|Home | About | Journals | Submit | Contact Us | Français|
Basal cell carcinomas (BCCs) have relative genomic stability and relatively benign clinical behavior but whether these two are related causally is unknown. To investigate the effects of introducing genomic instability into murine BCCs, we have compared ionizing radiation-induced tumorigenesis in Ptch1+/− mice versus that in Ptch1+/− mice carrying mutant Blm alleles. We found that BCCs in Ptch1+/− Blmtm3Brd/tm3Brd mice had a trend toward greater genomic instability as measured by array comprehensive genomic hybridization and that these mice developed significantly more microscopic BCCs than did Ptch1+/− Blm+/tm3Brd or Ptch1+/− Blm+/+ mice. The mutant Blm alleles also markedly enhanced the formation of rhabdomyosarcomas (RMSs), another cancer to which Ptch1+/− mice and PTCH1+/− (basal cell nevus syndrome) patients are susceptible. Highly recurrent but different copy number changes were associated with the two tumor types and included losses of chromosomes 4 and 10 in all BCCs and gain of chromosome 10 in 80% of RMSs. Loss of chromosome 11 and 13, including the Trp53 and Ptch1 loci, respectively, occurred frequently in BCCs, suggesting tissue-specific selection for genes or pathways that collaborate with Ptch deficiency in tumorigenesis. Despite the quantitative differences, there was no dramatic qualititative difference in the BCC or RMS tumors associated with the mutant Blm genotype.
Basal cell carcinomas (BCCs) comprise ~80% of all non-melanoma skin cancers and as such are the most common of human cancers. Despite their locally invasive nature, BCCs metastasize extraordinarily rarely (1,2), and the mechanism underlying this extremely low incidence is unclear. Suggested barriers to metastasis have included BCCs’ relative genomic stability, which contrasts with the nearly uniform genomic instability of more malignant, more metastatic visceral cancers (3). Such genomic instability may generate more mutations and hence more opportunities to produce cells able to overcome the barriers to more extensive local invasion and more widespread dissemination (4). However, there has been no direct assessment of the influence of the level of genomic instability on the behavior of BCCs and so the connection remains hypothetical.
One autosomal recessive hereditable condition with chromosomal instability is the Bloom syndrome (5), which is caused by mutations in the RecQ helicase BLM. Helicases unwind complementary strands of DNA for processes such as transcription of RNA, DNA repair and recombination and replication in which single-stranded DNA is required. Mutations in other ReqQ helicase family members underlie Werner syndrome (WRN), Rothmund–Thomson syndrome, (RAPADILINO) and Baller-Gerold syndrome (RECQ5) (6–13). Mutations in each of these three helicases lead to spontaneous chromosome instability, elevated mutations and elevated predisposition to cancers, findings which are common to Bloom, Werner and Rothmund–Thomson syndromes (14–16).
In this study, we assessed whether the loss of genetic stability induced by Blm dysfunction would affect tumorigenesis arising in the Ptch1+/− mouse model of hedgehog-driven carcinogenesis (17). Several mutant murine Blm alleles have been engineered (18–20), and the degree of genomic instability and the severity of cancer predisposition in mice carrying these alleles vary inversely according to the degree of loss of Blm protein function (21). Hence, we mated Ptch1+/− mice to mice carrying a hypomorphic Blm allele (Blmtm3Brd) (20) and assessed ionizing radiation-induced carcinogenesis in the compound mutant mice. We found that reduced Blm function indeed significantly increases tumorigenesis of BCCs and RMSs, two tumor types that are characteristic of Ptch1+/− mice and PTCH1+/− human (basal cell nevus syndrome) patients. Despite the increased tumorigenesis, tumors arising in Blm wild-type and deficient mice could not be distinguished histologically.
Ptch1+/− heterozygous mice (17) were bred with Blmtm3Brd/tm3Brd mice to produce Blm+/+ wild-type, heterozygous and homozygous mutant mice (20). We enrolled 90 mice in these studies—29 Ptch1+/− Blm+/+ mice, 30 Ptch1+/− Blm+/tm3Brd mice and 31 Ptch1+/− Blmtm3Brd/tm3Brd mice. Mice of all three tested genotypes were littermates and were of mixed FVBN/J, C57BL/6 and DBA/2 backgrounds. Mice were maintained in plastic cages with metal lids at 21–23°C, 50% humidity and on a 12 h white fluorescent light, from 34 W bulbs, and 12 h dark cycle. They had access to laboratory chow (5008; Purina, St Louis, MO) and tap water ad libitum. At the age of 2 months, all of the mice were exposed to 5 Gy of γ-radiation, using a Best Industries (Springfield, VA) 137Cs radiation device (half-value layer, 0.60 cm Pb; dose, 0.94 Gy/min).
At the age of 9 months, 1 cm2 of dorsal skin was excised, divided into 5 equal slivers and fixed in 10% buffered formalin for β-galactosidase staining. All biopsies were performed under anesthetic, using a 1:1 solution of 20 mg/ml xylazine and 100 mg/ml ketamine. During the study, all mice were observed for macroscopic tumor formation. Mice exceeding the animal welfare guidelines were euthanized. Tumors and a 1 cm2 piece of back skin lacking visible tumor were excised and fixed for histological assessments. For β-galactosidase staining, samples were fixed and stained with 5-bromo-4-chloro-3-indolyl-β-D galactopyranoside in iron buffer solution (22).
A single investigator, blinded to the study groups, counted the number of BCC-like tumors and measured the two-dimensional cross-sectional size of these tumors in five slivers of the dorsal skin taken from each mouse. Total BCC area was determined by summing the cross-sectional areas of all BCC-like tumors observed in the five sections of the 1 cm2 skin samples.
Array comprehensive genomic hybridization (CGH) and imaging was performed as described previously using arrays of 2870 mouse bacterial artificial chromosome and P1 clones (23) and a custom imaging system (24). Data analysis was carried out as described previously (25). Genomic DNA was isolated from paraffin-embedded blocks (six BCC samples and five RMS samples from separate mice) by standard techniques. In addition, genomic DNA was extracted from paraffin-embedded liver tissue from each mouse to be used as the normal reference. The tumor and reference DNAs were labeled by random priming to incorporate Cy3- and Cy5-deoxyuridine triphosphate, respectively. The array CGH data were deposited in the National Center for Biotechnology Information Gene Expression Omnibus database, accession number GSE19880.
SPSS version 11.5 for Windows software (SPSS, Chicago, IL) was used to analyze the data. Significance was assigned to P values equal to or <0.05. One-way analysis of variance and post hoc tests were used for bivariate analysis of quantitative data. Kaplan–Meier plots were used to determine the survival probabilities by occurrence of the macroscopic BCCs or RMSs, being BCC free or being RMS free. Log-rank test was used to compare the survival times among study groups. Quantitative data are reported as mean±1 SD. Survivals are reported as mean±1 SE. χ2 test (or Fisher's exact test, where necessary) was used to compare the differences of categorical variables among genetic groups.
Ptch1+/− mice of wild-type, heterozygous mutant, or homozygous mutant Blm genotypes were treated with ionizing radiation at age 8 weeks, biopsied at age 9 months to assess microscopic BCC-like tumor formation and observed subsequently for visible BCC and RMS development. Irradiated Ptch1+/− mice with homozygous or heterozygous Blmtm3Brd alleles had significantly shorter survival than did those with wild-type Blm alleles (Figure 1a).
Visible BCCs were present in similar numbers of mice in each of the three genotypes, and there were no significant differences in the visible BCC free survival (Figure 1b). The morphology of these tumors varied from BCC-like with almost no follicular differentiation to trichoblastoma-like with advanced follicular differentiation, and the histologic findings did not differ in Blm wild-type versus Blm mutant mice (Figure 2a). However, reduction of Blm gene function was associated with increased microscopic BCC-like tumor size, tumor number per unit area and total BCC involved area with a greater increase in Blmtm3Brd/tm3Brd compared with Blm+/tm3Brd animals. Both the larger size and increased numbers of microscopic BCC-like tumors in Blmtm3Brd/tm3Brd mice compared with Blm+/+ animals (P=0.01) were statistically significant (Figure 3). There was a trend toward enhanced microscopic BCC-like tumor formation in the Blm heterozygotes.
Metastasis of BCC-like tumors was not detected in any mice of the three studied genotypes.
Formation of RMS tumors was significantly enhanced in both Blm+/tm3Brd and Blmtm3Brd/tm3Brd genetic backgrounds suggesting that Blm is haploinsufficient for RMS development. Ten-fold more RMS tumors developed in Ptch1+/− mice carrying one or two mutant Blm alleles compared with Blm+/+ mice (32% of Blmtm3Brd/tm3Brd, 30% of Blm+/tm3Brd versus 3% of Blm+/+ mice; P=0.004 and 0.01 for homozygotes and heterozgotes versus wild-type, respectively). In addition, the age of onset of RMS was reduced in these genetic backgrounds compared with Ptch1+/− Blm+/+ animals (Figure 1c). Nevertheless, as for BCCs, we found no significant histologic differences in RMS tumors arising in mice of the three genotypes including features such as infiltrative growth, necrosis, mitotic activity, heterologous differentiation, atypical mitoses and vascular invasion (Figure 2b).
Genomic copy number was analyzed by array CGH on six BCC and five RMS tumors from Ptch1+/− mice carrying wild-type, heterozygous or knockout Blm alleles. Hierarchical clustering revealed two distinct clusters separating the BCC and RMS samples (Figure 4). All of the BCC samples had losses of the entire chromosome 13 (the site of the Ptch1 gene) as well as of distal chromosome 4. In addition, three BCCs, two from Blm+/tm3Brd mice and one from a Blmtm3Brd/tm3Brd mouse each had copy number loss at the Trp53 locus on chromosome 11. Losses involving portions of chromosome 10 were also present in three BCCs and included Gli1, one of the transcription factors mediating hedgehog signaling, in one case. In contrast, a gain of chromosome 10, including Gli1, was present in all five RMS tumors from Ptch1+/− mice with Blm+/tm3Brd or Blmtm3Brd/tm3Brd genotypes but not in the one RMS tumor arising in a Ptch1+/− Blm+/+ animal. The latter RMS harbored focal deletions at distal chromosomes 6 and 12. One of two RMS tumors arising in Ptch1+/− Blmtm3Brd/tm3Brd mice had partial loss of chromosome 13 including the Ptch1 locus. Overall, the differences in CGH-detectable changes according to Blm genotype did not reach statistical significance (i.e. not <0.05). However, the fraction of the genome altered was higher in the BCCs from Ptch1+/− Blmtm3Brd/tm3Brd mice (0.22 and 0.24) than was the fraction in tumors from the Ptch1+/− Blm+/tm3Brd or Blm+/+ mice (0.11, 0.13, 0.13 and 0.16), although the difference could not be statistically tested due to inadequate sample size. This finding suggests a possible contribution of Blm haploinsufficiency, perhaps via enhancement of Trp53 loss (present in three of these four tumors), to copy number genomic instability in BCCs. A similar analysis was not done for RMS because only one tumor with Blm wild-type genotype was analyzed.
We found that Blm deficiency reduced the survival of Ptch1+/− mice, largely via significantly enhanced RMS tumor formation, with reduced tumor latency and increasing tumor number. Blm deficiency also significantly enhanced the development of microscopic BCC-like tumors, albeit not of visible BCCs. Similar to our findings of mutant Blm alleles enhancing tumorigenesis in Ptch1+/− mice, mutant Blm alleles have been reported to enhance by 2- to 3-fold intestinal tumor formation in Apc +/min mice but, as in our study, did not alter the histology of the tumors (19,20). Similarly, Blmtm3Brd/tm3Brd mice develop far more hematologic malignancies than do Blm wild-type mice, an increase similar to that induced by ionizing radiation, and the incidence is enhanced in a multiplicative fashion by the two (26). Thus, genetically engineered mice with mutated Blm alleles develop cancers of the same types as occur in Blm+/+ mice. Similarly, human Bloom syndrome patients have an abnormally high incidence of cancers, and of 100 such cancers in 168 persons, the range of tumor types was not significantly different from that in sporadic cases (27). One Bloom syndrome patient with many BCCs has been reported (28), but no data on BLM sequences in BCCs have been published. Patients heterozygous for a mutant BLM allele have been reported to have an increased incidence of cancers of various types (29,30). Hence, Blm appears to act as a ‘caretaker’ with respect to many types of cancers as opposed to a ‘gatekeeper’ whose loss enhances the incidence of a more limited range of cancers.
Age and hair cycle phase at irradiation time point can affect murine Ptch1+/− BCC carcinogenesis, with increased BCCs if ionizing radiation is given specifically during anagen (31). In our study, we did not control specifically for follicular phase.
Bloom syndrome cells have an increased frequency of somatic mutations (32,33), micronucleus formation (34) and homologous recombination (35). Chromosome instability in this syndrome is characterized by a striking tendency for spontaneous exchanges between DNA strands. These exchanges occur either within chromosomes, termed sister-chromatid exchanges or between chromosomes at homologous sites (36). Sister-chromatid exchanges, the genetic hallmark of Bloom syndrome, cause chromosomal rearrangements such as duplications, deletions and translocations if they occur either unequally using identical sequences or between non-identical repeat sequences (37). The somatic recombination leads to homozygosity, hemizygosity or partial loss of the expression, e.g. of tumor suppressor genes in Bloom Syndrome cells (33). Mitotic recombination in Bloom Syndrome cells that are heterozygous for tumor suppressor genes may result in daughter cells homozygous or hemizygous for a mutant allele (38,39), and this increased tendency for loss of wild-type alleles may explain at least part of the cancer predisposition in Bloom syndrome patients (40).
Blm appears to function primarily in DNA repair, especially in reducing error-prone homologous recombination, e.g. following double-strand DNA breaks. Blm protein localizes to foci at double-strand DNA breaks with p-ATM, p-CHK1 and 53BP1 and is capable of binding directly to 53BP1 and CHK1(41). Cells lacking Blm protein appear to have increased activation of their endogenous DNA damage response with increased numbers of the latter such foci (42), and this finding suggests that Blm protein normally participates in repair of DNA damage and that its absence delays such resolution.
Like the Blm gene, hypomorphic alleles of the gene encoding another enzyme particularly active in double-strand DNA breakage repair, the Nijmegen breakage syndrome gene (Nbs1ΔB), also induced relatively little qualitative change in intestinal tumorigenesis in Apcmin/+min mice. Although they did not test for chromosomal instability, the authors of that study hypothesized that reduction in Nbs1 activity may not lead to genetic instability in epithelial cells of the intestine. (43). Indeed, knockdown of Blm protein expression in human colorectal carcinoma cell lines did not cause chromosomal aberrations (35,44). These observations are consistent with our findings of a lack of extensive DNA copy number aberrations in either BCC or RMS tumors in our Ptch1+/− mice with mutant Blm alleles (Figure 3); however, CGH does not detect all types of genome instability.
The BCC and RMS tumors differed markedly in their spectra of copy number alterations, including loss of chromosomes 4 and 13 in all BCCs and gain of chromosome 10 in 80% of RMS tumors. In addition, losses of chromosomes 10 and 11, including the Gli1 and Trp53 loci, respectively, were observed in 50% of BCCs (Figure 3). Since Gli1 is one of the transcriptional effectors of hedgehog signaling, selection for its loss in BCCs would not be expected. Indeed, in only one of the three cases was the entire chromosome lost, in the other two Gli1 was excluded from the large minimal region of deletion defined by the extent of the copy number loss. The marked difference we found in loss of chromosome 13, the site of the murine Ptch1 gene, between BCC and RMS tumors is consistent with past findings of frequent, albeit not uniform, retention of the wild-type gene in RMS tumors. Some evidence suggests that in such RMS tumors with both wild-type and mutant alleles, most of the expressed Ptch1 messenger RNA is encoded by the mutant allele, perhaps reflecting epigenetic silencing of the wild-type allele (45,46).
A role for Blm protein in maintaining genomic integrity is certain, but the mechanism by which Blm loss enhances BCC and RMS tumorigenesis in Ptch1+/− mice and the connection between genomic stability and tumor behavior remain open to conjecture (47,48).
National Institutes of Health (CA81888, CA84118).
We thank Levy Kopelovich for ongoing advice and encouragement and Isaak Khaimsky for animal husbandry.
Conflict of Interest Statement: None declared.