TSPAN12 Regulates Retinal Vascular
Development by Promoting Norrin- but Not
Wnt-Induced FZD4/b-Catenin Signaling
Harald J. Junge,1Stacey Yang,1,5Jeremy B. Burton,1,5Kim Paes,4Xiao Shu,1,6Dorothy M. French,2Mike Costa,3
Dennis S. Rice,4and Weilan Ye1,*
1Tumor Biology and Angiogenesis Department
3Department of Cancer Targets
Genentech Inc., 1 DNA Way, South San Francisco, CA 94080, USA
4Ophthalmology Department, Lexicon Pharmaceuticals Inc., 8800 Technology Forest, The Woodlands, TX 77381-1160, USA
5These authors contributed equally to this work
6Present address: Department of Nutritional Sciences and Toxicology, 119 Morgan Hall, University of California, Berkeley, Berkeley,
CA 94720-3104, USA
Mutations in the genes encoding the Wnt receptor
Frizzled-4 (FZD4), coreceptor LRP5, or the ligand
Norrin disrupt retinal vascular development and
cause ophthalmic diseases. Although Norrin is struc-
turally unrelated to Wnts, it binds FZD4 and activates
the canonical Wnt pathway. Here we show that the
tetraspanin Tspan12 is expressed in the retinal
vasculature, and loss of Tspan12 phenocopies
defects seen in Fzd4, Lrp5, and Norrin mutant
mice. In addition, Tspan12 genetically interacts with
Norrin or Lrp5. Overexpressed TSPAN12 associates
with the Norrin-receptor complex and significantly
increases Norrin/b-catenin but not Wnt/b-catenin
signaling, whereas Tspan12 siRNA abolishes tran-
scriptional responses to Norrin but not Wnt3A in
retinal endothelial cells. Signaling defects caused
by Norrin or FZD4 mutations that are predicted to
impair receptor multimerization are rescued by over-
multimers and TSPAN12 cooperatively promote mul-
timerization of FZD4 and its associated proteins to
elicit physiological levels of signaling.
Vascular diseases of the retina are a major cause of impaired
vision and blindness. Retinas in the majority of mammals are
linked to the activity of the canonical Wnt pathway (Fruttiger,
2007) that promotes accumulation of b-catenin and stimulates
Nusse, 2006). In humans, mutations in the Wnt-receptor
Frizzled-4 (Fzd4) and the coreceptor Lrp5 cause familial exuda-
tive vitreoretinopathy (FEVR) (Warden et al., 2007), an inherited
disease characterized by incomplete vascularization of the
peripheral retina (Berger and Ropers, 2001). Mutations in the
gene encoding the cysteine knot protein Norrin cause Norrie
disease, FEVR, retinopathy of prematurity (ROP), or Coat’s dis-
ease. The retinal vasculature is altered in each of these diseases
patients, targeted inactivation of Norrin, Lrp5, or Fzd4 in mice
results in similar retinal phenotypes characterized by a dramatic
reduction of intraretinal capillaries (Luhmann et al., 2005;
Xia et al., 2008; Xu et al., 2004). Norrin is a high-affinity ligand
for FZD4 that signals via stabilizing b-catenin and activates
manner (Xu et al., 2004). FZD4 is the only Norrin receptor among
the function of Wnts (Lobov et al., 2005). Norrin isa cysteine knot
protein that lacks the structural characteristics of Wnts, and it is
unclearhowFZD4/LRP5 can respondto twotypesofstructurally
distinct ligands and elicit similar intracellular changes (i.e., the
accumulation of b-catenin and activation of LEF/TCF-mediated
transcription). Here we report the finding that FZD4/LRP5
signaling induced by Norrin, but not by Wnt ligands, depends
on an additional membrane protein TSPAN12.
TSPAN12 is a member of the phylogenetically ancient tetra-
spanin family (Serru et al., 2000). All tetraspanins share common
structural features that distinguish them from other four trans-
membrane domain proteins (Garcia-Espana et al., 2008). Inves-
tigation of several prototypic tetraspanins has led to a model in
which tetraspanins organize specialized tetraspanin-enriched
microdomains (TEMs) that act as signaling platforms in the
plasma membrane (Boucheix and Rubinstein, 2001; Hemler,
2005). Through a genetic screen, we discovered that Tspan12
is expressed in the retinal vasculature and Tspan12 mutant
mice phenocopy a multitude of defects observed in Norrin,
Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc. 299
Fzd4, and Lrp5 mutant mice. Through a series of in vivo and
in vitro analyses, we demonstrated that Tspan12 specifically
regulates Norrin/b-catenin but not Wnt/b-catenin signaling by
modulating FZD4 multimerization.
Since the canonical b-catenin signaling pathway regulates
a plethora of important biological processes, one of the key
questionsin thefieldishowspatiotemporal control ofsignalacti-
vation is achieved (van Amerongen et al., 2008). In this study,
we show that Tspan12 expression is restricted to the vascula-
ture within the retina, whereas Norrin (Hartzer et al., 1999), Friz-
zled-4 (Wang et al., 2001), and Lrp5 (Figueroa et al., 2000) are
more broadly expressed in the retina. The requirement of
TSPAN12 for Norrin/b-catenin but not Wnt/b-catenin signaling
may enable only a subset of Frizzled-4-expressing cells to
respond to Norrin, thereby providing a hitherto unrecognized
mechanism to generate specific spatiotemporal patterns of
Targeting of the Tspan12 Gene
We carried out a large-scale reverse genetic screen in mice to
identify disease-related phenotypes by knocking out 475 genes
that encode putative transmembrane or secreted proteins
(T. Tang et al., personal communication). Through this effort, we
identified a line of mutant mice exhibiting retinal vascular defects
in the fluorescein angiography test (Figure S2A available online).
The gene mutated in this line encodes a previously uncharacter-
ized tetraspanin named TSPAN12 (synonyms: TM4SF12 and
Net-2). The murine Tspan12 gene (refseq NM_173007) contains
chromosome 7). The gene product is a highly conserved 4TM
protein with a predicted molecular weight (MW) of 35 kDa and
whose electrophoretic mobility is slightly higher than predicted
(Figure S1F and data not shown). The predicted topology
features two extracellular loops, with both N and C termini local-
ized intracellularly (Figure S1B). Of the 136 bp in exon 3, a 30
portion of 97 bp and additional intronic sequences were elimi-
nated by our targeting strategy (Figure S1C). Thus, the start
codon and a major portion of the first membrane-spanning
region were replaced by the targeting cassette (an IRES element
followed by the lacZ derivative BGeo as a reporter of TSPAN12
expression, and the puromycin resistance gene driven by
a PGK promoter). Gene targeting was confirmed by Southern
blot and PCR (Figures S1D and S1E). Loss of the start codon
abolished translation of TSPAN12 in the homozygous knockout
(KO or ?/?) mice, as evident by immunoprecipitation (IP) fol-
lowed bywestern blot from lysates of embryos using a TSPAN12
polyclonal antibody (Figure S1F).
Vascular Defects in the Tspan12 Mutant Mice
Tspan12?/?mice were viable and fertile. Analysis of LacZ
expression on P15 retinal sections of Tspan12?/?or Tspan12+/?
mice (Figures 1B and S3B) and in situ hybridization using
revealed expression of Tspan12 in the neonatal retinal vascula-
ture. LacZ expression was also detected in the neonatal menin-
The alteration of retinal vessels (Figures 1B and S2A) prompted
us to analyze the retinal vasculature further. In the murine retina,
a superficial vascular plexus in the nerve fiber layer (NFL)
develops from the central artery through a combination of
sprouting, migration, and remodeling between P0 and P8.
Subsequently, vessels sprout vertically into the retina and ramify
in the outer plexiform layer (OPL) and inner plexiform layer (IPL),
where two capillary beds are established, resulting in a three-
layered vascular architecture (Fruttiger, 2007). We examined
retinas between P5 and P12 and found that the centrifugal
outgrowth of the NFL vasculature was moderately delayed in
Tspan12?/?retinas (Figures 1C and 1D and data not shown).
At P11, vertical sprouts and OPL capillaries appeared in
Tspan12+/+mice, whereas both were completely absent in
Tspan12?/?mice (Figures 1E and 1F). The fact that Tspan12
expression was detected in the vasculature but not other retinal
tissues, together with the observation that the retinal histology
appeared normalatP11byhematoxylin and eosin(H&E) staining
(Figures 1G and 1H), suggests that the vascular defect is
primary. In adult Tspan12?/?mice, the OPL remains avascular,
confirming that the defect is not transient (Figures 1I and 1J).
The thickness of the outer nuclear layer in Tspan12?/?retinas
was consistently reduced in adult but not neonatal mice
(compare Figures 1E, 1F, 1I, and 1J), indicating that neural cells
were secondarily affected by the vascular defects.
The phenotypes in Tspan12?/?mice described above are
similar to those reported for Fzd4 (Xu et al., 2004), Lrp5
(Xia et al., 2008), and Norrin (Luhmann et al., 2005) mutant
mice. We therefore examined Tspan12?/?mice for additional
characteristics observed in Norrin or Fzd4 mutant mice. Microa-
neurisms extending from the NFL toward the inner nuclear layer
were found in Norrin mutant mice at P15 (Luhmann et al., 2005).
P16 retinas of Tspan12?/?mice contained microaneurisms that
were remarkably similar to those described in Norrin mutant
mice (Figure 2A). Strikingly, these lesions develop at similar
time points in Norrin?/?and Tspan12?/?mice. Since aberrant
formation of retinal vascular fenestrations is prominent in Fzd4
mutant mice (Xu et al., 2004), we examined expression of the
fenestrated endothelial marker MECA-32 (also called PLVAP)
in Tspan12?/?mice. Consistent with the fact that normal retinal
vessels are not fenestrated, MECA-32 was not detected in
Tspan12+/+retinal endothelial cells (ECs) (Figure 2B, top panel).
However, strong MECA-32 signal was observed in Tspan12?/?
retinal ECs (Figure 2B, bottom panel). Delayed hyaloid vessel
regression was reported in Norrin (Luhmann et al., 2005), Fzd4
(Xu et al., 2004) and Lrp5 (Lobov et al., 2005) mutant mice. We
found that hyaloid vessel regression was also significantly
delayed in Tspan12?/?mice at several developmental stages
examined (Figure 2C). Three additional parallels between
Tspan12 and Fzd4 and/or Norrin mutant mice were observed:
focal hemorrhage in the adult retinas was found in Tspan12
(Figure S2B) and Fzd4 (Xu et al., 2004) mutant mice; retinal glial
cell activation was reported in Norrin mutant mice (Luhmann
et al., 2005) and was also observed in Tspan12?/?retinas
(Figure S2C); vessel enlargement in the stria vascularis of the
inner ear was observed in Tspan12 (Figure S2D), Fzd4
(Xu et al., 2004), and Norrin (Rehm et al., 2002) mutant mice.
300 Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc.
Since Norrin and FZD4 are a ligand/receptor pair that, in
conjunction with the coreceptor LRP5, activate the canonical
Wnt pathway and promote accumulation of cytoplasmic b-cate-
that TSPAN12 may be required for Norrin/b-catenin signaling
lation is reported to be a readout of impaired canonical b-catenin
signaling (Liebner et al., 2008), the strong vascular expression of
H + E
Figure 1. Tspan12 Is Required for the Develop-
ment of Intraretinal Capillaries
(A and B) Double staining of PECAM (brown) and b-galac-
tosidase (blue) on postnatal day 15 (P15) retinal sections.
b-galactosidase-positive cells are only detected in the
vasculature of ?/? animals.
(C and D) IsolectinB4 (IB4) staining of whole-mount P6
(E and F) P11 retinal sections stained with IB4 (green) and
(G and H) P11 retinal sections stained with hematoxylin
(blue) and eosin (pink).
(I and J) Sections of adult retinas stained with IB4 (green)
and DAPI (blue). Arrows: location of the OPL. Arrowhead:
ONL= outer nuclear layer, INL =inner nuclearlayer, GCL=
ganglion cell layer. OPL, IPL, and NFL are defined in the
text. Scale bars = 100 mm.
MECA-32 in the Tspan12?/?but not+/+retina
(Figure 2B) suggests that Tspan12 is required
for b-catenin signaling in retinal ECs.
Tspan12 Genetically Interacts
with Norrin or Lrp5
To further investigate if TSPAN12 is functionally
linked to Norrin/b-catenin signaling, we gener-
ated compound mutant mice that lacked one
allele each of Tspan12 and Norrin or Tspan12
and Lrp5. Because the Norrin and Lrp5 alleles
reported here (Figure S4) have not been previ-
ously described, we examined if our Norrin or
Lrp5 homozygous mutants displayed the docu-
mented phenotypes. We found that both
Norrin?/yand Lrp5?/?mice indeed had a leaky
and aberrant retinal vasculature (Figure S5A),
lacked the IPL and OPL capillary beds, and
strongly upregulated MECA-32 (Figure S5B
and data notshown). We then crossed
Tspan12+/?mice with either Lrp5+/?or Norrin?/y
mice and analyzed postnatal retinal vascular
development in resulting wild-type, single-gene
heterozygous, or transheterozygous offspring.
First, we quantified the number of vertical
sprouts extending from the NFL vasculature
into the OPL (method described in Figure S6)
because lack of vertical sprouts appears to be
an early and possibly primary defect in Norrin,
Lrp5, Fzd4, and Tspan12 mutant mice. Loss of
one allele of Tspan12 or Lrp5 or Norrin caused
minimal to moderate (0%–15.5%) reduction of vertical sprouts,
whereas concurrent loss of one Tspan12 allele with either one
Norrin or one Lrp5 allele significantly reduced sprouting by
40%–50% (Figures 3A and 3B). The strong statistical signifi-
cance of the differences (p < 0.0001, Figures 3A and 3B)
between the Tspan12+/?;Lrp5+/?or Tspan12+/?;Norrin+/?trans-
heterozygous mice and their single-gene heterozygous litter-
mates indicates that Tspan12 genetically interacts with Lrp5
Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc. 301
or Norrin in the regulation of intraretinal capillary development.
In a different analysis, MECA-32 staining on retinal sections
from the same mice analyzed above revealed only sporadic
weak expression in Tspan12+/?, Norrin+/?, or Lrp5+/?retinas
(arrows, Figures 3C and 3D) but significant and broad expres-
sion in the Tspan12+/?;Norrin+/?or Tspan12+/?;Lrp5+/?trans-
heterozygous retinal vasculatures (Figures 3C and 3D). These
data argue that Tspan12 genetically interacts with Norrin or Lrp5
with regard to the transcriptional response to Norrin/b-catenin
Interestingly, the transheterozygous mice did not manifest
the full spectrum of phenotypes found in the homozygous
single-gene knockouts, suggesting that they are hypomorphic
Taken together, our in vivo analyses provide strong genetic
evidence that TSPAN12 may be involved in Norrin/b-catenin
Isolectin-B4 (IB4), P16 flatmount retinas
MECA-32 IB4 DAPI
Figure 2. Formation of Microaneurisms,
Aberrant Fenestration, and Delayed Hyaloid
Vessel Regression in Tspan12?/?Mice
(A) Confocal projections of IB4-stained NFL (left),
IPL (middle), and OPL (right) vasculatures in P16
(B) P15 retinal sections stained with MECA-32
(red), IB4 (green), and DAPI (blue).
(C) Hyaloid vessels isolated from P7 and P12
In all panels, Tspan12 genotypes are indicated on
the left. Scale bars = 100 mm (main panels) and
10 mm (insets). Arrows in (A) and (B): microaneur-
in Norrin/b-catenin signaling, we per-
HEK293 cells. Cells were transfected
with reporter constructs and plasmids
encoding FZD4 and LRP5, as well as
either a plasmid encoding TSPAN12 or
a negative control plasmid. Transfected
cells were then cultured in the presence
or absence of 10 nM recombinant Norrin.
In FZD4 and LRP5 cotransfected cells,
Norrin induced a 6.85- ± 0.037-fold
activation of reporter activity compared
to cells expressing the receptors but
not stimulated with Norrin. The addition
of TSPAN12 strongly
response to Norrin and resulted in a
19.78- ± 2.57-fold activation (Figure 4A),
whereas the negative control reporter
Fopflash was not activated under these
conditions (Figure S7A). Activity of the
internal control (Renilla luciferase under
a constitutive promoter) was virtually
identical in cells expressing TSPAN12 or vector control (not
shown). Overexpression of TSPAN12 alone, or coexpression of
TSPAN12 and LRP5 in the absence of FZD4, was not sufficient
to allow activation with Norrin. Coexpression of FZD4 and
TSPAN12 in the absence of exogenous LRP5 led only to a very
weak activation (Figure 4A). Thus, exogenous TSPAN12 strongly
Norrin are present, and it can do so over a wide range of Norrin
concentrations (Figure S7B).
We then investigated if TSPAN12 can enhance signaling
induced by other FZD ligands by transfecting cells with human
Wnt3a cDNA in the presence or absence of FZD4 and LRP5
cDNAs. Intriguingly, TSPAN12 did not enhance Wnt3a/b-catenin
signaling through endogenous Frizzleds and LRPs or overex-
pressed FZD4 and LRP5 (Figure 4B). We also examined if
TSPAN12 could enhance signaling initiated by Wnt5a, 7a, or
7b because Wnt7b is a FZD4 ligand that functions during hyaloid
302 Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc.
vessel regression (Lobov et al., 2005), and Wnt7a and 7b are
required for the vascularization of the spinal cord (Daneman
etal.,2009; Stenman et al.,2008). Againwe foundthat TSPAN12
enhanced only Norrin/b-catenin signaling but not signaling
induced by several Wnts (Figures S8A and S8B). In agreement
with the ligand specificity observed in vitro, we found that loss
of Tspan12 did not affect spinal cord vascular development
(Figure S9), a process regulated by Wnt7a and 7b (Daneman
et al., 2009; Stenman et al., 2008). Although we cannot formally
rule out that TSPAN12 enhances signaling initiated by other Wnt
ligands, our in vitro and in vivo analyses suggest that TSPAN12
functions selectively in Norrin/b-catenin signaling.
enhancement of Norrin/b-catenin signaling by substituting
TSPAN12with CD9,CD63, CD151,TSPAN7, TSPAN8,
TSPAN11, and TSPAN13. None of these TSPANs enhanced
Norrin/b-catenin signaling (Figure 4C and data not shown), con-
firming that TSPAN12 has specific functions that are distinct
from other tetraspanins. Finally, we coexpressed TSPAN12
with multiple Frizzleds that do not bind Norrin (Smallwood et al.,
2007). As expected, Norrin did not activate these Frizzleds, and
TSPAN12 also had no effect on them (Figures 4C and S8C).
When Wnt3a was used to activate this set of Frizzled receptors,
TSPAN12 did not enhance their activity either (Figure S8C).
The ability of TSPAN12 to enhance Norrin/b-catenin signaling
was verified by examining b-catenin stabilization in 293 cells
overexpressing FZD4 and LRP5 with or without TSPAN12.
Whereas recombinant Norrinincreased b-catenin in the absence
of TSPAN12,coexpressionofTSPAN12 led tofurtherincreaseof
b-catenin levels (Figure 4D). Again, TSPAN12 had no effect on
P15 retinal sections Isolectin B4 MECA-32 DAPI
P15 retinal sections Isolectin B4 MECA-32 DAPI
(% of wild-type average)
(% of wild-type average)
Retinal vertical sprouts in offspring
of Tspan12 +/- mated with Lrp5 +/-
Retinal vertical sprouts in offspring
of Tspan12 +/- mated with Norrin -/y
Figure 3. Tspan12 Genetically Interacts with Norrin and Lrp5
(A and B) Quantitative analysis of vertical sprouts in the retinas of postnatal mice from two mating setups indicated on top of the graphs. Each symbol represents
one animal. Group averages and standard errors are overlaid over individual data points.
(C and D) Immunofluorescence staining of IB4 (green), MECA-32 (red), and DAPI (blue) on retinal sections. White arrows indicate sporadic weak expression of
Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc. 303
+ + + + + +
+ + + + + +
+ + + + + + + + + + + +
β-catenin / β-actin ratio
Tspan12 (ratio to ctl)
(ratio to ctl.)
(ratio to ctl.)
Figure 4. TSPAN12 Enhances Norrin/b-Catenin but Not Wnt/b-Catenin Signaling
(A–C) Topflash assay in 293 cells transfected with plasmids indicated below the graph and stimulated with ligands indicated in each panel.
(D) Lysates of 293 cells expressing the indicated proteins and stimulated with Norrin or Wnt3a were probed with anti-b-catenin and anti-b-actin antibodies. The
density of each band was quantified, and the ratios between the b-catenin and b-actin band densities for each condition were calculated and plotted in the top
(E) Topflash reporter activities induced with (black bars) and without Norrin (white bars) in 293 cells transfected with FZD4 and LRP5 or stimulated by expressing
increasing concentrations of b-catenin in the presence or absence of TSPAN12 (gray bars).
(F) Quantitative real-time PCR of the Tspan12 message in human retinal ECs transfected with a control or Tspan12 siRNAs.
(G) Quantitative real-time PCR of Axin2 (left) and Meca-32 (right) messages in human retinal ECs treated with the indicated ligands and transfected with a control
or Tspan12 siRNAs. In (F) and (G), Tspan12, Axin2, and Meca-32 messages were normalized to the housekeeping gene Gapdh in all samples then calculated as
ratios to the averages of untreated cells.
Bars in all plots represent the mean of triplicate samples; error bars represent standard deviations (SD). Asterisks indicate a significant difference between a pair
of samples with a p value < 0.05.
304 Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc.
b-catenin levels induced by Wnt3a (Figure 4D). When we acti-
coexpression of TSPAN12 had no effect (Figure 4E). Taken
together, our findings suggest that TSPAN12 specifically
enhances Norrin signaling through FZD4 and LRP5, and that it
acts upstream of b-catenin stabilization.
TSPAN12 Is Required for Norrin/b-Catenin Signaling
in Retinal Endothelial Cells
Because we detected TSPAN12 expression in the retinal vascu-
lature (Figures 1B, S3A, and S3B), and activity of the b-catenin
pathway is documented in several types of central nervous
system (CNS) ECs including those of the retina (Phng et al.,
2009), brain (Liebner et al., 2008), and spinal cord (Daneman
et al., 2009; Stenman et al., 2008), we asked whether TSPAN12
is required for Norrin/b-catenin signaling in retinal ECs. We used
human retinal microvascular endothelial cells (HRMVECs) to
investigate the function of endogenous TSPAN12. PCR analysis
showed that Tspan12, Fzd4, and Lrp5 were expressed in
HRMVECs (data not shown), and siRNA-mediated silencing
efficiently reduced the endogenous Tspan12 message (Fig-
ure 4F). Stimulation of HRMVECs with either Norrin or Wnt3a
significantly increased levels of Axin2 mRNA, a target gene of
the canonical b-catenin pathway in many cell types, and signifi-
cantly decreased levels of Meca-32, a target gene suppressed
by canonical b-catenin signaling in CNS ECs in vitro and
in vivo (Liebner et al., 2008). Transfection with a pool of Tspan12
siRNA(Figure 4G)or anindependent singleTspan12 siRNA (data
not shown) in HRMVECs completely abolished the regulation of
with Wnt3a(Figure4G).Ourobservation thatMeca-32isupregu-
lated in the retinal vasculature of Norrin, Lrp5, and Tspan12
mutant mice (Figures 2B and S5B) in vivo, in conjunction with
the finding that Meca-32 upregulation is a direct consequence
of losing b-catenin signaling in CNS ECs (Liebner et al., 2008)
and HRMVECs (Figure 4G), provides strong evidence that
TSPAN12 Is a Component of the Norrin-Receptor
Given that TSPAN12 is a transmembrane protein and acts
upstream of b-catenin stabilization, it may function in the plasma
membrane by interacting with components of the Norrin-
receptor complex. In order to test this possibility, we first asked
if TSPAN12 colocalizes with FZD4 on the cell surface. To this
end, we transfected HeLa cells with N-terminally flag-tagged
FZD4 (flag is extracellular) and HA-tagged TSPAN12 (HA is intra-
cellular). Plasma membrane FZD4 was detected with a flag anti-
bodyon nonpermeabilized live cells on ice to prevent internaliza-
tion, andTSPAN12 was detected subsequently afterfixation and
expression on the surface of HeLa cells within numerous punc-
tae, and TSPAN12 significantly colocalized with FZD4-positive
punctae in the plasma membrane (Figures 5A and 5D). In con-
trast, CD9 was not found in FZD4-positive structures (Fig-
ure 5B). When FZD4 was substituted with FZD5 and coex-
pressed with TSPAN12 we found that TSPAN12 and FZD5
were mostly separated (Figures 5Cand 5D). A fraction of overex-
pressed TSPAN12 was also observed in the endoplasmic retic-
ulum (ER) (colocalized with Calnexin in separate experiments;
data not shown). These data suggest that when TSPAN12 is
transported to the plasma membrane, it colocalizes with FZD4
in punctate structures.
Next we asked if TSPAN12 is physically associated with the
Norrin-receptor complex in the plasma membrane. We coex-
pressed FZD4, LRP5, and TSPAN12 in 293 cells and used
FZD5 and CD9 as specificity controls for FZD4 and TSPAN12,
respectively. After incubation of these cells with conditioned
medium containing flag-Alkaline Phosphatase (AP)-Norrin,
Norrin-associated membrane proteins were mildly crosslinked
before Norrin was immunoprecipitated (IPed) with a flag anti-
body (Figure 5E). Norrin was bound to cells expressing FZD4/
LRP5 with and without TSPAN12 but was washed off from
cells expressing FZD5/LRP5/TSPAN12 or TSPAN12/LRP5 (Fig-
ure 5E, 2nd panel), hence Norrin efficiently pulled down FZD4
but not FZD5 (Figure 5E, top panel). TSPAN12 was co-IPed
with Norrin/FZD4 but not when FZD4 was substituted with FZD5
or when no Frizzled was present. In contrast, CD9, although ex-
pressed at a similar level as TSPAN12, was not co-IPed with
Norrin/FZD4 (Figure 5E, 3rd panel). LRP5 was present in the
FZD4 complex pulled down by Norrin (Figure 5E, bottom panel).
Thus, TSPAN12 is physically associated with the Norrin/FZD4/
LRP5 ligand-receptor complex. TSPAN12 also co-IPed with
Norrin when only FZD4 but no LRP5 was present (Fig-
ure S10A). When TSPAN12 and LRP5 were coexpressed in the
absence of FZD4, and TSPAN12 was IPed, no association with
LRP5 was detected (Figure S10B). Another LRP family member
VLDLR was absent from the complex containing Norrin, FZD4,
and TSPAN12 (Figure S10C), confirming that LRP5 is a specific
component of the ligand-receptor complex. Taken together,
our data suggest that TSPAN12 is a component of the Norrin-
receptor complex, and FZD4 is a pivotal molecule that mediates
the association of all other elements in this complex.
TSPAN12 Does Not Alter Ligand-Receptor Binding
The colocalization and association of TSPAN12 with the Norrin-
anisms by which TSPAN12 can enhance Norrin signaling: (1)
TSPAN12 binds Norrin directly and brings more ligands into
the receptor complex; (2) TSPAN12 enhances the ability of
FZD4 to bind Norrin; (3) TSPAN12 increases the surface concen-
tration of FZD4. To test these possibilities, we carried out the
Figure 5E shows that the first possibility is not supported by
our results; nonetheless, we tested it further by incubating
HeLa cells overexpressing FZD4, LRP5, or TSPAN12 with condi-
tioned medium containing flag-AP-Norrin and subsequently
detecting the receptor-bound flag-AP-Norrin by AP substrate
staining. Consistent with previous reports (Xu et al., 2004), we
found that flag-AP-Norrin efficiently bound to cells expressing
FZD4 but not LRP5. Importantly, Norrin did not bind to cells ex-
pressing TSPAN12 alone (Figure 5F), indicating that TSPAN12
does not bind to Norrin directly.
Next, HeLa cells overexpressing FZD4/LRP5, together with
TSPAN12 or vector control, were probed with several dilutions
of flag-AP-Norrin conditioned medium (Figure 5G). Binding of
Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc. 305
1 2 3 4 5 6 7 8
lysate lysate lysate
flag-FZD4, LRP5, vector
flag-FZD4, LRP5, TSPAN12
Dilutions of flag-AP-Norrin CM (%)
0 10 20 35 60 100
(OD 405 nm)
flag-FZD4, LRP5, vector
flag-FZD4, LRP5, TSPAN12
Dilution of anti-flag-HRP (X:1,000)
0 1 2 3.5 6 10
(OD 450 nm)
% of TSPAN12
colocalized with FZDs
% of FZDs colocalized with TSPAN12
Figure 5. TSPAN12 Is a Component of the FZD4-Receptor Complex but Does Not Alter Norrin/FZD4 Binding
(A–C) Deconvolution microscopic images of HeLa cells expressing the indicated proteins. White frames in the far right panels outline the areas shown in the three
panels on the left.
(D) Percentages of green voxels colocalized with red voxels (top panel) or red voxels colocalized with green voxels (bottom panel) in cells transfected with either
FZD4 + TSPAN12 (black bars, n = 6) or FZD5 + TSPAN12 (white bars, n = 6). Quantification was done using three-dimensional images of cells obtained by
deconvolution microscopy. Bars represent the mean of six samples; error bars represent SD.
(E)293cellsexpressingtheindicated proteins wereincubatedwithflag-AP-Norrin conditioned medium (CM),washed, andmildly crosslinked. ExtractswereIPed
with anti-flag antibody. Total cell lysates (in lanes labeled with ‘‘lysate’’) or proteins that were coprecipitated with Norrin (in lanes labeled with ‘‘IP flag’’) were
detected with anti-epitope tag antibodies indicated on the right. Proteins detected by each epitope tag antibody are indicated on the left.
(F) Binding of flag-AP-Norrin to HeLa cells expressing the indicated proteins was detected after formation of the purple AP reaction product.
(G) HeLa cells transfected with the indicated plasmids were incubated with several dilutions of flag-AP-Norrin CM, and the bound Norrin was quantified by an AP
(H) In an experiment parallel to that in panel G, surface FZD4 was measured with several dilutions of an HRP-conjugated anti-flag antibody and quantified with an
HRP substrate assay. Bars in (G) and (H) represent the mean of triplicate samples; error bars represent SD (see Figure S11).
306 Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc.
flag-AP-Norrin to cells was similar in the presence or absence of
expression of TSPAN12 does not alter the ability of FZD4 to bind
Finally, surface FZD4 levels from HeLa cells overexpressing
flag-FZD4/LRP5 in the presence or absence of TSPAN12 were
evaluated with an HRP-coupled anti-flag antibody and found
to be unaltered (Figure 5H).
TSPAN12 Enhances Receptor Clustering
ligand-receptor binding or receptor expression, and in light of
reports that oligomerization of Frizzleds and LRP5/6 is important
for their function (Cong et al., 2004), we went on to investigate
whether TSPAN12 regulates the organization of the Norrin-
receptor complex. Toward this aim, we took advantage of the
unique biochemical properties of Norrin. Norrin belongs to the
subgroup of cysteine knot proteins that form dimers via intermo-
lecular disulfide bonds (Vitt et al., 2001). In addition, it has been
reported that Norrin dimers can further assemble into higher-
molecular-weight multimers, and through reduction of the inter-
molecular disulfide bonds or mutation of the cysteine in position
To investigate if multimer formation plays a role in Norrin func-
tion, we mutated cysteine 95 to arginine as the C95R mutation
a b c
a b c
Lane a: Norrin-V5
Lane b: Norrin-C95R-V5
Lane c: Vector control
5 3050 1005 30 50 1005 30 50 1005 30 50 100
+ + + +
+ + + +
Wild-type C95RWild-type C95R
p = 4.8e-04
p = 4.6e-05
p = 0.0076
p = 0.0048
p = 0.013
p = 0.014
p = 0.039
Figure 6. TSPAN12 Rescues the Defect of Norrin-
(A) Western blots of extracellular matrix extracts from 293
cells transfected with the indicated plasmids under
reducing (left) or nonreducing (right) conditions using
(B) Topflash assay in 293 cells transfected with plasmids
encoding FZD4 and LRP5 and increasing amounts of plas-
mids encoding wild-type or C95R mutant Norrin, together
sent the mean of triplicate samples; error bars represent
has been reported in a human patient with reti-
nopathy (Isashiki et al., 1995). Wild-type Norrin
or Norrin-C95R tagged with the V5 epitope
were expressed in 293 cells and isolated from
the extracellular matrix (ECM). SDS-PAGE
under reducing conditions revealed that wild-
type Norrin and Norrin-C95R are both efficiently
expressed. Consistent with previous reports
(Perez-Vilar and Hill, 1997), analysis under
wild-type Norrin formed dimers and multimers.
In contrast, Norrin-C95R was predominantly
monomeric with a small fraction of dimers that
might have formed by alternative disulfide
bonding but formed no larger assemblies (Fig-
amounts of wild-type or mutant Norrin cDNAs
together with the receptors to induce Topflash
activity in the presence or absence of TSPAN12.
Expression of wild-type Norrin efficiently induced Norrin/b-cate-
nin signaling in a dose-dependent manner, and the presence of
TSPAN12 strongly enhanced this activity at all ligand concentra-
tions (Figure6B). Norrin-C95R,on theother hand,was inactivein
plasmid, although this mutant form of Norrin can still efficiently
bind FZD4 (Figure S11). Intriguingly, the presence of TSPAN12
partially restored the activity of Norrin-C95R, and at the highest
ligand concentration, signaling strengths of the wild-type and
mutant Norrins were comparable (Figure 6B). The rescue activity
of TSPAN12 in this experiment did not result from modulating
Norrin-C95R binding to FZD4 (Figure S11), suggesting that it
may function by compensating for the inability of monomeric
Norrin-C95R to promote receptor multimerization.
Since we were unable to express Norrin-C95R at high enough
concentration to carry out a biochemical analysis of receptor
clustering, we utilized the previously described mutation FZD4-
M157V, which strongly impairs Norrin/b-catenin signaling but
maintains the ability to bind Norrin (Xu et al., 2004). Aided by
structural information (Dann et al., 2001), the M157V mutation
has been proposed to affect Norrin-induced FZD4 dimerization
and consequently multimerization (Dann et al., 2001; Toomes
et al., 2004; Xu et al., 2004). Consistent with previous reports
(Xu et al., 2004), we found that signaling mediated by FZD4-
M157V wasseverely impaired.
Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc. 307
coexpression fully rescued the signaling defect of FZD4-M157V
Wethenutilized FZD4-M157V to directlyinvestigate the role of
TSPAN12 in FZD4 multimerization. 293 cells were transfected
with TSPAN12, control vector, or CD9 and cotransfected with
flag-FZD4 and gD-FZD4 or flag-FZD5 and gD-FZD5 or with
flag-FZD4-M157V and gD-FZD4-M157V. Cells were incubated
on ice with medium containing Norrin or no ligand. Cell lysates
were IPed with anti-flag antibody and probed for coimmunopre-
cipitation of gD-FZD4. To enable quantification of protein-
protein interactions, no crosslinking reagent was used in this
Similar baseline levels of association between gD-FZD4 and
flag-FZD4 or gD-FZD4-M157V and flag-FZD4-M157V were de-
tected (Figure 7B and data not shown). TSPAN12 increased
the amount of gD-FZD4 pulled down by flag-FZD4 compared
to samples without a tetraspanin (Figures 7B and 7C) or with tet-
raspanin CD9 (Figures S12A and S12B) but did not enhance the
multimerization of FZD5 (Figures S12C and S12D). Norrin alone
also increased the amount of gD-FZD4 pulled down by flag-
FZD4, and the combination of Norrin and TSPAN12 further
increased FZD4 clustering (Figure 7B, left panels and Fig-
ure 7C, open bars). Importantly, the M157V mutation severely
impaired the ability of Norrin to cluster gD-FZD4 with flag-
FZD4, whereas coexpression of TSPAN12 compensated this
defect (Figure 7B, right panels and Figure 7C, filled bars).
Together, these data indicate that TSPAN12 and Norrin both
promote FZD4 multimerziation and suggest that initiation of Nor-
rin/b-catenin signaling requires (1) factors that promote FZD4
multimerization and (2) activation of FZD4 by ligand binding.
TSPAN12 Is Required for FZD4/b-Catenin Signaling
Induced by Norrin but Not Wnts
The central conclusion from the present study is that the tetra-
spanin TSPAN12 is required for retinal vascular development
and Norrin/b-catenin but not Wnt/b-catenin signaling, thus
uncovering a mechanistic distinction between these two
signaling systems. Consistent with a specific role for TSPAN12,
Tspan12?/?mice lack the embryonic phenotypes (Figure S9)
that were found in mutants of several Wnts (Daneman et al.,
2009; Stenman et al., 2008; van Amerongen and Berns, 2006).
In addition, Tspan12?/?mice do not phenocopy Fzd4 mutant
phenotypes that are independent of Norrin, e.g., malnutrition
and impaired growth due to the lack of esophageal skeletal
muscle (Wang et al., 2001). Furthermore, Lrp5 mutant mice
whereas bone mass in Tspan12?/?mice is normal (data not
shown). Finally, although hyaloid vessel regression is delayed
in Tspan12?/?and Norrin?/?mice, the degrees of delay appear
to be less severe than what was observed in Fzd4 and Lrp5
mutant mice, likely due to the activity of Wnt7b mediated by
FZD4 (Lobov et al., 2005). Thus, the spectrum of phenotypes
seen in the Tspan12?/?mice isconsistent with our in vitro results
and indicates a specific role of TSPAN12 in Norrin/b-catenin
signaling but not Wnt/b-catenin signaling.
TSPAN12 and Norrin Share Similar Profiles
of Evolutionary Conservation
The human genome harbors 33 tetraspanins, and numerous tet-
raspanins are found in multicellular eukaryotes (Garcia-Espana
et al., 2008). TSPAN12 is highly conserved in evolution, and
Fold changegD-FZD4 / flag-FZD4
Figure 7. TSPAN12 Rescues FZD4-M157V Multimerization and
(A) Topflash assay in 293 cells transfected with plasmids indicated below the
graphs and stimulated with Norrin or no ligand.
(B) 293 cells transfected with the indicated plasmids were incubated on ice
with or without recombinant Norrin, lysed, and IPed with anti-flag antibody.
Membranes were probed consecutively with anti-gD and anti-flag antibodies.
(C) Ratios of gD-FZD4 and flag-FZD4 band densities were calculated and
plotted from triplicate experiments similar to the example shown in (B). Data
are normalized to the value represented by the first column in each graph.
SD. Asterisks indicate a significant difference between a pair of samples with
a p value < 0.05.
308 Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc.
a high degree of conservation extends to birds (94% identity
between Homo sapiens and Gallus gallus) and fish (70% identity
between Homo sapiens and Danio rerio). However, no TSPAN12
ortholog could be identified in flies or nematodes. Interestingly,
Norrin orthologs are also found only in birds (87% identity
between Homo sapiens and Gallus gallus) and fish (68.6% iden-
tity between Homo sapiens and Danio rerio) but not in flies and
nematodes (using NCBI Homologene). Although Tspan12 and
Norrin reside in different chromosomes, it appears possible
that these two molecules coevolved and became functionally
Regulation of FZD4 Clustering by Norrin and TSPAN12
Receptoroligomerization hasbeenreportedto playanimportant
role in Wnt/b-catenin signaling (Bilic et al., 2007; Carron et al.,
2003; Cong et al., 2004; Kaykas et al., 2004). Our study suggests
that FZD4 clustering is also required for Norrin/b-catenin
signaling, but the mechanisms to regulate receptor oligomeriza-
tion appear to be different between Norrin and Wnts. Data in this
report suggest that pre-existing FZD4 oligomers are insufficient
to transduce Norrin/b-catenin signaling, and enhancement of
FZD4 clustering via binding to Norrin multimers and association
with TSPAN12 is required to fully activate the system. Given that
TSPAN12 alone can enhance FZD4 multimerization, it is unclear
why TSPAN12 does not affect signaling induced by Wnts. It is
possible that Wntscan induce strong signaling on small receptor
clusters or organize receptor multimers via a distinct mecha-
nism, thus circumventing a requirement for TSPAN12. More
detailed structural and biochemical characterization of the
receptor complexes induced by Norrin or Wnts will be required
to explain why TSPAN12-mediated FZD4 multimerization regu-
lates Norrin but not Wnt activities.
TSPAN12 Is a Candidate Gene for FEVR
FEVR (familial exudative vitreoretinopathy) is a progressive
retinal vascular disease with autosomal dominant, autosomal
recessive, and X-linked recessive inheritance patterns. Several
disease-associated loci have been identified, corresponding to
the Fzd4, Lrp5, and Norrin genes (Warden et al., 2007). Interest-
ingly,a family with FEVRpatients wasrecently identified in which
linkage to any known EVR locus could be excluded (Toomes
et al., 2005). This report also suggested that only about 40% of
FEVR patients carry mutations in either Fzd4 or Lrp5. In light of
the phenotypic similarities of Tspan12, Fzd4, Lrp5, and Norrin
mutant mice, it seems likely that FEVR or related vascular
diseases of the retina could be caused by mutations in Tspan12.
Tissues were fixed with 4% PFA and embedded in OCT. LacZ, DAB, and IB4
staining was performed according to standard procedures. Antigen retrieval
using DAKO Target Retrieval Solution was done before staining with IB4
(Sigma) and MECA-32 (BD), which were detected with a secondary antibody
conjugated to Alexa-594 or Alexa-488 (Molecular Probes), respectively.
Hyaloid Vessel Preparation
Neonatal eyes were fixed in 4% PFA overnight. Cornea, lens, and iris were
removed. 1.5% low melting agarose was injected into the vitreous. The solid-
ified agarose was extricated and heat melted on a glass slide, washed with
warm water, air-dried, and imaged without mounting.
Expression vectors encoding multiple TSPANs, Frizzleds, Norrin, Wnts, and
LRPs were in the pCMV6XL vectors. cDNAs encoding epitope-tagged
proteins were generated by PCR and subcloned into pEGFPN1, pCMV, or
pRK5. Topflash and Fopflash plasmids and pRL-CMV (Renilla luciferase)
were from Promega. Norrin-C95R and FZD-M157V were generated by site-
Firefly and Renilla luciferase activity was measured using the Dual Stop and
Glo system (Promega). In 24-well plates, 160K 293 cells/well were transfected
with 400 ng DNA and 1.5 ml Fugene6 (Roche). The DNA mix contained 100 ng
each of FZD and LRP5 plasmids, 50 ng of TSPAN12 plasmid or vector control,
and 150 ng of reporter mix (105 ng Topflash or Fopflash, 30 ng pRL-CMV, and
15 ng pCAN-myc-Lef-1). Cells were stimulated 24 hr (hr) after transfection for
coexpressing the indicated amount of Norrin or Wnt plasmids. Reporter
activity was calculated as firefly/renilla activity in each well. Within each plot
all data were normalized to the datapoint represented by the first bar.
Knockdown of TSPAN12 in HRMVEC
HRMVECs (ACBRI) were transfected with 20 nM of pooled or single siRNA
(Dharmacon, for target sequences see Supplemental Data) with RNAimax.
After 24 hr, cells were split at 50 K into 6-well plates. Twelve hours later cells
were stimulated with 1.25 mg/ml Norrin or 0.3 mg/ml hWnt3a (R&D Systems)
for 48 hr. Total RNA was extracted using Trizol (Invitrogen) and digested
with DNase (Stratagene) and used for Axin2 and Meca-32 RT-qPCR.
b-Catenin Stabilization Assay
293 cells were transfected in 24-well plates with 100 ng each of FZD4, LRP5
plasmids, and TSPAN12 plasmid or vector control. After 24 hr cells were stim-
ulated with 500 ng/ml Norrin or 500 ng/ml Wnt3a (R&D Systems) for 12 hr and
beta-D-maltoside [Calbiochem], 3 mM MgCl2and 30 U/ml DNaseI) on ice for
1 hr. Lysates were cleared by centrifugation and 200 ml supernatant were incu-
bated for 1 hr at 4?C with Convacalin-A beads to adsorb junctional b-catenin.
Supernatants were cleared by centrifugation and processed for western blot
using anti-b-catenin (Transduction labs) and anti-actin (Novus) antibodies.
Immunofluorescence Staining and Protein Colocalization Analysis
HeLa cells grown on 4-well chamber sides (Labtek) were transfected for
36–40 hr, cooled on ice, and stained with rabbit anti-flag Ab (Sigma, 1:500)
on ice for 1 hr. After extensive washing, cells were briefly fixed with 4% PFA,
blocked and permeabilized with 10% goat serum, 0.1% Triton X-100 in PBS,
and stained with anti-HA or anti-myc antibody. Images were taken with a 603
oil lens using a deconvolution microscope.
Semiconfluent 293 cells in 175 cm2dishes were transfected with 15 mg of Friz-
zled, 6.3 mg of LRP5-HA, and 18 mg HA-TSPAN12 or 16.65 mg vector control
plus 1.35 mg HA-CD9 (to achieve similar expression levels of TSPAN12 and
CD9) using135 ul Fugene6 for 36–40 hr. Cells were washed (subsequent steps
were at 0?C) with culture medium and incubated for 1 hr with 40 ml CM con-
taining flag-AP-Norrin. After several washes with PBS, cells were subjected
to mild surface crosslinking with 40 ml of 0.02 mM DTSSP (Pierce) in PBS
for 1 hr (stopped with Tris [pH 7.5], 20 mM final). Cells were pelleted and
washed with TBS before lysis in 20 ml lysis buffer (see b-Catenin Stabilization
1 hr and preadsorbed with agarose beads. Subsequently, flag-AP-Norrin was
bound to anti-flag-beads (Sigma) overnight. After extensive washing with lysis
buffer, beads were eluted with 150 ml 13 LDS sample buffer (Invitrogen).
Eluates were analyzed by western blotting with HRP-conjugated primary anti-
bodies to detect flag and HA, and anti-rho antibody (ABR, clone 1D4) was
Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc. 309
detected using mouse Trueblot reagent (eBioscience). For immunoprecipita-
tion with anti-HA-affinity matrix (Roche), cells were transfected, lysed, and
processed as described above.
For coimmunoprecipitation of gD-FZD with flag-FZD, semiconfluent 293
cells in 25 cm2dishes were transfected with 390 ng of each FZD plasmid,
910 ng of LRP5, and 3380 ng of HA-TSPAN12 or vector control for 36 hr.
When Norrin was added, cells were placed on ice and incubated with 3 ml
of ice-cold medium containing 600 ng/ml Norrin or no ligand for 45 min. Cells
were lysed with the lysis buffer (see above) containing 600 ng/ml Norrin or no
ligand. After centrifugation at 280K g for 10 min, supernatants were incubated
for 60 min with anti-flag beads while rocking on ice. After three brief washes
with ice-cold wash buffer (100 mM salt, 50 mM Tris [pH8] and plus 10% v/v
lysisbuffer), proteins were eluted with 110 ml of 13 LDS sample buffer and pro-
cessed for western blotting using an anti-gD antibody and mouse Trueblot
reagent or HRP-conjugated anti-flag antibody.
Quantification of Plasma Membrane FZD4
HeLa cells overexpressing FZD4 were incubated on ice with HRP-anti-flag Ab
(Sigma) at dilutions from 1:1.000 to 1:10.000 in culture medium, washed, lysed
in TBS 1% Triton X-100, and mixed with an equal volume of TMB ELISA
substrate (Rockland). After several minutes the reaction was stopped with
an equal volume of 1M H3PO4and the reaction product was quantified at
Extracellular Matrix Extracts of Wild-Type and Mutant Norrin
293 cells weretransfected withplasmids encoding V5-tagged wild-type Norrin
or Norrin-C95R. After 4 days all cells were removed using an enzyme-free cell
dissociation reagent (Sigma). ECM wasextractedwith 6MGuanidinium hydro-
chloride/0.1 M NaAcetate (pH 5.5) (60 ml per cm2) for 30 min at room temper-
ature on a rocking platform. Four hundred microliters of ECM extract were
ethanol precipitated and washed. Pellets were dissolved in 100 ml LDS sample
buffer with and without DTT and analyzed by western blotting with anti-V5 Ab
If not mentioned otherwise, averages were calculated from triplicate experi-
ments. Groups were compared using a two-tailed, unpaired Student’s t test.
p values < 0.05 were considered significant.
Supplemental Data include 12 figures and can be found with this article online
The authors thank Michelle Nee, Kathy Hotzel, Richard Carano, Hani Bou-
Reslan, and Luis Rodriguez for excellent technical assistance; Bonnee Rubin-
feld, James Ernst, Alia Majeed, Marinella Callow, Grant Kalinowski, Fred de
Sauvage, Leon Parker, Paul Polakis, and Greg Plowman for advice and
reagents; Hok-Seon Kim, Dan Yasura, Manda Wong, Richard Vandlen, Yi
Xia, Joanne Hongo, and Kurt Schroeder for supporting the immunization
efforts; Melvin Chen, Alfred Wong, and Vida Ashgeri for animal care; Pamela
Crowder and Linda Rangell for inner ear histology; and Cecile Chalouini for
microscopy assistance. H.J.J., S.Y., J.B.B., D.M.F., M.C., and W.Y. are
employees and shareholders of Genentech Inc. K.P. and D.S.R. are
employees and shareholders of Lexicon Pharmaceuticals Inc. Data contrib-
uted by X.S. were generated when he was an employee of Genentech Inc.
Received: December 3, 2008
Revised: May 14, 2009
Accepted: July 30, 2009
Published: October 15, 2009
Berger, W., and Ropers, H.H. (2001). Norrie disease. In The Metabolic and
Molecular Bases of Inherited Disease, C.R. Scriver, A.L. Beaudet, W.S. Sly,
and D. Valle, eds. (New York: McGraw Hill), pp. 5977–5985.
Bilic, J., Huang, Y.L., Davidson, G., Zimmermann, T., Cruciat, C.M., Bienz, M.,
and Niehrs, C. (2007). Wnt induces LRP6 signalosomes and promotes dishev-
elled-dependent LRP6 phosphorylation. Science 316, 1619–1622.
Boucheix, C., and Rubinstein, E. (2001). Tetraspanins. Cell. Mol. Life Sci. 58,
Carron, C., Pascal, A., Djiane, A., Boucaut, J.C., Shi, D.L., and Umbhauer, M.
(2003). Frizzled receptor dimerizationis sufficient toactivate theWnt/beta-cat-
enin pathway. J. Cell Sci. 116, 2541–2550.
Cong, F., Schweizer, L., and Varmus, H. (2004). Wnt signals across the plasma
ing its receptors, Frizzled and LRP. Development 131, 5103–5115.
Daneman, R., Agalliu, D., Zhou, L., Kuhnert, F., Kuo, C.J., and Barres, B.A.
(2009). Wnt/beta-catenin signaling is required for CNS, but not non-CNS,
angiogenesis. Proc. Natl. Acad. Sci. USA 106, 641–646.
Dann, C.E., Hsieh, J.C., Rattner, A., Sharma, D., Nathans, J., and Leahy, D.J.
(2001). Insights into Wnt binding and signalling from the structures of two
Frizzled cysteine-rich domains. Nature 412, 86–90.
Figueroa, D.J., Hess, J.F., Ky, B., Brown, S.D., Sandig, V., Hermanowski-
Vosatka, A., Twells, R.C.J., Todd, J.A., and Austin, C.P. (2000). Expression
of the type I diabetes-associated gene LRP5 in macrophages, vitamin A
system cells, and the islets of Langerhans suggests multiple potential roles
in diabetes. J. Hystochem. Cytochem. 48, 1357–1368.
Fruttiger, M. (2007). Development of the retinal vasculature. Angiogenesis 10,
Garcia-Espana, A., Chung, P.J., Sarkar, I.N., Stiner, E., Sun, T.T., and Desalle,
R. (2008). Appearance of new tetraspanin genes during vertebrate evolution.
Genomics 91, 326–334.
Gordon, M.D., and Nusse, R. (2006). Wnt signaling: multiple pathways,
multiple receptors, and multiple transcription factors. J. Biol. Chem. 281,
Hartzer, M.K., Cheng, M., Liu, X., and Shastry, B.S. (1999). Localization of the
Norrie disease gene mRNA by in situ hybridization. Brain Res. Bull. 49,
Hemler, M.E. (2005). Tetraspanin functions and associated microdomains.
Nat. Rev. Mol. Cell Biol. 6, 801–811.
Isashiki, Y., Ohba, N., Yanagita, T., Hokita, N., Hotta, Y., Hayakawa, M., Fujiki,
K., and Tanabe, U. (1995). Mutations in the Norrie disease gene: a new muta-
tion in a Japanese family. Br. J. Ophthalmol. 79, 703–704.
Kaykas, A., Yang-Snyder, J., Heroux, M., Shah, K.V., Bouvier, M., and Moon,
Frizzled in the endoplasmic reticulum by oligomerization. Nat. Cell Biol. 6,
Liebner, S., Corada, M., Bangsow, T., Babbage, J., Taddei, A., Czupalla, C.J.,
Reis, M., Felici, A., Wolburg, H., Fruttiger, M., et al. (2008). Wnt/beta-catenin
signaling controls development of the blood-brain barrier. J. Cell Biol. 183,
Lobov, I.B., Rao, S., Carroll, T.J., Vallance, J.E., Ito, M., Ondr, J.K., Kurup, S.,
Glass, D.A., Patel, M.S., Shu, W., et al. (2005). WNT7b mediates macrophage-
induced programmed cell death in patterning of the vasculature. Nature 437,
Luhmann, U.F., Lin, J., Acar, N., Lammel, S., Feil, S., Grimm, C., Seeliger,
M.W., Hammes, H.P., and Berger, W. (2005). Role of the Norrie disease pseu-
doglioma gene in sprouting angiogenesis during development of the retinal
vasculature. Invest. Ophthalmol. Vis. Sci. 46, 3372–3382.
Perez-Vilar, J., and Hill, R.L. (1997). Norrie disease protein (norrin) forms disul-
fide-linked oligomers associated with the extracellular matrix. J. Biol. Chem.
310 Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc.
Phng, L.K., Potente, M., Leslie, J.D., Babbage, J., Nyqvist, D., Lobov, I., Ondr, Download full-text
J.K., Rao, S., Lang, R.A., Thurston, G., and Gerhardt, H. (2009). Nrarp coordi-
nates endothelial Notch and Wnt signaling to control vessel density in angio-
genesis. Dev. Cell 16, 70–82.
Rehm, H.L., Zhang, D.S., Brown, M.C., Burgess, B., Halpin, C., Berger, W.,
Morton, C.C., Corey, D.P., and Chen, Z.Y. (2002). Vascular defects and senso-
rineural deafness in a mouse model of Norrie disease. J. Neurosci. 22, 4286–
Serru, V., Dessen, P., Boucheix, C., and Rubinstein, E. (2000). Sequence and
expression of seven new tetraspans. Biochim. Biophys. Acta 1478, 159–163.
Smallwood, P.M., Williams, J., Xu, Q., Leahy, D.J., and Nathans, J. (2007).
Mutational analysis of Norrin-Frizzled4 recognition. J. Biol. Chem. 282,
Stenman, J.M., Rajagopal, J., Carroll, T.J., Ishibashi, M., McMahon, J., and
McMahon, A.P. (2008). Canonical Wnt signaling regulates organ-specific
assembly and differentiation of CNS vasculature. Science 322, 1247–1250.
Toomes, C., Bottomley, H.M., Scott, S., Mackey, D.A., Craig, J.E., Appukut-
tan, B., Stout, J.T., Flaxel, C.J., Zhang, K., Black, G.C., et al. (2004). Spectrum
and frequency of FZD4 mutations in familial exudative vitreoretinopathy.
Invest. Ophthalmol. Vis. Sci. 45, 2083–2090.
Toomes, C., Downey, L.M., Bottomley, H.M., Mintz-Hittner, H.A., and Ingle-
hearn, C.F. (2005). Further evidence of genetic heterogeneity in familial exuda-
tive vitreoretinopathy; exclusion of EVR1, EVR3, and EVR4 in a large auto-
somal dominant pedigree. Br. J. Ophthalmol. 89, 194–197.
van Amerongen, R., and Berns, A. (2006). Knockout mouse models to study
Wnt signal transduction. Trends Genet. 22, 678–689.
van Amerongen, R., Mikels, A., and Nusse, R. (2008). Alternative wnt signaling
is initiated by distinct receptors. Sci. Signal. 1, re9.
Vitt, U.A., Hsu, S.Y., and Hsueh, A.J. (2001). Evolution and classification of
cystine knot-containing hormones and related extracellular signaling mole-
cules. Mol. Endocrinol. 15, 681–694.
Wang, Y., Huso, D., Cahill, H., Ryugo, D., and Nathans, J. (2001). Progressive
cerebellar, auditory, and esophageal dysfunction caused by targeted disrup-
tion of the frizzled-4 gene. J. Neurosci. 21, 4761–4771.
Warden, S.M., Andreoli, C.M., and Mukai, S. (2007). The Wnt signaling
pathway in familial exudative vitreoretinopathy and Norrie disease. Semin.
Ophthalmol. 22, 211–217.
Xia, C., Liu, H., Cheung, D., Wang, M., Cheng, D., Du, X., Chang, B., Beutler,
B., and Gong, X. (2008). A model for familial exudative vitreoretinopathy
caused by LPR5 mutations. Hum. Mol. Genet. 17, 1605–1612.
Xu, Q., Wang, Y., Dabdoub, A., Smallwood, P.M., Williams, J., Woods, C.,
Kelley, M.W., Jiang, L., Tasman, W., Zhang, K., and Nathans, J. (2004).
Vascular development in the retina and inner ear: control by Norrin and Friz-
zled-4, a high-affinity ligand-receptor pair. Cell 116, 883–895.
Cell 139, 299–311, October 16, 2009 ª2009 Elsevier Inc. 311