Understanding the accumulation of mutational events in the development of cancer is important to elucidate novel therapeutic targets that may potentially halt the disease at early stages. For the childhood retinal cancer retinoblastoma, a step-wise model of genetic events begins with the loss of the retinoblastoma tumor suppressor gene (RB1), followed by the gains and losses of potential oncogenes and tumor suppressor genes, respectively.1–6 The loss of RB1 alone only leads to the benign retinal tumor retinoma7; thus, the events immediately preceding the transition from retinoma to retinoblastoma are key targets for therapies intended to halt progression to malignancy. Because infants who inherit the RB1 mutant allele from a parent can be identified before birth, there is a realistic opportunity to not only treat early tumors, but potentially prevent them.8
One event closely associated with this transition is the loss of the p75 neurotrophin receptor (p75NTR).5, 9 We have shown previously that human retina and retinoma, the benign precursor to retinoblastoma, express p75NTR, while retinoblastoma do not.5 The p75NTR protein has been implicated in a number of, often seemingly paradoxical, cellular functions, including apoptosis/survival, axonal extension/growth cone disassembly and differentiation/proliferation.10 In the retina, a major role of p75NTR is in developmental apoptosis.11–13 In the TAg-RB murine retinoblastoma model,14 retinal specific expression of Simian Virus 40 T-antigen (TAg) in specific cells of the inner nuclear layer under regulation of the LHβ promoter, results in the inactivation of pRB and related proteins. p75NTR is expressed in early, but not late stage, TAg-RB retinoblastoma tumors, and its presence in early TAg-RB tumors correlates with presence of activated caspase-3, suggestive of apoptosis.5
We studied the potential tumor suppressor function of p75NTR in retinoblastoma development by crossing the TAg-RB mice with existing p75NTR knockout mice. The E3KO mice were created by deletion of p75NTR exon 3.15 Their phenotype is marked by a 37% reduction of neurons16 and a decrease in apoptosis in the embryonic retina.12 When the E3KO mouse was found to express a p75NTR splice variant lacking the third exon, a new mouse model was created by deletion of p75NTR exon 4.17 The E4KO mice display a more severe phenotype than the E3KO mice, with a neuronal loss of ∼54% compared to wildtype.17 However, these mice are also not complete knockouts, since a truncated carboxy-terminal domain (CTD) p75NTR protein produced by a cryptic-start site induced as an artifact of the homologous recombination intended to delete p75NTR exon 418 is able to initiate apoptosis in vitro. Since there was no full p75NTR knockout mouse available, we generated TAg-RB mice with one or both mutant alleles of the E3KO and E4KO mice and assessed tumor development compared to TAg-RB mice.
We further tested the function of p75NTR in human retinoblastoma cell lines by adenoviral-mediated expression of the gene. The combined results from murine models and human cell lines indicate that p75NTR is a tumor suppressor in retinoblastoma development and an interesting target for therapy.
Material and methods
TAg-RB(C57/Bl6) mice14 (gift from the laboratory of Dr. Joan O'Brien), E3KO(mixed agouti/C57/Bl6)15 and E4KO (C57/Bl6) mice17 (gift from the laboratory of Dr. Philip A. Barker) were maintained and sacrificed using protocols approved by the Animal Care Committee of the Ontario Cancer Institute (OCI) which adhere to the EC Directive 86/609/EEC for animal experiments. TAg-RB mice were crossed with either p75NTR E3KO or E4KO mice, and TAg-positive progeny were backcrossed to the knockout parent to obtain F2 generation TAg-positive mice either homozygous or heterozygous for the p75NTR null allele (either E3KO or E4KO, respectively). TAg-RB mice of equivalent background strain were used as controls. Genotyping was performed as previously described.14, 15, 17 Mice (3–6 per genotype) were euthanized by CO2 asphyxiation at ages 3, 4 and 8 weeks.
The eyes of each mouse were removed, fixed overnight in 4% formaldehyde (from paraformaldehyde) and embedded in paraffin wax. For one eye of each mouse, 5 μm papillary-optic nerve sections were cut through the entire thickness of each eye, and every 60th section was studied (∼5–6 sections). Sections were deparaffinized in xylene and rehydrated in a series of ethanol solutions. Immunohistochemistry was performed using mouse anti-TAg (sc-147, Santa Cruz, Santa Cruz, CA) at a 1:200 dilution, secondary horse anti-mouse (Vector Laboratories, Burningame, CA) and visualized using a diaminobenzimide kit (Immunopure, Pierce Biotechnology, Rockford, IL).
Pellets from cell lines were immobilized in 2% agarose, then fixed, embedded and sectioned as above. Immunofluorescence on cell lines was performed using rabbit anti-p75NTR (no. G3231, Promega, Madison, WI) at 1:200 dilution and rabbit anti-TrkA (no. sc-118-G, Santa Cruz, Santa Cruz, CA) at 1:100 dilution as described earlier.5 Immunofluorescence on murine retinal sections was performed using mouse anti-TAg (sc-147, Santa Cruz, CA) at 1:200 dilution and rabbit anti-activated caspase-3 (AF835, Research & Diagnostic Systems, Minneapolis, MN) at 1:100 dilution as described earlier.5
WERI-Rb119 and Y7920 retinoblastoma cell lines were cultured in a 37°C incubator with 5% CO2 in Dulbecco's Modified Iscove's medium(Gibco/OCI no.12200) supplemented with 15% FetalClone III (Hyclone), 1 × penicillin/streptomycin, 10 mg/L insulin (Sigma) and 0.0004% β-mercaptoethanol.
Adenovirus encoding p75NTR and GFP (Adp75NTR) or GFP alone (AdGFP) were gifts from the laboratory of Dr. Philip A. Barker.21 Cells were infected with adenovirus at 200 multiplicity of infection (MOI) and incubated for 24 hr at 37°C and 5% CO2. Cells were then harvested, fixed for 1 hr in 4% formaldehyde (from paraformaldehyde), rinsed and resuspended in PBS, DAPI stained and observed under fluorescent microscope. The number GFP expressing cells in 6 fields of vision at 40 × were counted and scored based on presence or absence of pyknotic nuclei, as revealed by DAPI staining. Background apoptosis was assessed by counting the number of cells with pyknotic nuclei as a fraction of the total cells in each field of vision.
Image capture and analysis
Images of murine tumors were captured using a ScanScope CS slide scanner (Aperio Technologies, Vista, CA) at the Advanced Optical Microscope Facility (Ontario Cancer Institute/Princess Margaret Hospital). Images were analyzed using ImageJ software (http://rsb.info.nih.gov/ij/). The area of the retina and the stained tumor cells, respectively, were determined using the measure function of ImageJ, and percent tumor per eye was calculated as the mean of the 5–6 sections per eye. Each value for the average tumor area per eye was used to obtain a mean tumor area per genotype (Supplementary Fig. 1).
Images of cell lines were captured using a Coolsnap camera attached to a Leica DMLB microscope and processed using Photoshop 7.0.
Significance was determined using the Mann–Whitney or Kruskal–Wallis test (analysis of variance by ranks, comparing 2 or 3 groups, respectively) for mouse tumors, and χ2 test for adenovirus infected cell lines.
TAg-RBE3KO mice showed no difference in tumor area compared to controls
The TAg-RB (C57/Bl) mice were crossed to E3KO (mixed strain) mice to generate TAg-RB mice with normal, or 1 or 2 mutant copies of the p75NTR allele (mixed strain) and 3–6 mice of each genotype were assessed at 3, 4 and 8 weeks of age. TAg-RBE3KO knockout and heterozygous mice did not display a significant difference in tumor area as a percentage of retinal area, compared to the TAg-RB control mice, at any time point studied (Supplementary Fig. 2; results are summarized in Table I).
Table I. Percent Tumor Area PER Genotype for TAg-RBE3KO and TAg-RBE4KO Mice
Area of tumor as a percentage of retinal area
Results are mean values from 3 to 6 mice per genotype.
Indicates only one mouse was sampled). Significant difference was considered if p ≤ 0.05 by Kruskal–Wallis.
Mann–Whitney test. Significant differences are in bold.
TAg-RBE4KO mice had greater tumor area than controls
The E4KO (C57/Bl) mice were crossed to the TAg-RB (C57/Bl) mouse to generate TAg-RB mice with normal, or one or both mutant p75NTR alleles (C57/Bl). The 3-week-old TAg-RBE4KO knockout and heterozygous mice had significantly larger tumor area than the TAg-RB mice with wildtype p75NTR alleles (p = 0.05) (Fig. 1). Four-week-old mice followed the same pattern of larger tumors with loss of p75NTR alleles, as TAg-RB mice had 3.7% tumor area compared to 5.7% for the TAg-RBE4KO heterozygotes and 8.7% for the homozygous null mouse. However, since only one TAg-RBE4KO homozygous null mouse could be analyzed at the 4-week time point, a Mann–Whitney analysis was performed to test significance between the TAg-RB mice with wild-type and heterozygous p75NTR alleles, and the difference in tumor size was insignificant (p = 0.39).
By 8 weeks, the percent tumor area of mice with 1 or 2 mutant p75NTR alleles was again significantly larger than in the TAg-RB control mice (p = 0.02) (Fig. 1; results are summarized in Table I). Pairwise comparisons between TAg-RBE4KO +/+ with each of the +/− and −/− cohorts at each timepoint was consistent, showing significant difference for the 3 and 8 week timepoints (Supplementary Table I).
TAg-RbE4KO tumors formed throughout the inner nuclear layer, while TAg-RB tumors formed only in the periphery
In both the TAg-RB and TAg-RbE4KO mice, the TAg-positive cells were observed throughout the inner nuclear layer of the whole retina during the early timepoints (Figs. 2 and 3 week timepoint shown). In the TAg-RB mice at 8 weeks of age, tumor foci were observed in the anterior peripheral retina, but no TAg-RB stained cells were observed at the central, posterior pole of the eyes. However, in the TAg-RBE4KO mice (both +/− and −/−), at 8 weeks emerging tumor foci were observed in the inner nuclear layer throughout the eye, including at the central posterior pole (Fig. 2).
TAg-RbE4KO early tumors do not express activated-caspase-3
Our previous work showed that early TAg-RB tumors express p75NTR and activated caspase-3, but lose the expression of both as the tumors grow and progress.5 Thus, we wished to determine whether or not the absence of p75NTR from the onset of tumor progression in the TAg-RBE4KO mice affected expression of activated caspase-3. As expected, tumors from 3-week-old TAg-RB mice stained positive for activated caspase-3; however, tumors from mice lacking one or both copies of the p75NTR allele did not (Fig. 3).
TAg-RB mice of different background strain displayed variation in tumor penetrance
We observed a significant difference in tumor size of TAg-RB mice on mixed vs. pure C57/Bl background strain. TAg-RB mice with wildtype p75NTR alleles on mixed background strain had significantly larger tumor area (p = 0.05) than the equivalent on a pure C57/Bl background, at 3 and 8 weeks of age (Table II). The percent tumor area at 4 weeks of age was not significantly different (p = 0.29), as we observed much more variability in tumor area amongst the individual animals studied. This could potentially be a result of apoptosis occurring at this timepoint, the rate of which could vary among individual animals.
Table II. Percent Tumor Area for TAg-RB of Differing Background Strains
Area of tumor as a percentage of retinal area
Results are mean values from 3 to 6 mice per genotype. Significant difference was considered if p ≤ 0.05 by Mann–Whitney test. Significant differences are in bold.
Adenoviral-mediated p75NTR expression induced apoptosis in WERI, but not Y79, retinoblastoma cell lines
Six retinoblastoma cell lines (Y79, WERI, RB247, RB381, RB383 and RB1021) were immunostained for presence of p75NTR and TrkA (a tyrosine kinase receptor that p75NTR can work in concert with to promote cell survival, depending on cellular context10). While WERI, RB247, RB381, RB383 and RB1021 cell lines were negative for both proteins, Y79 cells retained expression of both TrkA and p75NTR (Supplementary Fig. 3).
The WERI and Y79 cell lines were infected with Adp75NTR or AdGFP for 24 hr, then scored under fluorescent microscope for presence of GFP and pyknotic nuclei (visualized by DAPI stain). The fraction of GFP-positive WERI cells infected by Adp75NTR that displayed pyknotic nuclei was 33.3%, while the fraction of AdGFP-infected cells that displayed pyknotic nuclei was significantly lower, at 6.78% (p = 0.0006) (Fig. 4; results summarized in Table III). Background apoptosis (percentage of uninfected cells with pyknotic nuclei) and infection efficiency (percentage of total cells infected, determined by GFP fluorescence) were not significantly different (p = 0.17 and p = 0.56, respectively) between Adp75NTR and AdGFP infected cells.
Table III. Adenoviral-Mediated Expression of p75NTR in the Weri and Y79 Retinoblastoma Cell Lines
p (χ2 test)
Results were pooled from three replicates; significance was measured by χ2 test. Significant differences (p ≤ 0.05) are in bold.
Infection efficiency (%)
% Background apoptosis
Infection efficiency (%)
% Background apoptosis
Equivalent experiments using the p75NTR-expressing Y79 retinoblastoma cell line yielded quite opposite results. Apoptosis induced by Adp75NTR was 5.45%, not significantly higher than the 3.22% induced by AdGFP (p = 0.44) (Fig. 4; results summarized in Table III). Background apoptosis and infection efficiency were also not significantly different (p = 0.46 and p = 1, respectively) between Adp75NTR and AdGFP infected cells.
Murine models of cancer allow the real-time tracking of tumor development in vivo, supporting the study of both early and late stages of disease progression. Human retinoblastoma specimens available for study are most often advanced tumors removed as therapy, so the use of murine models to study step-wise accumulation of genetic events resulting in cancer is important.
Many different retinoblastoma mouse models exist14, 22–26; however, they differ in their degree of similarity to human retinoblastoma. With the exception of the TAg-RB model, these models have not been documented to display complete tumor penetrance or presence of Flexner-Wintersteiner rosettes, a hallmark of retinoblastoma (though many display Homer Wright rosettes, a feature of neural tumors).22–24, 26 Of post-RB1 loss events associated with retinoblastoma, only the recent conditional Rb1; p130 double knockout26 shows NMyc amplification and 1q gain, similar to human retinoblastoma, while the TAg-RB model shows overexpression of DEK4 and KIF14 (Pajovic et al., manuscript in preparation) and loss of CDH112 and p75NTR5 expression. Arguably the murine retinoblastoma model that most closely mimics the human condition, by tumor penetrance, histology and molecular features, is the TAg-RB mouse model.14
Small early tumors in the TAg-RB model are visualized by immunohistochemical detection of large TAg, which is also presumed to represent the expression pattern of the small TAg since it is encoded by the same transgene and controlled by the same regulatory region. Although small tAg expression has not been specifically documented in Tag-RB tumors, the assumption is that it is expressed, since transformation of cells by large Tag is dependent on the coexpression of small tAg.27 At early timepoints TAg-positive clusters do not yet form rosettes characteristic of retinoblastoma, however we have observed continuous abnormal proliferation by immunohistochemical detection of Ki67 from the onset of TAg-RB expression at post-natal day 8 to timepoints beyond 8 weeks of age (Pajovic et al., manuscript in preparation), indicating that these TAg-expressing cells are the initiating stages of tumorigenesis. We have previously shown that the early TAg-positive clusters express p75NTR, but lose this expression as the tumors grow and progress.5 The loss of p75NTR is accompanied by concomitant loss of activated caspase-3 expression, which is abundant in early p75NTR-expressing TAg-RB clusters,5 suggesting that the effect of p75NTR in early tumors is to induce apoptosis and impede tumor progression.
We crossed the TAg-RB mouse with each of the 2 available p75NTR knockout models in an attempt to inhibit this potential early apoptotic event in retinoblastoma tumorigenesis. If our hypothesis is true, we would expect to see, at equivalent stages, larger tumor area within the retinas of the double transgenics than in the retinas of the TAg-RB mice with wild type p75NTR. Our results indicate that the TAg-RBE4KO, but not TAg-RBE3KO, mice do display a larger tumor area than the TAg-RB mice.
The TAg-RBE4KO mice presumably express the truncated CTD-p75NTR artifact18; yet, we found no evidence that it promotes apoptosis as observed in in vitro studies, since we observed an increase in tumor area as a percent of retina in these mice compared to TAg-RB mice with normal p75NTR (Fig. 1), regardless of whether one or both p75NTR alleles were knocked out. The heterozygous phenotype we observe may result from loss of heterozygosity or a dominant negative function of the allele.
The increase in tumor area in the absence of p75NTR is consistent with the apoptosis we have observed in early stage p75NTR-positive TAg-RB tumors.5 We showed that activated caspase-3 expression correlated with p75NTR expression in early TAg-positive cells but was reduced in the TAg-positive cells of 4-week-old TAg-RB mice.5 The absence of detectable activated caspase-3 in the TAg-RBE4KO +/− and −/− 3-week tumors (Fig. 3) strongly suggests that reduced apoptosis may well be the mechanism for the increased tumor area in TAg-RBE4KO mice, and also that loss of only one p75NTR allele is sufficient to observe this phenotype.
A previous suggestion that the E4KO mouse is actually a gain-of-function model in terms of apoptotic activity18, 28 is not consistent with our results. It is possible that the truncated p75NTR cooperates in apoptotic function at a very low level or only in specific tissues or developmental stages in the mouse, or possibly it is not proapoptotic at all in vivo. Alternatively, this protein artifact could contribute to oncogenesis in our model, resulting in larger tumors; however, there is no published evidence supporting such a role.
The TAg-RB mouse initiates tumor development at P8, when single TAg-positve cells expressing TAg emerge in the inner nuclear layer and proliferate in the otherwise mature and nonproliferative retina (Pajovic et al., manuscript in preparation). We observed that these TAg-positive cells expanded to fill a large part of the inner nuclear layer by 3 weeks of age (Fig. 2). The cells did not display specific anterior–posterior localization at this stage. However, by 8 weeks of age, large tumor foci were observed only in the peripheral retina. This suggests an underlying difference in cancer susceptibility between the central and peripheral retina. It is well documented that the peripheral retina develops later than the central retina. Perhaps the cells at the more mature central retina are already differentiated past the point of susceptibility to transformation, and like later cell types,29 degenerate when exposed to TAg. Indeed, our earlier data showed that TAg-RB tumor cells at 3 and 4 weeks of age that express p75NTR frequently undergo apoptosis5; perhaps the TAg-positive cells are also cleared more effectively from the central retina than from the peripheral retina.
Importantly, the key role of p75NTR in eliminating the TAg cells in central retina is clear from our data in the TAg-RBE4KO mice. TAg expression in the inner nuclear layer in TAg-RBE4KO heterozygotes and homozygous null mice involved both central and peripheral retina; yet, the 8-week mice developed tumors throughout the length of the inner nuclear layer, in both central and peripheral retina. In the absence of p75NTR, the central retina has lost the capacity to eliminate the TAg cells, and therefore also forms tumors. This is likely a contributing factor to the larger tumor area observed in these mice as compared to TAg-RB. In other mouse models, tumors originate in the peripheral retina for the conditional Rb1; p10722 and Rb1; p13026 double knockouts; however, this is because Rb1 expression is blocked solely in the peripheral retina.
TAg-RBE3KO mice, which presumably retain some residual function of p75NTR through the expression of the splice variant s-p75NTR, show similar tumor development to the TAg-RB mice with normal p75NTR alleles.5 However, the mixed background of these animals may have contributed to some variability in tumor development, and thus confounded our results.
We observed a variation in tumor size between TAg-RB mice with wildtype p75NTR on C57/Bl background as compared to the equivalent genotype and age on mixed strain background (Table II). It is well known that strain differences result in varying sets of modifier genes that can affect overall phenotype. When TAg-RB mice on a C57/Bl background were crossed to FVB mice and the F1 generation backcrossed to FVB mice for 2 generations, offspring displayed a 25% reduction in tumor incidence as compared to the original inbred C57/Bl strain.30 The TAg-RBE4KO and control TAg-RB are both C57/Bl, so strain background is not a factor in those experiments.
Murine models can give valuable insights into cancer development. However, because they do not precisely mimic the human condition, the information garnered from them may have limited applicability to human cancer. For example, we know that human retinoblastoma is a stepwise accumulation of genetic events that leads a normal retinal cell to a dysplastic, senescent state (retinoma) that may then acquire further insults to progress to malignancy (retinoblastoma).7 The TAg-RB model, although most similar in molecular and histological qualities to human retinoblastoma of the various murine models, does not manifest a benign, retinoma-like state, presumably because the TAg expression induces many early changes all at once, with loss of pRB, p130, p107, p53 and other proteins inactivated by TAg. Other models may more accurately mimic the earliest events, with loss of RB1 and another RB family member, but in a wider range of undefined retinal cell types, depending on the promoter used to accomplish tissue-specific knockout.
The TAg-RB mouse is the best studied retinoblastoma model for preclinical assays of therapy.31–33 It is important to know how retinoblastoma develops in TAg-RB mice, in which expression from the LHβ promoter of TAg may have serendipitously targeted a developing retinal cell very similar to the susceptible cell in human retina. Knowledge of how the murine model compares to human retinoblastoma is essential to interpret preclinical therapy for humans; an effective treatment for murine retinoblastoma may not be equally effective if the cell type and mechanism of retinoblastoma development in the mouse model are different than in humans.
To determine if p75NTR has a proapoptotic role in human retinoblastoma as suggested for TAg-RB retinoblastoma, we studied human retinoblastoma cell lines. Retinoblastoma cell lines represent an advanced stage of tumor progression, since most primary retinoblastoma fail to grow in vitro. Many of the genomic events associated with the cancer have already occurred in cell lines,2–4, 6, 34 and in fact, Y79, 1021 and 247c show chromosomal breakpoints at 17q21, the location of p75NTR, further implicating it as a tumor suppressor.35 The Y7920 and WERI-Rb-119 cell lines used in this study were the first retinoblastoma cell lines to be developed many years ago, and the best characterized. Other retinoblastoma cell lines (Supplementary Fig. 3), while useful for some studies, were excluded from these experiments because of high levels of background apoptosis and relative inefficient viral infection.
While in the murine study we eliminated p75NTR from the onset of tumor development and showed enhanced tumor progression, we forced p75NTR overexpression and measured apoptosis in advanced retinoblastoma cell lines. In the WERI cells, which did not express endogenous levels of p75NTR by immunohistochemistry (Supplementary Fig. 3) or qRT-PCR5 we saw a marked increase in apoptosis upon ectopic p75NTR expression, indicating that pathways downstream of p75NTR were intact for apoptosis. In p75NTR-nonexpressing cells, additional p75NTR could be useful as therapy. On the other hand, Y79 cells, which expressed p75NTR by immunohistochemistry (Supplementary Fig. 3), albeit at decreased levels compared to human retina by quantitative RT-PCR,5 were relatively unaffected by adenoviral-mediated expression of p75NTR. This suggests that these cells may have a defect in the pathway downstream of p75NTR that precludes the effectiveness of p75NTR to induce apoptosis. Alternatively, the ability of p75NTR to induce apoptosis in retinoblastoma may be related to the stage of tumor development or the combination of other post-RB1 loss mutational events at play in the individual tumor.2–4
p75NTR binds to all known neurotrophins,36–38 however, with highest affinity to the immature form of nerve growth factor (NGF), pro-NGF.39 Pro-NGF binding to p75NTR and its coreceptor sortilin leads to apoptosis40; however, other neurotrophin and coreceptor combinations could also lead to apoptosis, as occurs when cells are mistargeted to locations where “foreign” ligands potentially induce apoptosis.41 For these reasons, in this study we did not specifically test the effect of a particular p75NTR ligand on the apoptotic function of the receptor, as we could not exclude the possibility that these factors were present in the growth media.
In glioma, expression of p75NTR promotes invasion and metastasis,42 however, our data demonstrate that p75NTR is a tumor suppressor for retinoblastoma. Similarly for prostate cancer, therapies which induce expression of p75NTR have been shown to suppress tumor growth by activation of apoptosis and reduction of proliferation.43 The question remains if p75NTR induction would be similarly effective against retinoblastoma, potentially by initiating apoptosis, as it does in WERI cells. Perhaps a drug cocktail targeting a combination of candidate retinoblastoma oncogenes3, 4 and tumor suppressors2 and induction of apoptosis by p75NTR might have a greater effect. Drug cocktails such as these will best be studied in a murine model that mimics the progression of the human disease, by sequentially invoking post-RB1 loss events in a cancer-susceptible retinal cell.
This work was supported in part by grants to BLG from the National Cancer Institute of Canada with funds from the Terry Fox Run and the Canadian Cancer Society, the National Cancer Institute, NIH, the Keene Retinoblastoma Perennial Plant Sale, the Royal Arch Masons of Canada and the Canadian Retinoblastoma Society. A University of Toronto Vision Science Research Program Studentship supported HD. We thank Dr. Phillip A. Barker for the gifts of the p75NTR transgenic mice and adenoviruses and Dr. Joan O'Brien for the Tag-RB mice. We thank members of the Gallie lab for careful reading of the manuscript and helpful suggestions.