Tumor formation in mice with somatic inactivation of ...

Oncogene (2002) 21, 4635 – 4645 ª 2002 Nature Publishing Group All rights reserved 0950 – 9232/02 $25.00 www.nature.com/onc

Tumor formation in mice with somatic inactivation of the retinoblastoma gene in interphotoreceptor retinol binding protein-expressing cells Marc Vooijs1,5, Hein te Riele2, Martin van der Valk1,3 and Anton Berns*,1,4 1

Division of Molecular Genetics, The Netherlands Cancer Institute. Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands; Division of Molecular Biology, The Netherlands Cancer Institute. Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands; 3 Department of Experimental Animal Pathology, The Netherlands Cancer Institute. Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands; 4Centre for Biomedical Genetics, The Netherlands Cancer Institute. Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands 2

The retinoblastoma suppressor gene product Rb has been assigned a critical role in cell cycle regulation, the induction of differentiation, and inhibition of oncogenic transformation. Inheritance of a mutant RB allele in humans is responsible for bilateral retinoblastoma, a malignant tumor of the retina. Trilateral retinoblastoma (TRB) is a rare variant of familial retinoblastoma in which, in addition to retinal tumors, tumors develop from the pineal gland, an organ ontologically related to the retina. Germline inactivation of Rb in mice leads to midgestational lethality with defects in erythropoeisis and neurogenesis. This embryonic lethality prohibits the analysis of Rb function in selected cell types at later stages of development or in the adult. Here, we describe the Cre-LoxP mediated somatic inactivation of Rb in a subset of neuroendocrine cells, including photoreceptor cells. We observed neuroendocrine tumors of the pineal and pituitary gland. These tumors invariably showed inactivation of Rb and Trp53. Remarkably, loss of Rb in photoreceptor cells does not lead to retinoblastoma or any phenotypic changes, not even when photoreceptor cells are made deficient in Rb, p107 and Trp53. Our results highlight the important differences that exist in tumor susceptibility between mice and man (e.g pineal gland) and question the photoreceptor cell origin of human retinoblastoma. Oncogene (2002) 21, 4635 – 4645. doi:10.1038/sj.onc. 1205575 Keywords: Rb; p107; Trp53; conditional knockout; retinoblastoma; photoreceptor cells Introduction Retinoblastoma is a malignant tumor of the retina that is caused by inactivating mutations in the retinoblastoma susceptibility gene (RB) in humans. Inheritance of a

*Correspondence: A Berns; E-mail: [email protected] 5 Current address: Department of Molecular Biology and Pharmacology, Washington University School of Medicine, Box 8103, 660 S. Euclid Avenue, St. Louis, Missouri, MO 63110, USA; E-mail: [email protected]. Received 22 January 2002; revised 28 March 2002; accepted 15 April 2002

mutant RB allele is responsible for familial forms of this disease and increases the risk of developing secondary malignancies (Moll et al., 1997). Trilateral retinoblastoma (TRB) is a rare variant of familial retinoblastoma in which, in addition to bilateral retinoblastoma, tumors develop from the pineal gland (Jakobiec et al., 1977). Mutational analysis of human tumors has shown that components of the Rb pathway are found inactivated in virtually all tumor types (Weinberg, 1995). Rb encodes a nuclear phosphoprotein that integrates extracellular signals to regulate cell cycle progression. The hypophosphorylated, active form of Rb complexes with E2F/DP heterodimeric transcription factors and represses transcription directly by binding to promoters carrying E2F-binding sites. Cell cycle progression through G1 requires functional inactivation of Rb by phosphorylation. Consecutive phosphorylation of Rb by Cyclin D/CDK4/6 and Cyclin E/CDK2 results in the release of bound E2F/DP and promotes progression into S phase (Harbour and Dean, 2000). Another level of regulation is mediated through the CIP/KIP and INK4 families of proteins. These proteins act to assemble as well as to inhibit the Cyclin/CDK complexes upon mitogenic signaling (Sherr and Roberts, 1999). Rb can also actively promote differentiation of several cell types (Lipinski and Jacks, 1999). The Rb family of pocket proteins comprises three functionally related members; Rb, p107 and p130 with overlapping and distinct functions (Mulligan and Jacks, 1998). Of these, RB is predominantly found mutated in human tumors but mutations have also been found in p130 (Claudio et al., 2000). Mice homozygous for a null mutation in Rb show embryonic lethality with defects in hematopoiesis, neurogenesis, and terminal lens differentiation (Clarke et al., 1992; Jacks et al., 1992; Lee et al., 1992; Robanus-Maandag et al., 1994). Heterozygous Rb+/7 mutant mice primarily develop melanotroph tumors of the pituitary gland and thyroid hyperplasia with loss of the wild type Rb allele (Jacks et al., 1992). However, no retinoblastomas develop in Rb+/7 mice. Chimeric mice composed of Rb-deficient and wild type cells are viable and rapidly develop pituitary tumors and thyroid adenomas and show a defect in terminal differentiation of lens epithelial cells (Robanus-Maandag et al., 1994; Williams et al., 1994b). While chimeric mice do not

Neuroendocrine tumors in mice with somatic Rb inactivation M Vooijs et al

4636 Table 1

Tumor spectrum in mice with somatic Rb loss in IRBP-expressing cells

RbF19/F19 Mean lifespan in days (+s.e.m.) Anterior lobe tumor (%) Melanotroph tumor (%) Bi-pituitary tumor (%) Pituitary tumor (%) Pinealoblastoma (%) Other tumors (%) No tumors found (%) Number of mice

IC64 350 15 1 2 1

(137) (58) (4) (8) (4) 0 6 (23) 1 (4) 26

IC47 411 10 12 3 5

(107) (30) (36) (9) (15) 0 1 (3) 2 (6) 33

Trp53+/7:IC64

Trp53+/7:IC47

Trp537/7:IC64

264 (123) 32 (63) 1 (2) 0 4 (8) 8 (15) 6 (11) 5 (10) 51

278 (128) 5 (50) 0 3 (30) 1 (10) 0 1 (10) 0 10

88 (29) 2 (15) 0 0 0 8 (62) 11 (85) 2 (15) 13

Summary of histopathological findings in IC:RbF19/F19;Trp53 mutant mice. Mice were examined macroscopically and microscopically for the presence of tumors. The mean lifespan and the mean error are indicated and the incidence of each lesion is expressed as the percentile of the total number of animals in a specific group. Indicated are animals that developed both a melanotroph as well as an anterior lobe tumor (bi-pituitary tumor). Pituitary tumor indicates macroscopically a pituitary tumor was seen but the type was not confirmed histologically. In addition to the groups shown in Table 1, a cohort of Trp537/7 (n=58) mice was monitored for tumor development until moribund. Trp537/7 mice developed haemangiosarcomas (69%), lymphosarcomas (40%) and teratocarcinomas (10%) with an average latency of 96 (+/735) days and often carried several different tumors

develop retinoblastomas, Rb-deficient retinoblasts show a reduced contribution to the adult retina (Robanus-Maandag et al., 1994). The requirement for Rb function during retinal development is also supported by the phenotype of cyclin D1 null mice that suffer from retinal hypoplasia (Sicinski et al., 1995). Cyclin D1 is highly expressed in the embryonic retina and this correlates with phosphorylation of Rb. Photoreceptor cell degeneration in cyclin D1 null mice can be rescued by cyclin E when expressed from the cyclin D1 locus or in a p27KIP1-deficient background (Geng et al., 1999, 2001). This supports the notion that the retinal hypoplasia in cyclin D1 null mice is a defect in G1 cell cycle control that impinges on Rb regulation. The analysis of Rb+/+;Rb7/7 chimeras indicates that, in contrast to man, inactivation of Rb in mice does not lead to retinoblastoma and that additional mutations are required for tumor formation. One of these is the Rb-related tumor suppressor p107, as chimeric mice partly composed of cells lacking both p107 and Rb, develop retinoblastoma originating from the inner nuclear layer (INL) with amacrine cell characteristics (Robanus-Maandag et al., 1998). However, the majority of human retinoblastomas show characteristics of photoreceptor cell differentiation, express markers for cone- or rod photoreceptors and have an intact phototransduction cascade suggesting that they originate from the outer nuclear layer (ONL) of the retina (Bogenmann et al., 1988; GonzalezFernandez et al., 1992). Additionally the benign variant of retinoblastoma, retinocytoma, shows photoreceptor cell differentiation despite its location within the INL (Gallie et al., 1999; Margo et al., 1983). Furthermore, in mice, expression of SV40 large T-antigen (Tag) using the Interphotoreceptor Retinol Binding Protein (IRBP) promoter leads to retinoblastoma with photoreceptor cell characteristics (al-Ubaidi et al., 1992; Howes et al., 1994a). However, the pleiotropic effects of Tag make it difficult to dissect the requirement for the individual gene defects for retinoblastoma formation in mice. Oncogene

To further define the role of each of these tumor suppressor genes in retinoblastoma formation we studied the consequences of photoreceptor-specific inactivation of Rb in conjunction with defects in p107 and Trp53. This is now possible using the CreloxP and Flp-FRT methodology, which permits the spatially and temporally controlled somatic introduction of mutations in mice (Metzger and Feil, 1999). Utilizing these systems, we generated mice carrying a conditional allele for Rb and crossed these with mice expressing a cre transgene under the control of the IRBP gene promoter which is active from E13.5 onwards in rod- and cone photoreceptor cells of the outer nuclear layer (ONL) and in pinealocytes of the pineal gland (al-Ubaidi et al., 1992; Liou et al., 1990). Using this approach, we generated a mouse model for neuroendocrine tumors and found strong collaboration between Rb and p53 loss of function mutations in these tumors. Our results exemplify the utility of conditional mutagenesis in generating specific tumor prone phenotypes in mice at high incidence. Results Cre-mediated inactivation of Rb results in a null allele To circumvent the embryonic lethality associated with an Rb null mutation, we utilized the Cre-loxP system to achieve cell type specific inactivation of Rb (Sauer and Henderson, 1988). We flanked exon 19 comprising part of the A-pocket of Rb with LoxP sites (i.e. RbF19, Figure 1a) following a strategy similar to the generation of mice carrying the recognition sequences for FLP recombinase (FRT) around exon 19 (Vooijs et al., 1998). Southern blot analysis confirmed targeting at the Rb locus and the presence of both loxP sites surrounding exon 19 in embryonic stem cells (Figure 1b). Mice homozygous for the conditional Rb allele (RbF19/F19) were derived from these clones and were viable, fertile, and showed no spontaneous tumor predisposition similar to RbFRT


4637

Figure 1 Generation and characterization of the conditional Rb allele. (a) ‘Hit and Run’ strategy that introduces loxP sites into the murine Rb locus surrounding exon 19 (Vooijs et al., 1998). ES cells were electroporated with a BamH1 linearized Rb-F19 targeting vector and selected for HAT resistance (HATR, step I). Homologous recombination at the Rb locus resulted in HATR clones that carry a duplication of the targeted allele (Rb-dup). Spontaneous intrachromosomal recombination (step II) resulted in reversion to the wild type (at a.) or the conditional mutant RbF19 allele (at b.) with concomitant loss of the plasmid backbone and HPRT selectable marker leading to 6TGR resistance. (b) Predicted structure of the mouse Rb locus and the targeted LoxP modified Rb locus (RbF19). The presence of loxP sites disrupts the NheI and EcoRV restriction sites, and is used to identify targeted clones. Southern blot analysis on 6TGR ES cell clones shows detection of the 3’-loxP site using probe C on EcoRV digested ES cell DNA. A wild type 7 kb fragment and a novel 10 kb fragment for the 3’-loxP insertion are present (b). Lower panel shows identification of the 5’loxP insertion using probe B after an EcoRI/NheI double digest resulting in a 10 kb fragment for the RbF19 allele and a 4.5 kb fragment for the wild type allele. RI; EcoRI; Nh, NheI; RV, EcoRV

mice demonstrating that the RbF19 allele is phenotypically wild type. To address whether an exon 19 deleted Rb (RbD19) allele acts as a null allele, we deleted Rb in the germ line of mice by crossing RbF19/F19 mice with a Credeleter strain to obtain RbD19/+ mice (M Vooijs and A Berns, unpublished). In the offspring from RbD19/+ matings we never obtained RbD19/D19 mice suggesting that these mice fail to develop to term. Further analysis showed that RbD19/D19 embryos die in utero between day E13 and E15.5 with phenotypic aberrations reminiscent of a null mutation in exon 19 of Rb (Clarke et al., 1992, not shown). Conditional inactivation of Rb in the retina To direct Rb inactivation to photoreceptor cells we generated cre transgenic mice using the human IRBP promoter. Two lines of IRBP-cre (IC) transgenic mice: IC64 and IC47 respectively were used that have

a different transgene copy number (Akagi et al., 1997). First, we examined the efficiency and tissue-specificity of Cre recombinase by breeding IC64 and IC47 mice with the AcZl Cre reporter strain (Akagi et al., 1997). The AcZL reporter line allows in situ identification of Cre activity by excision of a floxed STOP cassette and concomitant activation of lacZ. X-gal staining of retinas from bi-transgenic offspring from both lines indicated a mosaic pattern of Cre recombinase activity in both the peripheral and the posterior part of the ONL at 2 weeks of age (Figure 2a,b). X-gal staining was only seen in the outer segments (OS) of the ONL and the outer plexiform layer (OPL), all of which harbor photoreceptor cells or their dendritic protrusions. No staining was seen in other cell types in the retina. To determine whether Cre was active in both rods and cones we generated bi-transgenic mice also carrying the recessive retinal degeneration (rd) allele resulting in the selective loss of rods but not Oncogene


4638

Figure 2 Cre recombinase activity and expression in tissues from IRBPcre mice (a – h). lacZ staining reflecting Cre recombinase mediated switching of a reporter gene in double transgenic IC64 : AcZL mice in (a) the peripheral and (b) posterior ONL of the retina and (c) in the rd/rd background at two months of age and (d) lacZ staining throughout the pineal gland of adult mice. Immunohistochemical staining for Cre in IC47 retinas at day E14.5 (e), at PND7 in peripheral (f), and posterior retina (g) and at 3 weeks (h) in IC47 : RbF19/+ mice. The effect of Rb inactivation in the retinas of IC47 : RbF19/F19 mice at PND7 (i – m). HE-staining (i) Cre expression in the ONL (j) lack of BrdU incorporation (k) and TUNEL labeling (l). Arrows indicate TUNEL labeled cells. Note apoptotic cell death in the INL and absence of TUNEL or BrdU positive cells in the posterior ONL. Expression of Cre results in the deletion of Rb in ONL of the retina at PND1 shown by microdissection PCR on the ONL of newborn IC47 : RbF19/+ and non-transgenic RbF19/F19 retinas (m). PCR fragment sizes are 678 bp (Rbwt), 746 (RbF19) and 321 bp (RbD19). RbF19/D19 is liver DNA isolated from a RbF19/D19 mouse. Note significant deletion of RbF19 in cre transgenics but not in non-transgenic RbF19/F19 mice. INL (Inner Nuclear Layer), OPL (Outer Plexiform Layer), ONL (Outer Nuclear Layer) and OS (Outer Segments). Sections are counterstained with nuclear fast red (a – c) and HE (d – k) or methyl-green (l). Original magnifications are a – d; 106 and e – i; 206

cones by 4 weeks of age (Carter-Dawson et al., 1978). As expected double transgenic IC : AcZL mice from 2 weeks onwards from both lines showed a progressive degeneration of rods from the ONL and a concomitant reduction in the number of X-gal stained cells when maintained on a rd/rd genetic background. However, some photoreceptor cells still stained at an age of several months demonstrating that Cre was also active in cones (Figure 2c) (Akagi et al., 1997). Both IRBPcre transgenic lines showed Cre activity in the pineal gland as monitored by X-gal staining in bitransgenic offspring albeit to a different extent: IC64 staining virtually all pinealocytes (Figure 2d) whereas IC47 showed a more scattered X-gal staining (not shown). In IC47 retinas, Cre was first observed at day E14.5 in the lower part of the ventricular layer and expression continued to increase up to birth, showing a scattered expression pattern throughout the ONL as shown by immunohistochemical staining using a Cre Oncogene

antibody (Figure 2e). One week after birth, photoreceptor cells in the peripheral- and in the posterior retina expressed Cre throughout the ONL, and this expression pattern was maintained for at least another 2 weeks after birth (Figure 2f – h). In addition, a very small number of cells within the OPL near the INL showed Cre staining. Line IC64 did not show expression at day E14.5 and started to express around birth (not shown). Rb function is not required for rod- or cone photoreceptor cell differentiation Loss of Rb function in mice has been reported to result in aberrant proliferation followed by apoptosis in several tissues including photoreceptors of the retina (Howes et al., 1994b; Robanus-Maandag et al., 1998). To investigate the consequences of Rb loss on photoreceptor cell development and tumor formation


4639

we generated IC47 : RbF19/F19 mice and compared the retinas of these mice with non-transgenic littermates at PND7 when the majority of photoreceptor cells are terminally differentiated (Young, 1985a). We concentrated on the analysis of IC47 because of the earlier onset of Cre expression in the retinas of these mice. A histological examination of retinas from control versus IC47 : RbF19/F19 mice did not reveal any differences in cyto-architecture (Figure 2i) nor in proliferation as determined by staining for BrdU (Figure 2k). At PND7 a limited but comparable number of BrdU stained cells was observed in the peripheral ONL retinas from IC47 : RbF19/F19 and wild type mice (not shown) where proliferation ceases at PND11. During normal retinal development, very few photoreceptor cells in the ONL are eliminated by apoptosis. In the ONL of IC47 : RbF19/F19 mice analysed at PND7, no evidence was seen for increased apoptosis compared to control littermates (Figure 2l). In contrast, TUNEL labeling readily identified apoptotic cells in the INL of control and IC47 : RbF19/F19 mice at this age reflecting normal apoptotic cell death (Young, 1984). Apoptosis in the INL decreased after this time point and could not be detected in the INL or the ONL of control and IC47 : RbF19/F19 mice three weeks after birth (not shown). A further comparison of retinas from transgenic and non-transgenic RbF19/F19 mice at day E14.5, PND1, PND7 and at 3 weeks after birth did not reveal any obvious differences in terms of morphology, proliferation, or apoptosis (not shown). We did not observe any retinal tumors in adult mice over 1 year old that succumbed from other tumors (see below). Using Cre staining we compared the number and location of Cre positive cells between the retinas of IC47 : RbF19/+ and IC47 : RbF19/F19 mice at PND7. We found that Cre was expressed in a stochastic fashion throughout the posterior- and peripheral ONL regardless of the genotype, indicating that no selection occurs against Cre expressing cells that carry a bi-allelic Rb conditional allele (compare Figure 2g and j). To ascertain that under conditions of reporter gene switching, Rb was also inactivated by Cre-mediated recombination, we performed a PCR analysis on the ONL from normal and transgenic IC47 : RbF19/+ newborn retinas obtained by microdissection from HE-stained sections. This analysis demonstrated that a 321 bp fragment indicative of exon 19 excision by Cre (yielding RbD19) could be readily amplified from IC47 : RbF19/+ but not from control non-transgenic retinas from RbF19/F19 mice (Figure 2m). All together these results indicate Cre expression in IC47 mice starts at the onset of photoreceptor differentiation and continues to be expressed at least until 3 weeks after birth in cones as well as rods mediating switching of both a reporter substrate and the RbF19 allele. To investigate the possibility that the absence of pathology in Rb-deficient photoreceptor cells is caused by functional compensation of the Rb-related family member p107 (Robanus-Maandag et al., 1998), we generated mice lacking both Rb and p107 in photoreceptor cells. As expected, the retinas of p107-deficient

mice showed comparable levels of proliferation and apoptosis as in wild type mice. Remarkably, inactivation of Rb in the ONL of the retinas of p107-deficient mice (IC47 : RbF19/F19;p1077/7) mice did not lead to any changes in proliferation or apoptosis at PND7. Accordingly, Cre expression in these retinas was comparable to control IC47 : RbF19/+;p1077/7 retinas at PND7 (not shown). Rb loss in IRBP-expressing cells leads to neuroendocrine tumors To investigate the consequence of somatic Rb inactivation on tumor formation in the retina and the pineal gland we examined a cohort of IC64 : RbF19/F19 (n=26) and IC47 : RbF19/F19 (n=33) mice, bi-weekly for the development of tumors. These mice with an average life span of 350 and 411 days respectively, did not develop retinoblastomas or pineal gland tumors but succumbed to pituitary gland tumors instead (Table 1 and Figure 4a). Tumors arose from the anterior- and intermediate lobe of the pituitary gland and infrequently both tumor types developed within one animal (Figure 3a – c). Anterior lobe pituitary tumors could be classified histologically into two groups, a rather monomorph cavernous type of tumor with a tendency to hemorrhage and a second consisting of highly pleiomorphic tumor cells demonstrating anisonucleosis, hyperchromatic nuclei and poikilonucleosis sometimes with concomitant mineralization and occasionally tumors were composed of both histological types (Figure 3d). To characterize the origin of the anterior lobe pituitary tumors we used immunohistochemistry against pituitary hormones expressed in the normal anterior lobe but did not find any evidence for hormone production in nine tumors tested. Intermediate lobe pituitary tumors expressed a-MSH demonstrating to their melanotroph origin as previously reported (not shown). Most anterior lobe tumors expressed Synaptophysin indicative of some degree of neuroendocrine differentiation (Figure 3e). Rb and Trp53 loss synergize in neuroendocrine tumor formation Many human tumors show mutations in both RB and TP53 pathways and in mice mutations in Trp53 and Rb strongly cooperate in tumor development in a wide range of tissues. To investigate if Trp53 loss collaborates with Rb loss on anterior lobe pituitary tumor formation we crossed IRBPcre : RbF19/F19 mice with Trp53 mutant mice. Both IC64 : RbF19/F19;Trp53+/ 7 (n=51) as well as IC47 : RbF19/F19;Trp53+/7 (n=10) mice developed tumors with a significantly reduced mean latency of 264 and 278 days, respectively (Table 1 and Figure 4a). In addition to anterior lobe tumors, IC64 : RbF19/F19;Trp53+/7 mice also developed tumors originating from the pineal gland (Table 1 and Figure 3f – k). Importantly, pituitary and pineal tumors did not arise spontaneously in Trp53+/7 mice that succumbed from tumors with a mean latency of 517 Oncogene


4640

Figure 3 Histopathology in IC64 : RbF19/F19 (a – e) and IC64 : RbF19/F19:Trp537/7 (f – k) mice. Shown are examples of an anterior lobe tumor (a) a bi-pituitary tumor (b) an intermediate lobe tumor (c) a biphasic anterior lobe tumor containing highly pleiomorphic nuclei and (d) synaptophysin immunostaining in anterior lobe tumors (e). Pineal gland tumors at low magnification (f), invasion into the surrounding brain parenchyma (g) and presence of Flexner-Wintersteiner rosettes (h). HE staining of pineal gland tumor (i), Synaptophysin (j) and GFAP (k) immunostaining on adjacent sections. Note that pineal gland tumors stain positive for Synaptophysin but not GFAP whereas remnants of the normal pineal gland (arrow) stain positive for both Synaptosphysin and GFAP. Magnification h; 1006, a, c; 406, d, e, g; 206, i – k; 106, b; 2.56

days (Figure 4a). Histological examination showed that pinealoblastomas locally invaded the brain and showed a highly undifferentiated aggressive and anaplastic cytological appearance and a high degree of pyknotic nuclei (Figure 3f – i). Within these tumors occasionally regions of well-differentiated areas could be found resembling the Flexner-Wintersteiner rosettes characteristic for human pinealoblastoma and retinoblastoma and indicative for photoreceptor cell differentiation (Figure 3h). Pineal tumors expressed Synaptophysin but not the glial marker GFAP (Figure 3j,k). In addition, tumors showed focal expression of IRBP and S-antigen (not shown) both also expressed in the normal pineal gland (Blackshaw and Snyder, 1997). Oncogene

Loss of heterozygosity of Trp53 in neuroendocrine tumors with Rb loss The fact that both strains of IC : RbF19/F19;Trp53+/7 mice showed a significantly reduced median survival compared to both groups of IC : RbF19/F19 mice (209 days vs 389 days, P50.001) indicated that Rb and Trp53 act synergistically to suppress tumor formation in the pituitary and the pineal gland. To substantiate this, we determined whether tumors had undergone loss of heterozygosity of the wild type Trp53 allele and Cre-mediated deletion of Rb. Both in anterior lobe tumors (10 out of 10) and in pinealoblastomas (4 out of 4) we found a major reduction or complete absence of fragments specific


4641

Figure 4 Tumor incidence in IRBPcre (IC) mice with conditional Rb inactivation and Trp53 mutation. Graph summarizing the tumor-free survival IC : RbF19/F19, IC : RbF19/F19;Trp53+/7, IC : RbF19/F19;Trp537/7 and non-transgenic Trp537/7 and Trp53+/7 mice (a). The median tumor free survivals were 389 days (IC : RbF19/F19), 209 days (IC : RbF19/F19;Trp53+/7), 86 days (IC64 : RbF19/F19;Trp537/7), 91 days (Trp537/7) and 516 days (Trp53+/7), respectively. Survival curves for IC64 and IC47 strains have been grouped. Number of mice used in each group is indicated. Note significant acceleration of tumor development in mice with combined mutations in Rb and Trp53. Representative PCR analysis to determine the loss of Rb (lanes 1 – 4), p107 (lanes 5 – 8) and Trp53 alleles (lane 9 – 12) in anterior lobe tumors (ALT, b) and pinealoblastomas (PB, c) from IC64 : RbF19/F19;p107+/7;Trp53+/ 7 mice. Note that Cre-mediated loss of Rb is observed in tumors but not normal (N) tissue and loss of the Trp53 but not the p107 wild type allele is only seen in tumors. Note Cre-mediated Rb loss in pinealoblastoma (PB4) but not in haemangiosarcoma (HS4) from the same IC64 : RbF19/F19;Trp537/7 animal (c)

for the wild type Trp53 allele as well as significant Cre-mediated loss of Rb as examined by allelespecific PCR (Figure 4b,c). Tumor formation was further accelerated in IC64 : RbF19/F19;Trp537/7 mice that developed pineal gland tumors (8 out of 13) and a small number of anterior lobe tumors (2 out of 13) with an average latency of 88 days (Table 1 and Figure 4a). Remarkably, no retinoblastomas developed in IC64 : RbF19/F19;Trp537/7 mice that also lacked one (7 out of 13) or both copies of p107 (2 out of 13), respectively. IC64 : RbF19/F19;Trp537/7 mice also developed several tumor types concordant with the tumor spectrum of Trp537/7 mice (n=56, Table 1). The earlier appearance of pineal- and anterior lobe tumors in IC64 : RbF19/F19;Trp537/7 mice further supports the synergistic effects of Rb and Trp53 mutations in these tumors. However, since the median tumor free survival of Trp537/7 vs IC64 : RbF19/F19;Trp537/7 mice did not markedly differ (91 days vs 86 days) we conclude that these mice primarily succumb to tumors that do not depend on Rb loss. We also investigated whether p107 loss contributed to the development of pineal gland or anterior lobe pituitary tumors. No effect was seen of p107 loss in IC64 : RbF19/F19;p1077/7 (n=13) mice on the rate of tumor development or on the tumor spectrum (not shown) consistent with the fact that we did not find loss of the p107 wild type allele in pineal- or anterior lobe tumors from IC64 : RbF19/F19;Trp53+/7;p107+/7 mice (Figure 4b,c).

Discussion Conditional mutagenesis The generation of knockout mice has provided insight into the role of Rb in specific developmental processes as well in suppressing cancer in mice (reviewed in Vooijs and Berns, 1999). However, the embryonic lethality of Rb null mice prohibits the phenotypic analysis of Rb-deficiency in adult mice. Unlike humans where RB heterozygosity invariably leads to retinoblastoma, in mice this leads to intermediate lobe pituitary tumors and thyroid hyperplasia with high incidence. Conceivably, Rb+/7 mutant mice succumb from these tumors before the onset of other malignancies in which RB1 loss is also found in man. This has prohibited a detailed analysis on the role of Rb in tumor onset and progression of specific tumor types in mice. Conditional somatic mutagenesis circumvents this problem as it permits the introduction of mutations in a subpopulation of cells such that specific tumors can be generated at high incidence. Furthermore, since one can choose conditions in which gene inactivation is occurring in only a small fraction of the cells it is possible to better mimic the onset of sporadic cancer in man. Loss of Rb function in photoreceptors Our first concern was to show that Cre-mediated excision of exon 19 of the RbF19 allele results in a null allele. Removal of exon 19 is expected to result in a truncation of Rb at exon 22 disrupting the B-pocket Oncogene


4642

essential for tumor suppressor activity. As expected, the truncated RbD19 allele phenocopies an Rb null mutation in mice (Clarke et al., 1992) consistent with loss of tumor suppressor function in vivo (Marino et al., 2000). To investigate the requirement for Rb function in photoreceptor cell development and tumorigenesis we disrupted Rb in cone- and rod photoreceptors of the retina using IRBP driven Cre recombinase expression in mice carrying conditional Rb alleles. In these mice Cre recombinase expression initiates in the embryonic retina at day E14.5 and continues to be active in rods and cones within the ONL up to at least PND 21 when all photoreceptors are terminally differentiated (Figure 2). The timing and tissue distribution of Cre expression in these mice resembles the normal pattern of endogenous IRBP expression in mice (Liou et al., 1994). Inactivation of Rb from day E14.5 onwards in IRBP-expressing cells did not lead to any obvious pathology in the retina. The time of onset and cell-type specificity of Cre expression reported here indicates that a significant proportion of rods as well as cones have become Rb-deficient during their genesis and persist in the adult retina (Carter-Dawson and LaVail, 1979; Young, 1985a,b). Thus, Rb function does not seem to be required to maintain the postmitotic state of photoreceptor cells nor did bi-allelic Rb loss predispose to retinal tumors. We showed that this could not be attributed to functional compensation by the Rb family member p107 since concurrent loss of Rb and p107 did not affect the proliferation or terminal differentiation of cone- or rod photoreceptor cells in the postnatal retina nor did it lead to the development of retinoblastomas. Our findings contrast with the occurrence of retinoblastoma in IRBP-Tag transgenic mice, the depletion of Rb (p107)-deficient photoreceptor cells in Rb7/7/Rb+/+ chimeric mice and the development of retinal dysplasia in Rb+/7;p1077/7 or Rb+/7;Trp537/7 mutant mice (al-Ubaidi et al., 1992; Lee et al., 1996; Robanus-Maandag et al., 1998; Williams et al., 1994a). Therefore, important differences exist between somatic Rb loss in our model and previously described models. These differences can be explained in several ways: (i) Rb loss in our model occurs in a very limited number of cells. Our data indicate that Cre is robustly expressed, mediates switching of a reporter substrate and deletion of Rb in the ONL of the retina at PND7, although it is difficult to predict the actual number of Rb-deficient cells. Based on our recent findings that the efficiency of recombination on the RbF19 allele in vivo is several fold higher than that of other conditional alleles tested including two reporter substrates we conclude that our reporter analysis underestimates the actual number of Rb deficient cells in the ONL of these mice (Vooijs et al., 2001). Furthermore, stochastic expression or silencing of the AcZl reporter transgene may underestimate the number of cells with Cre recombinase activity. In addition, Cre-mediated Rb inactivation will occur at a much higher frequency than the spontaneous loss of the wild type allele presumed to result in Oncogene

retinal dysplasia in Rb+/7;p1077/7 mice (Lee et al., 1996). Consequently, it is extremely unlikely that the lack of any pathology is due to an insufficient number of Rb-deficient photoreceptors. (ii) If Rb inactivation occurs relatively late, sufficient Rb may remain to assure terminal differentiation and prevent tumor formation during a critical time window. Because Rb is a relatively stable protein, cells that have undergone Cre-mediated excision of exon 19 may have to undergo a substantial number of divisions before full depletion of Rb occurs (Mihara et al., 1989). This could explain the normal development of Rb-deficient cones that become postmitotic already around the onset of Cre expression and may have sufficient Rb to complete terminal differentiation. However, rods then will still go through a major wave of expansion and as a result, a significant fraction of rods should be Rb-deficient around birth. In our view, this justifies the conclusion that terminal differentiation of rods can occur in the absence of functional Rb and/or p107 protein. Therefore, it is unlikely that loss-of-Rb blocks photoreceptor cell differentiation, which in turn could foster retinoblastoma development. (iii) Additional mutations are required to elicit photoreceptor cell tumors. Our results show that loss of Rb, p107 and Trp53 in photoreceptor cells does not lead to tumors whereas expression of Tag does (al-Ubaidi et al., 1992; Howes et al., 1994a). Formally, we cannot rule out that other functions of Tag may be required for tumor development. Among these may be inactivation of the Rb family member p130 not affected in our experiments. (iv) The IRBPexpressing cell is not the precursor cell for retinoblastoma. Human retinoblastomas are thought to originate from a progenitor capable of photoreceptor differentiation or Mu¨ller cell differentiation (Gonzalez-Fernandez et al., 1992; Nork et al., 1995). Interestingly lineage marking has shown that both cell types may share a common precursor in the rodent retina and that competence is maintained even after terminal mitosis (Cepko et al., 1996). Case reports on benign retinoblastoma suggest that these tumors may originate from the INL while showing resemblance with photoreceptor cells (Gallie et al., 1999; Margo et al., 1983). In keeping with this, chimeric mice deficient for both Rb and p107 develop retinoblastomas originating from the inner- but not the outer nuclear layer. If this hypothesis is correct, no tumors can be expected, as the INL cells that give rise to retinoblastoma in chimeric Rb7/7;p1077/7/Rb+/+;p107+/+ mice do not express IRBP and therefore, have not lost Rb in our system. Thus, it is possible that IRBP is not expressed in the retinoblastoma precursor cell but that expression of IRBP frequently seen in human retinoblastoma reflects the capacity of tumor cells to transdifferentiate. This is currently our favorite explanation and in accordance with the detailed analyses performed in chimeric mice (Robanus-Maandag et al., 1998). Targeting Rb loss to different retinal compartments is required to resolve these issues. (v) Finally, it has been reported that multiple loci in the FVB genetic background suppress retinoblastoma formation in mice


4643

predominantly caused by a homozygous mutation in rod cGMP-phosphodiesterase (rd) allele (Griep et al., 1998). To minimize the effects of genetic background on tumor predisposition in our mice that are of a mixed 129 Ola/FVB background, mice were also kept heterozygous for the rd allele. However, the presence of other modifier alleles from FVB or Ola129 may also contribute to the absence of phenotypic aberrations. Currently, we cannot exclude this possibility. Several of the explanations mentioned above can be experimentally tested and this will be the focus of future work. Rb and Trp53 loss synergize in tumor formation Despite the lack of retinoblastomas in mice with photoreceptor cell specific inactivation of Rb, mice developed pituitary- and pineal gland tumors. The development of pineal gland tumors is of interest as pinealoblastomas occur at low frequency as second primary tumors in patients with hereditary retinoblastoma. This syndrome, referred to as trilateral retinoblastoma (TRB) also includes other intracranial malignancies of primitive neuroectodermal origin (Jakobiec et al., 1977). The close phylogenetic relationship between the pineal gland and the photoreceptor cells of the retina has led to the suggestion that these organs share a common precursor, from which retinoblastoma may arise. Pinealoblastomas also arise in IRBP-Tag transgenic mice and in Rb+/7;Trp537/7 mice (Howes et al., 1994a; Williams et al., 1994a). We extend these observations by showing that Trp53 loss is rate-limiting in the development of pineal tumors in mice. The high incidence of pineal tumors in Rb conditional/Trp53 mutant mice in the absence of retinal tumors, illustrates a fundamental difference with the human condition: RB heterozygosity in man invariably leads to retinoblastoma but rarely to pineal tumors, whereas loss of Rb and Trp53 are rate-limiting in the development of pinealoblastoma but not for retinoblastoma formation in mice. Remarkably, TP53 is rarely found mutated in human pinealoblastoma (Tsumanuma et al., 1995) but invariably lost in the pineal tumors we have studied. In view of this, it would be worth re-examining whether other TP53 pathway components are mutated in pineal gland tumors in man. IRBP targeted Rb loss also resulted in the development of anterior lobe tumors with neuroendocrine characteristics. The development of these tumors was dramatically accelerated in the Trp53 heterozygous background and consistently showed loss of the remaining Trp53 wild type allele. The non-hormone producing pituitary gland tumors we found differed from the spontaneous anterior lobe tumors observed in a fraction of Rb+/7 mice (Nikitin et al., 1999). As Rb loss is a rate-limiting step in pituitary tumor development in mice, it is possible that Cre-loxP mediated biallelic Rb loss results in more advanced tumors that are less well differentiated. Interestingly, Trp53 loss is not involved in the development of intermediate lobe tumors in mice (Williams et al., 1994a). The accelerated

development of anterior lobe tumors but not intermediate lobe tumors in mice with combined loss of Rb and Trp53 further supports this. RB and TP53 gene inactivation has not been found to play an important role in the pathogenesis of common types of human pituitary tumors (Levy et al., 1994; Woloschak et al., 1994). Pituitary adenomas in humans often show silencing of p16ink4a by methylation instead (Woloschak et al., 1997). Therefore, different components of the same signaling pathway may be affected in murine vs human tumors. Remarkably, humans are the only species that spontaneously develop retinoblastoma. This is invariably associated with loss of Rb function. Retinoblastoma development in mice clearly requires additional mutations. This may be related to a different physiology and composition of the human versus the rodent retina. However, it is also possible that in the human retina one-or-more signaling pathways are partially deficient or constitutively active making the retina highly tumor prone during a particular developmental window. In such a window, tumor suppression might critically depend on the presence of Rb as the sole safeguard. Therefore, the identification of the complementing mutations required in mice to cause retinoblastoma might also point to pathways that are deficient in the human retina.

Materials and methods Targeting and mice Using a conditional targeting strategy loxP sites were inserted into the NheI and EcoRV sites, flanking exon 19 of Rb using a ‘hit and run’ strategy described in detail previously (Vooijs et al., 1998). Mice with targeted null alleles for Trp53 and p107 have been described (Donehower et al., 1992; RobanusMaandag et al., 1998). A 1.3 kb fragment containing the hIRBP promoter (a gift from G Liou) was cloned into a vector containing a nuclear localized cre gene (nlscre) and rabbit b-globin splicing and polyadenylation signals, isolated by NotI-SalI digestion and prepared for microinjection into fertilized FVB/N oocytes. Deletion of RbF19 alleles in the germ line was achieved by crossing with the Actincre deleter strain (M Vooijs and A Berns, unpublished results). Histology and immunohistochemistry Tissues were fixed in 4% buffered formalin and 5 mm sections were stained with haematoxilin – eosin (HE). Whole mount bgalactosidase staining was performed as described (Akagi et al., 1997). For immunohistochemical staining using DAB substrate, sections were blocked in methanol/3% H2O2 followed by blocking in PBS containing 10% normal goat serum and 1% BSA. Primary antibodies were incubated o/n at 48C in PBS, 1% BSA. Polyclonal antisera were obtained from Harlan (a-MSH, ACTH, b-Endorphin, LH/FSH or from DAKO (BrdU, Prolactin, Growth Hormone, Synaptophysin and GFAP). Rabbit antisera against b-TSH (AFP1274789) and rat-a-glycoprotein subunit (AFP-66P9986), human-S-antigen and monkey-IRBP were gifts. Immunostaining using a polyclonal antiserum against Cre (1 : 15 000, Oncogene


4644

Novagen) required antigen retrieval. TUNEL labeling was performed according to the manufacturer (ITK, Diagnostics). For S phase analysis by BrdU incorporation, mice were injected i.p. with 100 mg/gr bodyweight BrdU (PBS, 7 mM NaOH) and sacrificed 2 – 3 h later. Tissue DNA analysis and genotyping DNA from ES cells or tissues was isolated as described (Laird et al., 1991). P107 and Trp53 knockout mice were identified by PCR as described (Donehower et al., 1992; Robanus-Maandag et al., 1998). Rb floxed mice were analysed as described (Vooijs et al., 1998) yielding products of 678 bp (Rb+), 746 bp (RbF19) and 321 bp (RbD19), respectively. PCR fragment sizes for Trp53 are 430 bp (wt) and 390 bp (ko) and for p107, 560 bp (ko) and 216 bp (wt). PCR reaction and cycling conditions were similar for all genotypes except for Rb for which 9% glycerol was added and MgCl2 was 1.8 mM. The rd status in mice was determined as described (Pittler and Baehr, 1991). Microdissection of the ONL of the retina was performed using a laser capture microdissection microscope (Arcturus) on 10 mm HE-stained formalin-fixed sections using conditions described elsewhere (Vooijs et al., 1998).

Abbreviations Retinoblastoma (Rb); Interphotoreceptor Retinol Binding Protein (IRBP).

Acknowledgements We wish to thank B Wiggert (National Eye Institute, NIH, Bethesda, USA), T Shinohara (BWH, Harvard, Boston, USA) and A Parlow (NIDKD, NIH, Bethesda, USA) for antibodies and G Liou for the hIRBP promoter fragment (Medical College of Georgia, Augusta, USA). J Zevenhoven, D Hoogervorst, M Tjin-a-Koeng, K de Goeij and J Bulthuis for histo-technical assistance (Histology-core facility, NKI Amsterdam); N Bosnie, F van der Ahe´, A Zwerver, T Maidment and S Greven for animal husbandry, and Jos Jonkers for the analysis of p53+/7 mice. We thank E Danen, K Quon, S Ruiz and E Robanus-Maandag for useful comments on the manuscript. This work was supported through a program grant from the Netherlands Organization of Scientific Research (M Vooijs) and a grant from the Dutch Cancer Society (NKB/KWF to A Berns and M Vooijs).

References Akagi K, Sandig V, Vooijs M, van der Valk M, Giovannini M, Strauss M and Berns A. (1997). Nucleic Acids Res., 25, 1766 – 1773. al-Ubaidi MR, Font RL, Quiambao AB, Keener MJ, Liou GI, Overbeek PA and Baehr W. (1992). J. Cell Biol., 119, 1681 – 1687. Blackshaw S and Snyder SH. (1997). J. Neurosci., 17, 8074 – 8082. Bogenmann E, Lochrie MA and Simon MI. (1988). Science, 240, 76 – 78. Carter-Dawson LD and LaVail MM. (1979). J. Comp. Neurol., 188, 263 – 272. Carter-Dawson LD, LaVail MM and Sidman RL. (1978). Invest. Ophthalmol. Vis. Sci., 17, 489 – 498. Cepko CL, Austin CP, Yang X, Alexiades M and Ezzeddine D. (1996). Proc. Natl. Acad. Sci. USA, 93, 589 – 595. Clarke AR, Robanus-Maandag E, van Roon M, van der Lugt NM, van der Valk M, Hooper ML, Berns A and te Riele H. (1992). Nature, 359, 328 – 330. Claudio PP, Howard CM, Pacilio C, Cinti C, Romano G, Minimo C, Maraldi NM, Minna JD, Gelbert L, Leoncini L, Tosi GM, Hicheli P, Caputi M, Giordano GG and Giordano A. (2000). Cancer Res., 60, 372 – 382. Donehower LA, Harvey M, Slagle BL, McArthur MJ, Montgomery CA, Butel JS and Bradley A. (1992). Nature, 356, 215 – 221. Gallie BL, Campbell C, Devlin H, Duckett A and Squire JA. (1999). Cancer Res., 59, 1731 – 1735. Geng Y, Whoriskey W, Park MY, Bronson RT, Medema RH, Li T, Weinberg RA and Sicinski P. (1999). Cell, 97, 767 – 777. Geng Y, Yu Q, Sicinska E, Das M, Bronson RT and Sicinski P. (2001). Proc. Natl. Acad. Sci. USA, 98, 194 – 199. Gonzalez-Fernandez F, Lopes MB, Garcia-Fernandez JM, Foster RG, De Grip WJ, Rosemberg S, Newman SA and VandenBerg SR. (1992). Am. J. Pathol., 141, 363 – 375.

Oncogene

Griep AE, Krawcek J, Lee D, Liem A, Albert DM, Carabeo R, Drinkwater N, McCall M, Sattler C, Lasudry JG and Lambert PF. (1998). Invest. Ophthalmol. Vis. Sci., 39, 2723 – 2732. Harbour JW and Dean DC. (2000). Nat. Cell Biol., 2, E65 – E67. Howes KA, Lasudry JG, Albert DM and Windle JJ. (1994a). Invest. Ophthalmol. Vis. Sci., 35, 342 – 351. Howes KA, Ransom N, Papermaster DS, Lasudry JG, Albert DM and Windle JJ. (1994b). Genes Dev., 8, 1300 – 1310. Jacks T, Fazeli A, Schmitt EM, Bronson RT, Goodell MA and Weinberg RA. (1992). Nature, 359, 295 – 300. Jakobiec FA, Tso MO, Zimmerman LE and Danis P. (1977). Cancer, 39, 2048 – 2058. Laird PW, Zijderveld A, Linders K, Rudnicki MA, Jaenisch R and Berns A. (1991). Nucleic Acids Res., 19, 4293. Lee EY, Chang CY, Hu N, Wang YC, Lai CC, Herrup K, Lee WH and Bradley A. (1992). Nature, 359, 288 – 294. Lee MH, Williams BO, Mulligan G, Mukai S, Bronson RT, Dyson N, Harlow E and Jacks T. (1996). Genes Dev., 10, 1621 – 1632. Levy A, Hall L, Yeudall WA and Lightman SL. (1994). Clin. Endocrinol., (Oxf), 41, 809 – 814. Liou GI, Geng L, al UM, Matragoon S, Hanten G, Baehr W and Overbeek PA. (1990). J. Biol. Chem., 265, 8373 – 8376. Liou GI, Wang M and Matragoon S. (1994). Dev. Biol., 161, 345 – 356. Lipinski MM and Jacks T. (1999). Oncogene, 18, 7873 – 7882. Margo C, Hidayat A, Kopelman J and Zimmerman LE. (1983). Arch. Ophthalmol., 101, 1519 – 1531. Marino S, Vooijs M, van Der Gulden H, Jonkers J and Berns A. (2000). Genes Dev., 14, 994 – 1004. Metzger D and Feil R. (1999). Curr. Opin. Biotechnol., 10, 470 – 476.


4645

Mihara K, Cao XR, Yen A, Chandler S, Driscoll B, Murphree AL, T’Ang A and Fung YK. (1989). Science, 246, 1300 – 1303. Moll AC, Imhof SM, Bouter LM and Tan KE. (1997). Opthalmic. Genet., 18, 27 – 34. Mulligan G and Jacks T. (1998). Trends Genet., 14, 223 – 229. Nikitin AY, Juarez-Perez MI, Li S, Huang L and Lee WH. (1999). Proc. Natl. Acad. Sci USA, 96, 3916 – 3921. Nork TM, Schwartz TL, Doshi HM and Millecchia LL. (1995). Arch. Ophthalmol., 113, 791 – 802. Pittler SJ and Baehr W. (1991). Proc. Natl. Acad. Sci. USA, 88, 8322 – 8326. Robanus-Maandag E, Dekker M, van der Valk M, Carrozza ML, Jeanny JC, Dannenberg JH, Berns A and te Riele H. (1998). Genes Dev., 12, 1599 – 1609. Robanus-Maandag E, van der Valk M, Vlaar M, Feltkamp C, O’Brien J, van Roon M, van der Lugt N, Berns A and te R. (1994). EMBO J., 13, 4260 – 4268. Sauer B and Henderson N. (1988). Proc. Natl. Acad. Sci. USA, 85, 5166 – 5170. Sherr CJ and Roberts JM. (1999). Genes Dev., 13, 1501 – 1512. Sicinski P, Donaher JL, Parker SB, Li T, Fazeli A, Gardner H, Haslam SZ, Bronson RT, Elledge SJ and Weinberg RA. (1995). Cell, 82, 621 – 630.

Tsumanuma I, Sato M, Okazaki H, Tanaka R, Washiyama K, Kawasaki T and Kumanishi T. (1995). Noshuyo Byori, 12, 39 – 43. Vooijs M and Berns A. (1999). Oncogene, 18, 5293 – 5303. Vooijs M, Jonkers J and Berns A. (2001). EMBO Rep., 2, 292 – 297. Vooijs M, van der Valk M, te Riele H and Berns A. (1998). Oncogene, 17, 1 – 12. Weinberg RA. (1995). Cell, 81, 323 – 330. Williams BO, Remington L, Albert DM, Mukai S, Bronson RT and Jacks T. (1994a). Nat. Genet., 7, 480 – 484. Williams BO, Schmitt EM, Remington L, Bronson RT, Albert DM, Weinberg RA and Jacks T. (1994b). EMBO J., 13, 4251 – 4259. Woloschak M, Roberts JL and Post KD. (1994). Cancer, 74, 693 – 696. Woloschak M, Yu A and Post KD. (1997). Mol. Carcinog., 19, 221 – 224. Young RW. (1984). J. Comp. Neurol., 229, 362 – 373. Young RW. (1985a). Anat. Rec., 212, 199 – 205. Young RW. (1985b). Brain. Res., 353, 229 – 239.

Oncogene