mutations in Rfx5 cause the bare lymphocyte syndrome (Reith and Mach 2001). RFX3 has been shown to direct nodal cilium development and left–right asymmetry specification (Bonnafe et al. 2004). The functions of RFX1 and RFX2 remain unknown, but they have been implicated in the regulation of certain medically important genes, such as the gene encoding the interleukin-5 receptor α chain (IL-5Rα; Iwama et al. 1999).

RFX4 was originally identified in human breast cancer cells as a partial cDNA encoding a fusion of the amino-terminal half of the estrogen receptor and a short RFX-type DBD. This fusion is possibly as a result of an abnormal chromosomal translocation in breast cancers (Dotzlaw et al. 1992). More recently, two full-length RFX4 cDNAs that were expressed specifically in human testis were isolated, and named RFX4_v1 (for RFX4 variant transcript 1) and RFX4_v2 (Morotomi-Yano et al. 2002).

A new isoform of Rfx4 was identified recently in our laboratories by insertional mutagenesis (Blackshear et al. 2003). In these studies, transgenic mice were generated for cardiac-specific expression of a human epoxygenase gene (Seubert et al. 2004). One line of these mice developed an unexpected brain phenotype, with marked obstructive congenital hydrocephalus secondary to failure of formation of the subcommissural organ (SCO). Identification of the genomic sequences flanking the transgene indicated that the transgene was inserted into an intron in the mouse Rfx4 locus, and created a null allele for the expression of a novel brain specific isoform of Rfx4, termed Rfx4_v3. Interruption of both alleles (–/–) led to failure of dorsal midline brain structure formation and perinatal death. Rfx4_v3 mRNA was found primarily in the brain, where its regional expression during development was highly dynamic from the neural plate stages. The data indicated that Rfx4_v3 was critical for early brain development. Subsequently, another group also isolated a full-length Rfx4_v3 cDNA using a PCR approach, and demonstrated its expression in the suprachiasmatic nucleus, the central pacemaker site of the circadian clock (Araki et al. 2004).

Although Rfx4_v3 clearly has been implicated in early brain development, the molecular mechanisms underlying its function remain unknown. As RFX4_v3 is a presumed transcription factor, we speculated that decreased expression of this gene would interfere with the expression of downstream genes. To identify potential target genes for RFX4_v3, we hybridized mouse oligonucleotide microarray chips with samples derived from the Rfx4_v3 wild-type and null mutant mouse heads at embryonic day 10.5 (E10.5). We then validated outliers from the microarray experiments by the use of real-time quantitative PCR, and used the promoter of one of these outliers, Cx3cl1, to demonstrate direct trans-acting activity of RFX4_v3 on gene expression. These data suggest that RFX4_v3 may act upstream of some critical signaling pathways to regulate brain development.

Materials and methods

Mice and RNA isolation

The development of null Rfx4_v3 mice and their genotyping have been described elsewhere (Blackshear et al. 2003; Seubert et al. 2004). The mice were maintained on a pure C57BL/6 background, and Rfx4_v3 +/– mice were interbred to generate –/– mice. If the pregnant +/– mice were allowed to carry to term and deliver, the average litter size of these pregnancies was significantly smaller than litters from a control line, and only 11% of the born pups were –/–, suggesting substantial intrauterine or perinatal loss of the –/– pups. All of the –/– pups born died within 1 h of birth. If the pups were obtained between E8 and E18, the average size of those litters was similar to the control litters, and about 25% of the pups obtained were –/–, indicating no excess intrauterine mortality. All studies were approved by the NIEHS Animal Care and Use Committee.

Total RNA was isolated with the RNeasy Mini Kit (Qiagen, Valencia, CA, USA) from individual heads of Rfx4_v3 wild-type (+/+) or mutant (–/–) animals dissected at E10.5. Each embryo was genotyped from genomic DNA isolated from tail pieces.

Microarray analysis

Five pairs of samples were used for the microarray analysis. Four pairs of RNA samples were isolated from individual Rfx4_v3–/– or +/+ embryonic heads at E10.5. The fifth pair of samples contained pooled RNA extracted from multiple Rfx4_v3–/– or +/+ embryonic heads. Agilent mouse oligonucleotide microarray chips were used for the hybridization, containing more than 20 000 unique 60-mer oligonucleotides representing well-characterized mouse genes (see http://www.agilent.com/chem./dnasupport). Each RNA sample was amplified and converted to fluorescently labeled cRNA, either with cyanine 3 (Cy3) or cyanine 5 (Cy5), using the Agilent Fluorescent Linear Amplification Kit. There is routinely at least a 100-fold RNA amplification with use of this kit, according to the manufacturer's experiments. The fluorescently labeled cRNAs were mixed and hybridized simultaneously to the microarray chips. Each sample pair was hybridized to two arrays, employing a fluorescence reversal accomplished by labeling the –/– sample with Cy3 and the +/+ sample with Cy5 for one chip, and the –/– sample with Cy5 and the +/+ sample with Cy3 for another chip, to account for fluorophore incorporation bias. Chips were scanned with an Agilent dual-laser scanner (Agilent DNA Microarray Scanner with SureScan technology) using independent laser excitation of the two fluorescences at 532 and 633 nm wavelengths for the Cy3 and Cy5 labels, respectively.

Gene expression data, including images and gene expression markup language (GEML) files, were generated using the Agilent feature extraction software (v7.1), with defaults for all parameters. Data were then processed using the Rosetta Resolver system (version 3.2, build 3.2.2.0.33) (Rosetta Biosoftware, Seattle, WA, USA), which generated expression data such as p-values and error measurements. The Resolver system performed a squeeze operation that created ratio profiles by combining replicates while applying error weighting (in our experiments, the replication number is 2, as each pair of samples was hybridized to two chips). The error weighting consisted of adjusting for additive and multiplicative noise. A p-value was generated and propagated throughout the system, representing the probability that a gene was differentially expressed. The Resolver system then combined ratio profiles to create ratio experiments using an error-weighted average, as described in Stoughton and Dai (2002). We set the threshold of p-value as 0.01, and genes were considered ‘signature genes’ or ‘outliers’ if the p-value was less than 0.01. The final microarray outlier list contains genes that differed in at least four out of five pairs of biological samples. The entire data sets are available at the website: http://dir.niehs.nih.gov/microarray/zhang12/Zhang_RFX4_compiled.txt.

TaqMan^® real-time PCR and statistical analysis

To validate the microarray outlier list, TaqMan^® real-time PCR experiments were performed. RNA was treated with RQ1 RNase-free DNase (Promega, Madison, WI, USA), and reverse transcribed into first-strand cDNA using the ABI/Prism High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA, USA). The TaqMan PCR reactions were conducted in triplicate. The amplifications were performed as follows: 2 min at 50°C, 10 min at 95°C, then 40 cycles each at 95°C for 15 s and 60°C for 60 s in the ABI/Prism 7900 HT Sequence Detector System. The primer/probe sets used for the TaqMan assays are listed in Table 1. Twenty-one genes had pre-designed primer/probe sets available from ABI Assays-on-Demand™ gene expression products, and the assay IDs are listed. For the other three genes, primer/probe sets were designed by Primer Express software and synthesized by ABI. Results were normalized to an internal control transcript, that encoding glyceraldehyde-3-phosphate dehydrogenase (Gapd), using the TaqMan Rodent GAPDH Control Reagents (ABI/Prism). RNA that was treated with DNase but not reverse transcribed into cDNA was also used for TaqMan PCR reactions. No PCR products were detected from any of the RNA samples using all of the TaqMan primer/probe sets, indicating that genomic DNA contamination was not influencing the results.

Table 1.  Primer and probe information used for real-time PCR

GenBank ID	Gene	Primers and probe
Detailed description for each gene can be found in Tables 2 and 3. ^aEST: expressed sequence tag.
NM_031166	Idb4	Mm00499701_m1
AK014242	Fez	Forward: GGCGACTCAGTCATGGACAGT
		Reverse: GCTGGAGCAGTCGCTAGCA
		Probe: CTGCCTCAACGCGACCACCAAA
NM_013833	Rax	Mm00443434_g1
NM_007811	Cyp26a1	Mm00514486_m1
NM_010052	Dlk1	Mm00494477_m1
NM_011127	Prrx1	Mm00599934_m1
NM_010512	Igf1	Mm00439561_m1
NM_027280	Nkd1	Mm00471902_m1
NM_011707	Vtn	Mm00495976_m1
NM_010446	Foxa2	Mm00839704_mH
AK034131	Rfx4_v3	Forward: TCTCGCTCTCTCCTTCAGCTCTA
		Reverse: GGTGGGTGAGAGGGCAAAA
		Probe: CGCTTCTTCGCCTCTTTTCTTTCCACTAGTT
NM_022435	Sp5	Mm00491634_m1
AK053799	Bh1	Forward: CGGCTTGTGCACCCTACAG
		Reverse: CGCGTTGCTCTCCTGCTT
		Probe: CCAGCGTTTCTTCTCCTCATCACACCC
NM_009522	Wnt3a	Mm00437337_m1
AK044492	EST^a	Mm00620949_m1
NM_009573	Zic1	Mm00656094_m1
NM_008516	Lrrn1	Mm00493200_m1
NM_009142	Cx3cl1	Mm00436454_m1
AK052950	Fzd10	Mm00558396_s1
NM_138683	Rspondin	Mm00507076_m1
NM_133241	Mlc1	Mm00453827_m1
NM_009575	Zic3	Mm00494362_m1
NM_010836	Msx3	Mm00440331_m1
NM_009521	Wnt3	Mm00437336_m1

All real-time PCR data are expressed as mean + SEM. Student's t-test (two samples assuming equal variances, two-tail) was used to determine significant differences between groups.

Plasmid constructions

The full-length human RFX4_v3 cDNA was amplified by a PCR approach (forward primer 5′-GATGGATCCCCAAACATCAG-3′ and reverse primer 5′-TGGCTCGAGTTTAGCCCATC-3′) and inserted into the pcDNA3.1/MycHis vector (Invitrogen, Carlsbad, CA, USA).

To clone the 5′-upstream region of the mouse Cx3cl1 gene, PCR was carried out with mouse genomic DNA using forward primer 5′-GATACGCGTGACTCCCAGGATT-3′ (corresponding to bases 19821580-92 of NT_078575.2) and reverse primer 5′-TGGAGATCTGGCTGGAGTGCAG-3′ (bases 19822565-77). A 998-bp genomic fragment, −937 to +61 bp relative to the Cx3cl1 gene transcription initiation site (base 19822517 of NT_078575.2), was generated and cloned into the pGL-3 Basic vector (Promega). This Cx3cl1 promoter/luciferase construct was named −937/+61-Luc. Subclones with appropriate unidirectional deletions of the −937/+61-Luc construct were prepared using the Erase-a-Base System (Promega) according to the method described in manufacturer's protocol. Cx3cl1 promoter/luciferase constructs with mutated X-box 1 were made from the −937/+61-Luc plasmid using a QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA). The first mutation construct (X-box 1 mutant 1-Luc) was generated with primers F1 5′-GCAACTGGTTCCCTACGCGTCTGGGAGGAATCCAGC-3′ and R1 5′-GCTGGATTCCTCCCAGACGCGTAGGGAACCAGTTGC-3′, and the second mutation construct (X-box 1 mutant 2-Luc) was amplified with primers F2 5′-GCAACTGGTTCCCTCCCGGGCTGGGAGGAATCCAGC-3′ and R2 5′-GCTGGATTCCTCCCAGCCCGGGAGGGAACAGTTGC-3′. The underlined nucleotides represent the mutated ones. The primers for the mutagenesis studies correspond to bases 19822376-411 of NT_078575.2; the X-box 1 is located at bases 19822383-95.

Correct sequences for all constructs were confirmed using the ABI/Prism dRhodamine Terminator Cycle Sequencing Ready Reaction Kit.

Transient transfection and reporter assays

COS cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum and 100 units/mL penicillin/streptomycin. For reporter assays, cells were grown in 12-well plates at 60% confluence, and then transiently transfected with the RFX4_v3 expression vector or the control vector (0.5 μg), and a Cx3cl1 promoter plasmid (0.5 μg), using Superfect reagent (Qiagen). Plasmid pRL-SV40 (1 ng, Promega) was also co-transfected to normalize transfection efficiency. Transfection assays were performed in triplicate. Forty-eight hours after transfection, the cells were washed once with phosphate-buffered saline, and lysed in the Passive Lysis Buffer (Promega). Luciferase activities in the lysates were measured with Dual-Luciferase Reporter Assay Systems (Promega) according to the manufacturer's protocol.

Electrophoretic mobility shift assays (EMSA)

Oligonucleotides corresponding to the wild-type or mutant X-boxes were synthesized. The oligonucleotides were annealed as described (Morotomi-Yano et al. 2002) and end-labeled with [γ-³²P]ATP by using T4 polynucleotide kinase (New England Biolabs, Ipswich, MA, USA). The RFX4_v3 protein with a Myc tag was synthesized in vitro with a TNT T7 Quick Coupled Transcription/Translation System (Promega) according to the manufacturer's protocol. The protein-DNA complex was allowed to form for 20 min at 22°C in a 20-μL mixture containing the ³²P-labeled probe (0.2 ng, 150 000 cpm), 2 μL of in vitro translation products, 4 μg of poly(dI-dC), 80 μg bovine serum albumin, 40 mm KCl, 0.2 mm EDTA, 2 mm dithiothreitol, 10 mm Tris (pH 7.5), and 20% (v/v) glycerol. For competition analyses, an excess amount of unlabeled oligonucleotides was included in the binding reaction for 10 min prior to the addition of the probe. For supershift experiments, 1 μg of anti-Myc polyclonal antibody PRB-150P (Covance, Princeton, NJ, USA) was added to the binding reaction for 1 h. The samples were separated on a 4% (v/v) polyacrylamide gel in 0.5 × Tris borate EDTA buffer at room temperature (Morotomi-Yano et al. 2002). After the gel was dried, the DNA-protein complexes were visualized by autoradiography.

Nucleofection and chromatin immunoprecipitation (ChIP)

C2C12 (mouse myoblast/fibroblast) cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum and 100 units/mL penicillin/streptomycin. For nucleofection experiments, six 10-cm dishes of cells were trypsinized, re-suspended in 600 μL of Nucleofector Solution V (Amaxa, Gaithersburg, MD, USA) and mixed with 15 μg of RFX4_v3 expression plasmid. Cell mixtures were then placed in six Amaxa cuvettes and each cuvette of cells was given an electric pulse (program B-032) using the Nucleofecotor II device (Amaxa). After the nucleofection procedure, cells were plated and incubated for 20 h at 37°C. As a negative control, another six 10-cm dishes of C2C12 cells were also nucleofected with 15 μg of control plasmid. The transfection efficiency of C2C12 cells was more than 80%, and the cell viability was more than 70% 20 h post-nucleofection (data not shown).

ChIP experiments were performed using ChIP-IT™ Enzymatic kit (Active Motif, Carlsbad, CA, USA) according to the manufacturer's instructions. Basically, cells post-nucleofection were fixed in 1% formaldehyde, and chromatin was isolated and enzymatically sheared into 200–800 bp fragments (chromatin was incubated with enzymatic shearing cocktail for 8.5 min at 37°C). Sheared chromatin was pre-cleared with the protein G beads, and 10 μL of supernatant was saved as ‘input DNA’. Half of the remaining supernatant was incubated with negative control IgG (2 μg, Active motif) at 4°C overnight, and another half of the supernatant was incubated with 2 μg of anti-Myc polyclonal antibody PRB-150P (Covance). The chromatin-antibody mixture was then incubated with the protein G beads and, following the washing steps, DNA was eluted from the beads. After the reverse cross-linking step, DNA was treated with RNase A and proteinase K, and purified from the mini-columns. DNA was then amplified by real-time PCR using the ABI/Prism 7900 HT Sequence Detector System and Power SYBR Green PCR master mix. The PCR reactions were conducted in triplicate, and performed as follows: 2 min at 50°C, 10 min at 95°C, 40 cycles each at 95°C for 15 s and 60°C for 60 s, then followed by the dissociation stage of 95°C for 15 s, 60°C for 15 s, and 95°C for 15 s. The primers used to amplify the mouse Cx3cL1 promoter were: F1 5′-CCTAGGTTCTCCAGGGAAGG-3′, F2 5′-GCTCAGTGTCCGGCTTAGAC-3′, and R 5′-GGGGAGAGGAAGAGCCTGTA-3′, and their positions were indicated in Fig. 2(a). All real-time PCR data are expressed as mean + SEM. The ChIP experiments were repeated twice and similar results were obtained.

Figure 2.  Activation of mouse Cx3cl1 promoters by RFX4_v3 protein. (a) The 5′-upstream region of mouse Cx3cl1. The putative transcription start site is numbered +1, and the numbers on the left are the nucleotide positions relative to it. Consensus binding sites for the indicated transcription factors are underlined and labeled (→, binding sites on the ‘+’ strand of DNA; ←, binding sites on the ‘–’ strand). The translation start site ATG is also underlined. The promoter regions of nine different luciferase constructs used for the transient transfection assays are also shown (, 5′ ends of the promoters, with the numbers relating to the transcription start site; , 3′ ends of the promoters, which are all at nucleotide +61, just upstream of the ATG). The primers used for ChIP assays (Fig. 5) are underlined with the dotted lines. (b) X-box elements in the Cx3cl1 promoter. The sequences of the forward strand of the mouse Cx3cl1 X-box 1 element, the complementary strand of the mouse Cx3cl1 X-box 2 element (2r), the forward strand of the mouse Cx3cl1 X-box 3 element, and the forward strand of human Cx3cl1 X-box element are similar to the X-box consensus sequence and to the well-characterized X-box from the IL-5Rα enhancer (R, purine; Y, pyrimidine, N, any nucleotide). (c) Transient transfection assay of mouse Cx3cl1 promoters. Nine luciferase constructs containing different lengths of the 5′-upstream regions of mouse Cx3cl1 (details in Fig. 2a) and pGL3-Basic vector (promoter-less) were transiently transfected with control vector (pcDNA3.1/MycHis, white bars) or the RFX4_v3 over-expressing vector (RFX4_v3/pcDNA3.1/MycHis, gray bars). The firefly luciferase activities were normalized by renilla luciferase activities, and relative luciferase activities were calculated as a ratio of the activity from a construct containing −937/+61 of the promoter region and co-transfected with the control vector. The values represent the means and SEM. *Significantly different when comparing co-transfection with the control vector and the RFX4_v3 over-expressing vector of the same luciferase construct.

Results

Differentially expressed genes in E10.5 embryonic heads in Rfx4_v3–/– and +/+ mice

For this analysis, five independent pairs of samples were used, and each pair of RNA samples was assayed in duplicate. Four pairs of samples were isolated from individual embryonic heads at E10.5, whereas the fifth pair of samples contained pooled RNA from multiple Rfx4_v3–/– or +/+ embryos from different litters than the individual samples. We chose E10.5 as the time point to analyze as it is the last stage in development in which overall brain structure appears to be normal, with failure of midline structure formation and gross brain disruption occurring at later stages. The microarray results were analyzed, and the Resolver lists containing transcripts whose levels were different at p < 0.01 are available at http://dir.niehs.gov/microarray/zhang12/Zhang_RFX4_compiled.txt. The final microarray outlier list contained the common transcripts that differed in at least four of five biological replicates (Tables 2 and 3; only the transcripts with average fold changes greater than 1.2 are shown). Thirty-four genes showed increased expression, and 75 genes exhibited decreased expression in the Rfx4_v3–/– heads. Expression patterns of these 109 transcripts were evaluated using Rosetta Resolver cluster software with average gene expression ratio intensities at the 95% confidence level (Fig. 1). The expression patterns were very similar among the five pairs of samples.

Table 2.   Up-regulated genes (fold changes, p < 0.01) in Rfx4_v3–/– embryonic heads compared with +/+ heads (only transcripts with fold changes greater than 1.20 are shown)

Gene description	GenBank ID	Fold change^a
^a Mean ± SEM (for five pairs of samples).
RIKEN cDNA 3110069A13 gene	AK014242	1.50 ± 0.10
Mus musculus adult male testis cDNA, clone: 4932442L11	AK077061	1.47 ± 0.07
Forkhead box G1	NM_008241	1.43 ± 0.08
Cytochrome P450, 26, retinoic acid	NM_007811	1.41 ± 0.12
RIKEN cDNA 1190002H23 gene	NM_025427	1.41 ± 0.09
Retina and anterior neural fold homeobox	NM_013833	1.40 ± 0.05
Delta-like homolog (Drosophila)	NM_010052	1.40 ± 0.03
Inhibitor of DNA binding 4	NM_031166	1.38 ± 0.09
Scleraxis = basic helix-loop-helix transcription factor	S78079	1.37 ± 0.05
Mus musculus 16 days embryo head cDNA, clone: C130018M11	AK047877	1.36 ± 0.07
Distal-less homeobox 2	NM_010054	1.35 ± 0.09
Small inducible cytokine subfamily B (Cys-X-Cys), member 14	NM_019568	1.33 ± 0.11
Paired related homeobox 1	NM_011127	1.32 ± 0.06
LIM homeobox protein 8	D49658	1.32 ± 0.04
Hypothetical Metallo-dependent phosphatases structure containing protein	AK048421	1.30 ± 0.04
Somatostatin	NM_009215	1.30 ± 0.04
B-cell lymphoma/leukaemia 11B	NM_021399	1.29 ± 0.04
RIKEN cDNA 3110001N10 gene	NM_027397	1.26 ± 0.05
Neuronatin	NM_010923	1.24 ± 0.05
Hypothetical protein, MGC: 7475	NM_024474	1.24 ± 0.04
Retinol binding protein 1, cellular	NM_011254	1.24 ± 0.04
Insulin-like growth factor 1	NM_010512	1.24 ± 0.03
Up-regulated during skeletal muscle growth 4	NM_031401	1.24 ± 0.03
Goosecoid	NM_010351	1.22 ± 0.05
Integral membrane protein 2	NM_008409	1.22 ± 0.05
Coagulation factor II	NM_010168	1.21 ± 0.22
Jun oncogene	NM_010591	1.21 ± 0.04
Mus musculus phosphatidic acid phosphatase type 2B (Ppap2b)	NM_080555	1.21 ± 0.03
Mus musculus 10, 11 days embryo whole body cDNA, clone:2810035F16	AK012857	1.21 ± 0.02

Table 3.   Down-regulated genes (fold changes, p < 0.01) in Rfx4_v3–/– embryonic heads compared with +/+ heads (only transcripts with fold changes greater than 1.20 are shown)

Gene description	GenBank ID	Fold change^a
^a Mean ± SEM (for five pairs of samples).
Trans-acting transcription factor 5	NM_022435	−2.89 ± 0.19
Mus musculus 0 day neonate eyeball cDNA, homeobox protein B-H1	AK053799	−2.68 ± 0.21
Mus musculus adult retina cDNA, clone: A930016D02	AK044492	−2.45 ± 0.22
Wingless-related MMTV integration site 3A	NM_009522	−2.16 ± 0.10
RIKEN cDNA 2900024I10 gene	AK013587	−2.11 ± 0.17
Small inducible cytokine subfamily D, 1/[chemokine (C-X3-C motif) ligand 1]	NM_009142	−2.05 ± 0.03
RIKEN cDNA 3100002J23 gene	AK013920	−2.01 ± 0.13
Wingless-related MMTV integration site 3	NM_009521	−1.99 ± 0.88
Homeo box, msh-like 3	NM_010836	−1.89 ± 0.28
Zinc finger protein of the cerebellum 1	NM_009573	−1.85 ± 0.07
MHC (A.CA/J(H-2K-f) class I antigen	NM_019909	−1.84 ± 0.10
Leucine-rich repeat protein 1, neuronal	NM_008516	−1.79 ± 0.11
Histocompatibility 2, K region	U47328	−1.71 ± 0.07
Mus musculus 13 days embryo head cDNA, clone: 3110063L10	AK014234	−1.65 ± 0.05
Mus musculus thrombospondin type 1 domain containing gene	NM_138683	−1.64 ± 0.06
Megalencephalic leukoencephalopathy with subcortical cysts 1 homolog	NM_133241	−1.63 ± 0.11
Zinc finger protein of the cerebellum 3	NM_009575	−1.63 ± 0.04
Mus musculus WAP, FS, Ig, KU, and NTR-containing protein	BC026460	−1.60 ± 0.17
Forkhead box A2	NM_010446	−1.59 ± 0.10
Mus musculus cDNA sequence AB023957	NM_133237	−1.59 ± 0.08
Expressed sequence AI449438	BE372171	−1.59 ± 0.05
Mus musculus 15 days embryo head cDNA, Frizzled 10 Precusor	AK052950	−1.56 ± 0.12
Mus musculus 0 day neonate eyeball cDNA, clone: E130318K11	AK053890	−1.56 ± 0.10
RIKEN cDNA 2810434J10 gene	NM_027280	−1.52 ± 0.03
Oligodendrocyte transcription factor 3	NM_053008	−1.50 ± 0.10
RIKEN cDNA 0610010I23 gene	NM_024223	−1.50 ± 0.05
G protein-coupled receptor 56	NM_018882	−1.47 ± 0.08
Mus musculus 13 days embryo forelimb cDNA, clone:5930402A21	AK031068	−1.46 ± 0.06
Vitronectin	NM_011707	−1.46 ± 0.04
Mus musculus fatty acid binding protein 7, brain (Fabp7)	NM_021272	−1.45 ± 0.11
Mus musculus N-acetylneuraminate pyruvate lyase	BC022734	−1.42 ± 0.04
RIKEN cDNA 2700055K07 gene	NM_026481	−1.41 ± 0.09
Cadherin EGF LAG seven-pass G-type receptor	NM_009886	−1.41 ± 0.07
Glycoprotein m6b	NM_023122	−1.40 ± 0.11
Mus musculus cDNA clone IMAGE:5149318	BC025841	−1.40 ± 0.04
Tumor necrosis factor receptor superfamily, member 19	NM_013869	−1.39 ± 0.03
RIKEN cDNA 1700018O18 gene	AK006096	−1.36 ± 0.07
Mus musculus 3 days neonate thymus cDNA, clone:A630062B03	AK042133	−1.34 ± 0.08
RIKEN cDNA 4432405B04 gene	NM_026486	−1.34 ± 0.06
Mus musculus, clone MGC:12117 IMAGE:3710089	BC012437	−1.33 ± 0.06
Creatine kinase, brain	NM_021273	−1.32 ± 0.07
Mus musculus adult male diencephalon cDNA, regulatory factor X, 4	AK034131	−1.32 ± 0.06
Slit homolog 2 (Drosophila)	AF144628	−1.28 ± 0.11
Mus musculus 0 day neonate cerebellum cDNA, clone: C230063O06	AK082560	−1.28 ± 0.03
Pre-B-cell leukemia transcription factor 4	NM_030555	−1.27 ± 0.03
Mus musculus collectin subfamily member 12 (Colec12)	NM_130449	−1.26 ± 0.06
Mus musculus Emu1 gene (Emu1)	NM_080595	−1.25 ± 0.06
Neuronal guanine nucleotide exchange factor	NM_019867	−1.25 ± 0.04
Mus musculus chemokine-like factor super family 8 (Cklfsf8)	NM_027294	−1.25 ± 0.03
RIKEN cDNA 2900046G09 gene	BC003957	−1.24 ± 0.10
Mus musculus adult male hypothalamus cDNA, clone: A230098A12	AK039115	−1.24 ± 0.06
Expressed sequence AI851425	AW681374	−1.24 ± 0.07
Expressed sequence tags (ESTs)	BB657009	−1.24 ± 0.03
Mus musculus citrate lyase β-like (Clybl)	NM_029556	−1.23 ± 0.07
Mus musculus adult male hippocampus cDNA, clone: C630002O20	AK049842	−1.22 ± 0.04
Mus musculus, RIKEN cDNA 2610034M16 gene, clone IMAGE:5358286	BC027128	−1.22 ± 0.04
Mus musculus 15 days embryo male testis cDNA, clone: 8030446E18	AK078756	−1.22 ± 0.03
RIKEN cDNA 2700091C21 gene	AK012596	−1.21 ± 0.09
Mus musculus 10 days embryo whole body cDNA, clone: 2610100L16	AK011787	−1.21 ± 0.09
Inhibitor of DNA binding 1	AK008264	−1.21 ± 0.03
calcitonin/calcitonin-related polypeptide, α	X97991	−1.21 ± 0.03
Solute carrier family 4 (anion exchanger), member 1	NM_011403	−1.20 ± 0.17
Mus musculus brain stress early protein (Gbi) mRNA	AF481964	−1.20 ± 0.06

Figure 1.  Cluster analysis of microarray data. Data obtained from the final outlier list of average gene expression ratio intensities at the 95% confidence level were used for cluster analysis. Transcripts with increased expression (–/– vs.+/+) are shown in red, and transcripts with decreased expression are shown in green, with relative intensities corresponding to the degree of change. Columns represent data obtained from different sample pairs.

To verify the differentially expressed transcripts identified in the microarray analysis, real-time PCR experiments were performed. Twenty-four transcripts from Tables 2 and 3 were chosen for validation by real-time PCR on the basis of their relatively large fold changes by microarray analysis, and/or their known or presumed roles in brain development. Real-time PCR quantifications were normalized with respect to Gapd, but similar results were also obtained when the PCR results were normalized to the 18s RNA (data not shown). Validations were performed on two groups of samples, and very similar results were obtained from both groups. One group of samples consisted of the five pairs of RNA samples used for the microarray analysis (data not shown), and another group of samples consisted of individual RNA samples isolated from four –/– and four +/+ Rfx4_v3 heads at E10.5 from the same litter that had not been used for the microarray analysis (real-time, Table 4). Twenty-two of the 24 transcript changes tested were verified by real-time PCR (p < 0.05), with similar or larger fold changes when compared with the results obtained by microarray. For some outlier genes, the microarray assays underestimated the differences between the –/– and the +/+ samples. For example, Wnt3a and Msx3 transcripts were down-regulated 2.16- and 1.89-fold, respectively, in the –/– samples, as shown by microarray; however, they were found to be decreased by about 16- and 15-fold in the real-time PCR analysis. There are many potential reasons for discrepancies between the two approaches; however, the data from the real-time PCR are more accurate, as it is a quantitative test with a wide dynamic range, while the microarray experiments have a low dynamic range with many variables. Changes in two other transcripts, Prrx1 and Igf1, which had only minor fold increases in the microarray analysis (1.32- and 1.24-fold, respectively), were not confirmed by the real-time PCR experiments (p > 0.05).

Table 4.   Real-time PCR validation of selected genes in the microarray outlier list (fold change: −/− vs. +/+)

GenBank ID	Gene	Gene Description	Microarray^a	Real-time^a
^aMean  ±  SEM.
Wnt signaling
NM_009522	Wnt3a	wingless-related MMTV integration site 3A	−2.16 ± 0.10	−16.17 ± 1.74
NM_009521	Wnt3	wingless-related MMTV integration site 3	−1.99 ± 0.88	−4.93 ± 0.53
AK052950	Fzd10	Frizzled 10 precusor	−1.56 ± 0.12	−2.41 ± 0.14
NM_027280	Nkd1	RIKEN cDNA 2810434J10 gene	−1.52 ± 0.03	−1.98 ± 0.32
Retinoic acid signaling
NM_007811	Cyp26a1	cytochrome P450, 26, retinoic acid	1.41 ± 0.12	1.65 ± 0.20
BMP signaling
NM_010836	Msx3	homeo box, msh−like 3	−1.89 ± 0.28	−14.73 ± 1.58
Transcription factors (not in the Wnt, RA, or BMP pathways)
NM_031166	Idb4	inhibitor of DNA binding 4	1.38 ± 0.09	1.68 ± 0.22
AK014242	Fez	RIKEN cDNA 3110069A13 gene	1.50 ± 0.10	1.58 ± 0.10
NM_013833	Rax	retina and anterior neural fold homeobox	1.40 ± 0.05	1.90 ± 0.19
NM_010446	Foxa2	forkhead box A2	−1.59 ± 0.10	−1.51 ± 0.09
AK034131	Rfx4_v3	regulatory factor X, 4	−1.32 ± 0.06	−6.81 ± 2.18
NM_022435	Sp5	trans-acting transcription factor 5	−2.89 ± 0.19	−3.76 ± 0.43
AK053799	Bh1	homeobox protein B-H1	−2.68 ± 0.21	−2.98 ± 0.44
NM_009573	Zic1	zinc finger protein of the cerebellum 1	−1.85 ± 0.07	−2.20 ± 0.24
NM_009575	Zic3	zinc finger protein of the cerebellum 3	−1.63 ± 0.04	−2.08 ± 0.21
Transmembrane proteins (not in the Wnt, RA, or BMP pathways)
NM_010052	Dlk1	delta-like homolog	1.40 ± 0.03	1.63 ± 0.10
NM_008516	Lrrn1	leucine rich repeat protein 1, neuronal	−1.79 ± 0.11	−1.70 ± 0.18
NM_133241	Mlc1	Megalencephalic leukoencephalopathy with subcortical cysts 1 homolog	−1.63 ± 0.11	−2.28 ± 0.33
Cytokines
NM_009142	Cx3cl1	CX3C ligand 1	−2.05 ± 0.03	−2.32 ± 0.12
Other genes
NM_138683	Rspondin	Mus musculus thrombospondin type 1	−1.64 ± 0.06	−1.50 ± 0.25
NM_011707	Vtn	vitronectin	−1.46 ± 0.04	−1.69 ± 0.25
AK044492	EST	adult retina cDNA, clone: A930016D02	−2.45 ± 0.22	−2.78 ± 0.78
Unvalidated genes
NM_011127	Prrx1	paired related homeobox 1	1.32 ± 0.06	0.94 ± 0.20
NM_010512	Igf1	insulin-like growth factor 1	1.24 ± 0.03	1.14 ± 0.28

Transcript changes in the Wnt, retinoic acid and bone morphogenetic protein pathways

Many constituents of Wnt signaling pathways are expressed in the developing and mature nervous systems, and Wnt signaling plays diverse roles in both early and later stages of brain development (Patapoutian and Reichardt 2000). Wnt proteins (> 19 in mammals) belong to a large family of secreted signaling molecules. Wnt proteins bind to a similarly large family of seven-transmembrane domain proteins, Frizzled proteins, ultimately activating downstream target gene expression (Burden 2000; Nelson and Nusse 2004). During brain development, Wnt signaling controls patterning, neural differentiation, and apoptosis. Among the microarray outlier transcripts validated by real-time PCR, four belonged to the Wnt signaling pathway: Wnt3a, Wnt3, Frizzled-10, and Nkd1. WNT3 and WNT3A are secreted Wnt glycoproteins, and their mRNA levels were dramatically down-regulated in heads from the Rfx4_v3–/– mice, decreasing by approximately 5- and 16-fold, respectively (Table 4). Frizzled-10 (Fzd10) is a member of the frizzled gene family. Fzd10 transcript levels were also decreased by about 2.4-fold in the heads from the Rfx4_v3–/– animals (Table 4). NKD1 protein has been proposed to act as an inducible antagonist of Wnt signaling, in which NKD1 antagonizes Wnt signals via a direct interaction with the intracellular Wnt signaling component Dishevelled (Wharton et al. 2001). The level of Nkd1 mRNA was decreased by about twofold in the Rfx4_v3–/– heads (Table 4).

Retinoic acid (RA) is a key factor in the patterning of the developing central nervous system (Melton et al. 2004), activating gene transcription via specific nuclear receptors. RA levels require precise regulation by controlled synthesis and catabolism, and when RA concentrations deviate from normal in either direction, abnormal growth and development occur. RA-deficient animals exhibit defects in caudal hindbrain formation, neurite outgrowth, dorsoventral patterning, and neuron apoptosis (McCaffery et al. 2003). CYP26A1 is a cytochrome p450 protein that metabolizes RA into more polar hydroxylated and oxidized derivatives. Cyp26a1–/– mouse fetuses have lethal morphogenetic phenotypes mimicking those generated by excess RA administration, indicating that CYP26A1 may be essential in controlling RA levels during development (Niederreither et al. 2002). Cyp26a1 mRNA was expressed at higher levels in the Rfx4_v3–/– embryo heads (Table 4), suggesting that RA levels might be lower than normal as a result of higher catabolism, and that some of the abnormalities observed in the Rfx4_v3 mutant brains might be caused by RA deficiency.

Bone morphogenetic proteins (BMPs) are members of the large transforming growth factor-β superfamily, mediating their effects by binding to heteromeric cell surface type I and type II receptors and initiating intracellular signaling. BMPs may initiate different subsets of activities through different downstream mediators. Msx genes encode homeodomain transcription factors related to the Drosophila msh gene, and genes of this family have been implicated as downstream targets of the BMPs, as Msx members are induced where BMP signaling is active (Cornell and Ohlen 2000). Different MSX proteins may mediate distinct aspects of BMP signaling during brain development (Liu et al. 2004). Msx3 mRNA levels were decreased more than 14-fold in the Rfx4_v3–/– heads (Table 4), suggesting that disruption of Rfx4_v3 expression may interfere with some aspect of BMP signaling during brain morphogenesis.

Expression of transcription factors critical for brain development

Of the 22 transcripts whose levels were validated by real-time PCR, 10 were transcripts encoding transcription factors. Besides Rfx4_v3 itself and Msx3, four other differentially expressed transcripts encoded proteins that are known to play important roles during brain development.

Rax belongs to a subfamily of the paired-like homeobox genes. Targeted deletion of Rax in mice eliminates eye formation (Mathers et al. 1997). Mouse embryonic stem cells ectopically expressing Rax could differentiate into retinal neurons (Tabata et al. 2004). RAX can also interact with PAX6, another critical transcription factor in brain, and enhance its transactivation function (Mikkola et al. 2001). In the Rfx4_v3 mutant brains, Rax mRNA levels were increased by 1.9-fold (Table 4).

FOXA2 is a winged-helix transcription factor that up-regulates vitronectin expression during mouse neuroblastoma cell differentiation (Shimizu et al. 2002). Foxa2 mutant mice have defects in the specification of the forebrain and the anterior definitive endoderm (Hallonet et al. 2002). Foxa2 mRNA levels were decreased by 1.5-fold in the Rfx4_v3 -/– heads (Table 4).

ZIC1 and ZIC3 were initially identified as the zinc-finger proteins expressed in granule cells throughout the development of the cerebellum (Aruga et al. 1996). They belong to the GLI transcription factor subfamily. Zic1 mutant mice show cerebellar hypoplasia (Aruga et al. 1998), while mutations in Zic3 cause a complex syndrome of left-right axis inversion and central nervous system defects (Purandare et al. 2002). Levels of Zic1 and Zic3 transcripts were decreased by about twofold in the mutant brains (Table 4).

Other important neuronal regulators

MLC1 is a putative transmembrane protein expressed primarily in brain. Mutations within human MLC1 cause megalencephalic leukoencephalopathy with subcortical cysts (MLC) and catatonic schizophrenia. MLC1 protein possibly functions as a cation channel (Leegwater et al. 2002). In the Rfx4_v3 mutant mice, Mlc1 mRNA levels were significantly reduced by about twofold in the fetal heads (Table 4).

Cx3cl1, a chemokine highly expressed in the cortex, hippocampus, basal ganglia, and olfactory bulb, appears to play a role in the response to brain injury and infection (Tong et al. 2000; Soriano et al. 2002). Expression levels of Cx3cl1 mRNA were decreased by twofold in the heads of the Rfx4_v3 mutant mice (Table 4).

Identification of genes containing putative X-box elements in their proximal promoters

To determine if the 78 differentially expressed genes listed in Tables 2 and 3 might contain putative RFX binding sites, we first identified the human orthologous genes for 69 of the 78 genes in the NCBI HomoloGene database (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=homologene). We assumed that the cis-regulatory elements such as RFX binding sites were more likely to be conserved than non-functional elements during evolution, and therefore that these functional elements might be present at similar positions, with high sequence identity in both mouse and human promoters. It has been shown that comparative sequence analysis is effective in identifying functional elements (Xie et al. 2005). We then extracted both mouse and human promoter sequences for these 69 genes from the UCSC genome browser (http://genome.ucsc.edu/), each of which contained sequences 1 kb upstream from the transcription start site. We used matInspector (Quandt et al. 1995) to scan the set of promoter sequences for putative RFX binding sites, using the RFX1 position weight matrix (PWM) in the TRANSFAC database. For each orthologous pair, we required that a binding site was present in both human and mouse sequences with a score cut-off of 0.9, and shared at least 80% identity between the two species. Five of the 69 genes were found to contain putative RFX binding sites. As the original RFX1 position weight matrix corresponds to two half sites separated by 2 bp, we then deleted the two central columns in the matrix one by one to scan for two half sites separated by 1 and 0 base. Four and two genes were found, respectively. We also inserted a column of one base to the middle of the original RFX matrix to scan for two half sites separated by three bases. Two additional genes were found. Together, there were 11 genes that contained two putative half sites separated by 0–3 bases (15.9%, supplementary Table S1). To test if the potential RFX4_v3 target genes were enriched in these binding sites, we applied the same methodology and criteria to a set of 11 858 pairs of human/mouse orthologous promoter sequences. A total of 749 unique genes were found to contain two half sites separated by 0–3 bases (6.3%, supplementary Tables S2 and S3). This resulted in an odds ratio of 2.81, indicating that the putative RFX binding sites were enriched in the promoters of the RFX4_v3 outlier genes. A Fisher's exact test showed that this enrichment was significant (p = 0.004).

For the 22 validated outlier genes, the potential RFX binding sites in both 1 kb human and mouse promoter sequences are listed in Table 5. Only three genes had highly conserved RFX binding sites in both promoters (X-box sequences shared at least 80% identity between two species), and Cx3cl1 was one of them.

Table 5.   Putative X-box sequences in both human and mouse promoters of validated genes

GenBank ID	Gene	Human promoter X-box sequences	Mouse promoter X-box sequences
N/A, promoter sequences are not available from the UCSC genome browser. ^aX-box sequences share at least 80% identity between two species. The two half sites GTNRCC and RGYAAC are underlined.
NM_009522	Wnt3a	(+) cggggctggaaacc-409	(−) cttgctcttggcaac-640
NM_009521	Wnt3	(−) gtgtccgaggaacg-565
		(+) ggcttctggaaaca-435
AK052950	Fzd10
NM_027280	Nkd1		(−) tgtcactcgcaact-715
NM_007811	Cyp26a1	(+) cgtgggcagcaacc-71	(−) tggggcttgcaacc-860
			(−) cgggactaggaaca-226
NM_010836	Msx3	(+) tggcccgggcaaca-902	(−) ggccactagcaact-636
		(−) aggttcgagcaaca-595
		(+) ctagcggtgggaac-544
NM_031166	Idb4	(+) tgcgcaaggcaact-875	(−) ggcgccgggcaacc-470
AK014242	Fez	N/A
NM_013833	Rax	(+) gggagcgaggaact-854	(−) atcgctcttagcaac-899
NM_010446	Foxa2		(+) gtaaccttgaaaca-123
AK034131	Rfx4_v3	N/A	N/A
NM_022435	Sp5	(+) gtcgccagcaggaac-798	(+) tgtcccgagcaacc-148^a
		(+) tgtctcgagcaacc-360^a
AK053799	Bh1
NM_009573	Zic1	(+) ggaaactggcaaca-915
		(+) ataacctgggaact-865
NM_009575	Zic3	(+) gggagcgcgaaact-571	(−) aatagcaggaaaca-850
		(−) gtggctgcggggaac-331^a	(−) gaggctgtggggaac-286^a
NM_010052	Dlk1	(+) gccacggggcaact-558
NM_008516	Lrrn1	(+) gggcgccggcaacg-412
		(−) gcgacacaggcaac-159
NM_133241	Mlc1		(−) tatagcaggcaact-965
			(+) ggacgctgggaact-418
			(+) gctgccatgggaac-15
NM_009142	Cx3cl1	(+) gttccttggcaaca-150^a	(+) tgtttcaggaaact-427
			(−) ggcgcgatgggaac-171
			(+) ggttccctggcaac-122^a
NM_138683	Rspondin	(−) agtgacaaggaaca-585	(−) agacacctgcaacg-593
		(+) gcaacttgggaacg-28
NM_011707	Vtn	(−) ttcgctaagaaaca-577	(+) gggaactaggaact-41
AK044492	EST	(−) agcttcagggaact-861	(+) gccccaatggaaac-288

RFX4_v3 can activate the mouse Cx3cl1 promoter through the X-box 1

From the list of validated outliers, we chose Cx3cl1 as an example of a potential direct target for RFX4_v3 trans-activation because of its known high-level expression in the brain (Tarozzo et al. 2003), and the presence of highly conserved RFX binding sites in its mouse and human proximal promoters (Table 5, Figs 2a and b). There were three potential RFX binding sites within 500 bp upstream of the transcription start site (TSS) in mouse Cx3cl1 (X-box 1 referred to the X-box 122 bp upstream of the TSS, X-box 2 referred to the X-box 171 bp upstream of the TSS, and X-box 3 was 427 bp upstream), one X-box in human (150 bp upstream), and mouse X-box 1 and the human X-box were located at similar positions and shared high sequence identity, suggesting that the Cx3cl1 gene might be directly regulated by the RFX4_v3 protein through these X-boxes.

To characterize the effect of the RFX4_v3 protein on the mouse Cx3cl1 promoter, approximately 1 kb of the 5′-upstream region (−937/+61: 937 bp upstream and 61 bp downstream relative to the TSS) was fused to a firefly luciferase reporter (Luc), and transiently transfected with either the control vector or the RFX4_v3 over-expressing vector. The luciferase activity of the −937/+61 construct was increased by approximately 2–2.5-fold by co-expression of the RFX4_v3 protein, indicating that this promoter possessed RFX4_v3-dependent transcriptional activity (Fig. 2c). To further define potential responsive elements, constructs containing various promoter regions were generated by unidirectional deletions of the −937/+61-Luc plasmid. Activities of the constructs −715/+61, −402/+61, −284/+61, and −226/+61 were regulated similarly to the −937/+61 construct by the RFX4_v3 protein, suggesting that these deleted upstream regions did not contain responsive elements for RFX4_v3 and X-box 3 was not essential for the RFX4_v3 stimulation. Further deletions of X-box 2 (the −172/+61-Luc construct contained only two nucleotides of X-box 2, and −135/+61-Luc had the whole X-box 2 sequence removed) did not abolish the stimulatory effects of RFX4_v3. These data indicated that X-box 2 was not critical for the RFX4_v3 induction either. However, when the X-box 1 was deleted (−123/+61-Luc contained 2 nucleotides of X-box 1, and −88/+61-Luc contained no X-box 1 sequence), the induction by RFX4_v3 was markedly diminished, suggesting that RFX4_v3 functioned through the X-box 1 of the mouse Cx3cl1 promoter (Fig. 2c).

To further confirm that X-box 1 was essential for the RFX4_v3 regulation, constructs containing mutations in X-box 1 (X-box 1 mutants 1 and 2) were generated in the context of −937/+61-Luc (Fig. 3a). Both mutations totally eliminated the RFX4_v3 activation of the promoters (Fig. 3b). These results indicated that, at least in the 1 kb 5′-upstream region (937 bp upstream and 61 bp downstream of the TSS), X-box 1 was the major if not the only element that was responsive to RFX4_v3.

Figure 3.  Effects of mutations in X-box 1 element of mouse Cx3cl1 promoter activities. (a) Sequences of mouse X-box 1 mutant 1 and X-box 1 mutant 2. The mutated nucleotides are underlined. (b) Mutations in X-box 1 abolished RFX4_v3 induction of the mouse Cx3cl1 promoter. Luciferase constructs containing mutations in X-box 1 element in the context of the −937 to +61-bp fragment as well as the wild-type −937/+61 construct were co-transfected with control or RFX4_v3 over-expressing vectors. Relative luciferase values were calculated and presented as described in Fig. 2(c). *Significantly different when comparing co-transfection with the control vector and the RFX4_v3 over-expressing vector of the same luciferase construct.

Evidence for binding of the RFX4_v3 protein to X-box 1 of the Cx3cl1 promoter in vitro

To test whether RFX4_v3 directly bound to X-box 1 in vitro, EMSA was performed with an X-box 1 probe (Fig. 4a). The RFX4_v3 protein tagged with a Myc epitope was synthesized in vitro; EMSA showed that this protein could bind directly to the X-box 1 probe (Fig. 4b, lane 3). This binding could be competed with excess amounts of unlabeled oligonucleotides. As shown in Fig. 4(b), an X-box oligonucleotide from the IL-5Rα enhancer (Morotomi-Yano et al. 2002) and the Cx3cl1 X-box 1 oligonucleotide served as efficient competitors (Fig. 4b, lanes 4 and 5), whereas oligonucleotides corresponding to X-box 1 mutant 1, X-box 1 mutant 2, and X-box 2 did not disrupt the complex between the X-box 1 probe and the RFX4_v3 protein (Fig. 4b, lanes 6–8). We also verified the presence of RFX4_v3 protein in the DNA–protein complex, as the band was super-shifted by adding the anti-Myc antibody to the reaction (Fig. 4b, lane 9). These results demonstrated that RFX4_v3 could bind directly to the X-box 1 in the Cx3cl1 promoter in vitro, and that mutations in X-box 1 abolished RFX4_v3 binding. Although X-box 2 and X-box 3 contained potential RFX binding sites, they did not respond to RFX4_v3 in terms of induction of the gene. Mouse X-box 1 and the human X-box were highly conserved in terms of location and sequence similarity, and mouse X-box 1 was the only functional element out of the three mouse X-box sequences.

Figure 4.  Binding of the RFX4_v3 protein to the mouse Cx3cl1 X-box 1 sequence in vitro. (a) Sequences of oligonucleotides used in EMSA. The wild-type or mutant X-box sequences are in bold type. (b) The RFX4_v3 protein tagged with the Myc epitope was synthesized by the in vitro transcription/translation method. The double-stranded oligonucleotide for Cx3cl1 X-box 1 (1WT) was end-labeled with ³²P. EMSA was performed with various combinations of wild-type X-box 1 probe and non-radioactive competitors as indicated. Lane 1, ³²P-labeled probe only; lane 2, the vitro transcription/translation reagent without DNA template was incubated with the probe. The asterisk on the right indicates non-specific signals. Lane 3, in vitro synthesized RFX4_v3 protein was incubated with the probe. Lanes 4–8 are competition analyses using a 50-fold molar excess of unlabeled wild-type and mutant oligonucleotides, respectively. Lane 9, supershift assay with addition of the anti-Myc antibody to the reaction. The RFX4_v3/1WT complexes are shown by an arrow. S, RFX4_v3/1WT complexes super-shifted by the addition of anti-Myc antibody; F, unbound free probes.

Direct interaction of the RFX4_v3 protein with the mouse Cx3cl1 promoter in vivo

To further confirm that the Cx3cl1 gene was a direct target of the RFX4_v3 protein, and that RFX4_v3 could bind to the Cx3cl1 promoter in vivo, ChIP experiments were performed. As there was no available antibody against RFX4_v3 for the ChIP assay, mouse cells were transfected with either the Myc-tagged RFX4_v3 expression vector or the control vector. Exacts were prepared and RFX4_v3 was immunoprecipitated with an anti-Myc antibody, and its association with in vivo bound chromatin was examined. The portion of the mouse Cx3cl1 promoter region containing X-box 1 was amplified with two pairs of primers, F1 and R, and F2 and R; similar results were obtained with both pairs of primers (Fig. 5). When cells were transfected with the control vector, comparable amounts of Cx3cl1 promoter fragments were precipitated with control or anti-Myc antibody, whereas upon over-expression of the RFX4_v3 protein, precipitation of Cx3cl1 promoter fragments by the anti-Myc antibody was enriched about 4-fold compared with the negative control antibody, indicating that RFX4_v3 could bind to the mouse Cx3cl1 promoter in vivo. These data add further support to the demonstration that Cx3cl1 gene was a direct target of the RFX4_v3 protein.

Figure 5.  In vivo binding of the RFX4_v3 protein to the mouse Cx3cl1 promoter. C2C12 cells were nucleofected with either the empty control vector (pcDNA3.1/MycHis) or the RFX4_v3 expression vector (RFX4_v3/pcDNA3.1/MycHis). Twenty hours after transfection, sheared chromatin was prepared and precipitated with either the negative control IgG or the anti-Myc antibody. The co-immunoprecipitated DNA was amplified by real-time PCR with primers F1 and R (a) or primers F2 and R (b). Each precipitated DNA amount was normalized to the respective ‘input DNA’ amount, and relative DNA amount was calculated as a ratio of the DNA isolated from the cells transfected with control vector and precipitated with control IgG.

Discussion

A novel isoform of RFX4, RFX4_v3, was found to be critical for brain development (Blackshear et al. 2003). Whereas disruption of both Rfx4_v3 alleles severely altered brain morphogenesis with failure of dorsal midline structure formation, disruption of a single allele led to agenesis of the subcommissural organ and congenital hydrocephalus (Blackshear et al. 2003). Therefore, the RFX4_v3 protein may regulate dorsal patterning and other aspects of brain development, presumably by regulating the expression of downstream genes.

In the present study, we compared levels of transcripts in Rfx4_v3–/– and +/+ embryonic heads at E10.5, using microarray analysis followed by real-time PCR validation. We identified and confirmed 22 transcripts that were differentially expressed in the Rfx4_v3 mutant brains; these could represent either direct or indirect targets of RFX4_v3. Members of the RFX family of transcription factors are thought to bind to X-box consensus sequences in the promoters of targets and thereby regulate gene expression. One such X-box-containing promoter belonged to the gene encoding the chemokine CX3C ligand 1 (CX3CL1); this was selected for further study because both mouse and human Cx3cl1 proximal promoters contained conserved X-box consensus sequences. In the Rfx4_v3 mutant brain, Cx3cl1 message levels were decreased to about half of control levels, indicating that Cx3cl1 might be a downstream target of RFX4_v3 in the CNS. Promoter studies further demonstrated that Cx3cl1 could be directly regulated by RFX4_v3 through the X-box 1 that lies about 120 bp upstream of the transcription initiation site. These studies demonstrated that RFX4_v3 could act as an activating transcription factor, promoting Cx3cl1 expression after directly binding to its X-box 1. The X-box 1 of the mouse Cx3cl1 promoter was conserved across species: it shared high sequence similarity with other X-box elements in the proximal promoters of the human, rat, dog, and armadillo Cx3cl1 genes, suggesting that Cx3cl1 might be a conserved target gene of RFX4_v3 among other organisms (supplementary Fig. S1).

RFX4 proteins contain DNA binding and dimerization domains that are typical for most other RFX proteins, but lack the PQ, Q, and A regions that are believed to have trans-activating activities (Katan et al. 1997; Iwama et al. 1999). By using a chimeric protein of RFX1 and RFX4, the carboxyl-terminal domain of RFX4 was shown to be a possible transcriptional repression domain in in vitro reporter assays (Morotomi-Yano et al. 2002). RFX4 was therefore proposed to be the first mammalian member of the RFX family without transcriptional activation capacity; these authors suggested that RFX4 might need to interact with other RFX members to regulate gene expression. Our studies clearly show that RFX4_v3 can function as a transcriptional activator under certain circumstances, despite its apparent lack of a distinct activation domain.

CX3CL1 (also called fractalkine, neurotactin, and small inducible cytokine subfamily D1) is the only known member of the CX3C chemokine subfamily and is characterized by a unique Cys-X-X-X-Cys motif (Miller and Meucci 1999; Shiraishi et al. 2000). CX3CL1 is present in two distinct forms: as a membrane-bound protein that is thought to be involved in cell adhesion, or as a soluble, secreted molecule, generated by proteolytic cleavage, that can promote chemotaxis (Chapman et al. 2000). CX3CL1 acts by binding to a 7-transmembrane domain G-protein-coupled receptor, CX3CR1.

Unlike other chemokines, CX3CL1 is expressed predominantly in brain (Pan et al. 1997; Cotter et al. 2002), with the highest levels of its transcript detected in the hippocampus, cortex, and striatum (Tarozzo et al. 2003). Although it has been suggested that Cx3cl1 is expressed in microglia and astrocytes (Pan et al. 1997; Hatori et al. 2002), most studies indicate that, in the brain, neurons are the major cell type expressing Cx3cl1 (Maciejewski-Lenoir et al. 1999). The functions of CX3CL1 in CNS are still not well understood. In neuron cultures, CX3CL1 has been shown to modulate several cellular functions, including intracellular calcium concentrations and synaptic transmission (Meucci et al. 1998). Under homeostatic conditions, CX3CL1 promotes survival of neuronal progenitor cells and inhibits Fas-mediated cell death of brain microglia in vitro (Boehme et al. 2000; Krathwohl and Kaiser 2004). CX3CL1 also protects neurons from the neurotoxicity induced by human immunodeficiency virus (HIV) infection (Meucci et al. 2000). CX3CL1 is also thought to be involved in neuroinflammatory processes. CX3CL1 and its receptor CX3CR1 are dramatically up-regulated in neurons from brain tissue of patients with HIV-1 encephalitis (Tong et al. 2000). A marked change in the distribution of CX3CL1 at both the protein and the gene level also occurs after ischemia reperfusion injury in rat (Tarozzo et al. 2002), and Cx3cl1 knockout mice have reduced infarct volume and mortality following cerebral ischemia (Soriano et al. 2002).

Cx3cl1-deficient mice exhibited no obvious central nervous system phenotype (Cook et al. 2001) except susceptibility to cerebral ischemia–reperfusion injury (Soriano et al. 2002), indicating that Cx3cl1 may not play an important role during brain development, and that decreased expression of Cx3cl1 may not be a major cause of the neuronal phenotype seen in the Rfx4_v3 mutant mice. Given the roles of CX3CL1 in brain injury and inflammation, it is possible that RFX4_v3 may also be involved in those pathological processes, in addition to its known roles in midline brain structure formation.

Other transcripts whose levels were altered in the mutants are well-known regulators of brain development, such as components of the Wnt, RA, and BMP pathways (Altmann and Brivanlou 2001; Melton et al. 2004). Brain patterning and morphogenesis are complex processes, and require the precise, finely balanced control of cell specification and proliferation. This control is achieved through the complex interplay of multiple signaling systems, including BMP, Wnts, RA, fibroblast growth factors, and sonic hedgehog. These proteins operate as morphogens, directly acting to pattern adjacent tissues, or establishing local signaling centers that modulate the activities of other morphogens. We found that some transcripts in the Wnt, RA, and BMP pathways were differentially expressed when comparing the Rfx4_v3 +/+ and –/– heads, suggesting that the RFX4_v3 protein may directly or indirectly regulate these important signaling pathways. Rfx4_v3 deficiency also affected the expression of several important transcription factors, RAX, FOXA2, ZIC1 and ZIC3. These factors are also known to regulate many aspects of brain development, such as eye formation, specification of the forebrain and the anterior definitive endoderm, and cerebellar development. Therefore, RFX4_v3 may either directly or indirectly regulate downstream gene expression by modulating the expression levels of these transcription factors. Sp5, Bh1, Dlk1, and several other transcripts were also differentially expressed when comparing the Rfx4_v3 +/+ and –/– brains. Much future work will be required to understand the nature of the responses of these genes to Rfx4_v3 deficiency. Ultimately, these studies may lead to a better understanding of the molecular mechanisms underlying the function of the RFX4_v3 protein, as well as the general process of brain morphogenesis.

Acknowledgements

The authors wish to thank Drs Anton Jetten and Yuji Mishina for helpful suggestions during preparation of this manuscript. This research was supported by the Intramural Research Program of the NIH, National Institute of Environmental Health Sciences, Intra-disciplinary Award (Z01 ES101583) and an NIH Bench-to-Bedside Award.