Genetic Background Controls Tumor Development in Pten-Deficient Mice

PTEN (phosphatase and tensin homologue deleted from chromosome 10) or MMAC1 (mutated in multiple advanced cancers) is one of the most frequently mutated tumor suppressor genes in human cancers (1, 2). Germ line mutations of PTEN have been detected in three rare, autosomal-dominant disorders, Cowden disease, Lhermitte-Duclos disease, and Bannayan-Zonana syndrome (3, 4), which were originally identified as distinct syndromes. Although these disorders share similar pathologic features, such as benign tumors in multiple organs (hamartomas) and increased susceptibility to malignant cancers, the onset and spectrum of these clinical abnormalities are quite different. In some cases, patients with an identical mutation may have very different clinical symptoms and may be diagnosed with different diseases (5). Inactivation of Pten in mouse models, generated by three independent groups, has confirmed PTEN as a bona fide tumor suppressor (6–9). Although the targeted disruption centers on exon 5, which is known to contain the phosphatase domain, the three groups have subtle differences in excision ranging from exons 3 to 6 and consequently reported differences in the tumor spectra seen in mice. To clarify whether differences seen in both human patients and in animals models are due to (a) nature of individual mutations (e.g., hypomorphic variations in the Pten locus) or (b) genetic background/modifier genes, we generated two knockout mouse strains with different deletions and compared the consequences of Pten inactivation on 129/C57 and 129/BALB/c genetic backgrounds. Similar to previous reports, mice homozygous for both mutations die early during embryogenesis and mice with heterozygous Pten deletion develop tumors in various tissues. However, the onset and spectrum of tumor development shown in our study were quite different, and in particular, high incidence of prostate cancer and hemangioma formation was observed on 129/BALB/c background. Our study shows the crucial role of PTEN in prostate cancer formation and suggests that the genetic background or modifier gene(s) may have a significant effect on PTEN's biological functions.

Generation of Pten^+/− and Pten^Δ5/+ mice. Generation of Pten^+/− embryonic stem cells has been described previously (10). The Pten exon 5 deletion was generated by transiently transfecting a Cre recombinase–expressing vector into the conditional Pten knockout embryonic stem cell line (11) and subsequently selected for thymidine kinase resistant (tk^r) clones (Supplementary Fig. S1). Two classes of tk^r embryonic stem clones were generated: (a) conditional deletion clones with the exon 5 flanked by loxP sites (Pten^loxp/+; ref. 11) and (b) exon 5 deleted clones (Pten^Δ5/+). Embryonic stem cell clones carrying either Pten^+/− or Pten^Δ5/+ were injected into either C57/B6 or BALB/c blastocysts. Chimeric male mice were backcrossed to obtain heterozygous mice. F1 mice were then intercrossed for phenotype analysis.

Histology analysis and pathologic grading. Tumors were dissected and fixed in 10% buffered formalin followed by fixation in 70% alcohol. Tumor tissues were paraffin embedded, sectioned (4 μm), and stained with H&E.

PCR and Western blot analysis. To facilitate faster genotyping, a PCR strategy was developed using two primers: forward primer (derived from intron 4 sequence) 5′-ACTCAAGGCAGGGATGAGC-3′ and two reverse primers, P2 (derived from exon 5 sequence) 5′-AATCTAGGGCCTCTTGTGCC-3′ and P3 (derived from the vector sequence) 5′-GCTTGATATCGAATTCCTGCAGC-3′. These primers give rise to the amplified products of 1,100 bp for the loxp allele, 900 bp for the wild-type allele, and 300 bp for the Δ5 allele. To confirm Pten deletion, antibodies specific to the NH₂ terminus (Santa Cruz Biotechnology, Santa Cruz, CA) and COOH terminus (Cell Signaling Technologies, Beverly, MA) of the PTEN protein were used.

Generation of Pten^Δ5/+ mice. Embryonic stem cell clones carrying the Pten^Δ5/+ allele were generated by transiently expressing the Cre recombinase in Pten^loxp/+ embryonic stem cells (ref. 11; Fig. 1A). Germ line transmission of the Pten^Δ5 mutant allele was achieved from two independent embryonic stem cell clones. Because the deletion is specific for the exon 5, a PCR screening method was used to determine the mouse genotypes (for primers, see Fig. 1A). Intercrosses between heterozygous F1 mice yielded animals of three genotypes: Pten^+/+, Pten^+/Δ5, and Pten^Δ5/Δ5 (Fig. 1B). Pten^Δ5/Δ5 mice were embryonic lethal and died between embryonic days 9.5 to 10.5 (E9.5-10.5). More detailed morphologic analyses revealed phenotypes that have not been reported by the previous studies (6, 7, 9): mutant embryos exhibit a wavy neural tube (Supplementary Fig. S1, comparing B and D, areas pointed by black arrowheads), failure of neural tube closure (comparing A and C, white arrows), and pericardial bulging (A and C, black arrows) at E9.0-9.5, right before the onset of embryonic lethality. No PTEN protein could be detected using either anti-NH₂-terminal or anti-COOH-terminal antibodies (Fig. 1C), suggesting that exon 5 deletion leads to either a complete null mutation or a truncated protein that is very unstable.

Onset and spectrum of tumor formation in Pten^Δ5/+ mice. On the 129/BALB/c genetic background, the onset and spectrum of tumor development of Pten^Δ5/+ mice were significantly different from the previous reports, in almost all of which were done on the 129/C57 or 129/CD1 background (Table 1). We have surveyed >150 Pten^Δ5/+ mice, ranging from 1 to 14 months, and did detailed histopathologic analysis on 47 mice, covering tissues and organs, such as brain, thymus, lymph nodes, lung, liver, adrenal glands, pancreas, spleen, uterus, ovary, mammary glands, gastrointestinal tract, and prostate. Our findings can be summarized as follows: (a) the onset of tumor formation on the BALB/c background is significantly delayed versus other findings, which report spontaneous development of malignant tumors in <3.5 months (6, 7, 9). The majority of Pten^Δ5/+ heterozygous mice (90%) were tumor free during the first 6 months of their postnatal life, and only 50% of the animals had developed tumors by 10 months of age, as judged by histologic analysis (12). (b) The spectrum of tumors is also quite different (Table 1). Previously reported tumor incidences on the 129/C57 or 129/CD1 backgrounds were biased towards tissues and organs, such as lymphoid, endometrial, adrenal, thyroid, gastrointestinal, and mammary glands (6–9). In our aged animals, ranging from 7 to 14 months, the incidences of lymphoid and endometrial hyperplasia or neoplasia were rather low (25% and 9%, respectively). Heterozygous females were able to produce 6 to 7 litters, and their reproductive life span was similar to wild-type controls. Epithelial hyperplasia/carcinoma of the gastrointestinal tracts and tumors in the adrenal glands were not detected. On the other hand, among 21 male mice analyzed (age, 10-15 months), prostate abnormalities, ranging from hyperplasia to local invasion (Supplementary Fig. S2), were present in 95% heterozygotes. Furthermore, 30% of Pten^Δ5/+ female on the 129/BALB/c background also developed mammary tumors, ranging from 1 to 10 tumors per animal, similar to a previous report (ref. 8; Supplementary Fig. S3). In addition, hemangioma lesions were found in 45% of the Pten^Δ5/+ female uteri (see below).

Vascular abnormalities in Pten^Δ5/+ animals. In addition to hamartomas and increased susceptibility of malignant cancers, patients with inherited PTEN mutations also suffer from vascular malformations, such as angiomatosis and hemangiomas (13). Interestingly, vascular lesions, which resemble angiomatosis or hemangiomas, were observed in our Pten^Δ5/+ mice. Among 15 females (7-14 months) studied, seven of them had single or multiple solid masses in their uteri horns (Fig. 2A). The masses consisted of nonencapsulated aggregates of endothelium-lined, cavernous blood-filled spaces of variable sizes (Fig. 2B,, arrows). Narrow connective tissue septa separated these vascular spaces. The blood within the larger spaces was often thrombosed (Fig. 2B , Th). Occasionally, at the edges of these large cavernous spaces, there were small aggregates of capillaries, whereas in some areas, the tissue resembled arterioles or venules. These smaller vessels always seemed engorged and packed with masses of erythrocytes. These lesions were predominantly in the myometrium, involved the whole uterine wall, and protruded into the uterine lumen.

The vascular abnormalities were also found in the brain (Fig. 2C and D), mammary glands (Supplementary Fig. S3), spinal cord, and the body walls (data not shown), but with lower frequencies. Figure 2C shows a 14-month-old male mouse with vascular lesions in the brain similar to those described in the uterus. This mouse seemed ataxic with enlargement of the forehead. Grossly, the left olfactory bulb and frontal lobes of the cerebrum seemed to be replaced by masses of hemorrhagic-appearing tissue (Fig. 2D). Histologically, the mass consisted of aggregates of endothelium-lined, cavernous blood-filled spaces separated by narrow connective tissue septa (Fig. 2D,, white arrow). The blood within some spaces was thrombosed (Fig. 2D,, Th) and calcified (Fig. 2 D, Ca⁺⁺). Small aggregates of capillary arteriole- and venule-like vessels, similar to those seen in the uterus, were also present in the vicinity of the lesion.

Although most lesions were benign and isolated, malignant transformation was observed in three cases: two with local invasion and one with whole body metastasis. Figure 2E shows an example of liver metastasis. Similar lesions were also observed in the spleen, lymph nodes, intestine, lung, and many other organs in the same animal. Sectioning through the metastatic lesion in the liver reveals cord-like epithelial engulfing masses of erythrocytes. Thus, similar to what has been reported in humans, PTEN also controls vascular morphogenesis and possibly vascular functions in our knockout models.

Genetic background influences tumor development. Such phenotypic differences could be due to (a) hypomorphic properties of specific mutations created by individual research groups or (b) the influence of genetic background or modifier gene(s). To distinguish between these two possibilities, we compared the exon 5–specific deleted mouse strain with another heterozygous Pten knockout mouse strain generated in our laboratory, in which both transcriptional and translational start sites and exons 1 to 3 of the Pten gene were deleted (Pten^+/−; ref. 10). In the postnatal 7 months, the onset and spectrum of tumor development of Pten^+/− mice on BALB/c/129 background is similar to Pten^Δ5/+ mice but differ significantly to itself on C57/129 background (Fig. 3, columns 1 and 2). Interestingly, the tumor phenotypes observed in Pten^+/− mice on C57/129 background are very close to published reports (Table 1). Similarly, when we backcrossed BALB/c/129 Pten^Δ5/+ mice to C57/BL6 background, we found that 85% of Pten^Δ5/+ F2 female mice developed lymphoid hyperplasia and 56% with cystic endometrial hyperplasia (comparing to 25% and 9%, respectively, on the 129/BALB/c background). Furthermore, the onset of such abnormalities developed much earlier: 70% F2 heterozygous females developed lymphoid hyperplasia between 1 and 7 months compared with only 10% found in 129/BALB/c heterozygous females at similar age (Fig. 3, columns 3 and 4). This genetic study indicates that genetic background or modifier gene(s) play an important role in regulating PTEN's in vivo functions.

This study as well as other published studies observed increased tumor formation in animals harboring heterozygous mutations of Pten. However, tumor frequency, tumor spectrum, and onset of the tumors varied significantly. Multiple factors could contribute to the observed phenotypic differences. Although all published mutations were generated around exon 5 of the Pten gene, each group has employed slightly different strategies to create their specific mutation. Whether these different mutations code for truncated proteins that may display various hypomorphic or dominant-negative properties is currently unknown. In this study, we have directly compared two independent mutations in the murine Pten locus: one carries deletion from the promoter through exon 3, thus should be completely null (10), and the other with exon 5–specific deletion (this study). Biochemical analysis showed that no protein products could be detected from either type of mutations, suggesting that if a truncated protein is produced from either one of the deletion mutations, it must be very unstable. Furthermore, analyzing the onset and spectrum of tumor formation indicated that the major contributor to the phenotype variation was the genetic background, not the particular mutation generated.

Although, to date, vascular abnormalities in Pten^Δ5/+ animals on the 129/BALB/c background have not been reported, recent studies suggest that PTEN/phosphatidylinositol 3-kinase pathway controls normal vascular development and tumor angiogenesis (14–17), and AKT-1, the downstream effector of PTEN, regulates pathologic angiogenesis, vascular maturation, and permeability in vivo (18). Thus, the Pten knockout mice generated in this study may facilitate our understanding of PTEN-controlled angiogenesis in vivo.

Our finding that the genetic background or modifier gene(s) may play an essential role in modulating PTEN's biological functions is both biologically relevant and clinically significant. Strain-dependent phenotypes and modifying loci in tumorigenesis have been well documented in mouse models, such as the Mom-1 locus for intestinal polyposis syndromes in Min^+/- mice (19). The significant differences of the observed phenotypes on two different genetic backgrounds correspond well with the striking differences seen in the human familial disorders with reported PTEN mutations. At least two possibilities could account for this observation: (a) differences in the rate of PTEN loss or loss of heterozygosity, which is inherited by individual patient or mouse strain and (b) allelic differences in other disease-modifying genes that have not yet been identified. Further detailed genetic analysis and comparisons of the different Pten mutant mice will be valuable to uncover additional information in our understanding of PTEN's in vivo function and genotype/phenotype relationship.

Grant support: NIH grants CA98013 (H. Wu), Department of Defense grants PC991538 (H. Wu), Prostate Cancer Foundation (H. Wu), Howard Hughes Medical Institute (R. Lesche), Deutsche Forschungsgemeinschaft (R. Lesche), and Predoctoral Research Training Grant in Pharmacological Sciences and Prostate Cancer Foundation (D. Freeman).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank members of our laboratories for helpful comments, Dr. Robert D. Cardiff for pathologic evaluation and immunohistochemistry analysis, and Margy Blavin for article preparation.