© 2000 Oxford University Press Human Molecular Genetics, 2000, Vol. 9, No.

4 477–488

ARTICLE

Functional analysis of ARHGAP6, a novel GTPase-

activating protein for RhoA
Siddharth K. Prakash1, Richard Paylor1, Sarah Jenna5, Nathalie Lamarche-Vane5,
Dawna L. Armstrong3, Bisong Xu1, Michael A. Mancini2 and Huda Y. Zoghbi1,4,+

Departments of 1Molecular and Human Genetics, 2Cell Biology and 3Pathology and 4Howard Hughes Medical
Institute, Baylor College of Medicine, 1 Baylor Plaza, Houston, TX 77030, USA and 5Department of Anatomy
and Cell Biology, McGill University, Montreal, Quebec, Canada

Received 4 October 1999; Revised and Accepted 22 December 1999

Microphthalmia with linear skin defects (MLS) is an X-linked dominant, male-lethal syndrome characterized by
microphthalmia, aplastic skin and agenesis of the corpus callosum, and is caused by the deletion of a 500 kb
critical region in Xp22.3. Our laboratory isolated a novel rho GTPase-activating protein (rhoGAP) gene named
ARHGAP6 from the MLS region. ARHGAP6 contains 14 exons encoding a 974 amino acid protein with three
putative SH3-binding domains. Because exons 2–14 are deleted in all MLS patients, we hypothesized that
ARHGAP6 may be responsible for some of the phenotypic features of MLS. We pursued two approaches to study
the function of ARHGAP6 and its role in the pathogenesis of MLS: gene targeting of the rhoGAP domain in mouse
embryonic stem cells and in vitro expression studies. Surprisingly, loss of the rhoGAP function of Arhgap6 does
not cause any detectable phenotypic or behavioral abnormalities in the mutant mice. Transfected mammalian cells
expressing ARHGAP6 lose their actin stress fibers, retract from the growth surface and extend thin, branching
processes resembling filopodia. The ARHGAP6 protein co-localizes with actin filaments through an N-terminal
domain and recruits F-actin into the growing processes. Mutation of a conserved arginine residue in the rhoGAP
domain prevents the loss of stress fibers but has little effect on process outgrowth. These results suggest that
ARHGAP6 has two independent functions: one as a GAP with specificity for RhoA and the other as a cytoskeletal
protein that promotes actin remodeling.

INTRODUCTION system exquisite diversity and specificity, because a single

The rho family of small GTP-binding proteins provides a critical
link between membrane receptors, the cytoskeleton and the
nucleus. Overexpression of rho proteins in cultured cells creates
characteristic morphological alterations, including the extension
of filopodia, formation of stress fibers or the induction of
membrane ruffles (1). Conversely, rho inactivation by dominant-
negative mutations or enzymatic inhibitors rapidly causes
microfilamentous structures to collapse (2). Strict spatial and
temporal regulation of rho GTPases is critical for developmental
or physiological processes that require cytoskeletal plasticity.
GTPase-activating proteins (GAPs) are crucial components of
the regulatory machinery for rho GTPases. Rho-specific GAPs
(rhoGAPs) bind directly to activated GTPases, hydrolyze GTP to
GDP and switch off signal transduction. However, rhoGAPs are
not only inhibitory, but can also transduce separate signals or
function as accessory effectors of rho GTPases. The 140 amino
acid GAP domain is often only one segment of a large and
multifunctional rhoGAP (3). In mammals, the possible
combination of at least 14 rho GTPases and 12 rhoGAPs lends the
GTPase can interact differentially with several rhoGAPs (4).
RhoGAP domains isolated from N-chimerin, Graf and p190
elicit cytoskeletal changes in cultured cells, and the functions of
the full-length proteins apparently are eliminated by mutations
that abolish GAP activity (5–7). These observations suggest that
the primary cellular functions of many rhoGAPs are
accomplished through the inactivation of rho GTPases and that
defects caused by loss-of-function mutations in rhoGAPs and
constitutive activation of rho GTPases may be similar (8).
Other observations suggest that GAP proteins may have
additional signal transduction functions that are required for
normal development. Point mutations or small deletions in the
GAP domain of the NF1 gene cause neurofibromatosis. However,
patients with larger deletions that remove the entire gene display
learning impairments even in the absence of neuronal tumors (9).
Analysis of Nf1-deficient mice revealed a previously unknown
role for Nf1 in the morphogenesis of neural crest-derived tissues
(10). In vitro studies have implicated the N-terminus of NF1 in
signaling functions that do not require GAP activity (11,12). To

+To whom correspondence should be addressed. Tel: +1 713 798 6558; Fax: +1 713 798 8728; Email: hzoghbi@bcm.tmc.edu
478 Human Molecular Genetics, 2000, Vol. 9, No. 4
date, no other GAP gene has been inactivated in mice, and the
biological roles of most rhoGAP family members are not known.
ARHGAP6, which encodes a novel 974 amino acid rhoGAP, is
deleted in patients with microphthalmia with linear skin defects
(MLS) syndrome, a dominant, male-lethal disorder characterized
by eye, skin and central nervous system (CNS) malformations
(13). Direct mutational analysis of MLS patients is not possible
because ARHGAP6 is always contained within the Xp22 terminal
deletions that cause MLS syndrome. As an alternative approach to
determine whether loss of ARHGAP6 function contributes to
MLS, we generated mice with a targeted deletion of the rhoGAP
domain. Based on neurobehavioral and neuropathological
analysis, Arhgap6 mutant mice do not differ significantly from
their wild-type littermates and do not show features of MLS
syndrome. These results suggest that loss of the rhoGAP function
of ARHGAP6 does not contribute to the MLS phenotype.
Transfection studies show that ARHGAP6 inactivates rhoA
signaling and produces additional effects on the actin cytoskeleton
that are independent of its GAP activity.
RESULTS
Isolation of the ARHGAP6-coding sequence and
identification of novel splice variants
The published sequence of ARHGAP6 was 2.0 kb shorter than the
principal mRNA species (4.2 kb) and lacked a suitable Kozak
initiation site (14). To isolate additional coding sequence, a 275 bp
RT–PCR product containing exons 1–3 was hybridized to a
human fetal kidney cDNA library. Five positive clones (HFK-1 to
-5) were isolated, sequenced and assembled into a 5156 bp cDNA
contig. The complete ARHGAP6 human fetal kidney cDNA
sequence contains an 875 bp 5′-untranslated region (5′-UTR), a
2921 bp open reading frame (ORF) and a 1360 bp 3′-UTR.
ARHGAP6 is expressed ubiquitously as a 4.2 kb transcript with
additional 2.6, 5.0 and 6.0 kb isoforms in skeletal muscle and
kidney, as previously described (14). Sequence analysis of fetal
kidney, cerebellar and lung cDNAs suggests that the coding
region of ARHGAP6 undergoes extensive alternative splicing.
The HFK contig contains a 1462 bp 5′ sequence (exon 1a)
including a novel 800 bp ORF in-frame with exon 2. This
sequence is not present in HFK-2 (exon 1b) or in a fetal brain
cDNA (exon 1c), which contain unique, untranslated 5′ exons.
RT–PCR amplification of total RNA from lymphoblasts as well
as five adult human tissues shows that at least three alternative 5′
ends of ARHGAP6 are spliced to the invariant central nine exons
of the gene. The longest isoform is most abundant in all tissues
examined.
Alternative splicing at a cryptic splice donor site within exon 12
bypasses a stop codon and extends the ORF by 248 novel amino
acids. dbEST searches identified five expressed sequence tags
(ESTs) from fetal heart (accession no. 111376), pregnant uterus
(768489) and fetal liver/spleen libraries (997642, 292405 and
292407), which are identical to the 3′-UTR of the HFK sequence.
Expression of the alternative 3′ ends was compared by RT–PCR
using specific primers for each isoform. The short isoform with
the stop codon in exon 12 is widely expressed, but the longer
isoform containing exon 14 is not detectable in fetal brain (data
not shown). These results suggest that alternative splicing of the 3′
coding sequence is developmentally regulated and may play a role
in the function of the gene in the CNS.
The complete coding sequence of mouse Arhgap6 revealed that,
similarly to the human gene, exon 11 is alternatively spliced out
(14). The mouse Arhgap6 gene is also expressed as a prominent 4.2
kb transcript with fainter 2.4 and 6.0 kb isoforms at all embryonic
stages (data not shown). The 4.2 kb transcript appears to be the
solitary isoform in embryonic stem (ES) cells as well as in adult
brain, heart and lung, and a faint 6.0 kb transcript is visible in adult
kidney. The results suggest that the alternative splicing and
expression pattern of Arhgap6 are conserved in the mouse.
The longest isoform of ARHGAP6 encodes a 974 amino acid
protein with a 150 amino acid rhoGAP domain and three proline-
rich motifs with putative SH3-binding sites (15) (Fig. 1A). If exon
1a is alternatively spliced out, a 771 amino acid protein could be
translated from an initiation site in exon 2 (Fig. 1B). Alternative
splicing of exons 11 and 12 truncates the C-terminus of the protein
by 338 or 249 amino acids, respectively (Fig. 1B). Without the
stop codon-containing exons, the predicted protein sequence of
the mouse Arhgap6 gene is 84% identical and 88% similar to the
human sequence. The C-terminal coding sequence is not highly
conserved, with the exception of a polyproline motif and a region
with homology to the WASP protein family (amino acids 878–
974, in bold in Fig. 1A) that share >90% identity.
Characterization of the genomic structure
The exon–intron boundaries were sequenced using cosmid DNA
as genomic template or determined by computer alignment of the
cDNA contig with genomic sequence of three bacterial artificial
chromosome (BAC) clones, 602M16 (132 kb), 512P14 (125 kb)
and 590J6 (110 kb), which contain the 3′ and 5′ ends of the
ARHGAP6 gene, respectively. ARHGAP6 spans >500 kb of
genomic DNA and contains 14 exons (Fig. 1C). The intron–exon
boundaries and the sizes of the exons and introns in the newly
identified coding sequence were determined by sequence
alignment and deposited in GenBank (14). Exon 1a is entirely co-
linear with a 1.4 kb genomic sequence in BAC 590J6. A potential
promoter with a TATA box 860 bp from the 5′ end of the
transcript and a 423 bp CpG island in the intervening sequence,
marking the probable transcription initiation site, was identified
by sequence analysis. Exons 12 and 14 contain multiple stop
codons and polyadenylation signals. Each splice junction
conforms to the vertebrate consensus.
Exons 1a and 2 of ARHGAP6 are separated by two contiguous
BAC clones (27C22 and 285I15) and one gap of undetermined
size in the physical map (50–100 kb) totaling >280 kb of genomic
sequence (Fig. 1C). By Southern analysis, the ARHGAP6 gene
spans the proximal boundary of the MLS critical region, with
patient breakpoint BA325 located between exons 1a and 2. The 10
kb amelogenin gene (AMELX) is contained within the first intron
of ARHGAP6 ∼40 kb centromeric of exon 2 and is transcribed in
the opposite orientation (16). Exon 1b is separated from exon 2 by
a 9 kb intron with consensus splice sites, and exon 1c is separated
from the AMELX promoter by only 2930 bp of genomic sequence.
Generation of Arhgap6 null mice
To evaluate the in vivo function of Arhgap6 and its role in the
MLS phenotype, a portion of the Arhgap6 coding region was
deleted using homologous recombination in ES cells (Fig. 2A).
The targeting vector was designed to delete a 5.5 kb genomic
fragment containing exons 6–8, which encode the N-terminal 118
amino acids of the 140 amino acid rhoGAP domain (Fig. 2B).
 Human Molecular Genetics, 2000, Vol. 9, No. 4 479

Figure 1. Amino acid sequence and alternative splice variants of ARHGAP6. (A) Amino acid sequence of ARHGAP6 derived from conceptual translation of the longest
isoform. Exon boundaries are indicated by dividers (T); the triangles indicate the boundaries at which exons 1a (1), 11 (2) and 12 (3) are alternatively spliced out. The
rhoGAP domain (amino acids 410–562) is boxed and the novel sequences derived from the human fetal kidney cDNA clones are underlined. Three putative SH3-binding
domains (amino acids 85–98, 338–350 and 878–889) are shaded. (B) Alternatively spliced N- and C-termini of ARHGAP6. Sequence 1 is encoded by exon 1b, which is
spliced directly to exon 2 in one HFK clone. Exon 1b lacks a suitable initiation codon and, if exon 1a is not present, translation can initiate from M204 in exon 2 [in bold
in (A)]. Sequence 2 is encoded by exon 11, which contains a stop codon that truncates the full-length protein by 338 amino acids. Sequence 3 is encoded by the distal portion
of exon 12, which is translated if a cryptic splice site at bp 3051 is not used. The stop codon truncates the full-length protein by 249 amino acids. (C) The position of
ARHGAP6 relative to breakpoint BA325 (zigzag line), which defines the proximal boundary of the MLS critical region. Black squares represent STS markers used to array
a contig of PAC and BAC clones (gray rectangles). The position of the 14 exons of ARHGAP6 (thin vertical bars) in the genomic sequence and the transcriptional
orientation of the genes (arrows) are also indicated. Exon 1c, which was identified by sequence analysis of PAC 27C22, is marked with an asterisk. AMELX is the
amelogenin gene, HCCS is the human holocytochrome c-type synthetase gene and MID1 is the midline-1 gene. The GenBank accession numbers for the human isoforms
are NM001174, AF177664, AF012272, AF177663 and AF177665; the mouse sequence is AF012273.

High-percentage male chimeras (>60% agouti coat color) were To confirm that the homologous recombination event disrupts
obtained for three ES clones, which were transmitted onto a 129 Arhgap6 expression, northern analysis of total RNA from the
Sv/Ev background. All genotypes were observed in the expected brain, heart, lungs and kidneys of wild-type male mice and
Mendelian ratios (Fig. 2C). putative mutant male mice was performed (Fig. 3A). A 4.2 kb
480 Human Molecular Genetics, 2000, Vol. 9, No. 4
Figure 2. Targeted deletion of Arhgap6 by homologous recombination.
(A) Restriction map of the 30 kb Arhgap6 genomic locus between exons 5 and 9.
The rhoGAP domain is encoded by exons 6–9. The targeting vector contains exons
1 and 2 of an hprt minigene, a single loxP site (triangle) and a neomycin-selectable
marker flanked by 5.5 and 6.0 kb arms. (B) Integration of the vector into the
Arhgap6 locus results in the deletion of a 5.5 kb fragment containing exons 6–8.
Splicing around the inserted cassette is expected to create a frameshift between the
remaining exons 5 and 9 (indicated by *). Correct targeting was diagnosed by
Southern analysis of BamHI- or ApaI-digested genomic DNA with the appropriate
diagnostic probes (bars). (C) Southern analysis of targeted clones and germline
transmission of the mutation. Targeted XY ES cells (–) show a single 12 kb BamHI
fragment (left) or a 10.5 kb ApaI fragment (right). The wild-type (+) sizes are 14
and 22 kb, respectively. (D) (Left) Outcome of matings between chimeric males
and wild-type females to produce heterozygous female offspring (+/–) and wild-
type male offspring (+) in the F1 generation. These siblings were mated to produce
the F2 generation (right) including mutant male mice (–).
transcript was detected in each tissue of wild-type mice, but not in
Arhgap6 mutant mice. Mice with the targeted allele express a
3.8 kb transcript whose intensity is ∼50% of wild-type levels. This
may reflect mRNA instability or reduced transcription caused by
the targeting event. The 400 bp reduction is consistent with the
deletion of exons 6–8 in the rhoGAP domain. Hybridization of the
same blot with a probe from within the rhoGAP domain detects
the wild-type transcript but does not detect the mutant transcript
(data not shown). The mutation was also confirmed by RT–PCR
amplification of Arhgap6 from the same RNA samples with
primers flanking the deleted exons. A 728 bp product is amplified
from brain RNA of a wild-type male, but a 372 bp product is
visible in similar preparations from a heterozygous female or a
mutant male (Fig. 3B). Sequence analysis confirmed that the
smaller product lacks exons 6–8 as predicted. The deletion shifts
the reading frame and creates a stop codon 38 amino acids after
the splice junction between exons 5 and 9 (Fig. 3C). These results
demonstrate that the targeted allele of Arhgap6 encodes a
truncated protein that lacks the rhoGAP domain and the rest of the
coding region (Fig. 3D).
Arhgap6 mutant mice do not display histological or
behavioral abnormalities
Arghap6 mutant male and female mice were indistinguishable
from their wild-type littermates at birth and thereafter. There were
no detectable differences between the external appearance, cage
behavior and body weight of mutant (n = 15) and wild-type
(n = 13) animals. Two 3-month-old mutant males and two wild-
Figure 3. Northern and RT–PCR analyses of Arhgap6 mutant mice. (A) Northern
analysis of 20 µg of total RNA from a mutant male mouse (–) and a wild-type
littermate (+) using the Arhgap6 cDNA. In wild-type RNA, a 4.2 kb transcript is
detected. In the mutant, a 3.8 kb band due to the deletion of rhoGAP domain exons
6–8 is visible in all lanes at reduced levels compared with the endogenous
transcript. The 1.2 kb endogenous Hprt transcript is the loading control. (B) RT–
PCR with primers flanking the rhoGAP domain amplifies the expected 728 bp
product using RNA from the wild-type animal. The same primers produce a 372 bp
band in RNA from the mutant animal. No products were amplified from control
reactions without reverse transcriptase (–). (C) Sequence analysis of the PCR
products demonstrates that the shortened transcript is created by deletion of three
exons in the rhoGAP domain. In the transcript derived from the mutant allele, exon
5 is spliced directly to exon 9, creating a frameshift mutation (hatched) which
abolishes the rhoGAP domain and results in a premature stop codon (*) 38 amino
acids downstream. (D) Exons (rectangles) and transcript (line) of the targeted
Arhgap6 allele, in which exons 6–8 (hatched) are deleted, resulting in a frameshift
(gray line) and premature stop codon. This transcript encodes a 470 amino acid
protein that retains two putative SH3-binding domains (circles).
type littermates were evaluated histopathologically. Analysis of
brain, eyes, heart, lungs, liver, kidney, testes, intestine, skeletal
muscle and skin revealed no obvious abnormalities. No
abnormalities of ossification or skeletal morphogenesis were seen
in mutant E17 embryos (n = 4) compared with wild-type
littermates (n = 4) (data not shown).
Because the corpus callosum and septum pellucidum are
abnormal in several MLS patients, we evaluated hematoxylin and
eosin-stained serial coronal sections at 4–5 levels through the
corpus callosum, hippocampal commissure and dorsal fornix of
adult wild-type (n = 16) and mutant (n = 16) animals. Two
animals of each genotype were stained using the Sevier–Munger
method (17). No significant differences in the thickness or
contralateral projection of the fibers were observed. Normal
development of the cerebellum, cortex and hippocampus was
revealed by staining with cresyl violet and immunohistochemistry
using antibodies against glial fibrillary acidic protein, calbindin
and the GABA receptor β2 subunit in at least four animals of each
genotype. No lesions or reactive gliosis were detected in the
mutant animals in comparison with equal numbers of wild-type
controls. These results suggest that the architecture of the CNS is
grossly intact in Arhgap6 mutant animals. We also evaluated renal
function in view of the abundant expression of Arhgap6 in the
kidney. Routine chemistry and urinalysis were normal and there
was no evidence of proteinuria.
Thirteen mutant and seven wild-type animals were evaluated
using a behavioral test battery to screen for defects in several
domains of CNS function including basic sensory/motor abilities,
learning and memory (18). To test further the role of Arhgap6 in
cognition, a separate set of male mutant (n = 9) and wild-type
 Human Molecular Genetics, 2000, Vol. 9, No. 4 481

Figure 4. Behavioral analysis of Arhgap6 mutant mice. (A) Morris water task test results for wild-type (n = 7) and Arhgap6 mutant (n = 13) animals of a mixed C57Bl/6J
and 129 Sv/Ev background at age 6 weeks. The number of times the animals crossed the location of the platform (left) and total search time spent in each quadrant of the
pool (right) were recorded. (B) Rotarod test results for wild-type and Arhgap6 mutant animals. The longest continuous interval in which the mice maintained balance on
the rod (maximum performance) and the time of the first rotation of the mouse around the rod (1st rotation) were recorded for each of four trials. No differences between
wild-type and null mice were noted (P > 0.05).

(n = 9) mice were evaluated in a conditional alternation task (19),
which can be used to assess conditional working memory
processes. Basic sensory/motor responses, locomotor activity,
anxiety-related responses, sensorimotor gating and analgesia-
related responses of the mutant animals were not significantly
different from those of their wild-type littermates (P > 0.5).
Arhgap6 mutant mice did not display any defects in motor
coordination or skill learning deficits on the rotarod test (Fig. 4A).
In addition, the performance of Arhgap6 mutant mice in the
conditioned fear test, Morris water task (Fig. 4B) and conditional
alternation task was similar to that of the wild-type controls,
indicating that the processes mediating several forms of learning
and memory performance are not disrupted by the Arhgap6
mutation.
ARHGAP6 is a GAP for RhoA but not for Rac1 or Cdc42
To assess the GAP activity of ARHGAP6, a fragment of the
mouse Arhgap6 cDNA encoding the SH3-binding and rhoGAP
domains (residues 279–660 of the full-length protein) was
expressed as a glutathione (GST)–fusion protein in Escherichia
coli and purified through a GST–Sepharose column. The
recombinant protein was incubated with [γ-32P]GTP-loaded
RhoA, Rac1 and Cdc42 and assayed for its ability to stimulate the
hydrolysis of GTP. As shown in Figure 5A, ARHGAP6 activates
the intrinsic hydrolytic activity of RhoA but not of Rac1 or Cdc42.
In a parallel analysis, the rhoGAP domain of p50rhoGAP
displayed catalytic activity toward all three GTP-binding proteins,
as was demonstrated previously (20). The results suggest that
RhoA is the preferred in vitro substrate for ARHGAP6.
To determine whether ARHGAP6 shows in vivo specificity for
any member of the rho family, the full-length cDNA was
transfected as an N-terminal Xpress-tagged or C-terminal green
fluorescent protein (GFP)–fusion protein into COS-7 and HeLa
cells. Cells were scored for green fluorescence and double-labeled
with anti-actin antibodies (red) to visualize the cytoskeleton.
Agonists for RhoA [sphingosine-1-phosphate (SPP)] or Cdc42
(bradykinin) were added to the culture media and assayed for their
effects on the morphology of transfected and untransfected cells.
Serum-starved cells contain fewer actin stress fibers, correlating
with the reduced expression of RhoA. Exposure to SPP, which
rapidly restores the stress fibers, is a useful index for RhoA
activation. Whereas 44/51 untransfected cells formed intense
stress fibers after a 30 min exposure to SPP, all 50 ARHGAP6-
expressing cells that were examined did not form stress fibers
under the same conditions [χ2 = 61.3, P < 0.01 (Fig. 5D and E)].
Prevention of stress fiber formation after SPP treatment by
overexpression of ARHGAP6 demonstrates that the protein
hydrolyzes and inactivates RhoA. In a parallel experiment,
ARHGAP6 did not inhibit filopodial extension by bradykinin
(data not shown), supporting its specificity for RhoA in vivo and
confirming the results of the GAP assay.
Nine amino acids within the rhoGAP domain are conserved in
all rhoGAP family members, including an invariant arginine that
participates directly in the hydrolysis reaction and stabilizes the
GTP/GDP transition state. Mutations that replace the arginine
with another amino acid caused a 230- to 540-fold reduction in the
in vitro rate of GTP hydrolysis by p50rhoGAP and abolished the
morphological effects of two other RhoA-specific GAPs, Graf
and p190, in cultured cells (5,6,21). The conserved arginine
(R433) in ARHGAP6 was mutated to a glycine with a single base
change (CCG→CGG) in exon 6. To determine whether the
mutation affects rhoGAP activity in vivo, the mutant protein was
assayed for its ability to antagonize SPP-mediated stress fiber
482 Human Molecular Genetics, 2000, Vol. 9, No. 4
Figure 5. GTPase-stimulating activity of recombinant ARHGAP6. Equal
concentrations (0.2 µM) of GST–p50rhoGAP (amino acids 198–439) or GST–
ARHGAP6 (amino acids 279–660) were added to a GAP assay using 1.3 µM [γ-
32P]GTP-loaded RhoA (A), Cdc42 (B) or Rac1 (C). The percentage of [γ-32P]GTP
remaining as determined by a nitrocellulose filtration assay is recorded on the
vertical axis and each horizontal increment represents 5 min of incubation time.
Yellow, no GAP; green, p50rhoGAP; red, ARHGAP6. Note that ARHGAP6
demonstrates specificity for rhoA whereas p50rhoGAP displays activity toward
RhoA, Cdc42 and Rac1. (D–G) ARHGAP6 prevents the induction of stress fibers
by sphingosine-1-phosphate (SPP). (D) Expression of ARHGAP6–GFP (arrows)
in two serum-starved HeLa cells that were treated with SPP for 30 min.
Untransfected cells are indicated by the white arrowheads. (E) The same field
stained for actin stress fibers (red) showing reduced intensity of actin staining in
transfected cells. (F) Expression of the R433G mutant ARHGAP6 protein
(arrows) in serum-starved HeLa cells. (G) The same field stained for actin stress
fibers showing an indistinguishable response to SPP in cells expressing the mutant
protein.
formation in serum-starved HeLa cells. In contrast to the wild-
type protein, the R433G mutant failed to inhibit stress fiber
formation. Stress fibers formed in 69% of SPP-treated, transfected
cells compared with 73% (n = 70) of untransfected cells in the
same experiment, indicating that the R433G mutation effectively
abolishes the rhoGAP activity of ARHGAP6 [χ2 = 0.35, P > 0.05
(Fig. 5F and G)].
ARHGAP6 regulates the actin cytoskeleton
Microinjection of rho GTPases into mammalian cells remodels
the actin cytoskeleton to create stress fibers (RhoA), membrane
ruffles (Rac1) or filopodia (Cdc42). When grown in serum-
containing medium, HeLa or COS-7 cells expressing ARHGAP6
undergo similar morphological changes characterized by cell
retraction, loss of adhesion to the coverslip and the extension of
branched and beaded cytoplasmic processes. The protein appears
to be concentrated in bead-like protuberances near the tips of the
processes (arrows), and the extent of the cell body retraction is
revealed by confocal microscopy (Fig. 6A). Double-labeling
experiments with anti-actin antibodies indicate that actin filaments
form the core of the processes induced by ARHGAP6 (Fig. 6B).
These results suggested that disruption of the actin cytoskeleton
by ARHGAP6 may perturb cellular morphology. Actin stress
fibers are linked to the extracellular matrix through focal
adhesions, which anchor cells to the growth surface. The loss of
adhesion seen in transfected cells suggested that focal adhesions
may also be disrupted by overexpression of ARHGAP6. Focal
adhesions were visualized in transfected HeLa cells by co-staining
with a monoclonal anti-focal adhesion kinase (FAK) antibody.
The punctate staining pattern at the periphery of untransfected
cells disappears in cells expressing ARHGAP6 (Fig. 6C).
Overexpression of ARHGAP6 produced similar morphological
effects in 90% of HeLa or COS-7 cells (n = 50) examined,
although the length of the processes varied slightly between cell
types.
Function of N- and C-terminal domains of ARHGAP6
To determine whether rhoGAP activity is sufficient to create the
morphological effects, several deletion mutants were expressed in
HeLa and COS-7 cells. ARHGAP6 contains a central putative
SH3-binding motif and rhoGAP domain flanked by extensive
proline-rich N- and C-terminal domains of unknown function.
Transfection of the isolated SH3-binding and rhoGAP domains
results in a uniform cytoplasmic distribution of the protein without
any effect on the actin cytoskeleton (Fig. 6D). An N-terminal 454
amino acid peptide truncated between the second SH3-binding
domain and the rhoGAP domain co-localizes with cortical actin
fibers but does not affect cellular morphology (Fig. 6E). A
C-terminal deletion mutant, which is truncated immediately after
the rhoGAP domain, behaves similarly to the full-length protein
(Fig. 6F). These results suggest that rhoGAP activity is necessary
but not sufficient for the function of ARHGAP6 and implicate a
region near the N-terminus in the association of the protein with
the cytoskeleton.
To determine whether rhoGAP activity is essential for process
formation by ARHGAP6, the catalytically inactive rhoGAP
mutant was expressed in HeLa cells that were grown in serum-
containing medium. The mutant protein produced distinctive
morphological effects. Whereas stress fibers completely disappear
from cells transfected with the wild-type protein and the cells
retract from the coverslip (Fig. 6G), transfection of the R433G
construct produces beaded processes (arrows) but does not cause
depolymerization of stress fibers or retraction of the cells
(Fig. 6H). Eighty-six percent of HeLa cells (n = 48) transfected
with the wild-type construct exhibited the retracted phenotype
compared with only 18% (n = 50) of HeLa cells transfected with
the R433G mutant (χ2 = 54.06, P < 0.05). These results suggest
that ARHGAP6 has two independent functions as a GAP with
specificity for RhoA and as a cytoskeletal protein that promotes
actin remodeling. The point mutation selectively abolishes the
ability of ARHGAP6 to promote retraction of transfected cells
without affecting process outgrowth.
ARHGAP6 induces the active extension of cytoplasmic
processes
The cytoplasmic processes in ARHGAP6-transfected cells may be
created by retraction of actin stress fibers from extended positions,
leaving threads of cytoplasm attached to the cell body, or active
growth driven by polymerization of actin filaments. To
 Human Molecular Genetics, 2000, Vol. 9, No. 4 483

Figure 6. Transient expression of ARHGAP6 in HeLa and COS-7 cells. (A) Confocal fluorescence microscopy of HeLa cells using an FITC-conjugated secondary
antibody. Cells expressing ARHGAP6 (green) extend long, branching processes with distinctive globular protuberances (arrows). Adjacent untransfected cells do not form
processes and maintain a compact cuboidal morphology. (B) COS-7 cells expressing ARHGAP6 were counterstained with a monoclonal anti-actin antibody (red) to
visualize the cytoskeleton. The cytoplasmic processes contain a core of filamentous actin. (C) HeLa cells double-labeled for ARHGAP6 and focal adhesion kinase (FAK).
Untransfected cells show punctate accumulations of FAK at the periphery (arrow); this is lost in cells expressing ARHGAP6 (green). (D–H) Diagrams depict expressed
ARHGAP6 protein fragments with putative SH3-binding motifs in yellow, the rhoGAP domain in red and the remaining coding sequence in purple. (D) HeLa cells
transfected with a 380 amino acid central fragment of ARHGAP6 consisting of the second SH3-binding domain and rhoGAP domain show no morphological changes.
(E) An N-terminal fragment of ARHGAP6 (amino acids 1–425) truncated before the rhoGAP domain co-localizes with actin filaments but does not affect cell morphology.
(F) Expression of the N-terminus and rhoGAP domain without the C-terminus reconstitutes the complete activity of the full-length protein to promote process outgrowth
and perinuclear cell retraction. (G) HeLa cells expressing wild-type ARHGAP6–GFP display a typical rounded morphology and elongated actin-rich extensions. (H) Cells
expressing ARHGAP6 with the R433G mutation (X) extend similar cytoplasmic processes (arrows). However, cell retraction was reduced significantly and the cells remain
adherent to the coverslip with visible stress fibers.

distinguish between these possibilities, HeLa cells were
monitored by video microscopy in a thermoregulated flow
chamber between 6 and 9 h after transfection with the wild-type or
R433G ARHGAP6 cDNA. The video data indicate that the
extensions are created by the continuous protrusion of filopodial-
and lamellipodial-like structures (Fig. 7A). The ARHGAP6–GFP
fusion protein is concentrated at sites of dynamic actin
polymerization near the tips of the outgrowing processes.
Extrusion of processes was accompanied by refractive waves that
propagated across the length of the cells and by retraction of the
opposite surface of the cells. Cells expressing the GAP-deficient
mutant protein did not retract from the coverslip and exhibited a
distinctive morphology characterized by peripheral microspikes
and increased numbers of processes with greater tortuosity
(Fig. 7B). The rate of cell movement was similar in both cases.
Cells transfected with an empty GFP vector maintained a cuboidal
morphology and did not extend processes (data not shown). These
observations confirm that ARHGAP6 affects cell morphology
and motility by promoting the continuous elongation of processes
and simultaneous retraction of the cell body.
DISCUSSION
MLS syndrome is a dominantly inherited, male-lethal disorder
characterized by developmental defects of the brain, eyes and skin
(13). Most MLS patients are heterozygous for terminal deletions
of Xp22.3 that include ARHGAP6, a novel rhoGAP gene (22).
The longest isoform of ARHGAP6 encodes a 974 amino acid
protein with three potential SH3-binding motifs, the 140 amino
acid rhoGAP domain and a 371 amino acid C-terminus of
unknown function. ARHGAP6 is a good candidate to cause MLS
because it occupies one-third of the critical region, and mutations
in another rhoGAP gene (oligophrenin) were discovered recently
in patients with non-syndromic mental retardation (MRX) (22).
484 Human Molecular Genetics, 2000, Vol. 9, No. 4
Figure 7. Time-lapse video microscopy of HeLa cells expressing wild-type and R433G ARHGAP6–GFP fusion proteins. (A) Subconfluent HeLa cells expressing a wild-
type ARHGAP6–GFP fusion protein or (B) HeLa cells expressing ARHGAP6–GFP fusion protein with the R433G mutation were photographed every 3 min in a
thermoregulated chamber with circulating medium. Twenty optical sections were compiled to create each image. Cells were imaged at 3 min intervals 6–8 h after
transfection, and the time of the experiment in minutes is printed below each panel. The arrows point to dynamic filopodial projections. These cells are representative of
cells observed in five separate experiments.
To characterize ARHGAP6 as a candidate gene for MLS
syndrome, two approaches were taken: generation of mice with a
targeted mutation of Arhgap6 and functional analysis of the
recombinant protein in cultured cells. Our results suggest that the
rhoGAP activity of Arhgap6 is not involved in the pathogenesis of
MLS syndrome, but provide evidence that the N-terminus of the
protein plays a novel role in its regulation of cell morphology and
actin polymerization.
The coding sequence of ARHGAP6 is interrupted by a large
intron that spans at least 280 kb of genomic DNA between exons
1a and 2 and contains the amelogenin gene (AMELX). Other
vertebrate genes with variations of this unusual structure have
been described, including the NF1 gene and the human factor VIII
gene (23,24). Intriguingly, both of these genes are subject to
extensive alternative splicing, harbor multiple transcripts within a
single large intron and are targets of frequent chromosomal
rearrangements in the human disorders type 1 neurofibromatosis
and hemophilia. Several hypotheses for the significance of the
arrangement have been proposed, including antisense regulation
of the larger transcript and co-expression of genes with similar
functions (25).
With the apparent absence of abnormalities in the mutant
animals, Arhgap6 may be added to a growing list of putative
neurodevelopmental genes that unexpectedly failed to create
detectable defects when mutated in mice, including Hprt and
Ziprol, which encodes a zinc finger transcription factor involved
in cerebellar granule cell proliferation (26,27). Several loss-of-
function mutations in Drosophila, Caenorhabditis elegans and
yeast create no apparent defects; this is even more likely in
vertebrates given the abundance of homologous gene families or
potential compensatory pathways (28). The diversity of vertebrate
rhoGAPs suggests that functional redundancy may also modify
the phenotype of Arhgap6 mutants.
The targeted Arhgap6 allele putatively encodes a truncated
protein of 470 amino acids that lacks the rhoGAP domain and the
alternatively spliced C-terminus. Exons 1–5, which remain intact
in the mutant animals, encode two potential SH3-binding
domains. N-terminal regions of other GAP proteins such as
p120rasGAP, p190 and α1-chimerin have been implicated in
membrane interactions and signaling functions that do not depend
on intact GAP activity (7,29,30). The N-terminus of p190
undergoes tyrosine phosphorylation and interacts with
p120rasGAP to regulate ras signal transduction and cell growth
(5). ARHGAP6 also contains a tyrosine phosphorylation site in
addition to several protein interaction motifs outside the rhoGAP
domain. These observations suggest that targeted deletion of the
rhoGAP domain may not cause a complete loss of Arhgap6
function.
Mutations in the rho pathway involving a RacGAP
(oligophrenin) and an effector for Rac (PAK3) recently were
implicated in human MRX (22,31). ARHGAP6 is expressed in
fetal brain, and at least one MRX locus that overlaps the genomic
position of ARHGAP6 was identified recently in Xp22.31 (32).
These observations suggest that mutation of ARHGAP6 may
cause MRX, and its deletion as part of a contiguous gene defect
could contribute to mental retardation in MLS syndrome.
However, a thorough neurobehavioral study of Arhgap6 mutant
mice failed to reveal any abnormalities. The results of other
studies, including behavioral analysis of mice deficient for the
fragile X syndrome protein Fmr1p, suggest that mutations causing
mild or moderate mental retardation in humans may be difficult to
detect in mouse models (33,34). Without more sensitive assays or
evidence for the in vivo function of ARHGAP6, our results do not
exclude ARHGAP6 as a candidate gene to cause MRX or some
CNS manifestations of MLS syndrome.
Functional analysis of ARHGAP6 in cultured cells identified
several unique characterisitcs of this novel rhoGAP. Transfected
cells lose actin stress fibers, retract from the coverslip and extend
branching, beaded cytoplasmic processes. In contrast to other
rhoGAPs such as p190 and Graf, inactivation of the rhoGAP
domain does not abolish the morphological effects of ARHGAP6
outgrowth (5,6). The mutation specifically disrupts the ability of
ARHGAP6 to antagonize rhoA-mediated stress fiber formation
but does not significantly affect process outgrowth. These results
suggest that ARHGAP6 contains at least two actin-organizing
functions, which may be partially or completely independent of its
rhoGAP activity. Similar morphological effects were observed in
cells expressing N(α-1)-chimerin, a 38 kDa brain protein with in
vitro racGAP activity which, paradoxically, stimulates rac and
Cdc42 in vivo (7,35). GTPase stimulation or inhibition by

chimerins depends on lipid binding by an N-terminal domain,
which interacts with the racGAP domain (35). The N-terminus of
ARHGAP6 appears to target the rhoGAP domain to cytoskeletal
compartments near activated rho GTPases and is critical for
ARHGAP6 function. The rhoGAP domain of ARHGAP6 is
highly homologous to the chimerin family, and one could
envisage a similar dual role for ARHGAP6 in GTPase signaling.
The function of ARHGAP6 as a regulator of cell morphology
and the cytoskeleton is consistent with a role for this novel
rhoGAP in the pathogenesis of several neurodevelopmental
disorders that are linked to Xp22.3. The results presented here
show that the rhoGAP activity of ARHGAP6 does not contribute
to the pathogenesis of MLS syndrome. Generation of mutant mice
that lack the transcription initiation site or exon 1a, or deletion of
the entire coding region may be the only approach to resolve the
role of this gene in the MLS phenotype. Further mutation analysis
of ARHGAP6 and experiments to clarify its in vivo functions are
in progress.
MATERIALS AND METHODS
Southern analysis and library screening
Probes were radiolabeled using random oligohexamer priming
with [α-32P]dCTP (36) and purified through a Sephadex G-50
column. The mouse 129 Sv/Ev genomic library (Allan Bradley,
Baylor College of Medicine, Houston, TX) and the human fetal
kidney cDNA library (Clontech, Palo Alto, CA) screens were
performed according to our published protocol (37). The λFIX II
genomic library was screened with a 1158 bp PCR product
derived from a human adult retina cDNA clone using primers 5′-
GGTCAGTCCCCATCCAGAGTCTC-3′ and 5′-ATAATTT-
TCAATCATCTTTTGC-3′. The λgt11 human fetal kidney
library was screened with a subcloned 275 bp PCR product from
reverse-transcribed human fetal brain RNA with primers 5′-
CCGATACACATCGTAACTTTGACC-3′ and 5′-GTCTTT-
GTTTTTCTCCTTTCC-3′. A mouse brain cDNA library (Strata-
gene, La Jolla, CA) was arrayed in 96-well plates and screened
using two rounds of PCR with primers 5′-CCGAATACTCCA-
GAACCAGCTC-3′ and 5′-GCTATGATGGCCTGTGCTCTC-
3′ derived from the published mouse Arhgap6 sequence.
Sequence analysis
Size-selected fragments from subcloned human fetal kidney and
mouse brain cDNAs were isolated using the Exo III method (38)
and were sequenced with an ABI PRISM BigDye Cycle
Sequencing kit and an ABI 377 DNA sequencer (Perkin-Elmer
Applied Biosystems, Foster City, CA). Assembly of the cDNA
contig and sequence analysis were performed using the LINEUP,
BESTFIT, PILEUP and MOTIFS programs from the GCG
package (39) and the Sequencing Project Manager in the
Lasergene Navigator Suite (DNASTAR, Madison, WI).
Northern analysis and RT–PCR
A 730 bp PCR-generated probe containing exons 4–10 of a mouse
brain Arhgap6 cDNA was radiolabeled with [α-32P]dCTP and
hybridized to a mouse developmental northern blot (Clontech) as
well as northern blots containing RNA prepared from an Arhgap6
mutant animal and a wild-type littermate as previously described
(14). Commercial northern blots containing mRNA from E7–E17
Human Molecular Genetics, 2000, Vol. 9, No. 4 485
stage mouse embryos (#7762-2; Clontech) and from adult mouse
tissues (#7762-1; Clontech) were hybridized under the same condi-
tions. For RT–PCR, 1 mg of total RNA was reverse transcribed
using the Superscript II first-strand cDNA synthesis kit (Life Tech-
nologies, Gaithersburg, MD) with random hexamer primers
(Amersham Pharmacia Biotech, Uppsala, Sweden). PCRs were
performed in an MJD DNA engine tetrad (MJ Research, Water-
town, MA) with an initial denaturation step of 94°C for 3 min, 33
cycles of 94°C for 1 min, 60°C for 1 min, 72°C for 1 min, and a final
5 min extension step at 72°C. The alternative 5′ ends were amplified
using forward primers in exon 1a, 5′-GAGGCGGGCGCGGAG-
GGCAGTG-3′; exon 1b, 5′-AAGATTCACAGTAGCG-
GAGGTC-3′; or exon 1c, 5′-CCGATACACATCGTAACTT-
TGACC-3′, with a reverse primer in exon 2, 5′-GAGACTCT-
GGATGGGGACTGACC-3′. The alternative 3′ ends were ampli-
fied using a forward primer in exon 12, 5′-AGGA-
AGCCCTGACATGCTG-3′, or with reverse primers in exon 12,
5′-CCAGGCTCCAGTTACCCTTTC-3′ or exon 14, 5′-TGGA-
CATTGCCATCTGGTTTGG-3′. The reactions contained 10 mM
Tris–HCl pH 8.3, 50 mM KCl, 1.5 mM MgCl2, 250 µM dNTPs,
0.625 U of Taq polymerase and 1 µM each primer.
Generation of mice with a targeted mutation in Arhgap6
With a 1158 bp probe from adult human retina cDNA, two
overlapping positive clones were isolated from a mouse 129 Sv/
Ev genomic library. A restriction map of the region indicated that
the clones contain 30 kb of the mouse Arhgap6 locus including
exons 5–9. The Arhgap6 targeting vector was constructed by
inserting 6.0 kb EcoRI–SpeI and 5.5 kb BanI–SacI fragments of
cloned strain 129 Sv/Ev genomic DNA into the PL13 vector (a
gift from Allan Bradley). A homologous recombination event
inserts exon 1 and 2 of a hypoxanthine–guanine phosphoribosyl-
transferase (hprt) minigene, a loxP site and a neomycin-resistance
cassette while deleting 5.5 kb of the Arhgap6 locus including three
exons encoding the rhoGAP domain (codons 417–555 of the
murine Arhgap6 cDNA sequence). The targeting vector was
linearized by digestion with KpnI (25 µg) and electroporated into
HPRT-deficient AB2.1 129 Sv/Ev ES cells using a GenePulser
(Bio-Rad, Hercules, CA) set at 300 V and 500 µF. Cells were
plated onto a G418-resistant embryonic fibroblast feeder layer.
After 24 h, selection was initiated for 10 days using 175 µg/ml
G418 and 0.2 µM FIAU. A total of 288 surviving colonies were
transferred in triplicate into 96-well plates. Cells from one plate of
each clone were trypsinized and stored in cryo-preservation
medium at –80°C. DNA from surviving ES clones was prepared
and analyzed according to our published protocols (40). Three ES
cell clones carrying the mutant allele were injected into the inner
cell mass of C57BL/6J blastocysts. Chimeric mice with a high
contribution to coat color from 129 Sv/Ev were bred for germline
transmission. For genotyping, genomic DNA isolated from ES
cells or mouse tails was digested with BamHI for hybridization
with the 5′ diagnostic probe or with ApaI for the 3′ diagnostic
probe.
Histological analysis
Tissues were fixed in 4% formaldehyde, dehydrated and
embedded in paraffin. Sections were cut at 10 µm thickness,
mounted, stained with hematoxylin and eosin or with silver using
the Sevier–Munger method (17), and examined at 20–63×
magnification using an Axiophot microscope (Carl Zeiss,
486 Human Molecular Genetics, 2000, Vol. 9, No. 4
Thornwood, NY) Skeletons were prepared from E17.5 embryos
and stained with alizarin red and alcian blue as described (41).
Immunohistochemical staining was performed on mouse brain
sections according to our published protocol (42).
Neurobehavioral tests
Animals from a C57Bl/6J–129 Sv/Ev hybrid genetic background
were analyzed at 3–4 months of age. Mice were evaluated using a
battery of behavioral tests in the order presented: neurological
screen for gross sensory/motor deficits, 30 min open-field test, 10
min light–dark box test, four trials of rotarod testing using an
accelerating rotarod, acoustic startle response (120 dB) and pre-
pulse inhibition of the startle response, habituation of the acoustic
startle response, contextual and auditory-cued conditioned fear,
hidden platorm version of the Morris water task, and hot plate
analgesia test. Tests were separated by 1 week intervals, except for
the conditioned fear and Morris tests which were separated by 2–
4 weeks. Details of the testing equipment and procedures were
described previously (18). The conditional alternation test was
performed in a water-filled T maze using the methods described
by Paylor et al. (19). One- and two-way analyses of variance tests
were used to analyze most of the behavioral results. Newman–
Keuls and simple effects tests were used for follow-up
comparisons. The data from the neurological screen were
analyzed using Fisher’s exact probability tests.
cDNA expression constructs
A human cDNA containing the complete coding sequence of
ARHGAP6 was constructed by ligation of two overlapping fetal
kidney clones (HFK-3 and -4). The N-terminal coding sequence
was amplified from this template using primers 7258 (5′-
AGGGATCCAGAGCCTGCTCCACAGCGTC-3′) and 7260
(5′-AGAGCTCGAGGCAAGGAAAGTTGTAAAGC-3′) and
subcloned as a BamHI–XhoI fragment into pcDNA3.1
(Invitrogen, Carlsbad, CA) for expression as an Xpress-tagged
fusion protein (pcDNA-Nterm). To express the full-length protein
(pcDNA-FL), two restriction fragments derived from the C-
terminal coding sequence of HFK-3 were ligated sequentially into
pcDNA-Nterm. A 1.2 kb EcoRV–XbaI fragment of pcDNA-FL
containing exons 13 and 14 was subcloned into a separate
pcDNA3.1 vector to create pcDNA-Cterm. A 1.1 kb EcoRI
restriction fragment from a mouse fetal brain cDNA clone
containing exons 4–10, including the SH3 and rhoGAP domains,
was also subcloned into pcDNA3.1 to create pcDNA-GAP. For
production of a bacterial GST–Arhgap6 fusion protein, the same
mouse cDNA was subcloned into the pGEX 5X-3 vector
(Amersham Pharmacia Biotech). The GFP–Arhgap6 fusion
construct was created by subcloning the N-terminal Xpress tag
and coding sequence from pcDNA-FL into the pEGF-N1
expression vector (Clontech) in three overlapping restriction
fragments. This construct expresses 963 amino acids of the
ARHGAP6 coding sequence fused to GFP at a SalI site 11 amino
acids from the C-terminus. The integrity of all constructs was
verified by DNA sequencing. Western analysis of protein extracts
from COS-7 and HeLa cells was performed using our previously
published protocols (42) and antisera to either the Xpress, FLAG
or GFP epitopes.
GTPase activation assay
Rac1, Cdc42 (G25K isotype) or RhoA were expressed as GST–
fusion proteins in E.coli and purified from glutathione–Sepharose
beads by thrombin cleavage as described previously (43). The
purified proteins were dialyzed against 15 mM Tris pH 7.5,
150 mM NaCl, 5 mM MgCl2, 0.1 mM dithiothreitol and
concentrated by centrifugation with a Centricon-10 spin column
(Amicon, Beverly, MA). GST–p50rhoGAP (amino acids 198–
439) and GST–ARHGAP6 (amino acids 279–660) were eluted
from the beads with 5 mM reduced glutathione and concentrated
with a Centricon-10 column. The assay was performed using 5 µg
of purified Cdc42, Rac1 or RhoA, 0.2 µM of purified GST–
Arhgap6 (amino acids 279–660) and GST–p50rhoGAP fusion
proteins (amino acids 198–439) and 10 µCi of [γ-32P]GTP (6000
Ci/mmol; NEN Life Sciences, Boston, MA) as previously
described (20). Concentrations of GST–fusion proteins were
determined by spectrophotometry and SDS–PAGE. Similar data
were obtained in three independent experiments.
Cell culture and transfection
HeLa and COS-7 cells (Amercian Tissue Culture Collection,
Manassas, VA) were maintained in OptiMem (Life Technologies,
Gaithersburg, MD) containing 5% fetal bovine serum in an
atmosphere of 5% CO2 at 37°C. For transfections, 105 cells were
plated onto glass coverslips in 6-well plates, transfected with
Lipofectamine (Life Technologies) using 1 µg of plasmid DNA
per well and incubated with the transfection mixture for 4 h
without serum or antibiotics. Cells were rinsed and incubated in
serum-containing medium for an additional 24–48 h before
fixation and staining. For growth factor treatment, cells were
plated in serum for 4–7 days and then serum starved for 16 h. Cells
were trypsinized, resuspended in serum-free medium containing
0.5 mg/ml soybean trypsin inhibitor (Sigma, St Louis, MO),
rinsed once and replated onto fibronectin-coated glass
coverslips (Becton Dickinson, Bedford, MA) for transfection as
above. At 8–12 h post-transfection, cells were stimulated with
100 ng/ml bradykinin (Sigma) for 10 min at 37°C or with 5 µM
SPP (Sigma) for 30 min at 37°C. Immediately after the incubation
period, the cells were fixed and processed for immunofluorescent
staining. Transfected and untransfected cells were scored for
stress fibers or filopodia by double-labeling with anti-actin
antibodies.
Immunofluorescence
Cells were rinsed twice in calcium- and magnesium-free
phosphate-buffered saline (PBS) and fixed using 4%
formaldehyde in PEM buffer (50 mM K-PIPES, 5 mM EGTA,
2 mM MgCl2, pH 6.8) with 0.2% Triton X-100 for 10 min at room
temperature. After post-fixation in ice-cold methanol, the cells
were rinsed three times in PBS and allowed to rehydrate in PBS at
room temperature for at least 30 min. Cells were incubated with
the primary antibodies anti-FAK at 1:2000 dilution (Sigma),
monoclonal anti-actin JLA20 at 1:200 (Sigma) or anti-epitope
antibodies [anti-Xpress 1:500 (Invitrogen) or anti-FLAG 1:500
(Eastman Kodak, Rochester, NY)] in PBS for 1 h at room
temperature or overnight at 4°C. Following three washes in PBS,
the secondary antibodies [anti-rabbit or anti-mouse IgG
conjugated with either fluorescein isothiocyanate (FITC), Texas
Red or biotin (1:500; Vector Laboratories, Burlingame, CA)]

were applied to the cells for 1 h at room temperature. Biotin-
conjugated antibodies were detected by a third incubation with
Cy3-conjugated Extravidin (1:300; Sigma). Nuclei were
counterstained with 4,6-diamidino-2-phenylindole (DAPI, 1 µg/
ml). The coverslips were inverted onto slides and mounted with
Vector Shield medium (Vector Laboratories) for visualization.
Live video microscopy
Live microscopy was performed on cells transfected with a wild-
type ARHGAP6–GFP fusion construct, the mutated (R433G)
ARHGAP6–GFP construct or an empty GFP vector as the
control. Cells were grown on 40 mm coverslips in 60 mm plates
and transfected with 2 µg of each test plasmid using
Lipofectamine as above. Following a 4 h incubation with the
transfection mixture, cells were allowed to recover for 6–9 h and
were transferred to a live cell chamber (Bioptechs, Butler, PA).
Cells were maintained at 37°C in Dulbecco’s modified Eagle’s
medium with 5% stripped fetal bovine serum, which was
recirculated using a peristaltic pump. In order to minimize photo
damage, cells were imaged at suboptimal exposure using neutral
density filters, to allow only 32% of the total light, and 1 s
exposure times. Optical sections were limited to ∼20 per cell and
images were taken at 3 min intervals for 1–2 h or overnight at 10
min intervals.
ACKNOWLEDGEMENTS
We would like to thank C. Tranh and N. Chao for assistance with
the mutational analysis and library screening, and B. Antalffy for
preparation of sections for histopathological analysis. S.J. is a
recipient of the ‘Association pour la Recherche contre le Cancer’
(ARC) and N.L.-V. is a junior scholar from Fonds de la Recherche
en Santé du Quebec. This work was also supported by the Howard
Hughes Medical Institute (H.Y.Z.), the Medical Scientist Training
Program (S.P.) and resources from the Mental Retardation
Research Center and the Child Health Research Center at Baylor
College of Medicine.
REFERENCES
1. Ridley, A.J. (1995) Rho-related proteins: actin cytoskeleton and cell cycle.
Curr. Opin. Genet. Dev., 5, 24–30.
2. Yamamoto, M., Morii, N., Ikai, K. and Imamura, S. (1994) Effect of
botulinum C3 exoenzyme on cell growth and cytoskeleton organization in
transformed human epidermal cells in culture: a possible role for rho protein
in epidermal cells. J. Dermatol. Sci., 8, 103–109.
3. Lamarche, N. and Hall, A. (1994) GAPs for rho-related GTPases. Trends
Genet., 10, 436–440.
4. Ridley, A.J. (1994) Signal transduction through the GTP-binding proteins
Rac and Rho. J. Cell Sci., 18(suppl.), 127–131.
5. Tatsis, N., Lannigan, D.A. and Macara, I.G. (1998) The function of the p190
Rho GTPase-activating protein is controlled by its N-terminal GTP binding
domain. J. Biol. Chem., 273, 34631–34638.
6. Taylor, J.M., Macklem, M.M. and Parsons, J.T. (1999) Cytoskeletal
changes induced by GRAF, the GTPase regulator associated with focal
adhesion kinase, are mediated by Rho. J. Cell Sci., 112, 231–242.
7. Kozma, R., Ahmed, S., Best, A. and Lim, L. (1996) The GTPase-activating
protein N-chimaerin cooperates with Rac1 and Cdc42Hs to induce the
formation of lamellipodia and filopodia. Mol. Cell. Biol., 16, 5069–5080.
8. Mizuno, T., Kaibuchi, K., Ando, S., Musha, T., Hiraoka, K., Takaishi, K.,
Asada, M., Nunoi, H., Matsuda, I. and Takai, Y. (1992) Regulation of the
superoxide-generating NADPH oxidase by a small GTP-binding protein
and its stimulatory and inhibitory GDP/GTP exchange proteins. J. Biol.
Chem., 267, 10215–10218.
Human Molecular Genetics, 2000, Vol. 9, No. 4 487
9. Silva, A.J., Frankland, P.W., Marowitz, Z., Friedman, E., Lazlo, G., Cioffi,
D., Jacks, T. and Bourtchuladze, R. (1997) A mouse model for the learning
and memory deficits associated with neurofibromatosis type I. Nature
Genet., 15, 281–284.
10. Brannan, C.I., Perkins, A.S., Vogel, K.S., Ratner, N., Nordlund, M.L., Reid,
S.W., Buchberg, A.M., Jenkins, N.A., Parada, L.F. and Copeland, N.G.
(1994) Targeted disruption of the neurofibromatosis type-1 gene leads to
developmental abnormalities in heart and various neural crest-derived
tissues. Genes Dev., 8, 1019–1029.
11. Johnson, M.R., DeClue, J.E., Felzmann, S., Vass, W.C., Xu, G., White, R.
and Lowy, D.R. (1994) Neurofibromin can inhibit Ras-dependent growth
by a mechanism independent of its GTPase-accelerating function. Mol.
Cell. Biol., 14, 641–645.
12. Pomerance, M., Thang, M.N., Tocque, B. and Pierre, M. (1996) The Ras-
GTPase-activating protein SH3 domain is required for Cdc2 activation and mos
induction by oncogenic Ras in Xenopus oocytes independently of mitogen-
activated protein kinase activation. Mol. Cell. Biol., 16, 3179–3186.
13. Lindsay, E.A., Grillo, A., Ferrero, G.B., Roth, E.J., Magenis, E., Grompe, M.,
Hulten, M., Gould, C., Baldini, A., Zoghbi, H.Y. et al. (1994) Microphthalmia
with linear skin defects (MLS) syndrome: clinical, cytogenetic, and molecular
characterization. Am. J. Med. Genet., 49, 229–234.
14. Schaefer, L., Prakash, S. and Zoghbi, H.Y. (1997) Cloning and
characterization of a novel rho-type GTPase-activating protein gene
(ARHGAP6) from the critical region for microphthalmia with linear skin
defects. Genomics, 46, 268–277.
15. Mayer, B.J. and Gupta, R. (1998) Functions of SH2 and SH3 domains.
Curr. Topics Microbiol. Immunol., 228, 1–22.
16. Toyosawa, S., O’Huigin, C., Figueroa, F., Tichy, H. and Klein, J. (1998)
Identification and characterization of amelogenin genes in monotremes,
reptiles, and amphibians. Proc. Natl Acad. Sci. USA, 95, 13056–13061.
17. Smith Jr, D.M. and Haggitt, R.C. (1983) A comparative study of generic
stains for carcinoid secretory granules. Am. J. Surg. Pathol., 7, 61–68.
18. Paylor, R., Nguyen, M., Crawley, J.N., Patrick, J., Beaudet, A. and Orr-
Urtreger, A. (1998) Alpha-7 nicotinic receptor subunits are not necessary
for hippocampal-dependent learning or sensorimotor gating: a behavioral
characterization of Acra7-deficient mice. Learning Memory, 5, 302–316.
19. Paylor, R., Baskall, L. and Wehner, J. (1993) Behavioral dissociations
between C57BL/6 and DBA/2 mice on learning and memory tasks: a
hippocampal-dysfunction hypothesis. Psychology, 21, 11–26.
20. Lamarche-Vane, N. and Hall, A. (1998) CdGAP, a novel proline-rich GTPase-
activating protein for Cdc42 and Rac. J. Biol. Chem., 273, 29172–29177.
21. Graham, D.L., Eccleston, J.F. and Lowe, P.N. (1999) The conserved
arginine in rho-GTPase-activating protein is essential for efficient catalysis
but not for complex formation with Rho·GDP and aluminum fluoride.
Biochemistry, 38, 985–991.
22. Billuart, P., Bienvenu, T., Ronce, N., des Portes, V., Vinet, M.C., Zemni, R.,
Crollius, H.R., Carrie, A., Fauchereau, F., Cherry, M. et al. (1998)
Oligophrenin-1 encodes a rhoGAP protein involved in X-linked mental
retardation. Nature, 392, 923–926.
23. Levinson, B., Kenwrick, S., Gamel, P., Fisher, K. and Gitschier, J. (1992)
Evidence for a third transcript from the human factor VIII gene. Genomics,
14, 585–589.
24. Habib, A.A., Gulcher, J.R., Hognason, T., Zheng, L. and Stefansson, K.
(1998) The OMgp gene, a second growth suppressor within the NF1 gene.
Oncogene, 16, 1525–1531.
25. Viskochil, D., Cawthon, R., O’Connell, P., Xu, G.F., Stevens, J., Culver,
M., Carey, J. and White, R. (1991) The gene encoding the oligodendrocyte-
myelin glycoprotein is embedded within the neurofibromatosis type 1 gene.
Mol. Cell. Biol., 11, 906–912.
26. Engle, S.J., Womer, D.E., Davies, P.M., Boivin, G., Sahota, A., Simmonds,
H.A., Stambrook, P.J. and Tischfield, J.A. (1996) HPRT–APRT-deficient
mice are not a model for Lesch–Nyhan syndrome. Hum. Mol. Genet., 5,
1607–1610.
27. Yang, X.W., Wynder, C., Doughty, M.L. and Heintz, N. (1999) BAC-mediated
gene-dosage analysis reveals a role for Zipro1 (Ru49/Zfp38) in progenitor cell
proliferation in cerebellum and skin. Nature Genet., 22, 327–335.
28. Magdaleno, S.M. and Curran, T. (1999) Gene dosage in mice—BAC to the
future. Nature Genet., 22, 319–320.
29. Bryant, S.S., Briggs, S., Smithgall, T.E., Martin, G.A., McCormick, F.,
Chang, J.H., Parsons, S.J. and Jove, R. (1995) Two SH2 domains of p120
Ras GTPase-activating protein bind synergistically to tyrosine
phosphorylated p190 Rho GTPase-activating protein. J. Biol. Chem., 270,
17947–17952.
488 Human Molecular Genetics, 2000, Vol. 9, No. 4
30. van der Geer, P., Henkemeyer, M., Jacks, T. and Pawson, T. (1997)
Aberrant Ras regulation and reduced p190 tyrosine phosphorylation in cells
lacking p120-Gap. Mol. Cell. Biol., 17, 1840–1847.
31. Allen, K.M., Gleeson, J.G., Bagrodia, S., Partington, M.W., MacMillan,
J.C., Cerione, R.A., Mulley, J.C. and Walsh, C.A. (1998) PAK3 mutation in
nonsyndromic X-linked mental retardation. Nature Genet., 20, 25–30.
32. Bar-David, S., Lerer, I., Sarfaty, C.K., Kohan, Z.G., Meiner, V., Zlotogora,
J. and Abeliovich, D. (1996) Localization of two X-linked mental
retardation (XLMR) genes to Xp: MRX37 gene at Xp22.31–p22.32 and a
putative MRX gene on Xp22.11–p22.2. Am. J. Med. Genet., 64, 83–88.
33. Dutch–Belgian Fragile X Consortium (1994) Fmr1 knockout mice: a model
to study fragile X mental retardation. Cell, 78, 23–33.
34. D’Hooge, R., Nagels, G., Franck, F., Bakker, C.E., Reyniers, E., Storm, K.,
Kooy, R.F., Oostra, B.A., Willems, P.J. and De Deyn, P.P. (1997) Mildly
impaired water maze performance in male Fmr1 knockout mice.
Neuroscience, 76, 367–376.
35. Ahmed, S., Kozma, R., Hall, C. and Lim, L. (1995) GTPase-activating
protein activity of N(alpha 1)-chimaerin and effect of lipids. Methods
Enzymol., 256, 114–125.
36. Sambrook, J., Fritsch, E.F. and Maniatis, T. (1989) Molecular Cloning: A
Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring
Harbor, NY.
37. Banfi, S., Servadio, A., Chung, M.-y., Kwiatkowski Jr, T.J., McCall, A.E.,
Duvick, L.A., Shen, Y., Roth, E.J., Orr, H.T. and Zoghbi, H.Y. (1994)
Identification and characterization of the gene causing type 1
spinocerebellar ataxia. Nature Genet., 7, 513–519.
38. Henikoff, S. (1990) Ordered deletions for DNA sequencing and in vitro
mutagenesis by polymerase extension and exonuclease III gapping of
circular templates. Nucleic Acids Res., 18, 2961–2966.
39. Butler, B.A. (1998) Sequence analysis using GCG. Methods Biochem.
Anal., 39, 74–97.
40. Ben-Arie, N., Bellen, H.J., Armstrong, D.L., McCall, A.E., Gordadze, P.R.,
Guo, Q., Matzuk, M.M. and Zoghbi, H.Y. (1997) Math1 is essential for
genesis of cerebellar granule neurons. Nature, 390, 169–172.
41. Kimmel, C.A. and Trammell, C. (1981) A rapid procedure for routine
double staining of cartilage and bone in fetal and adult animals. Stain
Technol., 56, 271–273.
42. Skinner, P.J., Koshy, B.T., Cummings, C.J., Klement, I.A., Helin, K.,
Servadio, A., Zoghbi, H.Y. and Orr, H.T. (1997) Ataxin-1 with an expanded
glutamine tract alters nuclear matrix-associated structures. Nature, 389,
971–974.
43. Self, A.J. and Hall, A. (1995) Purification of recombinant Rho/Rac/G25K
from Escherichia coli. Methods Enzymol., 256, 3–10.