Download Identification of Hedgehog Pathway Components by RNAi in

RESEARCH ARTICLES
Identification of Hedgehog
Pathway Components by RNAi in
Drosophila Cultured Cells
Lawrence Lum,1 Shenqin Yao,1 Brian Mozer,2
Alessandra Rovescalli,2 Doris Von Kessler,1 Marshall Nirenberg,2
Philip A. Beachy1*
Classical genetic screens can be limited by the selectivity of mutational targeting, the complexities of anatomically based phenotypic analysis, or difficulties in subsequent gene identification. Focusing on signaling response to the
secreted morphogen Hedgehog (Hh), we used RNA interference (RNAi) and a
quantitative cultured cell assay to systematically screen functional roles of all
kinases and phosphatases, and subsequently 43% of predicted Drosophila
genes. Two gene products reported to function in Wingless ( Wg) signaling were
identified as Hh pathway components: a cell surface protein (Dally-like protein)
required for Hh signal reception, and casein kinase 1␣, a candidate tumor
suppressor that regulates basal activities of both Hh and Wg pathways. This
type of cultured cell– based functional genomics approach may be useful in the
systematic analysis of other biological processes.
The secreted protein signal Hedgehog (Hh)
elicits cellular proliferation and differentiation responses during normal embryonic development, and inappropriate pathway activation can contribute to tumorigenesis. In
Drosophila, this pathway is regulated by a
series of repressive interactions between protein components that ultimately result in gene
activation mediated by the transcription factor Cubitus interruptus (Ci) (Fig. 1A ) (1, 2).
Ci is regulated by a cytoplasmic complex
consisting of the kinesin-like protein Costal 2
(Cos2), the serine-threonine kinase Fused
(Fu), and Suppressor of fused [Su(fu)], a
protein that lacks known functional motifs.
This complex prevents activation and nuclear
localization of Ci and stimulates its proteolytic processing to a truncated form (Ci75)
that represses gene targets. The activity of
this complex is suppressed by Smoothened
(Smo), a seven-transmembrane protein, and
Smo activity in turn is suppressed by catalytic
action of the transporter-like protein Patched
(Ptc). Hh protein releases these sequential
repressive interactions by binding and inactivating Ptc, thus permitting Smo-mediated
suppression of the regulatory complex and
releasing Ci for activation of target genes.
Classical genetic approaches to the study of
embryonic processes such as Hh signaling have
been subject to limitations imposed by the selectivity of mutagenesis methods, by the diffi1
Department of Molecular Biology and Genetics,
Howard Hughes Medical Institute, Johns Hopkins University School of Medicine, Baltimore, MD 21205,
USA. 2Laboratory of Biochemical Genetics, NHLBI,
NIH, Bethesda, MD 20892, USA.
*To whom correspondence should be addressed. Email: [email protected]
culty of identifying mutations whose zygotic
phenotypes are cloaked by maternal contributions, and by difficulties in identifying mutated
genes. The Hh pathway suffers from the additional complication that its embryonic loss-offunction phenotype is similar to that controlled
by the signaling pathway regulated by the secreted protein Wingless (Wg), thus hindering
correct assignment of gene function. Like Hh,
Wg signaling employs a series of repressive
interactions in which receptor activity antagonizes a cytoplasmic complex that in turn causes
degradation of Armadillo (Arm), a key component in the activation of the Wg transcriptional
response (Fig. 1A). Indeed, certain components
of these pathways are shared, including the
F-box protein Slimb (Sli) (3) and glycogen
synthase kinase 3 (GSK3␤ or Sgg) (4, 5), each
with dual roles in the proteolytic degradation of
Arm and in the proteolytic formation of Ci75.
In addition, a casein kinase 1 (CK1) activity
triggers these proteolytic events (5), although
the actual CK1 family member that functions in
Ci75 formation remains to be identified.
RNAi in Hh and Wg cultured cell assays. To identify additional components in Hh
signal response while also circumventing the
difficulties of classical genetic screens, we used
a cultured cell assay (6) and RNA interference
(RNAi)–mediated disruption of gene function
(7, 8) to systematically screen Drosophila
genes. This assay is based on transfection of
wing imaginal disc– derived cl-8 cells with a
control reporter and a Hh-responsive luciferase
reporter. Unlike Drosophila S2 cultured cells,
simply bathing cl-8 cells in double-stranded
RNA (dsRNA) had no effect on Hh pathway
response (9). However, transfection of dsRNA
together with both reporter constructs affected
pathway response in such a way that RNAi
targeting of the positive regulatory components
Smo, Fu, and Ci inhibited response to the Hh
signal (Fig. 1B), and targeting of the negative
regulatory components Cos2 and Ptc resulted
either in basal activation or enhanced responsiveness to Hh (10, 11). Consistent effects on
the protein levels of several pathway components, including Smo and Ptc, were also observed in S2 cells after treatment with dsRNA
(Fig. 1B, inset). RNAi in Drosophila cultured
cells thus provides a functional test for gene
products of known or predicted sequence. The
Hh signaling assay in cl-8 cells is quantitative
and is specific for cellular response, because the
addition of exogenous Hh protein eliminates
the requirement for functions involved in Hh
protein synthesis or distribution.
We also found that multiple dsRNA species
could be combined in these transfection experiments, facilitating large-scale screening and tests
of gene interactions and epistasis. RNAi of
Su(fu) in cl-8 cells, for example, produced little
effect on Hh response but reversed the reduction
in responsiveness elicited by RNAi of Fu (Fig.
1B), thus mimicking phenotypic suppression of
fu mutations by Su(fu) mutations in flies (12). In
addition, combined RNAi of Cos2 and Ci yielded a loss of responsiveness (Fig. 1C), indicating
that Ci is epistatic to Cos2, in agreement with
phenotypic analysis of mutant combinations
(13). Ci remained epistatic to Cos2 even when
the amount of Cos2 dsRNA was 10 times greater
than that of Ci dsRNA (Fig. 1C). This indicates
that the RNAi machinery was not overwhelmed
by the transfection of excess dsRNA, even at
levels 10-fold higher than those routinely used.
To facilitate independent testing of gene
function in response to Wg and Hh, we also
established a Drosophila cultured cell–based
reporter assay for Wg response based on stabilization of Arm in the presence of Wg protein
(10, 14). In contrast to S2 cells, both embryoderived Kc cells and cl-8 cells were responsive
to Wg (Fig. 1D, inset) (9, 15), but the Kc cells
consistently displayed a stronger transcriptional
response (9). RNAi of positively acting components Frizzled 1 and 2 receptors (Fz1 and Fz2),
Arm, or Pangolin [Pan, the Drosophila T cell–
specific transcription factor (Tcf) homolog]
caused loss of Wg responsiveness, and RNAi of
GSK3␤ increased the response to Wg (Fig. 1D).
Taken together, RNAi of 11 known components
of the Hh and Wg pathways produced the effects predicted by classical genetic analyses,
indicating that RNAi in these assays can provide
a rapid and reliable indication of gene function.
Genomewide RNAi screen for kinases
and phosphatases in Hh signaling. To test
this assay system and establish high-throughput methods for dsRNAi synthesis and
screening, a library was prepared containing
dsRNAs corresponding to all kinases and
phosphatases predicted from the completed
Drosophila genome sequence (fig. S1A and
table S2) (10, 16). The initial screen identi-
www.sciencemag.org SCIENCE VOL 299 28 MARCH 2003
2039
RESEARCH ARTICLES
fied several dsRNA pools that showed either
a gain in basal reporter activity in the absence
of Hh (Fig. 2, top panel), as expected for
components with a predominantly negative
regulatory role in the pathway, or a reduction
in average fold induction (the ratio of reporter
activity in the presence and absence of Hh;
Fig. 2, bottom panel), as expected for components with either positive or negative pathway-regulatory roles (10). When pooled
RNAs and individual dsRNAs were rescreened, three dsRNAs emerged that had
reproducible effects on pathway activity.
These dsRNAs targeted the putative serinethreonine kinase Fu and cAMP-dependent
kinase 1 (PKA-C1), two known regulators of
Hh signaling (1, 2). CK1␣, not previously
implicated in Hh signaling, was also identified in the screen, because it was required to
maintain the regulated state of the pathway.
RNAi of CK1␣ elevated basal reporter activ-
ity in the absence of Hh (Fig. 2, left in inset)
and expanded the segmental domain of Wg
expression in embryos injected with CK1␣
dsRNA (Fig. 2, right in inset) (10).
Systematic screen of Drosophila Gene
Collection Release 1. The screen was extended with a dsRNA library based on the
Drosophila Gene Collection Release 1
(DGCr1), which contains approximately 43%
of the predicted genes in the Drosophila genome (Fig. 3A and fig. S1B) (10). When pooled
RNAs and individual dsRNAs were rescreened,
20 dsRNAs with reproducible effects on apparent pathway activity were identified, and these
effects were confirmed with dsRNA from distinct sets of gene-specific primers (10). Nineteen of these genes [not including thread (Fig.
3)] were classified into three groups (Fig. 3B).
Class I genes encoding proteins with known
roles in Hh response include Smo, Cos2, PKA,
and combgap, which encodes a zinc finger pro-
Fig. 1. RNAi in Drosophila cultured cell assays for Hh and Wg signaling.
(A) Schematic view of Hh and Wg signaling pathways. For clarity, only
components tested by RNAi in (B) and (D) are shown. Green or red text
denotes predominantly positive or negative effects of a particular component in the pathway response, respectively. (B) Effects of RNAi of
known pathway components on Hh response in cl-8 cells (10). Su(fu)
RNAi in combination with Fu RNAi reverses the loss of Hh response
caused by Fu RNAi alone. yfp, yellow fluorescent protein. (Inset) Western
blot analysis of Ptc protein levels after ptc dsRNA treatment. RNAi of ptc
in S2 cells, which do not express Ci, results in a decrease in Ptc protein
levels and stabilization of Smo protein (42). A Western blot of kinesin
2040
tein that regulates Ci expression (17). Our
screen identified all genes within DGCr1 that
are known to regulate expression of Ptc, the Hh
pathway target on which our reporter is based.
Our screen thus is highly reliable and is not
affected by pooling of dsRNAs.
The 11 genes in class II encode individual
components of complexes with broad cellular
roles that are not restricted to Hh signaling. This
group includes nine genes encoding proteins
that likely affect ribosome function, which,
when targeted by RNAi, reduced reporter activity, probably because of loss of luciferase
expression in response to Hh. Two additional
proteins, DeltaCOP and BetaCOP, are involved
in retrograde vesicle trafficking from the Golgi
complex to the endoplasmic reticulum. RNAi
targeting of these proteins increased basal reporter activity, but the effects were weaker and
more variable than that produced by RNAi of
Cos2. Other components of the COP1 complex
heavy chain (Khc) is shown as a control for protein loading. (C) Epistasis
analysis of Ci relative to Cos2 in cl-8 cells. RNAi of Ci inhibited Hh
induction of reporter activity, whereas RNAi of Cos2 increased basal
reporter activity. The epistatic effect exerted by RNAi of Ci on RNAi of
Cos2 prevailed, even at levels of cos2 dsRNA 10-fold higher than those
of ci dsRNA. (D) Effects on Wg response of RNAi of pathway components
in Kc cells. RNAi effects of known Wg pathway components on Wg
response were monitored with an assay similar to that described for Hh
response using the Wg-responsive reporter Super TopFlash (10). The
inset shows the accumulation of Arm in Kc cells after 2 hours of
incubation in the presence of Wg (in duplicate).
28 MARCH 2003 VOL 299 SCIENCE www.sciencemag.org
RESEARCH ARTICLES
also produced similar variable effects when targeted (9). How these proteins function in a
pathway-regulatory trafficking event remains to
be determined.
Class III comprises four genes with previously unrecognized roles in the Hh pathway,
two of which have been characterized in other
contexts. Dally-like protein (Dlp) is a member
of the glypican family of heparan sulfate proteoglycans (HSPGs) and is required for Hh
responsiveness in cl-8 cell– based assay and
maintenance of normal Wg expression in embryos (Fig. 3A, inset). CK1␣ was also identified in the kinase and phosphatase library
screen as a negative regulator of Hh pathway
activity. A potential role for HSPGs in Hh
signaling is suggested by binding of the Hh
signaling domain to heparin (18) and by the
requirement in Hh signaling for tout velu (ttv),
which encodes a heparan sulfate polymerase
that is important in synthesis of the glycosaminoglycan (GAG) chains of HSPGs (19). To
characterize the requirement for Dlp in Hh signal response, we examined the roles all four
known or predicted Drosophila HSPGs, including a second glypican, Dally, Syndecan (Sdc),
and Perlecan (Pcan). Only RNAi of Dlp affected Hh signal response (Fig. 4A) (20), and the
Fig. 2. An RNAi screen of Drosophila kinases and phosphatases in Hh signaling. The effects of
dsRNA corresponding to all known and predicted kinases and phosphatases in the Drosophila
genome are plotted as basal luciferase activity (top) and fold induction by Hh (bottom). STDEV,
standard deviation. Data points represent averages of three independent Hh signaling assays using
pools of three dsRNAs, with each dsRNA targeting a separate transcript. Cutoffs for selection of
dsRNA pools for identification of the gene of interest are indicated by the horizontal dotted lines
(10). Data points in which a single gene could be identified are colored as indicated in the figure
and labeled with the gene name. Controls included in the screen were dsRNA targeting Smo, Cos2,
and a genomic noncoding sequence (700 base pairs in length). Fu and PKA-C1, kinases that are
known to affect the pathway, were identified in this screen. In addition, a dsRNA pool targeting
CK1␣ and CK1ε induced basal reporter activity (red dot in upper panel). (Inset) Examination of
individual dsRNAs in this pool revealed that a decrease in the expression of CT6528, a transcript
encoding CK1␣, resulted in activation of the Hh pathway (29). Injection of this ck1␣ dsRNA into
preblastoderm Drosophila embryos resulted in an expansion of Wg expression domains at the
extended germ band stage, consistent with a role in regulating basal Hh pathway activity.
effect was comparable to RNAi of Smo. Expression of Dlp also increased the response to
Hh, comparable to the increase caused by expression of Ci (9).
Dlp, a cell surface HSPG, is required for
reception of the Hh signal. The glypican
Dlp has 14 conserved cysteines that are predicted to form an N-terminal globular domain and
several putative GAG modification sites next to
a consensus glycosylphosphatidylinositol (GPI)
attachment sequence. A monoclonal antibody
directed against the juxtamembrane portion of
the protein (10) revealed Dlp at the surface of
cl-8 cells (Fig. 4B). The sensitivity of Dlp to
treatment with heparinase III, but not chondroitinase ABC, indicates modification by heparan sulfate (Fig. 4B). Treatment with heparinase
III reduced the broad electrophoretic mobility of
Dlp to a more compact band that migrates faster
than the predicted molecular weight of the mature protein (⬃78 kD), suggesting that, like
other glypicans, maturation of Dlp may involve
proteolytic processing (21, 22).
The heparan sulfate modification of Dlp
could involve activity of the Ttv heparan
sulfate polymerase, whose protein targets are
unknown. Loss of Ttv function, however,
primarily affects movement of the Hh protein
signal through target tissues (18, 23), whereas
our assays show that Dlp plays a cell-autonomous role in response to the Hh signal.
Although it remains possible that Dlp could
also have a role in extracellular Hh transport,
perhaps alongside of other HSPGs, the cellautonomous function of Dlp in signal response is distinct from that of Ttv targets in
Hh transport. Consistent with an extracellular
transport role for the GAG chains elaborated
by Ttv, we failed to observe an effect on Hh
response when RNAi of Ttv or its relatives
Sotv (Dext2) and Botv (Dext3) were applied,
either singly or in combinations (9). Similarly, although RNAi of Dally or Dlp produced
segment polarity defects in embryos, there
was no effect in the Kc cell Wg assay (9),
suggesting that the previously reported roles
of these proteins in Wg signaling (24–26)
may be largely restricted to non–cell-autonomous extracellular effects on Wg transport.
To further investigate the mechanism of
Dlp action on Hh signal response, we examined RNAi of Dlp in combination with RNAi
of other pathway components. Dlp function
was not required for the increased basal pathway activity produced by loss of Cos2 (Fig.
4C). Moreover, the requirement for Dlp in Hh
response was suppressed by RNAi of Ptc,
suggesting that Dlp may act upstream or at
the level of the Ptc receptor (Fig. 4C). Given
its localization, Dlp could function to concentrate Hh on the surface of responding cells,
perhaps aiding in the delivery of Hh to Ptc.
Preliminary biochemical analysis shows that
a soluble fusion protein containing the extracellular domain of Dlp can associate with a
www.sciencemag.org SCIENCE VOL 299 28 MARCH 2003
2041
RESEARCH ARTICLES
tagged form of Hh (9). Also consistent with
such a role, RNAi of Dlp did not block signal
response when Hh was expressed in responding cells (Fig. 4D). In this case, mature Hh is
anchored to the plasma membrane via its
lipid modifications (27, 28). Circumvention
of the requirement for Dlp by expression of a
tethered Hh signal in responding cells suggests that Dlp may function normally to concentrate Hh released from producing cells.
A dual role for CK1␣ in regulation of
Hh and Wg pathway activity. In contrast
to the positive role of Dlp in Hh response, our
initial observations from the kinase-phospha-
tase and DGCr1 library screens suggested that
CK1␣ may control basal pathway activity
(Figs. 2 and 3). A role for kinases other than
PKA in regulation of Ci processing was proposed (6), and CK1 sites essential for processing Ci to Ci75 were reported (5). However, no
CK1 gene required for Hh pathway regulation
has been identified, although the gene encoding
CK1ε was proposed on the basis of the effects
of CK1ε overexpression in imaginal discs (5).
To investigate the role of CK1␣ in Hh pathway
regulation, RNAi of CK1␣ and CK1ε in cl-8
cells were compared. Three dsRNAs containing
distinct portions of the CK1␣ open reading
frame (ORF) increased basal reporter activity
(29). DsRNA corresponding to either the catalytic region or the extracatalytic region of CK1ε
had no effect (Fig. 5A). In contrast, overexpression of either CK1␣ or CK1ε suppressed Hhinduced pathway activation (Fig. 5B). These
results indicate that both CK1␣ and CK1ε affect basal pathway activity when overexpressed, but that physiological regulation depends primarily on CK1␣.
Both CK1␣ and CK1ε have been implicated
in Wnt pathway signaling in vertebrates, largely
on the basis of overexpression studies (30). In
the Wg signaling assay, only overexpression of
Fig. 3. RNAi screen of
43% of predicted Drosophila genes for components of the Hh pathway. (A) A screen of the
DGCr1 dsRNA library.
The upper panel shows
relative basal luciferase
activity and the lower
panel shows relative
fold induction, as described in the Fig. 2 legend. (Inset) Dlp was
identified as a positive
component in pathway
activation in a dsRNA
pool that also included
dsRNA targeting Mbt
and Mus210 transcripts.
Injection of dlp dsRNA
into Drosophila embryos resulted in decreased
Wg expression, consistent with a reduction in
response to Hh signaling. (B) Genes and gene
products identified in
the screen are grouped
into three different
classes: known pathway
components and regulators (class I), components of large multifunctional
protein
complexes that are
not likely to be involved exclusively in
Hh signaling (class II),
and genes with previously unrecognized
roles in the Hh pathway (class III). The apparent effect of RNAi
of thread, an antiapoptosis gene, is unreliable because of a dramatic reduction in control reporter activity, indicative
of cell death.
2042
28 MARCH 2003 VOL 299 SCIENCE www.sciencemag.org
RESEARCH ARTICLES
CK1ε elevated basal pathway activity (Fig. 5D,
right panel). In contrast, RNAi targeting revealed
that only loss of CK1␣ function increased basal
pathway activity (Fig. 5E, left panel), indicating
that physiological regulation of basal Wg pathway activity depends on CK1␣. The increase in
pathway activity produced by RNAi of CK1␣
was blocked by RNAi of Arm, consistent with
the proposed role of CK1␣ in Arm degradation
(31, 32). Similarly, Ci function is required for
basal activation of the Hh pathway by RNAi of
CK1␣, whereas other positively acting components Smo and Fu were not (Fig. 5C, left
panel). The increase in basal reporter activity
produced by RNAi of CK1␣ was enhanced by
RNAi of Sli, which is thought to be involved in
targeting Ci for processing to Ci75 (Fig. 5C,
right panel). These results suggest that CK1␣
acts downstream of Smo and Fu, and at or
upstream of Ci. The role of CK1␣ in maintaining the quiescent state of both Wg and Hh
pathways suggests that these pathways might
be simultaneously activated upon loss of CK1␣
activity (Fig. 5D, table).
In addition to Dlp and CK1␣, class III
components with previously unrecognized
roles in Hh signaling include caupolican,
one of a cluster of homeodomain genes
with roles in specifying wing vein and sensory organ pattern (33). A closely related
gene within the cluster, araucan, also reduced Hh-induced reporter activity when targeted by RNAi (9). Curiously, both of these
genes have been characterized as targets rather than mediators of Hh signaling. The remaining gene in class III, CG9211
(CT26314), encodes a putative cell surface
protein that shares features with a subfamily
of the immunoglobulin (Ig) Superfamily, including fibronectin type III and Ig domain
repeats (34).
Implications for human disease. The
identification of signaling pathway components
in Drosophila can have implications for human
disease. For example, the role of CK1␣ in regulating basal activity of both Wg and Hh signaling pathways suggests that it could act as a
tumor suppressor in colon cancer, basal cell
carcinoma, rhabdomyosarcoma, or medulloblastoma. These tumors are associated with inappropriate activity of one or the other pathway,
except medulloblastoma, which is associated
with the activation of either (2). In the case of
Dlp, GPC4 and GPC6 are the most closely
related of the six mammalian glypican family
members (35). GPC6 maps to 13q32 (36), a
human chromosomal locus whose deletion
(13q32 syndrome) is associated with defects,
including holoprosencephaly (HPE), anogenital
malformations, and an absent thumb (37); all of
these malformations are consistent with loss of
varying degrees of Sonic hedgehog signaling
(38, 39). If GPC6 levels are limiting in mammalian Hh responsiveness, then loss of GPC6
function may play a role in 13q32 syndrome
malformations, possibly alongside other HPE
genes in or near this region (40). Finally, mutation of CDO, the mammalian homolog of
CG9211, results in a form of HPE (41), consistent with a role for CDO in signaling.
An RNAi-based approach to functional
genomics in a Drosophila cultured cell assay
has provided a rapid screen that is sufficiently
sensitive to detect known maternal and zygotic
functions in Hh signaling, as well as the functions of previously unknown pathway components. This screen, together with RNAi-based
functional characterization, has allowed identi-
Fig. 4. A role for the
cell surface HSPG Dlp
in Hh response. (A) A
specific role of Dlp in
Hh signaling. Of four
characterized
Drosophila HSPGs (Sdc,
Pcan, Dally, and Dlp),
only RNAi of Dlp resulted in loss of pathway responsiveness
(left panel). Similar
effects were observed
with dsRNA corresponding to the Dlp
5⬘ UTR. YFP and Smo
dsRNAs were negative and positive controls. Conversely, expression of Dlp but
not of Dally or YFP
resulted in increased
responsiveness to Hh
(10) (right panel). (B)
Dlp is a cell surface–
localized,
heparan
sulfate–modified protein. (Top) A diagram of the Dlp precursor indicating the 14 conserved
cysteines in the globular domain, several putative GAG attachment
sites, a hydrophobic GPI attachment sequence, and the region recognized by the monoclonal antibody to Dlp (10). (Left) Immunofluorescence analysis of nonpermeabilized or detergent-permeabilized cl-8
cells expressing Dlp and green fluorescent protein (GFP) (10). Epitopes
for both Dlp and GFP were accessible in permeabilized cells, whereas
only Dlp protein was detected in nonpermeabilized cells. (Right) Analysis of the GAG modifications of Dlp in S2 cells. Control Arm protein (left lane) or Dlp (right lanes) was immunoprecipitated and treated before
Western blot analysis with heparinase III (HepIII), chondroitinase ABC (ChABC), and the reducing agent dithiothreitol (DT T) as indicated (10). The
vertical line indicates migration of the heparinase III–treated juxtamembrane subunit of Dlp in DT T-treated samples. Heparinase III–treated Dlp
migrates substantially more slowly in the absence of DT T (arrow), suggesting that Dlp undergoes proteolytic processing to yield a disulfide-linked
product. Background immunoglobulin G bands are indicated by asterisks. (C) Mapping of Dlp activity within the Hh pathway. Pathway activation
resulting from RNAi of Ptc or Cos2 was not affected by RNAi of Dlp, placing Dlp function upstream of both components. (D) Dlp is not required when
Hh is expressed within responding cells. Hh responsiveness in cl-8 cells was measured when Hh was supplied either in conditioned medium or by
transfection of a construct for the expression of full-length Hh. Hh response was not affected by loss of Dlp in Hh-transfected cells, demonstrating
that delivery of membrane-tethered Hh directly into responding cells circumvents the requirement for Dlp.
www.sciencemag.org SCIENCE VOL 299 28 MARCH 2003
2043
RESEARCH ARTICLES
Fig. 5. CK1␣ regulates basal activity of both
the Hh and Wg pathways. (A) Specificity of
dsRNA targeting CK1␣. (Left) Schematic diagram of the different dsRNAs targeting CK1␣
and its closest homolog CK1ε. Highly homologous regions and nucleotides corresponding
to the catalytic domains are indicated. CT
numbers corresponding to transcripts and
lengths of exon sequences targeted by RNAi
are noted for dsRNAs included in the kinase/
phosphatase library (thin line); only the length
of the targeted sequence is noted for dsRNAs
derived from a separate synthesis (thick line) (10). (Right)
Transfection of three out of four dsRNAs targeting CK1␣ but
neither of the two dsRNAs targeting CK1ε results in pathway
activation. Three additional dsRNAs targeting CK1ε (left panel,
thin lines that are not numbered) were included in the screen
described in Fig. 2 and had no effect on pathway responsiveness. (B) Expression of CK1␣ or CK1ε suppresses the Hh pathway. CK1␣, CK1ε, or known pathway components were expressed in cl-8 cells and pathway response was measured.
Expression of both CK1␣ and CK1ε blocked pathway response.
(C) Mapping of CK1␣ activity within the Hh pathway. Pathway
components were targeted by RNAi individually or in combination with RNAi of CK1␣. Only Ci was epistatic to CK1␣. RNAi
of Sli, an F-box protein that regulates Ci proteolytic processing,
in combination with RNAi of CK1␣ increased both basal and Hh-induced pathway activity more than did either individually. (D) CK1␣ also acts
as a negative regulator of the Wg pathway. RNAi of CK1␣ resulted in increased basal activity of the Wg pathway, whereas RNAi of Arm resulted
in loss of pathway responsiveness. RNAi of Arm was epistatic to RNAi of CK1␣, consistent with CK1␣ action upstream of Arm. Expression of
CK1␣ did not suppress Wg pathway response, and expression of CK1ε activated the pathway. (Right) Summary of the effects of RNAi or
overexpression (OEX ) of CK1␣ or CK1ε in Hh and Wg pathways.
fication of CK1␣ and Dlp in Hh response. In the
case of Dlp, the influence of possible effects on
extracellular signal transport was avoided, because the assay monitors only responses within
target cells. Furthermore, although both of these
genes were previously implicated in Wg signaling, the use of a quantitative Hh signaling assay
avoided the complexities of phenotypic analysis
and elucidated roles in the Hh response. Activity
assays for Hh, Wg, and other signaling pathways in a setting amenable to RNAi-based functional genomics approaches should make possible the systematic identification of components
within each pathway, thereby providing the basis for a more complete view of the mechanisms
of signaling. In addition, a fuller knowledge of
common, as well as unique, components in Hh,
Wg, and other pathways should lead to a better
understanding of the relationships and regulatory cross-talk between pathways.
1.
2.
3.
4.
2044
References and Notes
P. W. Ingham, EMBO J. 17, 3505 (1998).
J. Taipale, P. A. Beachy, Nature 411, 349 (2001).
J. Jiang, G. Struhl, Cell 86, 401 (1996).
J. Jia et al., Nature 416, 548 (2002).
5. M. A. Price, D. Kalderon, Cell 108, 823 (2002).
6. C. H. Chen et al., Cell 98, 305 (1999).
7. J. C. Clemens et al., Proc. Natl. Acad. Sci. U.S.A. 97,
6499 (2000).
8. A. Fire et al., Nature 391, 806 (1998).
9. L. Lum et al., data not shown.
10. Materials and methods are available as supporting
material on Science Online.
11. In contrast to Ptc inactivation by mutation, dsRNA
directed against Ptc does not cause constitutive pathway activation, probably because RNAi-induced loss
of ptc mRNA and protein is balanced by a consequent
increase in transcription of the still intact ptc gene,
resulting in sufficient levels of Ptc protein to maintain
pathway regulation. In S2 cells, which lack Ci and
consequently this regulatory feedback, Ptc-directed
RNAi treatment results in a decrease in Ptc protein
and an increase in Smo protein similar to that produced by Hh stimulation (Fig. 1B, inset).
12. T. Preat, Genetics 132, 725 (1992).
13. A. J. Forbes, Y. Nakano, A. M. Taylor, P. W. Ingham,
Development (Suppl.) 1993, 115 (1993).
14. M. T. Veemon et al., Curr. Biol., in press.
15. F. van Leeuwen, C. H. Samos, R. Nusse, Nature 368,
342 (1994).
16. D. K. Morrison, M. S. Murakami, V. Cleghon, J. Cell
Biol. 150, F57 (2000).
17. G. L. Campbell, A. Tomlinson, Development 127,
4095 (2000).
18. J. J. Lee et al., Science 266, 1528 (1994).
19. I. The, Y. Bellaiche, N. Perrimon, Mol. Cell 4, 633
(1999).
20. Injection of dally dsRNA used in cell-based assays
resulted in dramatically weakened or in some cases
loss of Wg staining, consistent with the loss of naked
cuticle associated with loss of Dally.
21. K. Watanabe, H. Yamada, Y. Yamaguchi, J. Cell. Biol.
130, 1207 (1995).
22. H. H. Song, J. Filmus, Biochim. Biophys. Acta 1573,
241 (2002).
23. Y. Bellaiche, I. The, N. Perrimon, Nature 394, 85
(1998).
24. G. H. Baeg, X. Lin, N. Khare, S. Baumgartner, N.
Perrimon, Development 128, 87 (2001).
25. X. Lin, N. Perrimon, Nature 400, 281 (1999).
26. M. Tsuda et al., Nature 400, 276 (1999).
27. Z. Chamoun et al., Science 293, 2080 (2001).
28. J. A. Porter et al., Cell 86, 21 (1996).
29. Of the two dsRNAs targeting CK1␣ included in the
kinase-phosphatase library, only one of these targeting the ORF resulted in a gain in basal reporter
activity. The other dsRNA targets the 5⬘ untranslated
region (UTR) and is ineffective, perhaps as a result of
possible alternative splicing of the transcript.
30. R. M. McKay, J. M. Peters, J. M. Graff, Dev. Biol. 235,
388 (2001).
31. C. Liu et al., Cell 108, 837 (2002).
32. S. Yanagawa et al., EMBO J. 21, 1733 (2002).
33. J. L. Gomez-Skarmeta, J. Modolell, Genes Dev. 10,
2935 (1996).
34. J. S. Kang et al., EMBO J. 21, 114 (2002).
35. B. De Cat, G. David, Semin. Cell Dev. Biol. 12, 117
(2001).
28 MARCH 2003 VOL 299 SCIENCE www.sciencemag.org
RESEARCH ARTICLES
36. S. Paine-Saunders, B. L. Viviano, S. Saunders, Genomics 57, 455 (1999).
37. S. Brown, J. Russo, D. Chitayat, D. Warburton, Am. J.
Hum. Genet. 57, 859 (1995).
38. M. Ramalho-Santos, D. A. Melton, A. P. McMahon,
Development 127, 2763 (2000).
39. C. Chiang et al., Nature 383, 407 (1996).
40. S. A. Brown et al., Nature Genet. 20, 180 (1998).
41. F. Cole, R. S. Krauss, Curr. Biol. 13, 411 (2003).
42. N. Denef, D. Neubuser, L. Perez, S. M. Cohen, Cell
102, 521 (2000).
43. We thank A. Kaykas and R. Moon for providing the
Super TopFlash reporter; F. Weiss-Garcia and the
Sloan-Kettering Hybridoma Core Facility (NY) for
development of monoclonal antibodies; R. Gong
for technical assistance; S. Celniker for help with
annotation of DGCr1; R. L. Johnson and S. M. Cohen
for antibodies; S. Zusman at Genetic Services (MA)
for help with dally dsRNA embryo injections; and J.
Taipale for critical review of the manuscipt. Supported by grants and fellowships from NIH and a
Life Sciences Research Foundation Fellowship (L.L).
P.A.B. is an investigator of the Howard Hughes
Medical Institute.
Supporting Online Material
www.sciencemag.org/cgi/content/full/299/5615/2039/
DC1
Materials and Methods
Fig. S1
Tables S1 and S2
References
11 December 2002; accepted 6 March 2003
R EPORTS
Polymer Replicas of Photonic
Porous Silicon for Sensing and
Drug Delivery Applications
Yang Yang Li,1 Frédérique Cunin,1 Jamie R. Link,1 Ting Gao,1
Ronald E. Betts,1 Sarah H. Reiver,1 Vicki Chin,2
Sangeeta N. Bhatia,2 Michael J. Sailor1
Elaborate one-dimensional photonic crystals are constructed from a variety of
organic and biopolymers, which can be dissolved or melted, by templating the
solution-cast or injection-molded materials in porous silicon or porous silicon
dioxide multilayer (rugate dielectric mirror) structures. After the removal of the
template by chemical dissolution, the polymer castings replicate the photonic
features and the nanostructure of the master. We demonstrate that these
castings can be used as vapor sensors, as deformable and tunable optical filters,
and as self-reporting, bioresorbable materials.
Synthesis of materials using nanostructured
templates has emerged as a useful and versatile
technique to generate ordered nanostructures
(1). Templates consisting of microporous membranes (2, 3), zeolites (4), and crystalline colloidal arrays (5–7) have been used to construct
elaborate electronic, mechanical, or optical
structures. Porous Si is an attractive candidate
for use as a template (8) because the porosity
and average pore size can be tuned by adjusting
the electrochemical preparation conditions that
allow the construction of photonic crystals,
dielectric mirrors, microcavities, and other
optical structures (9). For many applications,
porous Si is limited by its chemical and mechanical stability. The use of porous Si as a
template eliminates these issues while providing the means for construction of complex
optical structures from flexible materials that
are compatible with biological systems or
harsh environments.
Multilayered porous Si templates containing
nanometer-scale pores are prepared (10) by an
1
Department of Chemistry and Biochemistry, University of California, San Diego, 9500 Gilman Drive,
Department 0358, La Jolla, CA 92093– 0358, USA.
2
Department of Bioengineering, University of California, San Diego, 9500 Gilman Drive, Department 0412,
La Jolla, CA 92093– 0412, USA.
anodic electrochemical etch of crystalline silicon
wafers with the use of a pseudosinusoidal current-time waveform, according to published procedures (9, 11–16). The thickness, pore size, and
porosity of a given layer is controlled by the
current density, duration of the etch cycle, and
etchant solution composition (17). The multilayer templates possess a sinusoidally varying
porosity gradient, providing sharp features in the
optical reflectivity spectrum (Fig. 1) that approximate a rugate filter (18). The porous Si is converted to porous SiO2 by thermal oxidation, and
the oxidized nanostructure (fig. S1) (10) is used
as a template for solution-cast or injection-molded thermoplastic polymers.
Removal of the porous SiO2 template
from the polymer or biopolymer imprint by
chemical dissolution provides a freestanding
porous polymer film with the optical characteristics of the photonic crystal master (figs.
S2 to S4). Reflection spectroscopy (Fig. 1)
and scanning electron microscopy (SEM)
(Fig. 2) confirm that the photonic structure of
the porous Si master is retained in the polymer casting. The sharp optical reflectivity
feature expected of a rugate filter is observed
in both the template and the polymer casting
(Fig. 1), confirming that the process replicates the microstructure. Cross-sectional
SEM measurements (Fig. 2 and fig. S5) corroborate the optical data.
Vapor dosing experiments confirm that the
microporous nanostructure is retained in the
castings. The position of the spectral feature for
a rugate filter depends on the periodicity and
refractive index gradient of the structure. When
porous Si multilayers are exposed to condensable vapors such as ethanol or hexane, microcapillary condensation in the nanometer-scale
pores produces an increase in the average refractive index of the matrix and a spectral red shift of
the photonic feature (11, 19, 20). The shift of the
spectral peak correlates with partial pressure of
the analyte in the gas stream, following the
Kelvin equation for condensible vapors (15, 19–
21). Dose-response curves for ethanol vapor for
the porous Si template and for the polystyrene
Fig. 1. Reflectivity spectra of an oxidized porous Si rugate film (top) and a polystyrene film
cast from the porous Si template (bottom). The
spectral peaks correspond to the second-order
diffraction peak of the template and the second- and third-order diffraction peaks of the
imprint. The porous Si template was etched
using a sinusoidal current varying between 38.5
and 192.3 mA/cm2, with 70 repeats and a
periodicity of 8 s. The total thickness of the
porous Si film is 40 ␮m. The reflected light
spectra were obtained using an Ocean Optics
SD2000 charge-coupled device spectrometer
using tungsten light illumination. Spectra are
offset along the y axis for clarity.
www.sciencemag.org SCIENCE VOL 299 28 MARCH 2003
2045