Retinoblastoma Has Properties
of a Cone Precursor Tumor and Depends
Upon Cone-Specific MDM2 Signaling
Xiaoliang L. Xu,1,2,3Yuqiang Fang,3Thomas C. Lee,5Douglas Forrest,6Cheryl Gregory-Evans,7Dena Almeida,1,2
Aihong Liu,1,2Suresh C. Jhanwar,3,4David H. Abramson,2,8and David Cobrinik1,2,9,*
1Dyson Vision Research Institute
2Department of Ophthalmology
Weill Cornell Medical College, New York, NY 10021, USA
3Department of Pathology
4Department of Medicine
Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA
5The Vision Center, Childrens Hospital Los Angeles, Los Angeles, CA 90027 USA
6National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA
7Department of Clinical Neuroscience, Imperial College London, London SW7 2AZ, UK
8Ophthalmic Oncology Service, Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA
9Present address: Sloan-Kettering Institute, Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA
Retinoblastomas result from the inactivation of the
RB1 gene and the loss of Rb protein, yet the cell type
in which Rb suppresses retinoblastoma and the
circuitry that underlies the need for Rb are undefined.
Here, we show that retinoblastoma cells express
markers of postmitotic cone precursors but not
markers of other retinal cell types. We also demon-
strate that human cone precursors prominently
express MDM2 and N-Myc, that retinoblastoma cells
require both of these proteins for proliferation and
survival, and that MDM2 is needed to suppress
ARF-induced apoptosis in cultured retinoblastoma
expression was regulated by the cone-specific
RXRg transcription factor and a human-specific
RXRg consensus binding site, and proliferation
required RXRg, as well as the cone-specific thyroid
hormone receptor-b2. These
support for a cone precursor origin of retinoblastoma
and suggest that human cone-specific signaling
circuitry sensitizes to the oncogenic effects of RB1
Retinoblastoma is a childhood retinal tumor that has provided
numerous insights into human cancer biology. For instance, reti-
noblastoma was one of the first malignancies to be recognized
as having hereditary features, and it engendered the character-
ization of one of the first tumor suppressors genes to be cloned,
protein, Rb, was found to have crucial roles in cell-cycle control
tivated in most if not all human cancers (Weinberg, 1995;
Cobrinik, 2005; Skapek et al., 2006). However, despite the
general significance of Rb in cancer biology, the basis for its
Retinoblastomas are thought to result from the inactivation of
RB1, either with somatic inactivation of both RB1 alleles, or with
a germline RB1 mutation and somatic inactivation of the second
allele in hereditary cases. Biallelic RB1 mutations may initially
result in the production of benign retinomas, with subsequent
genetic changes mediating malignant transformation (Dimaras
et al., 2008). Germline RB1 mutations predispose to an average
of five retinoblastomas, usuallybilaterally, and with mostforming
in the first year (Abramson and Gombos, 1996), but predispose
to only an ?0.5% chance of developing other cancers per year
(Kleinerman et al., 2005). The exceptionally high rate of retino-
blastoma arising from the minuscule retinal cell population
implies that the tumors derive from a cell type that is unusually
sensitive to the loss of Rb function.
One way that a cell may be sensitized to the loss of Rb is by
having signaling circuitry that fails to respond to Rb loss with
appropriate countermeasures. Indeed, at least two cellular
responses to Rb loss impede tumorigenesis in other settings,
but evidently fail to do so in the cells from which retinoblastomas
arise. First, in diverse settings including explanted mouse
retinas, Rb loss is compensated by increased expression of
the Rb-related p107 (Donovan et al., 2006). Second, loss of Rb
can deregulate E2F transcription factors and elicit E2F-depen-
dent apoptosis (Chen et al., 2007). Among other effects, deregu-
lated E2Fs induce expression of a CDKN2A isoform that
encodes ARF, which inhibits MDM2 and promotes p53-medi-
1018 Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc.
and premalignant retinomas have greatly increased CDKN2AARF
expression (Laurie et al., 2006; Chen et al., 2007; Dimaras et al.,
2008), it appears that the ARF-induced apoptotic response may
be impaired in early stages of retinoblastoma tumorigenesis.
To gain insight into the circuitry that sensitizes to Rb loss, we
opment. As retinoblastomas form as early as 21 weeks of gesta-
tion (Maat-Kievit et al., 1993), we initially sought clues to this
circuitry by evaluating Rb’s developmental expression pattern.
We found that Rb is expressed in a cell cycle-dependent manner
precursors of the different retinal neurons. However, Rb was
detected in older, maturing retinal precursors, and at exception-
ally high levels in maturing cone precursor cells (Lee et al., 2006).
Rb’s prominent expression in cone precursors was intriguing,
as cultured retinoblastoma cells generally express cone markers
(Bogenmann et al., 1988), but not glial markers as was once
believed (Virtanen et al., 1988), and the tumors have cone but
not rod phototransduction activities (Hurwitz et al., 1990). In
addition, retinoblastomas are topographically distributed across
1994). Nevertheless, the significance of the cone circuitry has
been unclear, as cone features were not consistently detected
in retinoblastomas in situ, and cells with properties of other
retinal cell types are also present in the tumors (Gonzalez-Fer-
nandez et al., 1992; Nork et al., 1995). Moreover, mice with tar-
geted loss of Rb and Rb-related proteins produced tumors
with amacrine or horizontal cell, but not cone cell, features
(Robanus-Maandag et al., 1998; Chen et al., 2004; Dannenberg
et al., 2004; MacPherson et al., 2004; Ajioka et al., 2007). Thus,
in the current study, we examined the relevance of cone-specific
signaling circuitry to retinoblastoma tumorigenesis.
Widespread Expression of L/M Cone Photoreceptor
Markers in Retinoblastoma Tumors
To assess the cellular phenotypes of retinoblastoma cells, we
stained a panel of tumors with retinal cell type-specific markers.
The panel included tumors displaying a range of differentiation
states, with and without prior chemotherapy, and from bilaterally
and unilaterally affected patients (Table S1 available online).
As mature photoreceptor features were previously detected in
only a subset of retinoblastomas, we initially examined whether
the tumors more generally express proteins that are character-
istic of immature photoreceptor precursors. These included
CRX, which is specific to cones, rods, and bipolar cells (Bibb
et al., 2001); RXRg, which is specific to cones and ganglion cells
(Mori et al., 2001); TRb2, which is specific to cones (Ng et al.,
2001); and NRL, which is specific to rods (Swain et al., 2001)
(Figures 1A–1D and S1).
CRX, RXRg, and TRb2 were detected in the vast majority of
cells in each of 40 tumors, including >95% of cells in ten quanti-
tatively evaluated samples, whereas NRL generally was not de-
tected (Figures 1E–1H and 1P, and Table S2). CRX, RXRg, and
differentiated cells as well as in proliferating cells that express
Ki67 or phosphorylated histone H3 (Figures 1I–1L and data not
shown). In keeping with their neoplastic status, CRX+, RXRg+,
and TRb2+tumor cells lacked detectable Rb (Figures 2A and
2B), whereas CRX+, RXRg+, and TRb2+cone precursors in the
central retina had prominent Rb expression (Figures 1A–1D and
S1). Inaddition, the vast majority of cellsin each of 22 retinoblas-
tomas expressed cone-specific arrestin (Figures 1P and S2).
As coexpression of CRX, RXRg, TRb2, and cone arrestin was
toma cells resemble short wavelength-sensitive S cones, or the
developmentally distinct long and medium wavelength-sensitive
was detected in >95% of cells in each of the tumors, including
proliferating Ki67+cells and each of ?4000 CRX+cells evaluated
in a costaining analysis (Figures 1M, 1P, and S3). In contrast, S
opsin was detected in only 0.1%–2% of cells in 16 (73%) of the
1N and 1P). As S and L/M opsins are transiently coexpressed in
developingL/Mbut not S cones (Cornishet al., 2004) (Figure1O),
these findings indicate that retinoblastomas largely consist of
cells that have an L/M cone precursor phenotype.
Cells within Retinoblastomas that Lack Cone Markers
Express Rb and/or Retain Wild-Type RB1 Alleles
We next examined whether retinoblastoma cells also express
markers of retinal neurons other than cones. In these analyses,
we costained samples for cell type-specific markers and for
Rb, in order to identify nonneoplastic Rb+cells. We detected no
tumor cells expressing the amacrine and horizontal cell-specific
syntaxin, the ganglion cell-specific Brn-3b, or the horizontal,
amacrine, bipolar, and progenitor cell-specific Prox1 (Table
S2). Occasional tumors had clusters of cells that expressed the
retinal progenitor cell- and bipolar cell-specific Chx10 or the
cell-specific Pax6. However, the Chx10+and Pax6+cells coex-
pressed Rb (Figure S4) and were concluded to derive from the
normal retina. Similarly, a cluster of cells expressing the rod-
specific rhodopsin was detected in one tumor, yet these cells
alsocoexpressed Rb(FigureS4).Onesamplehadahighly differ-
entiated retinoma-like region in which the majority of cells
expressed cone markers and <1% expressed rhodopsin
tic, asno rhodopsin+cellscostained for Ki67.Thus, in thisseries,
neoplastic retinoblastoma cells expressed markers of cones but
not other retinal neurons.
We also examined retinoblastomas for expression of glial and
progenitor cell markers. Each of 20 tumors had cells that coex-
pressed glial fibrillary acidic protein (GFAP) and nestin (Table
S2). The vast majority of these cells coexpressed Rb (Figures
2A–2F), consistent with their being nonneoplastic astrocytes or
Mu ¨ller glia. However, Rb was not detected in rare GFAP+cells
(Figure 2K), or in rare cells that lacked both GFAP and the
cone marker CRX (Figures 2A–2C, arrowhead).
To assess whether the rare CRX–,Rb–or GFAP+,Rb–cells
derive from RB1-mutated tumor cells, we examined whether
they retained wild-type RB1 alleles. We first identified tumors
and recorded the positions of CRX–,Rb–and GFAP+,Rb–cells,
and then defined their RB1 status using two-color fluorescence
Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc. 1019
in situ hybridization (FISH) for RB1 (at 13q14) and for a 13q34
locus that served as a FISH positive control.
As expected, nearly all (89% ± 7.6%) CRX+,Rb–retinoblastoma
presence of only one RB1 signal in ?10% of CRX–,Rb+cells
(Figure 2M), reflects the limited sensitivity of the FISH assay.
Figure 1. Cone Precursor Markers in Devel-
oping Retina and Retinoblastoma Tumors
(A–D) Photoreceptor precursor markers (red) and
Rb (green) in gestational week 16 human retina.
Note coexpression of Rb with cone markers
CRX, RXRg, and TRb2 but not with the rod marker
NRL (arrows; see also Figure S1).
(E–L) Cone but not rod precursor markers (red) in
retinoblastomas without (E-H) or with (I-L) Ki67
costaining. The specificity of TRb2 staining was
confirmed by knockdown analyses (Figure S1).
(M) Coexpression of L/M opsin (red) and Ki67
(green) in retinoblastoma.
(N and O) Coexpression of L/M opsin and S opsin
(arrows) in retinoblastoma (N) and in week 18 peri-
foveal L/M cone precursors, but not in S cone
precursors (arrowheads, O).
(P) Mean percentage of cells expressing each
marker in representative sections of ten tumors
(see Table S1). Error bars indicate the standard
Scale bars represent 50 mm in (A)–(N) and 20 mm
Importantly, nearly all (93% ± 2.2%)
CRX–,Rb–cells in both tumors also had
two RB1 signals (Figure 2M). The similar
proportion of CRX–,Rb+and CRX–,Rb–
cells with two RB1 signals implied that
most, and potentially all, of the CRX–cells
retained two RB1 alleles. Many of these
may be microglia, as all tumors had cells
that expressed the CD68 marker, and
approximately one-half of these lacked
detectable Rb (Table S2 and Figure S6).
Similarly, GFAP+,Rb–cells generally had
two RB1 signals (Figures 2J–2L).
In summary, Rb–cells that lacked the
cone-related CRX, or that expressed the
glial marker GFAP, generally had two RB1
alleles and were nonneoplastic, whereas
Rb–cells that expressed CRX retained
one RB1 allele and were the neoplastic
Neoplastic Cone Precursor-like
Cells Propagate Retinoblastoma
While the above studies showed that reti-
noblastoma cells generally express cone
markers, it remained possible that rare cells that lack a cone
phenotype might propagate the tumors. To address this possi-
tinal space of nude mice (Figure 3A), and examined the propaga-
tion of cone-like and non-cone-like cells, using human nuclear
antigen (HuNu) as a human cell marker.
As expected, virtually all cells in the original human tumors
were HuNu+. ?2.7% of the cells lacked CRX (Figures 3B and
1020 Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc.
plastic cells as described above. Engrafting this population
resulted in tumors that had human HuNu+,CRX+tumor cells
Figure 2. Retinoblastoma Cells Express Cone but Not Glial Markers
(A–F) Rb expression in retinoblastoma glia.
(A–C)Rb (white)inendothelial cellsand GFAP+perivascularglia(green,arrow),
but not in CRX+tumor cells (red). Rare cells lack Rb, CRX, and GFAP (arrow-
(D–F) Rb (white) in nonperivascular cells that coexpress GFAP (green) and
nestin (red, arrows).
(G–M) RB1 FISH of CRX–,Rb–and GFAP+,Rb–cells. Sections were costained
for Rb and either CRX (G and H) or GFAP (J and K), and then probed by
FISH (I and L), and nuclei with two control 13q34 signals (green) examined
for RB1 (red). Boxed regions in (G) and (J) are magnified in (H) and (I) and in
(K) and (L), respectively. The inset in (L) shows the RB1 FISH for the cell in
the center of the image. Arrows in (G)–(L) indicate CRX+Rb–tumor cells (white),
nonneoplastic Rb+cells (green), and CRX–,Rb–or GFAP+,Rb–cells (yellow).
(M) The percentage of cells with the indicated CRX and Rb staining that have
2, 1, or 0 RB1 signals. Error bars indicate standard deviation for six sections
from two tumors.
Scale bars represent 50 mm in (A), (G), and (J) and 20 mm in (D).
Figure 3. Cone Precursor-like Cells Propagate Retinoblastoma
(A) Human retinoblastoma and mouse subretinal xenografts (arrows).
(B) Original retinoblastoma (top) and secondary xenograft (bottom) costained
with human nuclear antigen (HuNu, green) and either CRX, RXRg, or TRb2
(red). Most cells in the original tumor costained with HuNu and CRX (yellow).
Occasional green nuclei (arrows) represent HuNu+,CRX–cells. Green HuNu+
human cells lacking CRX, RXRg, or TRb2 were not detected in xenografts.
(C) Percentage of human cells that lack CRX in samples used in primary,
secondary and tertiary xenografts (,) and average mass of primary,
secondary, and tertiary xenografts (-). Error bars indicate the standard
Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc. 1021
and mouse HuNu–,CRX–cells, but no HuNu+,CRX–cells among
an estimated more than 8000 cells examined (Figure 3C).
Secondary grafts produced tumors of similar size, in which
each of at least 8000 HuNu+cells expressed CRX, TRb2,
RXRg, and L/M opsin (Figures 3B and 3C and data not shown),
implying that cone-specific proteins were expressed in each of
at least 32,000 engrafted cells. Likewise, cells from secondary
grafts produced similar-sized tumors in tertiary grafts, and cells
from tertiary grafts produced tumors in a fourth round. Although
cone markers were detected in each of ?32,000 engrafted cells,
re-engraftment of as few as 100 cells generated tumors in three
out of four eyes.
Tumors also developed after engrafting two retinoblastomas
that consisted entirely of cone-like L/M opsin+cells after 2
months in culture. In these cultures, CRX, TRb2, RXRg, and
L/M opsin were detected in each of at least 8000 HuNu+cells,
and tumors developed after engrafting 1000 cells (the least
attempted) in each of four eyes (Figure S7). Thus, two ap-
proaches with three tumors demonstrated that cone pre-
cursor-like cells propagate retinoblastoma in an orthologous
Figure 4. Species-Specific Retinal Tumor
Phenotypes and Rb Expression Patterns
(A and B) Human retinoblastoma (A) and a mouse
Rb1?/?,p130?/?retinal tumor (B) stained for cone
markers CRX and TRb2 (red), or for amacrine and
horizontal cell markers syntaxin and Pax6 (green),
and with DAPI (blue), with antibodies that stain the
appropriate mouse and human
(C and D) Prominent Rb (green) in TRb2+cone
precursors (arrows) in human week 18 fovea (C),
butnotinmouseP12central retina(D).Labels indi-
cate positions of cone (‘‘C’’), bipolar (‘‘B’’), Mu ¨ller
(‘‘M’’), and amacrine (‘‘A’’) cells, and an antibody-
independent signal (‘‘*’’). Note that the P12 cone
precursors in (D) are appropriately positioned in
the outer nuclear layer (Rich et al., 1997; Ng
et al., 2009).
Scale bars represent 50 mm.
The Cone Phenotype of Human
but Not Mouse Retinal Tumors
Correlates with Robust Rb
Expression in Human but Not
Mouse Cone Precursors
While the above studies showed that
cone markers were prominent in human
retinoblastomas, amacrine and horizontal
cell markers were prominent in mouse
retinal tumors that resulted from loss of
Rb and Rb-related proteins (Robanus-
Maandag et al., 1998; Chen et al., 2004;
MacPherson et al., 2004; Dannenberg
et al., 2004; Ajioka et al., 2007). To
address this discrepancy, we compared
retinal cell marker expression in tumors
from the two species. In this analysis,
the cone markers CRX and TRb2 were highly expressed in
human retinoblastomas but not detected in Rb/p130-deficient
mouse tumors, whereas the horizontal and amacrine markers
Pax6 and syntaxin were prominent in the mouse tumors but
not detected in human retinoblastomas (Figures 4A, 4B, and S8).
The cone phenotype of human but not mouse retinal tumors
was notable given that Rb was prominent in maturing human
cone precursors (Lee et al., 2006) but was not detected in imma-
ture mouse cone precursors from E14 to P5, nor in mature
mouse cones (Spencer et al., 2005, and data not shown).
Thus, we determined whether Rb was induced during human
but not mouse cone maturation, by evaluating human foveae
at fetal week 18 (which follows the onset of L/M opsin expres-
sion at week 14.5 but precedes emergence of inner and outer
segments [Hendrickson and Provis, 2006]) and mouse retinas
sion and the emergence of cone outer segments at P11 [Szel
et al., 1993]). In a side-by-side comparison, there was an intense
Rb signal in human TRb2+cone precursors but only a faint signal
in mouse cone precursors, whereas Rb signals were similar in
human and mouse Mu ¨ller and bipolar cells (Figures 4C, 4D,
1022 Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc.
and S9). Thus, Rb appeared to be robustly expressed during the
postmitotic maturation of human but not mouse cones, corre-
lating with the L/M cone phenotype of human but not mouse
cells at P12 and P42, in horizontal but not amacrine cells at P5,
did not exceed that of Mu ¨ller cells (Figure S10 and data not
shown). Thus, in contrast to the situation in humans, mouse
retinal tumors are jointly suppressed by Rb and Rb-related
proteins and resemble retinal cell types that have relatively
modest Rb expression.
MDM2 and N-Myc Are Highly Expressed in Maturing
Human Cone Precursors
The retinoblastoma cone precursor phenotype and the Rb
expression pattern suggested that Rbmight have an antiprolifer-
ative role in cone precursor cells. We reasoned that if Rb was
required to prevent proliferation, then cone precursors might
responses that often accompany the loss of Rb function. To
address this possibility, we examined human retinas for expres-
sion of MDM2, which can abrogate E2F- and ARF-mediated
responses by inhibiting p53 (Kowalik et al., 1998; Lomazzi
et al., 2002), using an antibody (SMP14) that is specific to full-
length p53-binding MDM2 isoforms.
Remarkably, MDM2 was highly expressed in foveal cone
precursors at gestational weeks 16, 18, and 21 (Figures 5A,
5B, and 5I and data not shown). The MDM2 signal diminished
in progressively more peripheral and less mature cone precur-
sors (Figures 5F–5I), suggesting that MDM2 levels increase
during cone cell maturation. Concordantly, MDM2 was promi-
nent in mature cones in tumor-associated retinas from retino-
blastoma patients, including cones in the far periphery (Figures
5C, 5D, and S11). MDM2 was detected at lower levels in hori-
zontal cells but generally was not detected in other retinal cell
types (Figure 5J and data not shown). MDM2 was also highly
expressed in each of eight retinoblastomas (Figure 5E), despite
that each of the tumors had a near diploid MDM2 copy number
We next examined MDM2 expression in mouse retinas,
wherein combined loss of Rb1 and related genes appears to
elicit amacrine or horizontal cell tumors. Little or no MDM2 was
detected in immature cones from E14 to P5, in maturing cone
precursors at P8, P12, and P16, or in mature cones at P42
(Figures 5K and 5M and data not shown). While we cannot rule
out that the lack of Mdm2 signal in mouse cones results from
epitope masking, this seems unlikely as Mdm2 was readily de-
tected in Prox1+horizontal cells and at lower levels in Prox1+
amacrine cells at P14 and P42, as well as in the Prox1+tumor
cells in Rb/p130-deficient mice (Figures 5K–5O). However,
Mdm2was notdetected inimmature amacrineor horizontalcells
at E14 or P0 (data not shown). Thus, MDM2 was highly ex-
pressed in the human and mouse cells that resemble the retinal
tumors in the two species, and particularly during their postmi-
As tumorigenesis requires proliferative as well as survival
signaling, we also examined whether postmitotic cone precur-
sors express proliferation-related proteins. While human cone
precursors lacked detectable E2F1 and cyclins D1, A2, and B,
they highly expressed N-Myc. N-Myc was prominent in perifo-
veal cone precursors at weeks 16 and 18, and in more peripheral
positions at weeks 18 and 21 (Figures 5P–5S and data not
shown) but was not detected in mature cones of the postnatal
human retina or in mouse cone precursors at P0, P5, or P12
(data not shown). The N-Myc staining appeared to be specific,
as two N-Myc antibodies stained cone precursor nuclei in similar
retinal positions (data not shown). N-Myc was also prominent in
each of eight retinoblastomas that had a near diploid MYCN
copy number (Figures 5T and S12).
MDM2 and N-Myc Are Required for Retinoblastoma Cell
Proliferation and Survival
The prominent expression of MDM2 and N-Myc in cone precur-
sors and retinoblastoma cells suggested that these proteins
might be relevant to retinoblastoma development and propaga-
tion. To evaluate this possibility, we used lentiviral short hairpin
RNA (shRNA) vectors to decrease MDM2 or N-Myc expression
in Y79 and early passage RB176, RB177 or RB178 cells, which
lack MDM2 and MYCN amplification (Figure S12).
Three shRNAs were found to decrease MDM2 mRNA and
at 5 days after transduction (Figures 6A and 6B). Over the next
5 days, each MDM2 shRNA impaired cell proliferation and
induced apoptosis, and two shRNAs increased the proportion of
G0/G1 cells (Figures 6C–6E), whereas expression of an shRNA-
findings suggest that constitutively high MDM2 expression is
crucial for retinoblastoma cell proliferation and survival.
Similarly, shRNAs that diminished MYCN expression by
?40%–50% impaired proliferation and promoted cell death in
retinoblastoma cultures (Figure 6G).
MDM2 Suppresses ARF-Induced Apoptosis
in Retinoblastoma Cells
As MDM2 is a major target of the ARF-mediated oncogenic
stress response, we hypothesized that MDM2 might impede
ARF-induced apoptosis in retinoblastoma cells. To test this
idea, we determined whether a reduction in ARF could diminish
the need for MDM2. We first produced RB177 derivatives that
stably expressed either of two CDKN2AARF-specific shRNAs or
a scrambled shRNA control (Figure 6H and data not shown).
The cells were then transduced with an shRNA against MDM2
(Figure 6I) and examined for proliferation and TUNEL staining.
Interestingly, the ARF-directed shRNAs increased RB177 prolif-
eration and abrogated the antiproliferative and proapoptotic
effects of the MDM2 shRNA (Figure 6J), indicating that MDM2
did indeed suppress ARF-induced apoptosis.
A Human-Specific RXR Element and the Cone-Specific
RXRg Regulate Retinoblastoma Cell MDM2 Expression
The expression of MDM2 in human cones and retinoblastoma
cells suggested that cone-specific circuitry might direct MDM2
expression during retinoblastoma tumorigenesis. To address
this possibility, we searched the MDM2 P1 and P2 promoters
for cone-specific control elements that were identified in an
Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc. 1023
Figure 5. MDM2 and N-Myc Expression in Human Cone Precursors
(A–J) MDM2 (green) in human gestational week 18 (A and B) or week 21 (F–J) retina, in an uninvolved retina from a 4-year-old retinoblastoma patient (C and D), or
inaretinoblastoma tumor(E),andcostainedforL/Mopsin(A–Dand F–I,red) or Prox1(J,red). Thesamesectionwasimaged withfixed parameters in(F)–(I). White
and yellow arrows indicate MDM2+cones and horizontal cells, respectively.
(K–O) Mdm2 (green) and Prox1 (red) in mouse P12 (K and L) or P42 (M and N) retina, or in a Rb1?/?,p130?/?mouse retinal tumor (O). Boxed regions in (K) and (M)
are shown at higher magnification in (L) and (N). White and yellow arrows indicate Mdm2+horizontal and amacrine cells, respectively.
(P–T) N-Myc (green) in week 18 parafovea (P–R) or mid-periphery (S), or in human retinoblastoma (T), with TRb2 costaining (Q, red). Arrows indicate N-Myc+
Scale bars represent 20 mm for all panels, except for (E) and (O), where they represent 10 mm.
1024 Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc.
unbiased bioinformatics analysis (Danko et al., 2007). This re-
vealed an element in the human but not mouse P2 promoter
that matched a cone-specific RXR-like element (Danko et al.,
2007) at each of six invariant positions, and matched the
consensus RXRg homodimer binding site (Dowhan et al., 1994)
at 14 of 15 positions (Figure 7A). We then examined whether
this human-specific element promotes MDM2 expression, using
anMDM2-P2-Luc reportergeneandamutantDRXR versionthat
had two human-to-mouse nucleotide substitutions. Notably, the
two substitutions significantly diminished MDM2-P2 promoter
activity in retinoblastoma cells (Figures 7A and 7B).
We next examined whether RXRg is important for MDM2
expression. We found that each of two shRNAs that diminished
RXRg expression also reduced expression of MDM2, yet the
effect was transient and MDM2 RNA subsequently increased
(Figure 7C, a and b). The initial effect of RXRg knockdown was
enhanced by the p53 inhibitor, pifithrin-a p-nitro, cyclic (Fig-
ure 7C, c), and by p53 knockdown (data not shown), suggesting
that p53 opposed the reduction in MDM2 expression. Moreover,
antibodies enriched for MDM2 sequences surrounding the P2
RXR site, but not for sequences in exon 12 (Figure 7D). We
conclude that a human-specific MDM2 promoter element and
the cognate cone-specific RXRg transcription factor contribute
to retinoblastoma cell MDM2 expression.
The Cone-Specific Transcription Factors RXRg
and TRb2 Are Required for Retinoblastoma Cell
Proliferation and Survival
As RXRg was found to promote MDM2 expression and is impli-
cated in numerous aspects of cone biology, we examined
whether this factor could contribute to retinoblastoma tumori-
genesis. Concordantly, RXRg knockdown dramatically impaired
the proliferation and survival of Y79, RB176, and RB177 cells
(Figure 7E). Moreover, RXRg had effects beyond regulating
MDM2, as the shRNA-induced decline in MDM2 was followed
by increased expression of MDM2 (Figure 7C) as well as other
p53 target genes, including CDKN1A and 14-3-3s (data not
We also examined the role of the cone-specific thyroid
hormone receptor b2 isoform (TRb2) (Ng et al., 2001) using
shRNAs directed against sequences that are specific to TRb2
or that are shared with TRb1. These shRNAs impaired prolifera-
tion of Y79,RB139, and RB176cells and suppressed tumorigen-
esis when shRNA-transduced Y79 cells were engrafted to the
mouse subretinal space (Figure 7F–G). We conclude that the
cone-specific TRb2 and RXRg contribute to retinoblastoma
cell proliferation and survival.
Germline RB1 mutations have long been known to predispose
to retinoblastoma, yet the basis for the selective predilection to
retinoblastoma, as opposed to other tumors, has not been
defined. In this study, we evaluated whether the underlying
cellular circuitry of retinoblastoma cells might sensitize to the
oncogenic effects of RB1 mutations.
Retinoblastoma Cells Resemble L/M Cone
Several lines of evidence indicated that retinoblastoma cells
resemble neoplastic L/M cone precursors.
First, the vast majority of cells in all tumors expressed cone-
specific proteins, as well as proteins that are concurrently ex-
pressed only in cones, such as CRX and RXRg. These proteins
are not merely associated with the cone phenotype, but include
transcription factors such as CRX, RXRg, and TRb2 that are
numerous cone cell features (Furukawa et al., 1999; Ng et al.,
totransduction activities were earlier detected in cultured retino-
blastomas (Bogenmann et al., 1988; Hurwitz et al., 1990), we
know of only two studies that evaluated cone-specific protein
expression in retinoblastomas in situ. One study reported wide-
spread expression of cone transducin-a (Rodrigues et al.,
1992), whereas the second showed that antibodies against
with an anti-human L/M opsin antibody, on the contrary, demon-
strate that retinoblastomas consistently exhibit features of
maturing L/M cone precursors.
Second, rare cells that expressed markers of retinal neurons
other than cones were found to coexpress Rb and were
concluded to derive from the surrounding normal retina. Simi-
larly, cells that expressed glial and progenitor markers generally
ally retained two RB1 alleles. These findings imply that the vast
majority and potentially all tumor cells that lack cone markers
Finally, cone precursor-like cells propagated retinoblastoma
tumors, with engraftment of as few as 100 cells sufficient to
initiate tumorigenesis. Thus, cone-like cells appear to propagate
the tumors in the absence of a distinct stem cell population.
We also confirmed that the human retinoblastoma phenotype
of the combined loss of Rb and Rb-related proteins. The distinc-
tion between the mouse and human tumors appeared to be
absolute, as we found no evidence of human retinoblastoma
cells that lacked cone markers, and no evidence of mouse tumor
cells that expressed cone markers.
A Role for Cone Precursor MDM2 Expression
in Retinoblastoma Tumorigenesis
To address whether cone precursor circuitry might contribute to
retinoblastoma, we examined whether cones have distinctive
circuitry that could impede an ARF-mediated apoptotic
response to Rb loss. We found that the ARF target, MDM2,
was expressed at exceptionally high levels in human cone
precursors and suppressed ARF-induced apoptosis in retino-
blastoma cells. The MDM2 expression and resistance to ARF
signaling seem to ensue at least in part from cone precursor
circuitry, as human MDM2 promoter activity depended upon
an RXR promoter element and on the cognate RXRg protein.
Importantly, cone precursor MDM2 expression provides a
rationale for the lack of classical p53 pathway mutations in reti-
noblastoma tumors. In most cancers, the p53 pathway is
Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc. 1025
Figure 6. Roles for MDM2 and N-Myc in Retinoblastoma Cells
(A) qRT-PCR of MDM2 RNA, 5 days after infection of Y79, RB177, or RB178 cells with lentivirus encoding MDM2 shRNAs or a scrambled control.
(B) Immunoblot analysis of MDM2 and g-tubulin (as a control) in Y79 cells on day 5 after infection.
(C) Growth of Y79, RB177, and RB178 cells after puromycin selection, plating on day 5, and counting on days 5 and 10 after infection.
(D) The proportion of G0/G1 cells in Y79 and RB177 cultures, 9 days after infection.
(E) The percentage of TUNEL+Y79 and RB177 cells, 8 days after infection (top) and representative TUNEL staining (bottom).
(F) RB177 cells transduced with sh380-resistant MDM2 cDNA or a vector control and reinfected with pLKO-shMDM2-380 or a scrambled shRNA control. Left:
qRT-PCR analysis of MDM2 RNA, 4 days after reinfection. Middle: cell numbers after plating on day 4 and counting on day 10 after reinfection. Right: percentage
of TUNEL+cells 11 days after reinfection.
(G) qRT-PCR analysis of MYCN RNA on day 5, and cell growth after plating on day 4 and counting on days 4, 10, and 15 after transduction of RB177 or RB176
with MYCN shRNAs.
1026 Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc.
inactivated by mutation of TP53 or CDKN2AARFor by amplifica-
tion of MDM2. Moreover, these changes appear to be necessary
for the tumors to avert ARF-mediated responses to oncogenic
stress, rather than to avert the DNA damage response (Christo-
phorou et al., 2006; Efeyan et al., 2006). Thus, it has been
puzzling that CDKN2A and TP53 mutations have not been de-
tected in retinoblastoma, and that MDM2 amplification is rare
(Kato et al., 1996; Schlamp et al., 1997; Laurie et al., 2006).
Our findings suggest that the ARF-mediated oncogenic stress
response isinactivated in retinoblastomas because of the robust
expression of MDM2, in the absence of MDM2 amplification.
Nevertheless, retinoblastoma cells retain an effective p53-medi-
ated DNA damage response (Kondo et al., 1997) and are sensi-
tive to the MDM2 antagonist Nutlin-3a (Elison et al., 2006), as ex-
pected for cells that rely on robust MDM2 expression.
It has also been proposed that the p53 pathway is impaired in
retinoblastoma as a consequence of copy number gains of the
MDM2-related gene, MDM4 (also called MDMX) on chromo-
some 1q32.1 (Laurie et al., 2006). However, while MDM4 was
shown to impede the p53-mediated radiation response, it was
of radiation, or to oppose ARF-mediated apoptotic signals.
Moreover, a recent study suggested that a gene other than
MDM4 may be a more general target of chromosome 1q gains
(Dimaras et al., 2008), and MDM4 amplification was rare in our
samples (Figure S12). Thus, while MDM4 copy number gains
may contribute to some retinoblastomas, our findings suggest
that robust expression of the unamplified MDM2 has a more
general role in averting the ARF-mediated oncogenic stress
response, and could act during early stages of tumorigenesis
before Rb-deficient cells acquire additional genetic aberrations.
Notably, MDM2 was prominent in human but not mouse
cones, yet was prominent in the horizontal cells of both species
and in Prox-1+amacrine cells in mice. The high MDM2 levels in
mouse horizontal and amacrine cells is intriguing in light of
evidence that mouse retinal tumors that arise due to combined
Rb1 family mutations derive from apoptosis-resistant cells
(Chen et al., 2004); specifically from horizontal cells in some
genetic backgrounds (Ajioka et al., 2007) and potentially from
amacrine cells in others (Robanus-Maandag et al., 1998; Chen
et al., 2004). Thus, MDM2 appears to be well situated to
contribute to mouse horizontal or amacrine cell tumors upon
combined loss of Rb and Rb-related proteins, and to contribute
to human cone precursor-like retinoblastomas upon loss of Rb.
However, MDM2 evidently does not predispose to horizontal
or amacrine cell tumors upon loss of only Rb, possibly because
of the redundant actions of p107 and p130 in these cell types.
Roles for Additional Cone-Specific Circuitry
We also found that human cone precursors prominently ex-
pressed N-Myc and that retinoblastoma cells require N-Myc
for proliferation and survival. The apoptotic response to N-Myc
knockdown is consistent with the need for persistent Myc
expression in numerous mouse tumor models, and might reflect
the altered expression of diverse Myc target genes, yet co-
ntinued CDKN2AARFexpression, when Myc is incompletely sup-
pressed (Shachaf et al., 2008). Interestingly, MYCN was ampli-
fied in a subset of mouse retinal tumors but generally is not
amplified in human retinoblastomas (Corson and Gallie, 2007;
MacPherson et al., 2007). Thus, while further study is required,
these observations are consistent with a role for cone
precursor-related N-Myc expression in retinoblastoma tumori-
This study also revealed roles for the cone cell transcription
factors RXRg and TRb2 in retinoblastoma cells. As these factors
regulate opsin expression in cone precursors, but are not ex-
pressed in proliferating retinal cells (Ng et al., 2001; Roberts
et al., 2005), their postmitotic roles appear to be co-opted for
proliferative purposes during retinoblastoma tumorigenesis.
Moreover, our data indicate that RXRg promotes MDM2 expres-
sion in retinoblastoma cells and suggest that RXRg could like-
wise contribute to MDM2 expression in human cone precursors
via an RXRg recognition element in the human MDM2 promoter.
As the RXRg element was present in the human but not mouse
promoter, this element could contribute to the uniquely human
proclivity to develop cone-like retinoblastoma tumors. Likewise,
ourdata suggest thathorizontaland amacrinecell circuitry might
promote Mdm2 expression and tumorigenesis in mouse retino-
Implications for the Retinoblastoma Cell of Origin
As cone precursor circuitry was crucial to retinoblastoma cell
proliferation, it is of interest to consider whether this circuitry is
adopted during the course of tumorigenesis or is continuously
present because of the tumor’s origin from cone precursor cells.
A cone-related origin of retinoblastoma was previously sug-
gested by the L/M cone-like distribution of the tumors over the
surface of the retina (Munier et al., 1994). This distribution is diffi-
cult to reconcile with an origin from cells that are unrelated to
cones but is consistent with an origin either from cone precur-
sors or from cone-directed retinal progenitor cells. Our results
provide support for a cone precursor but not progenitor origin,
as we did not detect neoplastic cells that have progenitor
features in 40 retinoblastoma tumors.
It was also suggested that retinoblastomas might arise from
cells that reside in the retinal inner nuclear layer (INL), rather
than from outer nuclear layer (ONL) cells such as cones, based
on a potentially nascent tumor that was adjacent to the INL
and had INL-like nuclear morphology (Gallie et al., 1999).
However, as the tumor was almost entirely within the ONL, and
its nuclear morphology has uncertain significance, this sample
is also consistent with a cone precursor origin.
A cone precursor origin is also supported by the robust
expression of Rb, MDM2, and N-Myc—and the abrupt decline
in p27 (Lee et al., 2006)—during cone precursor maturation,
(H) RB177 cells transduced with pLKO-shARF-331 or a scrambled control, and analyzed for CDKN2AARFRNA (top) and ARF protein (bottom).
(I)qRT-PCR analysis of MDM2 RNA and CDKN2AARFRNA, 4 daysafterreinfection of RB177-shARF and controlcells with pLKO-shMDM2-377or control shRNA.
(J) Cell numbers and TUNEL staining 5 days after reinfection with MDM2 or control shRNA vectors.
Error bars indicate the standard deviation, and (*) and (#) indicate p < 0.01 and p = 0.016, respectively.
Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc. 1027
1028 Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc.
and with Rb’s ability to reassert growth control when restored to
cone precursor-like retinoblastoma cells (Cobrinik et al., 2006).
Thus, our findings support a model in which a lack of Rb permits
the aberrant proliferation of N-Myc+cone precursors, robust
response, and cone factors such as RXRg and TRb2 further
promote cell proliferation and survival. Moreover, as Rb loss
can elicit mitotic instability (Hernando et al., 2004), the aberrantly
proliferating cells could rapidly acquire additional cytogenetic
changes that lead to malignancy (Figure 7H). As Rb, MDM2,
and N-Myc are prominent in maturing but not nascent cone
precursors, this model implies that contextual features that are
specific to human cone maturation could impose the require-
ment for Rb tumor suppressor function.
Finally, a cone precursor origin accords with the prediction
that human and mouse retinal tumors derive from different intrin-
sically death-resistant retinal precursors (Pacal and Bremner,
2006) and suggests that cell type-specific MDM2 expression
may underlie the death-resistant phenotypes. Thus, while this
derive from cone precursors within the developing human retina,
the ability to reconcile the distinct features of the human and
In summary, this study shows that retinoblastoma cells
express numerous components of the cone precursor signaling
circuitry and rely upon this circuitry for their proliferation and
survival. The findings appear to be consistent with a cone
precursor origin of retinoblastoma and suggest that elements
noblastoma therapy. More generally, the findings suggest that
cell type-specific as well as species-specific signaling circuitry
sensitizes specific cells to specific oncogenic mutations.
Immunostaining and FISH
Retinas and retinoblastoma samples were prepared, sectioned, and immuno-
for RB1 and a 13q34 hybridization control, and cells with immunostaining
profiles of interest and two 13q34 signals examined for hybridization to RB1,
as described in the Supplemental Data.
Two hundred thousand retinoblastoma cells were injected into the subretinal
space of nude mice 1 day after explant and examined 60 days after primary
and secondary grafts, 48 days after tertiary grafts, or 4–6 months after grafts
of 104, 103, or 102cells. Sections were stained for HuNu with a FITC-conju-
gated second antibody and for cone markers with a Cy3-conjugated second
antibody. More than 8000 HuNu+cells were examined per marker.
Cells infected with pLKO lentiviral shRNA vectors were selected for 48–72 hr
to coverslips for immunostaining or TUNEL,or analyzedby FACS, qRT-PCR, or
immunoblotting. RB177 cells transduced with pLKO shCDKN2AARFconstructs
or controls were selected, reinfected with pLKO-shMDM2-377 or a control,
counted on days 1 and 5, and analyzed by qRT-PCR on day 4 and by TUNEL
on day 5 after infection. RB177 cells were transduced with UINZ-MDM2-
sh380R or the UGINZ vector, selected in G418, and transduced with pLKO-
shMDM2-380 as detailed in the Supplemental Data.
Cells were transfected with pRL-TK (Promega), P2-Luc, and P2-Luc-DRXR
(as described in the Supplemental Data), and the firefly:Ranilla luciferase ratio
was determined at 72 hr.
Chromatin precipitation (ChIP) assays were performed with RXRg antibodies
H-105 (Ab1) and Y-20 (Ab2) (Santa Cruz) and qRT-PCR as described in the
Supplemental Data include Supplemental Experimental Procedures, 13
figures, and two tables and can be found with this article online at http://
We thank J. Nathans and A. Swaroop for antibodies, H. te Riele for mouse
tumors, J. Blaydes for plasmids, M. Pappas for compiling Table S1, and
A. Koff for facilitating the completion of this study. This work was supported
by the Starr Foundation Tri-Institutional Stem Cell Initiative, Research to
Prevent Blindness, and the Fund for Ophthalmic Knowledge.
Received: May 14, 2008
Revised: January 9, 2009
Accepted: March 13, 2009
Published: June 11, 2009
Figure 7. Roles for Cone-Specific Transcription Factors in Retinoblastoma Cell MDM2 Expression, Proliferation, and Survival
(A) A human MDM2 P2 promoter element (second row) with identity to acone-specific RXR element (Danko et al., 2007)at each of six invariantpositions (top row,
from murine sequences (fifth row, aligned as in Figure S13). Human-to-mouse substitutions in P2-Luc-DRXR (fourth row) are underlined.
(B) Top: luciferase activities after transfection of P2-Luc and P2-Luc-DRXR in two experiments. Bottom: P2-Luc structure.
after transduction in the presence or absence of 60 nM pifithrin-a p-nitro, cyclic (EMD Biosciences) for the last 6 hr (c).
(D) Top: ChIP analysis of RB177, RB176, and Y79 assayed by qRT-PCR directly (input) or after immunoprecipitation with two RXRg antibodies. Values are the
ratio of the MDM2 PCR products R1, R2, or ex12 to an HNF4a product. Bottom: the human MDM2 locus, showing positions of the RXR site and PCR products.
(E) qRT-PCR analysis of RXRg RNA on day 4 and cell growth after plating on day 4 after transduction of Y79, RB176, or RB177 with RXRg shRNAs.
(F) qRT-PCR analysis of TRb2 RNA at day 5 and cell growth after plating on day 5 and counting on days 6, 11, and 16 after transduction of Y79, RB176, or RB139
cells with the indicated TRb shRNAs.
(G) Mean tumor mass 50 days after engrafting Y79 cells transduced with shTRb2, shTRb1+2, or scrambled shRNAs.
(H) Potential role of cone precursor signaling proteins in retinoblastoma tumorigenesis. Proteins in red are highly expressed during human cone precursor
Error bars indicate standard deviation, and asterisks indicate p < 0.05.
Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc. 1029
Abramson, D.H., and Gombos, D.S. (1996). The topography of bilateral retino-
blastoma lesions. Retina 16, 232–239.
Ajioka, I., Martins, R.A., Bayazitov, I.T., Donovan, S., Johnson, D.A., Frase, S.,
Cicero, S.A., Boyd, K., Zakharenko, S.S., and Dyer, M.A. (2007). Differentiated
horizontal interneurons clonally expand to form metastatic retinoblastoma in
mice. Cell 131, 378–390.
Bibb, L.C., Holt, J.K., Tarttelin, E.E., Hodges, M.D., Gregory-Evans, K., Ruth-
erford, A., Lucas, R.J., Sowden, J.C., and Gregory-Evans, C.Y. (2001).
Temporal and spatial expression patterns of the CRX transcription factor
and its downstream targets. Critical differences during human and mouse
eye development. Hum. Mol. Genet. 10, 1571–1579.
Bogenmann, E., Lochrie, M.A., and Simon, M.I. (1988). Cone cell-specific
genes expressed in retinoblastoma. Science 240, 76–78.
Chen, D., Livne-bar, I., Vanderluit, J.L., Slack, R.S., Agochiya, M., and
Bremner, R. (2004). Cell-specific effects of RB or RB/p107 loss on retinal
development implicate an intrinsically death-resistant cell-of-origin in retino-
blastoma. Cancer Cell 5, 539–551.
Chen, D., Opavsky, R., Pacal, M., Tanimoto, N., Wenzel, P., Seeliger, M.W.,
Leone, G., and Bremner, R. (2007). Rb-mediated neuronal differentiation
through cell-cycle-independent regulation of E2f3a. PLoS Biol. 5, e179.
Christophorou, M.A., Ringshausen, I., Finch, A.J., Swigart, L.B., and Evan, G.I.
(2006). The pathological response to DNA damage does notcontributeto p53-
mediated tumour suppression. Nature 443, 214–217.
Cobrinik, D. (2005). Pocket proteins and cell cycle control. Oncogene 24,
Cobrinik, D., Francis, R.O., Abramson, D.H., and Lee, T.C. (2006). Rb induces
aproliferativearrestand curtailsBrn-2expressioninretinoblastoma cells.Mol.
Cancer 5, 72.
Cornish, E.E., Xiao, M., Yang, Z., Provis, J.M., and Hendrickson, A.E. (2004).
The role of opsin expression and apoptosis in determination of cone types in
human retina. Exp. Eye Res. 78, 1143–1154.
Corson, T.W., and Gallie, B.L. (2007). One hit, two hits, three hits, more?
Genomic changes in the development of retinoblastoma. Genes Chromo-
somes Cancer 46, 617–634.
Danko,C.G.,McIlvain, V.A.,Qin,M.,Knox,B.E.,and Pertsov, A.M. (2007).Bio-
informatic identification of novel putative photoreceptor specific cis-elements.
BMC Bioinformatics 8, 407.
Dannenberg, J.H., Schuijff, L., Dekker, M., van der Valk, M., and te Riele, H.
(2004). Tissue-specific tumor suppressor activity of retinoblastoma gene
homologs p107 and p130. Genes Dev. 18, 2952–2962.
Dimaras, H., Khetan, V., Halliday, W., Orlic, M., Prigoda, N.L., Piovesan, B.,
Marrano, P., Corson, T.W., Eagle, R.C., Jr., Squire, J.A., et al. (2008). Loss
of RB1 induces non-proliferative retinoma; increasing genomic instability
correlates with progression to retinoblastoma. Hum. Mol. Genet. 17, 1363–
Donovan, S.L., Schweers, B., Martins, R., Johnson, D., and Dyer, M.A. (2006).
Compensation by tumor suppressor genes during retinal development in mice
and humans. BMC Biol. 4, 14.
Dowhan, D.H., Downes, M., Sturm, R.A., and Muscat, G.E. (1994). Identifica-
tion of deoxyribonucleic acid sequences that bind retinoid-X receptor-gamma
with high affinity. Endocrinology 135, 2595–2607.
Efeyan, A., Garcia-Cao, I., Herranz, D., Velasco-Miguel, S., and Serrano, M.
(2006). Tumour biology: Policing of oncogene activity by p53. Nature 443, 159.
Elison, J.R., Cobrinik, D., Claros, N., Abramson, D.H., and Lee, T.C. (2006).
Small molecule inhibition of HDM2 leads to p53-mediated cell death in retino-
blastoma cells. Arch. Ophthalmol. 124, 1269–1275.
Friend, S.H., Bernards, R., Rogelj, S., Weinberg, R.A., Rapaport, J.M., Albert,
D.M., and Dryja, T.P. (1986). A human DNA segment with properties of the
gene that predisposes to retinoblastoma and osteosarcoma. Nature 323,
Furukawa, T., Morrow, E.M., Li, T., Davis, F.C., and Cepko, C.L. (1999). Reti-
nopathy and attenuated circadian entrainment in Crx-deficient mice. Nat.
Genet. 23, 466–470.
Gallie, B.L., Campbell, C., Devlin, H., Duckett, A., and Squire, J.A. (1999).
Developmental basis of retinal-specific induction of cancer by RB mutation.
Cancer Res. 59, 1731s–1735s.
Gonzalez-Fernandez, F., Lopes, M.B., Garcia-Fernandez, J.M., Foster, R.G.,
De Grip, W.J., Rosemberg, S., Newman, S.A., and VandenBerg, S.R. (1992).
Expression of developmentally defined retinal phenotypes in the histogenesis
of retinoblastoma. Am. J. Pathol. 141, 363–375.
Hendrickson, A., and Provis, J. (2006). Comparison of development of the
primate fovea centralis with peripheral retina. In Retinal Development, E. Ser-
nagor, S. Eglen, B. Harris, and R. Wong, eds. (Cambridge: Cambridge Univer-
sity Press), pp. 126–149.
Hernando, E., Nahle, Z., Juan, G., Diaz-Rodriguez, E., Alaminos, M., Hemann,
M., Michel, L., Mittal, V., Gerald, W., Benezra, R., et al. (2004). Rb inactivation
control. Nature 430, 797–802.
Hurwitz, R.L., Bogenmann, E., Font, R.L., Holcombe, V., and Clark, D. (1990).
Expression of the functional cone phototransduction cascade in retinoblas-
toma. J. Clin. Invest. 85, 1872–1878.
Iaquinta, P.J., and Lees, J.A. (2007). Life and death decisions by the E2F tran-
scription factors. Curr. Opin. Cell Biol. 19, 649–657.
Kato, M.V., Shimizu, T., Ishizaki, K., Kaneko, A., Yandell, D.W., Toguchida, J.,
and Sasaki, M.S. (1996). Loss of heterozygosity on chromosome 17 and muta-
tion of the p53 gene in retinoblastoma. Cancer Lett. 106, 75–82.
Kleinerman, R.A., Tucker, M.A., Tarone, R.E., Abramson, D.H., Seddon, J.M.,
Stovall, M., Li, F.P., and Fraumeni, J.F., Jr. (2005). Risk of new cancers after
radiotherapy in long-term survivors of retinoblastoma: an extended follow-
up. J. Clin. Oncol. 23, 2272–2279.
Kondo, Y., Kondo, S., Liu, J., Haqqi, T., Barnett, G.H., and Barna, B.P. (1997).
Involvement of p53 and WAF1/CIP1 in gamma-irradiation-induced apoptosis
of retinoblastoma cells. Exp. Cell Res. 236, 51–56.
Kowalik, T.F., DeGregori, J., Leone, G., Jakoi, L., and Nevins, J.R. (1998).
E2F1-specific induction of apoptosis and p53 accumulation, which is blocked
by Mdm2. Cell Growth Differ. 9, 113–118.
in retinoblastoma. Nature 444, 61–66.
Lee, T.C., Almeida, D., Claros, N., Abramson, D.H., and Cobrinik, D. (2006).
Cell cycle-specific and cell type-specific expression of Rb in the developing
human retina. Invest. Ophthalmol. Vis. Sci. 47, 5590–5598.
Lomazzi, M., Moroni, M.C., Jensen, M.R., Frittoli, E., and Helin, K. (2002).
Suppression of the p53- or pRB-mediated G1 checkpoint is required for
E2F-induced S-phase entry. Nat. Genet. 31, 190–194.
Maat-Kievit, J.A., Oepkes, D., Hartwig, N.G., Vermeij-Keers, C., van Kamp,
I.L., and van de Kamp, J.J. (1993). A large retinoblastoma detected in a fetus
at 21 weeks of gestation. Prenat. Diagn. 13, 377–384.
MacPherson, D., Sage, J., Kim, T., Ho, D., McLaughlin, M.E., and Jacks, T.
(2004). Cell type-specific effects of Rb deletion in the murine retina. Genes
Dev. 18, 1681–1694.
Murine bilateral retinoblastoma exhibiting rapid-onset, metastatic progression
and N-myc gene amplification. EMBO J. 26, 784–794.
Mori, M., Ghyselinck, N.B., Chambon, P., and Mark, M. (2001). Systematic im-
munolocalization of retinoid receptors in developing and adult mouse eyes.
Invest. Ophthalmol. Vis. Sci. 42, 1312–1318.
Munier, F.L., Balmer, A., van Melle, G., and Gailloud, C. (1994). Radial asym-
metry in the topography of retinoblastoma. Clues to the cell of origin.
Ophthalmic Genet. 15, 101–106.
and Forrest, D. (2001). A thyroid hormone receptor that is required for the
development of green cone photoreceptors. Nat. Genet. 27, 94–98.
1030 Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc.
Ng, L., Ma, M.,Curran,T.,and Forrest, D.(2009). Developmental expression of
thyroid hormone receptor beta2 protein in cone photoreceptors in the mouse.
Neuroreport 20, 627–631.
Nork, T.M., Schwartz, T.L., Doshi, H.M., and Millecchia, L.L. (1995). Retino-
blastoma. Cell of origin. Arch. Ophthalmol. 113, 791–802.
Pacal, M., and Bremner, R. (2006). Insights from animal models on the origins
and progression of retinoblastoma. Curr. Mol. Med. 6, 759–781.
Rich, K.A., Zhan, Y., and Blanks, J.C. (1997). Migration and synaptogenesis
of cone photoreceptors in the developing mouse retina. J. Comp. Neurol.
Robanus-Maandag, E., Dekker, M., van der Valk, M., Carrozza, M.L., Jeanny,
J.C., Dannenberg, J.H., Berns,A., and te Riele,H. (1998). p107 is a suppressor
of retinoblastoma development in pRb-deficient mice. Genes Dev. 12,
Roberts, M.R., Hendrickson, A., McGuire, C.R., and Reh, T.A. (2005). Retinoid
X receptor gamma is necessary to establish the S-opsin gradient in cone
photoreceptors of the developing mouse retina. Invest. Ophthalmol. Vis. Sci.
Rodrigues, M.M., Rajagopalan, S., Lee, L., Nair, C.N., Advani, S.H., Donoso,
L., Chader, G.J., and Wiggert, B. (1992). Retinoblastoma: messenger RNA
for interphotoreceptor retinoid binding protein. Curr. Eye Res. 11, 425–433.
Schlamp, C.L., Poulsen, G.L., Nork, T.M., and Nickells, R.W. (1997). Nuclear
exclusion of wild-type p53 in immortalized human retinoblastoma cells.
J. Natl. Cancer Inst. 89, 1530–1536.
Shachaf, C.M., Gentles, A.J., Elchuri, S., Sahoo, D., Soen, Y., Sharpe, O.,
Perez, O.D., Chang, M., Mitchel, D., Robinson, W.H., et al. (2008). Genomic
and proteomic analysis reveals a threshold level of MYC required for tumor
maintenance. Cancer Res. 68, 5132–5142.
Skapek, S.X., Pan, Y.R., and Lee, E.Y. (2006). Regulation of cell lineage spec-
ification by the retinoblastoma tumor suppressor. Oncogene 25, 5268–5276.
Spencer, C., Pajovic, S., Devlin, H., Dinh, Q.D., Corson, T.W., and Gallie, B.L.
(2005). Distinct patterns of expression of the RB gene family in mouse and
human retina. Gene Expr. Patterns 5, 687–694.
Swain, P.K., Hicks, D., Mears, A.J., Apel, I.J., Smith, J.E., John, S.K., Hen-
drickson, A., Milam, A.H., and Swaroop, A. (2001). Multiple phosphorylated
isoforms of NRL are expressed in rod photoreceptors. J. Biol. Chem. 276,
Szel, A., Rohlich, P., Mieziewska, K., Aguirre, G., and van Veen, T. (1993).
Spatial and temporal differences between the expression of short- and
middle-wave sensitive cone pigments in the mouse retina: a developmental
study. J. Comp. Neurol. 331, 564–577.
Virtanen, I., Kivela, T., Bugnoli, M., Mencarelli, C., Pallini, V., Albert, D.M., and
Tarkkanen, A. (1988). Expression of intermediate filaments and synaptophysin
show neuronal properties and lack of glial characteristics in Y79 retinoblas-
toma cells. Lab. Invest. 59, 649–656.
Weinberg, R.A. (1995). The retinoblastoma protein and cell cycle control. Cell
Cell 137, 1018–1031, June 12, 2009 ª2009 Elsevier Inc. 1031