Defining the Functional Domain of Programmed Cell
Death 10 through Its Interactions with
Christopher F. Dibble1, Jeremy A. Horst3, Michael H. Malone1, Kun Park2, Brenda Temple1, Holly
Cheeseman2, Justin R. Barbaro2, Gary L. Johnson1, Sompop Bencharit1,2*
1Department of Pharmacology, School of Medicine, and the Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, North Carolina, United
States of America, 2Department of Prosthodontics and the Dental Research Center, School of Dentistry, University of North Carolina, Chapel Hill, North Carolina, United
States of America, 3Department of Microbiology, School of Medicine, and Department of Oral Biology, School of Dentistry, University of Washington, Seattle, Washington,
United States of America
Cerebral cavernous malformations (CCM) are vascular abnormalities of the central nervous system predisposing blood
vessels to leakage, leading to hemorrhagic stroke. Three genes, Krit1 (CCM1), OSM (CCM2), and PDCD10 (CCM3) are involved
in CCM development. PDCD10 binds specifically to PtdIns(3,4,5)P3 and OSM. Using threading analysis and multi-template
modeling, we constructed a three-dimensional model of PDCD10. PDCD10 appears to be a six-helical-bundle protein
formed by two heptad-repeat-hairpin structures (a1–3 and a4–6) sharing the closest 3D homology with the bacterial
phosphate transporter, PhoU. We identified a stretch of five lysines forming an amphipathic helix, a potential PtdIns(3,4,5)P3
binding site, in the a5 helix. We generated a recombinant wild-type (WT) and three PDCD10 mutants that have two (D2KA),
three (D3KA), and five (D5KA) K to A mutations. D2KA and D3KA mutants hypothetically lack binding residues to
PtdIns(3,4,5)P3 at the beginning and the end of predicted helix, while D5KA completely lacks all predicted binding residues.
The WT, D2KA, and D3KA mutants maintain their binding to PtdIns(3,4,5)P3. Only the D5KA abolishes binding to
PtdIns(3,4,5)P3. Both D5KA and WT show similar secondary and tertiary structures; however, D5KA does not bind to OSM.
When WT and D5KA are co-expressed with membrane-bound constitutively-active PI3 kinase (p110-CAAX), the majority of
the WT is co-localized with p110-CAAX at the plasma membrane where PtdIns(3,4,5)P3 is presumably abundant. In contrast,
the D5KA remains in the cytoplasm and is not present in the plasma membrane. Combining computational modeling and
biological data, we propose that the CCM protein complex functions in the PI3K signaling pathway through the interaction
between PDCD10 and PtdIns(3,4,5)P3.
Citation: Dibble CF, Horst JA, Malone MH, Park K, Temple B, et al. (2010) Defining the Functional Domain of Programmed Cell Death 10 through Its Interactions
with Phosphatidylinositol-3,4,5-Trisphosphate. PLoS ONE 5(7): e11740. doi:10.1371/journal.pone.0011740
Editor: Andreas Hofmann, Griffith University, Australia
Received April 30, 2010; Accepted July 1, 2010; Published July 23, 2010
Copyright: ? 2010 Dibble et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The study was supported by the National Institutes of Health (NIH) grant HL092338 (to S.B.) GM068820 and GM068820-S1 (to G.L.J.), and T32GM80079
(to C.F.D.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: Sompop_Bencharit@dentistry.unc.edu
Cerebral cavernous malformations (CCM) are congenital or
sporadic vascular disorders of the central nervous system (CNS)
are abnormally large harmatomous vascular lesions formed by a
single layer of capillary endothelial cells without the support of brain
parenchyma [1–2,18–22]. Ruptured CCM lesions can cause
hemorrhagic stroke and are often associated with seizures, recurrent
headaches, and focal neurological defects (2–4). Three CCM loci
(OSM or CCM2), and 3q25.2–27 (PDCD10 or CCM3). Mutations
in these CCM loci cause loss of function of these proteins and result
in CCM [7–16,17,23].
CCM3, the smallest of the CCM proteins, is a 25 KDa protein
composed of 212 amino acids. It was originally identified as TF-1
cell apoptosis related gene-15 (TFAR15), since it is up-regulated
with the induction of apoptosis by serum withdrawal in TF-1
human premyeloid cells [17,23]. It was subsequently renamed
PDCD10 (programmed cell death 10) as it was thought to be
involved in apoptotic responses [17,23]. PDCD10 is the third and
latest CCM gene identified [17,23–24]. The N-terminal region of
PDCD10, which in some CCM patients is the site of an in-frame
deletion of an entire exon encoding from L33 to K50, was found to
be the binding site for the oxidant stress response serine/threonine
kinase 25 (STK25) and the mammalian Ste20-like kinase 4 (MST4)
. Similar to earlier observations, PDCD10 was found to
function in apoptotic pathways since overexpression of PDCD10
induces apoptosis through the caspase 3 pathway . Further-
more, PDCD10 may be regulated through phosphorylation and
dephosphorylation, since it can be phosphorylated by STK25 and
dephosphorylated by binding to the phosphatase domain of Fas-
associated phosphatase-1 . Recently, we showed that all three
CCM proteins (Krit1, OSM, and PDCD10) form a complex in the
cell and that PDCD10 binds directly to OSM independently of the
OSM-Krit1 interaction . We also showed that PDCD10 binds
to both phosphatidylinositol bis- or tris-phosphates, but seems to
PLoS ONE | www.plosone.org1 July 2010 | Volume 5 | Issue 7 | e11740
have the highest affinity to phosphatidylinositol-3,4,5-trisphosphate
(PtdIns(3,4,5)P3) . However, it is not known which part of
PDCD10 binds to PtdIns(3,4,5)P3 or OSM because there is
currently no structural data available for PDCD10. Creating a
structural model of PDCD10 is, therefore, a critical first step to
provide insight into the PDCD10 structure-function relationship.
The interaction of PDCD10 and PtdIns(3,4,5)P3 suggests that
PDCD10 may function in concert with phosphatidylinositol-3-
kinase (PI3K), the enzyme that catalyzes the formation of
PtdIns(3,4,5)P3at the plasma membrane . PI3K activation by
growth factors including vascular endothelial growth factor (VEGF)
is known to be crucial in angiogenesis. Thus, a relationship between
PDCD10 and PI3K would be evidence that CCM development
may result from dysregulation in the PI3K pathway through
PDCD10-PtdIns(3,4,5)P3interaction. In this study we attempted to
define the functional domain of PDCD10 that is important in
PtdIns(3,4,5)P3 binding by using molecular modeling combined
with site-directed mutagenesis.
Homology modeling allows for identification of critical amino
acid residues for protein-protein interaction and protein-ligand
interaction . The usefulness of homology is inversely
dependent on the evolutionary distance between the target and
templates . The structural conservation between the target and
template, as well as the correctness of the template alignment, are
among the most important factors in generating homology models.
Generating a homology model for PDCD10 is therefore
challenging because of the low structural conservation with
available templates and low sequence identity between PDCD10
and the templates [31–33]. Our homology models and biophysical
data predict that PDCD10 is likely a dimeric six-helical protein
composed of two trihelical heptad-repeat structures. We identified
the amphipathic helix and Lys residues essential for PDCD10-
PtdIns(3,4,5)P3 and PDCD10-OSM interaction. Finally, we
demonstrated that using membrane-bound constitutively active
PI3K (p110-CAAX), the wild-type (WT) PDCD10 co-localizes
with the p110-CAAX mostly at the plasma membrane, while the
mutants lacking Lys residues remain in the cytoplasm.
Threading analysis and homology modeling of PDCD10
The structure of PDCD10 was first examined using 3D-Jury to
generate meta-predictions for PDCD10 . Unlike Krit1 or
OSM, which are large scaffolding proteins composed of multiple
functional domains, PDCD10 seems to have a compact single
domain structure. The structures predicted by threading analysis
to be most similar to PDCD10 include a six-helical bundle protein
of no known function (1xwm), a five-helical bundle structure of
vinculin (1rke), a four-helical bundle structure of the FAT domain
of focal adhesion kinase-FAK (1pv3), and a four-helical bundle
structure of the SNARE complex (1sfc) (Table 1). Most of these
helical bundle proteins, similarly to PDCD10, are highly
conserved throughout evolution . Like Krit1 and OSM, these
helical bundle proteins also function mainly in protein localization
to the cytoskeleton and cellular membranes [35–37]. Based on this
threading analysis, PDCD10 is likely to be a helical bundle protein
composed of four to six helices. The compact structure of
PDCD10 is uniquely distinct from Krit1 and OSM and suggests
that PDCD10 may be an adaptor protein.
To further define the functional domain of PDCD10, multi-
target modeling was used to generate a three dimensional model
[38–45]. The model of PDCD10 shows a double heptad-repeat-
hairpin structure and the potential interactive surfaces of
PDCD10. The overall theoretical structure of PDCD10 is an a-
helical structure with over 68% of the amino acids in a-helical
conformations (Fig. 1A–B, supplemental structural model data).
This structure is composed of a six-helix bundle formed by two
three-helix bundles connected by a 15residue loop. The N-
terminal three-helix bundle is composed of a1, a2, and a3, while
the C-terminal three-helix bundle is composed of a4, a5, and a6.
The interface between these two three-helix bundles forms a
hydrophobic core, which is mediated mainly through hydrophobic
residues located in a1–a4 and a3–a6. These two three-helix
bundles are structural repeats known as heptad-repeats, a common
random hydrophobic/hydrophilic repeat in coiled-coil structures
[46–48] (Fig. 2A–B). The superposition of the two structural
repeats of the three-helix bundle, a1–a3 with a4–a6, results in an
r.m.s. deviation of about 5 A˚ over 85 equivalent Ca atoms
(Fig. 2A–B). The major difference between these two repeats is
that a1 is about 13 amino acids longer than a4. The other two
helices, of the helical repeats a2/a4 and a3/a6, can be
superimposed. Similar to the PhoU structures, superposition of
the two repeats suggested that the protein sequences in these two
repeats might be evolutionarily related  (Fig. 2A–B). Based on
superimposition of all generated models, the C-terminal portion of
the homology model appears to be more correct, while the N-
terminal portion is less certain. The predicted ligand binding site is
likely to be in the very end of the C-terminal portion. We therefore
focused on the use of the homology model and ligand interaction
in the C-terminal portion and did not attempt to predict other
ligand interactions in the N-terminal portion.
PDCD10 contains an unusually large number of highly flexible
sidechains for a small protein: 21 lysine and 21 glutamate residues
out of 212 total amino acids. Lysine and glutamate residue clusters
are known to exhibit a highly variable surface due to flexible
conformations [50–54]. This allows for structural flexilibity of the
protein molecule that, in combination with hydrophobic areas, is
often present in an interactive surface of a ligand or protein
binding partner. The lysine residues in PDCD10 are conserved
Table 1. Threading analysis using 3D-Jury.
(Jscore)PDB Hit Name of Protein
39.11 1sum_B PhoU protein homologue
36.89 1rke_AHuman vinculin
28.331pv3_A FAT domain of FAK
27.44 1k04_A FAT domain of FAK
27.331ktm_A FAT domain of FAK
a-catenin and b-catenin chimera
22.67 2bid_APro-apoptotic protein BID
21.331ddb_APro-apoptotic protein BID
21.331k40_APro-apoptotic protein BID
17.78 1k04_A FAT domain of FAK
16.56 1wph_AMetal binding protein
15.33 1sfc_D SNARE complex
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org2 July 2010 | Volume 5 | Issue 7 | e11740
and located among conserved hydrophobic residues, in particular
the C-terminal region (Fig. 1A, 2C). Our PDCD10 theoretical
model suggests that this C-terminal region of PDCD10 may form
an amphipathic helix that potentially binds to membrane
phospholipids. Note that the majority of proteins with an
amphipathic helix capable of binding to phospholipids often have
lysine residues predominantly interspersed with hydrophobic
residues, for example, the epsin ENTH domain (Protein Data
Bank (PDB) code 1H0A), and the AP180/CALM ANTH domain
(PDB code 1HFA) [55–57]. In the case of PDCD10, almost 10%
of its amino acid content consists of scattered lysine residues. This
distribution of lysine residues is seen in protein structures that
interact with inositolphosphate ligands, including Drosophila
melanogaster amphiphysin (PDB code 1URU) , endophilin-A1
BAR domain (PDB code 1ZWW), CIP4 (Cdc42-interacting
protein-4), F-BAR domain (PDB code 2EFK), and IMD domain
from IRSp53/missing-in-metastasis (PDB code 1Y2O) [58–62].
Lysine residues inside the bend of the cytoplasmic membrane
indicates that the protein is sensing or stabilizing a highly curved
membrane  .There are five conserved lysine residues in a5 and
a flexible loop that connects this helix with the last helix, a6,
including K169, K172, K179, K183, and K186 (Fig 2C). Surface
potential analysis of the theoretical structure showed that these
lysine residues form a cluster of positive charges interspersed with
the hydrophobic residues (Fig 3A–C). These observations, in
combination with our previous data showing that PDCD10
selectively binds to phosphatidylinostiol bis- and trisphosphates,
but binds PtdIns(3,4,5)P3 with strongest affinity, led us to
hypothesize that this area could be a potential amphipathic helix
that may play a role in PtdIns(3,4,5)P3binding .
Defining PtdIns(3,4,5)P3binding site of PDCD10
To determine whether the five lysines forming an amphipathic
helix in our theoretical structure are important in PtdIns(3,4,5)P3
binding, we generated three mutants designed to eliminate
important PtdIns(3,4,5)P3binding lysine residues with two, three
Figure 1. PDCD10 Model. A. Multiple sequence alignment of mouse, human, and zebra fish PDCD10 shows a highly conserved primary structure.
B/C. A three dimensional model for PDCD10 shows a double heptad-repeat-hairpin structure.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org3 July 2010 | Volume 5 | Issue 7 | e11740
and five lysine-to-alanine mutations; D2KA (K169A, and K172A),
D3KA K179A, K183A and K186A), D5KA (K169A, K172A,
K179A, K183A, and K186A). The mutations in D2KA are
located in the N-terminal portion of the amphipathic helix (a5),
while the D3KA ones are in the C-terminal portion of the helix.
The D5KA mutations would cover all possible lysines in the
amphipathic helix. We expressed these recombinant mutant
proteins along with wild-type PDCD10 with an N-terminal
6xHis-tag in bacteria as previously described . We investigated
the phospholipid binding potential of the PDCD10 recombinant
proteins using Membrane Lipid Array and PIP-Arrays (Echelon,
In the Membrane Lipid Array, WT PDCD10 binds weakly but
specifically to phosphoserine (Fig. 4A). However, in PIP-Arrays, WT
PDCD10 selectively binds PtdIns(3,4,5)P3with high affinity, and
phosphatidyl bisphosphate (phosphatidylinositol 3,4 bisphophate
(PtdIns(3,4)P2), phosphatidylinositol 3,5 bisphophate (PtdIns(3,5)P2),
and phosphatidylinositol 4,5 bisphophate (PtdIns(4,5)P2) with
moderate but lower affinity (Fig. 4B–C). We further screened for
important lysine residues that are essential forPtdIns(3,4,5)P3binding
using mutant proteins. The D5KA mutation completely abolishes
PtdIns(3,4,5)P3and phosphatidylinositol bisphosphate binding, while
the D2KA and D3KA mutants maintain similar phosphatidylinositol
bisphosphates and PtdIns(3,4,5)P3binding selectivity (Fig. 4A–C).
Interestingly, the D3KA seems to have lower affinity to phosphati-
dylinostol bisphosphates than the WT and the D2KA. These results
three lysines are sufficient for specific phosphatidylinositol bis- and
OSM interaction with D5KA, a PDCD10 mutant lacking
OSM (CCM2) interacts with Krit1 (CCM1) through a
canonical PTB domain/NPXY motif interaction. Recently, we
demonstrated that PDCD10 interacts with the OSM-Krit1
Figure 2. Heptad repeats and the amphipathic helix. A. Sequence alignment of the N-terminal and the C-terminal three-helix bundles. B.
Superimposition of the N-terminal and C-terminal three-helix bundles shows that these two heptad-repeat-hairpin structures are related in the three
dimensional structures. C. A diagram and primary sequence of the proposed amphipathic helix and the PIP binding site.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org4 July 2010 | Volume 5 | Issue 7 | e11740
complex by interacting with OSM . However, the mode of
OSM-PDCD10 interaction is not known. The phosphatidylinosi-
tol bis- and tris-phosphate binding study indicated that there are
five lysines located on the area of the a5 amphipathic helix that
function as a PtdIns(3,4,5)P3binding site (Fig. 2A,C). Based on our
model, these five lysines reside in a surface exposure. For a
relatively small protein like PDCD10, we expected that there may
be redundancy in PtdIns(3,4,5)P3and protein-protein interaction.
We therefore further investigated if these lysines in the
amphipathic helix (a5) are also involved in interactions with
OSM. Recombinant purified WT and D5KA PDCD10 coupled to
sepharose beads were used to pull-down FLAG-tagged overex-
pressed OSM and Krit1 (Fig. 4C). Interestingly, these pull-down
experiments showed that the mutant fails to interact with OSM.
These data suggests that the five lysine residues in this C-terminal
region of PDCD10 are not only important in PDCD10-
PtdIns(3,4,5)P3 interaction, but also important in PDCD10-
Secondary and tertiary structures of PDCD10
To be sure that the mutation of lysines in D5KA does not alter
the overall protein structure, we further examined the secondary
and tertiary structures of the WT and D5KA mutant using purified
recombinant proteins. Consistent with the modeling data, the
Figure 3. Proposed PtdIns(3,4,5)P3binding site and dimeric interface. A. Ribbon model of PDCD10 showing the proposed PIP binding Lys
residues. B. Surface-potential model of PDCD10 showing the surface positive potential charge area (blue) and negative potential charge area (red). C.
Superimposition of the ribbon model and the surface-potential model. Note the proposed PIP binding Lys residues. D. Superimposition of the ribbon
model and the surface-potential model shows the potential dimeric interface. Note that the dimeric interaction surfaces are mutually exclusive to the
PIP binding site.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org5 July 2010 | Volume 5 | Issue 7 | e11740
circular dichroism (CD) spectra of the WT and D5KA appears to
be of a-helical proteins. While the CD spectra of the WT and
D5KA mutant were almost identical, the Tm of the WT is about
7uC lower than the one of the D5KA ((Fig. 5A–B). These results
suggest that while the D5KA maintains similar secondary structure
to the WT, it has lower molecular stability, perhaps as a result of
five K-to-A mutations. In addition to examining the secondary
structure, we used HPLC size-exclusion chromatography coupled
with multi-angle laser light scattering (SEC-MALS) to determine
the native molecular weights of the WT and D5KA PDCD10
proteins (Table 2). The results showed that both proteins are
slightly larger than 50 KDa and therefore form a dimeric complex
in solution. This dimeric form is the only species found in both
WT and D5KA recombinant protein. The dimeric interface is
therefore mutually exclusive from the phospholipid and OSM
binding surface. The surface potential model of PDCD10 showed
Figure 4. PDCD10 interactions with phospholipids and OSM. A. The Membrane Lipid Array shows that PDCD10 binds exclusively to
phosphotidylserine with weak binding. B. PIP Arrays show the relative PIP binding affinity for the WT and three mutant PDCD10 proteins. C. Pull-
down experiments, using recombinant purified WT and D5KA PDCD10 coupled to sepharose beads to pull-down FLAG-tagged overexpressed OSM
and Krit1, showing that only the WT binds to the OSM-Krit1 complex.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org6 July 2010 | Volume 5 | Issue 7 | e11740
a highly negative charge area in the first trihelical heptad repeat
and a highly positive charge area in the second trihelical heptad
repeat (Fig 3D). The optimal docking area method indicates that
these areas may play a role in the dimerization of PDCD10 .
Cellular co-localization of PDCD10 and membrane-bound
The fact that PDCD10 seems to have the highest affinity with
PtdIns(3,4,5)P3 led us to believe that PDCD10 may function
with phospholinositol-3-kinase (PI3K). PI3K is a master kinase
that catalyzes the phosphorylation of PtdIns(4,5)P2to produce
PtdIns(3,4,5)P3. PI3K catalyzes PtdIns(3,4,5)P3formation upon
activation by growth factors such as VEGF . We used the
catalytic subunit of PI3K (lacking a regulatory domain) with a
COOH-terminal plasma membrane targeting sequence to
produce a PI3K construct that is constitutively active (p110-
CAAX) [63–65]. When WT PDCD10, D5KA, and p110-CAAX
were co-transfected, WT PDCD10 and p110-CAAX were co-
localized, while the D5KA was not (Fig 6). This experiment
confirms our biophysical data and suggests that PDCD10 binds
to PtdIns(3,4,5)P3 and may function in the PI3K signaling
Figure 5. Circular dichroism and fraction denaturation. A. CD spectra of the WT and D5KA PDCD10 showing identical CD spectra. B. Fraction
denaturation experiments show the lower Tm of D5KA compared to the WT.
Table 2. Determination of quaternary structure of wild-type
and mutant D5KA ccm3 in solution using size exclusion
chromatrography/multi angle laser light scattering (SEC-MAL).
Determined MwEstimate MwEstimated Mw
Wild-type 5.354 6104
D5KA 5.054 6104
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org7 July 2010 | Volume 5 | Issue 7 | e11740
CCM is a unique example of a genetic disorder that results from
the dysfunction of three non-catalytic signaling proteins. Krit1-
OSM-PDCD10 form a complex in the cell and CCM lesions
appear to result from a loss of integrity of this complex. While
Krit1 and OSM proteins both have known structural domains and
functionally link to several important signaling pathways, includ-
ing integrin signaling, the p38 MAP kinase pathway, and RhoA-
rho kinase regulation, PDCD10 lacks known structural domains.
PDCD10 was thought to relate to apoptosis; however, there is no
clear apoptotic pathway known. Defining a functional domain of
PDCD10 is therefore important in learning how this protein
functions and contributes to CCM development. Combining
molecular modeling and site-directed mutagenesis, it appears that
PDCD10 is a six helical bundle protein composed of a two heptad
repeat, and a5 is an important amphipathic helix housing five
lysine residues essential for PtdIns(3,4,5)P3 binding. Unlike
phosphatidylinositol specific domains, for instance C1, PH, PX
and FYVE, amphipathic helices often only demonstrate high
specificity to certain phosphatidylinositol molecules under specific
circumstances, such as when the membrane area possesses a highly
curved region . We speculate that this may be the case for
PDCD10 since there seem to be multiple lysine residues beyond
the amphipathic helix similar to some phosphatidylinositol binding
proteins that use these outside amphipathic helix lysine residues to
interact with the curved membrane region [57–62]. We further
show that the D5KA, a mutant lacking these five important K
residues, does not bind to PtdIns(3,4,5)P3in vitro. While the WT
and D5KA both appear to have the same helical secondary/
tertiary structure, and are dimeric in solution, the D5KA does not
bind OSM. In addition, when WT PDCD10 and D5KA are co-
expressed with p110-CAAX, a membrane-bound/constitutively
active PI3K, WT is co-localized with p110-CAAX at the plasma
membrane. However, D5KA stays in the cytoplasm. The results
suggest that PDCD10 may function with PI3K.
A recent study shows that PDCD10 functions in VEGF-
dependent translocation of vascular endothelial growth factor
receptor 2 (VEGFR2) and further, suggests that the C-terminal
portion of PDCD10 is important in PDCD10-VEGFR2 interac-
tion (66). VEGF is an upstream regulator of PI3K. It is therefore
plausible that PDCD10 may play a role in VEGF-PI3K signaling.
Interestingly, colocalization of PDCD10 and PI3K in our study is
almost identical to the colocalization of PDCD10 and VEGFR2
. It is possible that this colocalization is a result of PDCD10
and VEGFR2 interactions. It was shown that PDCD10 interacts
with VEGFR2 through its C-terminal region . This region
overlaps with the PtdIns(3,4,5)P3binding where we located the five
PtdIns(3,4,5)P3binding lysines. These lysines are also important in
PDCD10-OSM interaction. Furthermore, genetic studies demon-
strate that this region is predisposed to frameshift mutations that
often cause early termination of the protein resulting in CCM
. We therefore speculate that PDCD10-OSM and PDCD10-
VEGFR2 interactions may be regulated by the availability of
PtdIns(3,4,5)P3generated by PI3K.
Based on ourfindings andrecentstudies,wecomposed a signaling
model for PDCD10 (Fig 7). We propose that PDCD10 functions
closely with VEGFR2 and PI3K. Upon activation of VEGFR2 by
Figure 6. Co-localization of PDCD10 and p110-CAAX. Overexpression of the WT and D5KA PDCD10 together with p110-CAAX shows that WT
PDCD10 and p110-CAAX colocalize to the membrane while the D5KA stays only in the cytoplasm. A. WT in Mcherry. B. D5KA in CFP. C. p110-CAAX in
FITC. D. A composite picture of WT (Mcherry) and p110-CAAX (FITC). E. A composite picture of D5KA (CFP) and p110-CAAX (FITC). F. A composite
picture of WT (Mcherry) and D5KA (CFP). G. A composite picture of WT (Mcherry), p110-CAAX (FITC), and D5KA (CFP). H. A composite picture of D5KA
(CFP) and p110-CAAX (FITC). A light cyan mask was used to enhance the visualization of D5KA localization.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org8 July 2010 | Volume 5 | Issue 7 | e11740
VEGF, VEGFR2 binds to dimeric PDCD10, translocates to the
membrane, and becomes activated. As a result of VEGFR2
activation, PI3K is activated and favors the catalysis of PtdIns(3,4)P2
or PIP2to PtdIns(3,4,5)P3or PIP3(Fig 7). PtdIns(3,4,5)P3goes on
to bind PDCD10 and AKT. There may be an equilibrium
between PDCD10-PtdIns(3,4,5)P3and PDCD10-OSM/Krit1, as
well as a equilibrium between PDCD10-PtdIns(3,4,5)P3and AKT-
PtdIns(3,4,5)P3. PtdIns(3,4,5)P3and OSM seem to have the same
interactive site on the PDCD10 dimer, and it is possible that
VEGFR2 may also share the same binding site. It is therefore
plausible that PDCD10 may regulate the function of these three
important signaling molecules at the same time using simple
We propose that CCM development may result from the
dysregulation of the VEGF/PI3K signaling pathway through
PDCD10- PtdIns(3,4,5)P3 interaction. Future experiments will
need to be completed to define the role of PDCD10 in PI3K
signaling. PI3K can be activated by several growth factors,
including VEGF. Characterizing the role of PDCD10 during
PI3K activation by VEGF, as well as downstream effectors such as
AKT1, will be important in understanding CCM development
and how we may treat this condition by modulating the activity of
these kinases. Linking the role of PDCD10 in the VEGF/PI3K
pathway to other CCM proteins will elucidate the individual roles
of CCM proteins, as well as the CCM protein complex . We
recently showed that OSM plays a role in regulating the
degradation of the small GTPase RhoA and that the CCM
endothelial phenotype can be rescued with knockdown or
inhibitors of ROCK-RhoA effectors . Interestingly, activation
of PI3K can also increase RhoA activation . Further
experiments will also be needed to determine if PDCD10 functions
in concert with OSM and Krit1 in signaling the regulation of
RhoA and PI3K.
Materials and Methods
Generating a three-dimensional model of PDCD10 using
PDCD10 structure prediction.
structure prediction protocol was recently tested in the critical
assessment of techniques for protein structure prediction (CASP8;
,http://predictioncenter.org.). Not only did this method rank as
one of the best, but all predictions with normalized RAPDF scores
better than 255 (described below) were within 3.5 A˚ngstrom root
mean squared deviation of the corresponding experimental PDB
structure, indicating correct overall topology. Additionally, the
refinement method consistently improved initial models, in some
cases resulting in models within the accuracy of the corresponding
experimental structures themselves.
The initial comparative modeling templates
were identified using the secondary structure enhanced profile-
profile threading alignment (LOMETS) and the four part iterative
threading assembly refinement protocol of I-TASSER [38–39].
Ten different protein structure prediction servers were sampled,
each producing at least five models. All those conforming, at least
in part, to the selected templates (1kil, 1s35, 1sum, 1txd, 2boq,
2i0m, 2of3) were used in further analysis, along with five models
produced by PROTINFO  from the two full length templates
Iterations of ENCAD energy minimization and SCWRL3.0
sidechain optimization were applied to increase the sampled
conformational space between models, such that variation and
The following protein
Figure 7. Proposed PDCD10 signaling model.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org9 July 2010 | Volume 5 | Issue 7 | e11740
coverage were sufficient for clustering analysis [41–42]. With the
resultant model set, an iterative density calculation was applied,
which cycles between a cluster density calculation and removal of
outliers. Centroids for the five largest sub-clusters were then taken
as the five input models for refinement.
RAPDF and consensus based constraint selection.
the five initial models, a set of consensus interatomic distances was
derived as all atom-atom pair distances which occur within a 0.5
A˚ngstrom window for at least four of the five models. A residue
specific all atom probability discriminatory function (RAPDF)
was used to score the consensus distances. Here the philosophy
was that the probabilities derived from a Bayesian analysis of
distances observed in a structurally non-redundant database of
experimentally derived protein models versus random are likely to
be useful to build models similar to the native state protein
conformation . A batch-by-batch method compiled the final
distance set, starting with the consensus distances having the
highest RAPDF scores, resulting in a single interatomic distance
for each possible residue pair in the protein. Each distance was
weighted for importance in model building by the RAPDF score
and whether the distance was observed in four or all of the five
input models. Finally, three constraint sets were built using
different maximal distance cutoffs (12 A˚, 16 A˚, 20 A˚).
Model building and final selection.
sets were each used in fifty rounds of CYANA restrained torsion
angle dynamics simulations, for which a Ramachandran plot-like
distribution of torsion angles observed in the non-redundant
structure database was used to prescribe probabilities for torsion
angles . Each round produced twenty all-atom models, with a
total of three thousand conformations created. Half of the
conformations were filtered by sequentially applying RAPDF, a
van der Waals energy term, the hydrophobic compactness factor,
and an electrostatics term. The resulting models were minimized
by ENCAD and SCWRL3.0, and subsequently the iterative
density calculation was again applied to cyclically remove outliers
and re-cluster, and finally select the centroid for each of the five
largest clusters. A new set of interatomic distances were obtained
from the resulting five models, and used in a second round of
consensus modeling which produced the final tertiary structure
Homodimer interface site prediction.
interface sites were identified by applying the optimal docking
area method (ODA; ,http://www.molsoft.com/oda.), which
segregates surface patches and applies atomic desolvation calcu-
lations parameterized with octanol/water transfer experiments
adjusted to protein-protein interactions .
The three constraint
Examination of phospholipid and OSM binding using
recombinant PDCD10 proteins
recombinant PDCD10 protein was described previously .
Full-length murine PDCD10 was amplified using PCR from a
mouse fibroblast cDNA library and cloned into pMCSG7-His.
Recombinant murine 6xHis-PDCD10 was expressed in BL21 cells
and purified by nickel affinity chromatography. Three mutants
were generated using site-directed mutagenesis (QuikChange,
Stragetagene) including two K-to-A mutations (D2KA:K169A, and
K172A), three K-to-A mutations (D3KA:K179A, K183A and
K186A), and five K-to-A mutants (D5KA: K169A, K172A,
K179A, K183A, and K186A). Similar to the WT protein, mutant
proteins were expressed with N-terminal 6xHis-tag in BL21DE3
cells. Membrane lipid arrays and PIP arrays were purchased from
Echelon Biosciences. Membranes were blocked in 0.1% ovalbumin
in TBS-T for one hour then incubated with 1 mg/mlof recombinant
protein for two hours. After washing unbound protein using TBS-T,
bound protein was detected by immunoblotting with an anti-His
antibody (Santa Cruz Biotechnology).
transfected into HEK293 cells (the American Type Culture
Collection (ATCC) Rockville, MA, USA) using lipofectamine.
Twenty-four hours after transfection, cells were harvested in a
non-ionic detergent containing lysis buffer, and total protein
concentration was determined by the Bradford method. 10 mg
His-tag recombinant WT and mutant D5KA of PDCD10 were
bound to CNBr-activated sepharose beads (GE Biosciences) and
incubated with 500 mg of cell lysate for 16 hours at 4uC. Beads
were collected by centrifugation and washed 36with lysis buffer.
Washed beads were then mixed with 30 ml 26SDS-PAGE buffer
and analyzed on a 10% polyacrylamide gel. FLAG-tagged and
His-tag proteins were detected by immunoblotting similarly as
previously described .
OSM and were
Examination of the tertiary and secondary structures of
Size-exclusion chromatography-multi angle laser light
0.2 to 0.4 mg/ml of purified His-tag
recombinant WT and D5KA PDCD10 were dialyzed in 10 mM
phosphate buffer, pH 7.0 and 500 mM NaCl. The absolute
molecular weight of each protein was determined using high-
composed of Wyatt DAWN EOS light scattering instrument
interfaced to an Amersham Biosciences Akta FPLC, Wyatt
Optilab refractometer, and Wyatt dynamic light scattering
module at the UNC-CH Macromolecular Interaction Facility
using methods similar to ones previously described [69–70].
Circular dichroism (CD), and thermal denaturation
CD and thermal denaturation experiments were
conducted using an Applied Photophysics PiStar-180 CD
spectropolarimeter. WT or D5KA PDCD10 (0.15 mg/ml; in
10 mM phosphate buffer, pH 7.0) was used. CD data was
collected for each protein and ranged in wavelength from 185 to
260 nm.For the thermaldenaturation
temperature was increased from 25uC to 90uC while monitoring
at 222 nm. Plots of fraction denatured versus temperature were
produced by defining the upper and lower temperature baselines
as 0 and 100%, respectively.
Examination of PDCD10-PI3K colocalization in cell
COS7 cells (the American Type Culture Collection
(ATCC) Rockville, MA, USA) were maintained in DMEM
(LifeTechnologies) with 10% FBS, 100 U/ml penicillin, and
100 mg/ml streptomycin at 37uC with 7% CO2. Recombinant
human IL-1b was obtained from PeproTech. Polyclonal anti-Myc
antibody was obtained from Santa Cruz Biotechnology and FITC
goat anti-rabbit antibody was obtained from Invitrogen. pBabe-
p110-CAAX-Myc was generously provided by Dr. Channing Der
(University of North Carolina).
COS7 cells were transfected with WT PDCD10-mCherry,
D5KA-ECFP, and p110-CAAX-Myc, and were plated on 22 mm
square glass coverslips in a 6-well plate using Lipofectamine2000
asrecommendedby the manufacturer
24 hours, the cells on coverslips were fixed for 20 minutes with
4% paraformaldehyde in PBS at 25uC. Permeabilization was
conducted using 0.1% Triton X-100 in PBS for 10 min.
Nonspecific binding was blocked by incubation of coverslips for
1 h in 10% goat serum in PBS. The coverslips were incubated
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org 10July 2010 | Volume 5 | Issue 7 | e11740
with polyclonal anti-Myc antibody (Invitrogen) for 1 h and washed
with PBS. Bound primary antibodies were visualized by
incubation with FITC goat anti-rabbit antibody for 1 h, washed,
and mounted on glass slides. Imaging was performed using a Zeiss
Axiovert 200 M inverted microscope with a 125-W xenon arc
lamp (Sutter Instrument Company, Novato, CA), digital CCD
camera (CoolSNAP HQ, Roper Scientific, Tucson, AZ), and
Slidebook 5.0 software (Intelligent Imaging Innovations, Denver,
CO). An objective (636 Oil 1.25-numerical aperture, Plan-
Neofluar, Zeiss) was coupled with immersion oil to the bottom
face of glass coverslips. The images were obtained at 50 and 10 ms
exposure with 262 binning, respectively. For section analyses, the
background images from three planes were taken for each of the
three channels (CFP [a band-pass excitation filter of 436/20 nm, a
455DCLP band beamsplitter, and a band-pass emission filter of
480/40 nm], YFP [a band-pass excitation filter of 500/20 nm, a
515DCLP band beamsplitter, and a band-pass emission filter of
535/30 nm], and Cy5 [a band-pass excitation filter of 620/60 nm,
a 660DCLP band beamsplitter, and a band-pass emission filter of
700/75 nm]; Chroma). The three planes were deconvolved using
the nearest neighbor’s algorithm.
The authors thank Ashutosh Tripathy, John Sondek, Brant Hamel, Jeff
Duffy, Asya Borikova, Noah Sciaky, Stephanie Hicks, and Channing Der
for technical support and experimental assistance. We also thank Mike
Border for proof-reading of the manuscript.
Conceived and designed the experiments: JAH MM GLJ SB. Performed
the experiments: CFD JAH MM KP BRST HC JB SB. Analyzed the data:
CFD JAH MM BRST HC SB. Contributed reagents/materials/analysis
tools: GLJ SB. Wrote the paper: CFD SB.
1. Russel DS, Rubinstein LJ (1989) Pathology of tumors of the nervous system, 5th
ed. Williams and Wilkins, Baltimore. pp 730–736.
2. Rigamonti D, Hadley MN, Drayer BP, Johnson PC, Hoenig-Rigamonti K, et al.
(1988) Cerebral cavernous malformations: incidence and familial occurrence.
N Engl J Med 319: 343–347.
3. Robinson JR, Awad IA, Little JR (1991) Natural history of the cavernous
angiomas. J Neurosurg 75: 709–714.
4. Labauge P, Laberge S, Brunereau L, Levy C, Maciazek J, et al. (1998)
Hereditary cerebral cavernous angiomas: clinical and genetic features in 57
French families. Lancet 352: 1892–1897.
5. Otten P, Pizzolato GP, Rilliet B, Berney J (1989) 131 cases of cavernous angioma
(cavernomas) of the CNS, discovered by retrospective analysis of 24,535
autopsies. Neurochirurgie 35: 82–83 and 128–131.
6. Del Curling O, Jr., Kelly DL, Jr., Elster AD, Craven TE (1991) An analysis of
the natural history of cavernous angiomas. J Neurosurg 75: 702–708.
7. Dubovsky J, Zabramski JM, Kurth J, Spetzler RF, Rich SS, et al. (1995) A gene
responsibl for cavernous malformations of the brain maps to chromosome 7q.
Hum Mol Genet 4: 453–458.
8. Craig HD, Gunel M, Cepeda O, Johnson EW, Ptacek L, et al. (1998) Multilocus
linkage identifies two new loci for a Mendelian form of stroke, cerebral
cavernous malformation, at 7p15-13 and 3q25.2-27. Hum Mol Genet 7:
9. Dupre N, Verlaan DJ, Hand CK, Laurent SB, Turecki G, et al. (2003) Linkage
to the CCM2 locus and genetic heterogeneity in familial cerebral cavernous
malformation. Can J Neurol Sci 30: 122–128.
10. Zhang J, Clatterbuck RE, Rigamonti D, Dietz HC (2000) Mutations in KRIT1
in familial cerebral cavernous malformations. Neurosurgery 46: 1272–1277.
11. Zhang J, Clatterbuck RE, Rigamonti D, Chang DD, Dietz HC (2001)
Interaction between Krit1 and Icap1 infers perturbation of integrin beta1-
mediated angiogenesis in the pathogenesis of cerebral cavernous malformation.
Hum Mol Genet 10: 2953–2960.
12. Zawistowski JS, Serebriiskii IG, Lee MF, Golemis EA, Marchuk DA (2002)
KRIT1 association with the integrin-binding protein ICAP-1: a new direction in
the elucidation of cerebral cavernous malformations (CCM1) pathogenesis.
Hum Mol Genet 11: 389–396.
13. Whitehead KJ, Plummer NW, Adams JA, Marchuk DA, Li DY (2004) Ccm1 is
required for arterial morphogenesis: implications for the etiology of human
cavernous malformations. Development 131: 1437–1448.
14. Denier C, Goutagny S, Labauge P, Krivosic V, Arnoult M, et al. (2004)
Mutations within the MGC4607 gene cause cerebral cavernous malformations.
Am J Hum Genet 74: 326–337.
15. Verlaan DJ, Laurent SB, Rochefort DL, Liquori CL, Marchuk DA, et al. (2004)
CCM2 mutations account for 13% of cases in a large collection of kindreds with
hereditary cavernous malformations. Ann Neurol 55: 757–758.
16. Liquori CL, Berg MJ, Siegel AM, Huang E, Zawistowski JS, et al. (2003)
Mutations in a gene encoding a novel protein containing a phosphotyrosine-
binding domain cause type 2 cerebral cavernous malformations. Am J Hum
Genet 73: 1459–1464.
17. Bergametti F, Denier C, Labauge P, Arnoult M, Boetto S, et al. (2005)
Mutations within the Programmed Cell Death 10 Gene Cause Cerebral
Cavernous Malformations. Am J Hum Genet 76: 42–51.
18. Zabramski JM, Henn JS, Coons S (1999) Pathology of cerebral vascular
malformations. Neurosurg Clin N Am 10: 395–410.
19. Eerola I, Plate KH, Spiegel R, Boon LM, Mulliken JB, et al. (2000) KRIT1 is
mutated in hyperkeratotic cutaneous capillary-venous malformation associated
with cerebral capillary malformation. Hum Mol Genet 9: 1351–1355.
20. Couteulx SL, Brezin AP, Fontaine B, Tournier-Lasserve E, Labauge P (2002) A
novel KRIT1/CCM1 truncating mutation in a patient with cerebral and retinal
cavernous angiomas. Arch Ophthalmol 120: 217–218.
21. Davenport WJ, Siegel AM, Dichgans J, Drigo P, Mammi I, et al. (2001) CCM1
gene mutations in families segregating cerebral cavernous malformations.
Neurology 56: 540–543.
22. Clatterbuck RE, Cohen B, Gailloud P, Murphy K, Rigamonti D (2002)
Vertebral hemangiomas associated with familial cerebral cavernous malforma-
tion: segmental disease expression. Case report. J Neurosurg 97: 227–230.
23. Guclu B, Ozturk AK, Pricola KL, Bilguvar K, Shin D, et al. (2005) Mutations in
apoptosis-related gene, PDCD10, cause cerebral cavernous malformation 3.
Neurosurgery 57: 1008–1013.
24. Chen PY, Chang WS, Chou RH, Lai YK, Lin SC, et al. (2007) Two non-
homologous brain diseases-related genes, SERPINI1 and PDCD10, are tightly
linked by an asymmetric bidirectional promoter in an evolutionarily conserved
manner. BMC Mol Biol 8: 2.
25. Voss K, Stahl S, Hogan BM, Reinders J, Schleider E, et al. (2009) Functional
analyses of human and zebrafish 18-amino acid in-frame deletion pave the way
for domain mapping of the cerebral cavernous malformation 3 protein. Hum
Mutat 30: 1003–11.
26. Chen L, Tanriover G, Yano H, Friedlander R, Louvi A, et al. (2009) Apoptotic
functions of PDCD10/CCM3, the gene mutated in cerebral cavernous
malformation 3. Stroke 40: 1474–81.
27. Voss K, Stahl S, Schleider E, Ullrich S, Nickel J, et al. (2007) CCM3 interacts
with CCM2 indicating common pathogenesis for cerebral cavernous malfor-
mations. Neurogenetics 8: 249–56.
28. Hilder TL, Malone MH, Bencharit S, Colicelli J, Haystead TA, et al. (2007)
Proteomic identification of the cerebral cavernous malformation signaling
complex. J Proteome Res 6: 4343–4355.
29. Jiang BH, Liu LZ (2009) PI3K/PTEN signaling in angiogenesis and
tumorigenesis. Adv Cancer Res 102: 19–65.
30. Wang K, Horst JA, Cheng G, Nickle DC, Samudrala R (2008) Protein meta-
functional signatures from combining sequence, structure, evolution, and amino
acid property information. PLoS Comput Biol 4: e1000181.
31. Ginalski K (2006) Comparative modeling for protein structure prediction. Curr
Opin Struct Biol 2006;16: 172–7.
32. Martı ´-Renom MA, Ilyin VA, Sali A (2001) DBAli: a database of protein
structure alignments. Bioinformatics 2001;17: 746–7.
33. Kryshtafovych A, Milostan M, Szajkowski L, Daniluk P, Fidelis K (2005) CASP6
data processing and automatic evaluation at the protein structure prediction
center. Proteins 2005;61 Suppl 7: 19–23.
34. Ginalski K, Elofsson A, Fischer D, Rychlewski (2003) 3D-Jury: a simple
approach to improve protein structure predictions. Bioinformatics 19:
35. Izard T, Evans G, Borgon RA, Rush CL, Bricogne G, et al. (2004) Vinculin
activation by talin through helical bundle conversion. Nature 427: 171–175.
36. Prutzman KC, Gao G, King ML, Iyer VV, Mueller GA, et al. (2004) The focal
adhesion targeting domain of focal adhesion kinase contains a hinge region that
modulates tyrosine 926 phosphorylation. Structure 12: 881–891.
37. Sutton RB, Fasshauer D, Jahn R, Brunger AT (1998) Crystal structure of a
SNARE complex involved in synaptic exocytosis at 2.4 A resolution. Nature 395:
38. Wu S, Zhang Y (2007) LOMETS: A local meta-threading-server for protein
structure prediction. Nucleic Acids Research 35: 3375–3382.
39. Zhang Y (2008) I-TASSER server for protein 3D structure prediction. BMC
Bioinformatics, vol 9. 40 p.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org11July 2010 | Volume 5 | Issue 7 | e11740
40. Hung L-H, Ngan S-C, Liu T, Samudrala R (2005) PROTINFO: New Download full-text
algorithms for enhanced protein structure prediction. Nucleic Acids Research
41. Levitt M, Hirshberg M, Sharon R, Daggett V (1995) Potential energy function
and parameters for simulations of the molecular dynamics of proteins and
nucleic acids in solution. Comp Phys Comm 91: 215–231.
42. Canutescu AA, Shelenkov AA, Dunbrack RL (2003) A graph theory algorithm
for protein side-chain prediction. Protein Science 12: 2001–2014.
43. Samudrala R, Moult J (1997) An all-atom distance-dependent conditional
probability discriminatory function for protein structure prediction. J Mol Biol
44. Guntert P (2004) Automated NMR structure calculation with CYANA. Methods
Mol Biol 278: 353–378.
45. Fernandez-Recio J, Totrov M, Skorodumov C, Abagyan R (2005) Optimal
docking area: a new method for predicting protein-protein interaction sites.
Proteins 58: 134–143.
46. Parry DA, Fraser RD, Squire JM (2008) Fifty years of coiled-coils and alpha-
helical bundles: a close relationship between sequence and structure. J Struct
Biol 163: 258–69.
47. Gruber M, Lupas AN (2003) Historical review: another 50th anniversary–new
periodicities in coiled coils. Trends Biochem Sci 2003 Dec;28(12): 679–85.
48. Becker Y (2007) HIV-1 gp41 heptad repeat 2 (HR2) possesses an amino acid
domain that resembles the allergen domain in Aspergillus fumigatus Asp f1
protein: review, hypothesis and implications. Virus Genes 34: 233–40.
49. Oganesyan V, Oganesyan N, Adams PD, Jancarik J, Yokota HA, et al. (2005)
Crystal structure of the ‘‘PhoU-like’’ phosphate uptake regulator from Aquifex
aeolicus. J Bacteriol 187: 4238–44.
50. Saraf MC, Moore GL, Maranas CD (2003) Using multiple sequence correlation
analysis to characterize functionally important protein regions. Protein Eng 16:
51. Derewenda ZS (2004) Rational protein crystallization by mutational surface
engineering. Structure (Camb) 12: 529–535.
52. Baud F, Karlin S (1999) Measures of residue density in protein structures. Proc
Natl Acad Sci USA 96: 12494–12499.
53. Conte LL, Chothia C, Janin J (1999) The atomic structure of protein-protein
recognition sites. J Mol Biol 285: 2177–2198.
54. Pickett S, Sternberg M (1993) Empirical Scale of Side-Chain Conformational
Entropy in Protein Folding. J Mol Biol 231(3): 825–839.
55. Ford MG, Mills IG, Peter BJ, Vallis Y, Praefcke GJ, et al. (2002) Curvature of
clathrin-coated pits driven by epsin. Nature 419: 361–6.
56. Ford MG, Pearse BM, Higgins MK, Vallis Y, Owen DJ, et al. (2001)
Simultaneous binding of PtdIns(4,5)P2 and clathrin by AP180 in the nucleation
of clathrin lattices on membranes. Science 291: 1051–5.
57. Lemmon MA (2008) Membrane recognition by phospholipid-binding domains.
Nature Rev Mole Cell Biol 9: 99–111.
58. Itoh T, De Camilli P (2006) BAR, F-BAR (EFC) and ENTH/ANTH domains in
the regulation of membrane-cytosol interfaces and membrane curvature.
Biochim. Biophys. Acta 1761: 897–912.
59. Peter BJ, Kent HM, Mills IG, Vallis Y, Butler PJ, et al. (2004) BAR domains as
sensors of membrane curvature: the amphiphysin BAR structure. Science 303:
60. Weissenhorn W (2005) Crystal structure of the endophilin-A1 BAR domain.
J Mol Biol 351: 653–661.
61. Shimada A, Niwa H, Tsujita K, Suetsugu S, Nitta K, et al. (2007) Curved EFC/
F-BAR-domain dimers are joined end to end into a filament for membrane
invagination in endocytosis. Cell 129: 761–72.
62. Millard TH, Bompard G, Heung MY, Dafforn TR, Scott DJ, et al. (2005)
Structural basis of filopodia formation induced by the IRSp53/MIM homology
domain of human IRSp53. EMBO J 24: 240–50.
63. Rodriguez-Viciana P, Warne PH, Khwaja A, Marte BM, Pappin D, et al. (1997)
Role of phosphoinositide 3-OH kinase in cell transformation and control of the
actin cytoskeleton by Ras. Cell 89: 457–467.
64. McFall A, Ulku ¨ A, Lambert QT, Kusa A, Rogers-Graham K, et al. (2001)
Oncogenic Ras blocks anoikis by activation of a novel effector pathway
independent of phosphatidylinositol 3-kinase. Mol Cell Biol 21: 5488–99.
65. Murphy GA, Graham SM, Morita S, Reks SE, Rogers-Graham K, et al. (2002)
Involvement of phosphatidylinositol 3-kinase, but not RalGDS, in TC21/R-
Ras2-mediated transformation. J Biol Chem 277: 9966–75.
66. He Y, Zhang H, Yu L, Gunel M, Boggon TJ, et al. (2010) Stabilization of
VEGFR2 signaling by cerebral cavernous malformation 3 is critical for vascular
development. Sci Signal 3: ra26.
67. Borikova AL, Dibble CF, Sciaky N, Welch CM, Abell AN, et al. (2010) Rho
kinase inhibition rescues the endothelial cell cerebral cavernous malformation
phenotype. J Biol Chem 285: 11760–4.
68. Papakonstanti EA, Ridley AJ, Vanhaesebroeck B (2007) The p110delta isoform
of PI 3-kinase negatively controls RhoA and PTEN. EMBO J 26: 3050–61.
69. Gidh AV, Decker SR, Vinzant TB, Himmel ME, Williford C (2006)
Determination of lignin by size exclusion chromatography using multi angle
laser light scattering. J Chromatogr A 1114: 102–10.
70. Oliva A, Llabre ´s M, Farin ˜a JB (2001) Comparative study of protein molecular
weights by size-exclusion chromatography and laser-light scattering. J Pharm
Biomed Anal 25: 833–41.
PDCD10 Interacts with PIP3
PLoS ONE | www.plosone.org12 July 2010 | Volume 5 | Issue 7 | e11740