A Rb1 promoter variant with reduced activity contributes to osteosarcoma susceptibility in irradiated mice

We have previously shown that an interval on murine chromosome Mmu 14 (72.3 – 118 Mbp) contains the principal locus conferring susceptibility to alpha-radiation induced osteosarcoma in four different mouse strains [20, 21]. (compound LOD score 4.4). Due to incomplete penetrance of such a susceptibility gene, confounding causes of death after irradiation, and the inherent stochastic nature of radiation tumorigenesis a more precise mapping of this QTL using linkage-analysis was not feasible. We have therefore applied a complementary method to better define the position of the susceptibility gene, using somatic allelic imbalances mapping across this locus in osteosarcoma of (BALB/cHeNhg × CBA/Ca)F1 mice. After 227Thorium incorporation at the age of 100 days, 17 out of 80 female F1 mice developed osteosarcoma that were suitable for molecular analysis (median latency time = 553 days, range from 290 to 766 days). In 13 out of 17 of these tumors we observed allelic loss or imbalance for at least one marker between D14Mit234 and D14Mit97, i.e. within the principle OS susceptibility locus on chromosome 14 [21] (Figure 1). In 10/13 cases the maternal (BALB/c) allele was affected, whereas the paternal CBA allele was lost in the remaining 3 cases. This preferential loss of 10 BALB/c versus 3 CBA alleles (p = 0.014, relative to a total of 34 parental chromosomes) is consistent with the assumption that tumors in heterozygous F1 gain a growth advantage when they somatically lose an allele that confers relative resistance. We observed before that in the progeny of (BALB/cHeNhg × CBA/Ca) F1 mice backcrossed to female BALB/cHeNhg (further referred to as BBC) the BALB/c allele at this locus is associated with relative OS resistance [21]. The smallest common interval of overlapping allelic imbalances spans 2.654 Mbp (between D14Mit90 and D14Mit 225, Figure 1) and encompasses 12 known genes, 1 protein-coding gene with unknown function, 2 non-coding RNA genes and 1 pseudogene (Additional file 1). To nominate candidate genes from this genomic interval, we used PosMed, a search engine based on the entire biomedical literature to find links between a set of genes and a given phenotype [22]. Here we identified 7 genes in the D14Mit90/D14Mit225 interval (Rb1, Esd, Cysltr2, Htr2a, Fndc3a, Ltm2b, Lpar6) with a potential function in osteosarcoma susceptibility (Additional file 2). Of these 7 genes, only Rb1 is directly associated with osteosarcoma as suggested by 12 studies (p = 5 × 10⁻⁵³) whereas the other 6 candidates are linked to osteosarcoma indirectly, either via additional genes acting in-trans or through comorbidity (next best p-value 1.1 × 10⁻²⁰ for Esd). It is also worth mentioning that according to the Mouse Genome Informatics database, some modifier loci for other traits have been mapped to this interval (such as Modor3 for ocular retention, Myci2 for myocardial infarction, Hwq1 for heart weight). But since they are not defined as transcribed sequences, we limited our screen to genes and other transcribed elements as validated in the current mouse genome release (NCBI GRCm38).

The link between RB1 and osteosarcoma is mainly based on missense and nonsense mutations in the germline which predispose for retinoblastoma in children and OS in adolescents [23]. Following high-dose radiotherapy for primary retinoblastoma in these patients, the risk of a secondary OS is potentiated [24]. We therefore analysed the pattern of allelic imbalances or loss-of-heterozygosity (LOH) across this minimal deleted interval, using polymorphic microsatellite markers and an intragenic SNP at the Rb1 3’UTR (Additional file 3). LOH breakpoints were found between the distal (D14Mit225) markers and Rb1 in three tumors, two of which also had breakpoints in the proximal interval between D14Mit90 and Rb1 (Figure 1).

We further reasoned that any susceptibility gene for osteosarcoma must be expressed in the relevant target cells in order to influence the tumor risk. We therefore measured the mRNA expression in murine osteoblasts of all genes and none-coding elements as derived from the current release of the mouse genome in this interval (Additional file 1). We found that except for three of them with undetectable expression (Nudt15, Med4, miRNA687hom), the relative expression level in the MC3T3 preosteoblastic cell-line and in-vitro explanted osteoblasts varied between 0.15 fold (Cysltr2) and 3.7 fold (Ltm2b) relative to whole embryo mRNA (Additional file 4).

The classical view of genetic predisposition by a germline loss-of-function RB1 allele followed by somatic losses of the wt allele during progression of retinoblastoma is that eventually the tumor expresses the inherited allele which confers high susceptibility in the germline.

We therefore expected to find a gene variation in the CBA allele of Rb1 which can explain the association with the higher OS risk. Sequencing the coding region of Rb1 from the BALB/cHeNhg and CBA/Ca strain, we found only nucleotide variations in the 3’ UTR, but as they did not map to known mRNA stability motifs they were not considered of major functional relevance [25]. Missense and non-sense RB1 mutations, which account for about 80% of heritable retinoblastoma and osteosarcoma cases in patients [26‐28], can therefore be ruled out in our case. In addition to coding sequence RB1 mutations associated with highly penetrant phenotypes in humans, a small subset of cases exhibit retinoblastoma predisposition with incomplete penetrance, sometimes caused by mutations that affect RB1 promoter methylation sites [9, 29‐31]. An association of these low-penetrance gene mutations with OS predisposition, however, was never found, but might have been missed due to the scarcity of such families and the general low OS incidence.

In fact, we found a CBA/Ca specific TCGCCC hexanucleotide deletion in the Rb1 promoter, located 271 nt upstream of the ATG start and 78 nt upstream of the binding sites for Sp1, ATF and E2F in the promoter core (Figure 2a,b) [32]. In-silico analysis using the tool MathInspector [33, 34] to analyze changes in TF binding sites predicted that this hexanucleotide InDel could delete a WT1 TF binding site and disrupt a SP1-SP1 module (Additional file 5) [35].

To test the impact of these predicted TF binding site alterations on gene expression, we applied both an in-vitro promoter reporter assay as well as direct quantification of Rb1 transcripts in whole embryos of BALB/cHeNhg and CBA/Ca mouse strains. Rb1 promoter fragments encompassing the ubiquitous SP1, ATF and E2F binding sites together with the site of the Δ(TCGCCC) variation from both mouse strains (Additional file 6) were able to activate a CAT reporter gene in an orientation-dependent and reproducible manner (Figure 3a) after transient transfection of ROS17/2.8 rat osteoblasts. Quantification of the reporter activity from the two promoter variants confirmed that the CBA–specific Δ(TCGCCC) deletion produced less CAT-activity (2.42 AU, CI 2.19 – 2.65) when compared to the BALB/c allele (4.38 AU, CI 4.19 – 4.57, Figure 3b). This 1.8 fold difference is highly significant (p = 0.0003, t-test). An inverse orientation of the Rb1 promoter fragment shows only 21% activity and extracts from mock-transfected cells have less than 8.3% basal activity relative to the BALB/c construct. A difference was also found in the average Rb1 gene expression of day 16 whole embryo extracts, being lower in CBA/Ca (183 AU, CI 138 – 228) than in the BALB/cHeNhg strain (232 AU, CI 165 – 299, p = 0.038, t-test, Additional file 7). Whereas the 1.28 fold reduced mRNA expression in whole-embryo might be a more realistic value in view of the natural genomic context of the Rb1-gene, the 1.8 fold lower promoter activity in transfected osteoblasts could reflect some bone-specific regulation of the Rb1 promoter.

It is interesting to note that although the essential TF-binding sites in the core promoter of the Rb1 gene are not affected by this alteration, suggesting that transcription is still possible, two Rb1 control elements map within a 20 bp interval around this hexanucleotide indel [36]. Recently it was shown that these RCEs and other undiscovered sequence elements in-cis of the RB1 core promoter site are crucial for transcriptional fine-tuning in a tissue-specific manner. They are also involved in regulating the RB1 protein homeostasis through feedback loops involving all 3 Rb family proteins members, including a RB1 autosuppression [37].

If for practical purposes we consider a 1.5 fold lower Rb1 expression in homozygous CBA/Ca osteoblasts (i.e. the average between the reporter assay and the in-vivo embryo mRNA level), then the overall Rb1-expression from both alleles in heterozygous BALB/CBA osteoblasts would be reduced to about 83% as compared to 100% in homozygous BALB/BALB osteoblasts (after loss of the BALB/c allele in OS, expression in malignant osteoblasts could further drop to approximately 67%).

The impact of this reduced Rb1 expression in normal osteoblast cells for the risk to develop osteosarcoma after Th227-incorporation can be inferred from re-genotyping the BALB/cHeNhg × (BALB/cHeNhg × CBA/Ca) backcross [21] for the Rb1 promoter variant (Figure 2c). Out of 169 backcrossed mice, 93 inherited the BALB/CBA genotype at the Rb1 promoter, of which 23 (or 24.7%) developed an OS. Of the 76 mice inheriting the BALB/BALB genotype, only 10 were diagnosed with a tumor (13.2%, p = 0.016, Fisher’s exact test). In addition to increased tumor incidence, the Δ(TCGCCC) promoter variant in the CBA-allele is also linked to a shorter tumor latency (p = 0.0007, Log-Rank test) (Figure 2c).

The canonical function of RB1 as a prototype tumor-suppressor gene is based on its ability to regulate the G1/S cell-cycle transition by sequestration of E2F-transcription factors. In bilateral retinoblastoma, this gatekeeper function is completely abolished after loss-of-heterozygosity in carriers of a RB1 germline mutation. In osteosarcoma, although frequently affected by RB1 allelic losses [38], complete absence of gene expression is rarely seen, indicating that a residual function of the wt RB1 protein is compatible with or maybe even accelerates OS growth (Additional file 8). A residual expression was also shown in the majority of other cancer cell lines and tumors except retinoblastoma [26, 39], suggesting that reduced expression, but not complete absence of RB1 promotes tumorigenesis

Apart from the canonical RB1 function as a key cell cycle regulator, recent studies also point to its involvement in non-canonical pathways such as regulation of pancentromeric heterochromatin [40], mitotic checkpoint control [41] and telomere maintenance [42, 43]. We have recently shown that Rb1 haplo-insufficiency results in a more rapid telomere attrition and a higher number of anaphase-bridges and micro-nuclei after gamma irradiation in murine osteoblasts [43].

One could make the point that these non-canonical functions of RB1 are all involved in the maintenance of genome integrity during cell division. An impaired function of either of these mechanisms could lead to the accumulation of genetic or chromosomal defects, in particular those which are transmitted from parental to daughter cells. A defect in the canonical RB1 cell-cycle regulation, on the other hand might disturb the growth rate of a cell itself, but not necessarily increase genomic instability in the progenitor cells. A transmission of chromosomal instability from parent to daughter cells due to an impaired RB1 function was also suggested by discovering its involvement in the regulation of centromeric heterochromatin, a key factor for a proper chromosome segregation during mitosis [44]. In particular the genotoxic stress by ionizing irradiation would require an efficient cellular surveillance mechanism for genetic and chromosomal stability to minimize the risk of malignant transformation [45].

We hypothesize that a reduced RB1 expression renders normal cells more vulnerable to secondary genetic changes, and that the resulting loss of genome stability is a driving factor for tumorigenesis in itself. Rb1 haploinsufficiency in murine ES cells was shown to increase the rate of genetic instability [46], demonstrating that a reduced gene expression can easily compromise genome maintenance. Such a reduced capability of cells to adequately respond to chromosomal defects could have severe consequences after radiation exposure, considering that radiation-induced genotoxicity is mainly due to the generation of DNA strand breaks.

The majority of RB1 germline mutations found in congenital retinoblastoma patients cause partial or complete loss of the gene (41%) or frame shift and nonsense mutations (42%) [26‐28]. The remaining cases, some of which show incomplete penetrance of the condition, are assumed to carry germline mutations affecting regulatory elements of the RB1 gene. It would therefore be worth systematically testing non-syndromic OS patients for the presence of low-penetrance RB1 promoter variants. Recent efforts to screen the human genome for common variants have shown that the RB1 promoter not only harbours several SNPs, but at least two InDel variants. A hexanucleotide variation (rs2092879) 1.5 kbp upstream of the human RB1 promoter core was shown to alter TF binding and promoter activity in vitro [47], whereas a 23bp InDel variant (rs36230211) is located at a similar distance to the transcription start (221 bp proximal) as the murine InDel variation described here (Figure 4). Considering that the human 23bp InDel removes part of the SP1 site from the core promoter, and that the structure of transcription factor binding sites in this region is highly conserved from mouse to human [32, 36] one can assume that carriers of the human InDel variant are also affected by a reduced RB1 expression. The observations that a sufficient RB1 expression is instrumental for cells to prevent structural chromosomal defects from being transmitted to the next cell generation suggests, that a reduced RB1 level might result in a steady accumulation of mutations and genomic instability in normal cells. This would cause a higher incidence of OS in the group of older patients, but at the same time become critical in the event of a genotoxic stress as caused by radiation. It might therefore be interesting to search for the presence of the above mentioned human RB1 promoter variants in older OS patients and in those presenting with a post-radiation, therapy associated tumor.

All animal experiments were approved by the state government (license no. RegOB 211-2531-49/00 and 211-2531-56/06) and carried out in accordance with the federal animal welfare guidelines. (BALB/cHeNhg × CBA/Ca)F1 and F1 mice backcrossed to female BALB/cHeNhg (further referred to as BBC) mice were bred from HMGU Neuherberg (formerly GSF) breeding stocks. For tumor induction, 80 (BALB/cHeNhg × CBA/Ca)F1 and 169 BBC mice (all female) were injected i.p. at the age of 100 days with a single dose of 185Bq/g ²²⁷Th (as Thorium Citrate) [19]. Mice were housed 5 to a cage and examined 5 days a week for the development of tumors or other life-threatening conditions. Bone tumors were diagnosed radiologically at necropsy followed by histological examination after EDTA decalcification. All malignant diseases were classified by trained clinical and veterinary pathologists, using H&E and van-Gieson staining. Typical histologies and X-ray images of tumors are deposited in the PATHBASE database (http://​eulep.​pdn.​cam.​ac.​uk, Acc.No. 3410-3414, Acc.No. PB94713 ff).

Osteosarcoma-free survival curves, corrected for all competing causes of death, were calculated using the Kaplan-Mayer method. Significance of differences were tested with the Statistika 6.0 software (StatSoft, Tulsa OK) using Cox’s F- and LogRank-Test for tumor-free survival, two-tailed Fisher’s exact test for fraction of tumor-bearing mice, one-tailed Chi-square test of a four-field contingency table for the allelic imbalances and t-test for the gene-expression and reporter assays.

DNA for genotyping of mice was extracted from tail-tips and from macro-dissected tumor-sections for measuring allelic imbalances. Genomic PCR was done using primers for polymorphic microsatellite markers and for the Δ(TCGCCC) Rb 1 promoter variant (Additional file 9). Reactions were carried out in 96 well plates using a GeneAmp 9700 thermocycler (Applied Biosystems, Foster City, Ca, USA) followed by 90min electrophoresis on 1.5% or 3% agarose/TBE gels.

Allelotyping of 17 osteosarcomas diagnosed in (BALB/cHeNhg × CBA/Ca)F1 mice and of the paired normal tissue was done for polymorphic microsatellite markers on chromosome 14 (see above) by allel-specific PCR and for a SNP in the Rb1 3’ UTR (3090nt) by genomic sequencing. PCR-amplified microsatellite markers were run for 90min electrophoresis in 3% agarose/TBE/EtBr gels, the PCR products were visualized under UV illumination, captured as 8 bit TIF-files on an electronic DocuGel V system (Mitsubishi, Japan) and quantified using the software package Intelligent Quantifier (BioImage, Jackson MI). To validate linearity and sensitivity of this assay, standard curves of the allelic ratios were generated from BALB/CBA genomic DNA containing different percentages of pure BALB/c or pure CBA/Ca DNA (Additional file 10).

The smallest common region affected by allelic imbalances in osteosarcoma from (BALB/cHeNhg × CBA/Ca)F1 mice as defined by flanking microsatellite markers was submitted to a semantic web association study using PosMed [22]. The search parameters were set as: search=“gene”, condition=“genomic interval”, species=“mouse”, keyword=“osteosarcoma susceptibility” genome sequence version=“NCBIm37”. search settings=“all documents”. With a cut-off p-value of 0.01 (with Bonferroni correction), the list of gene hits was ranked in the order of increasing p-value.

mRNA (1μg per sample) was extracted from the MC3T3-E1 preosteoblastic cell line (ATCC code CRL-2593™) and from primary murine osteoblasts grown in-vitro[43] using RNAeasy mini kit (Quiagen GmbH, Hilden, Germany) and reverse transcribed using SuperscriptIII (Life Technologies, Carlsbad, CA). Quantitative RT-PCR was done in 96well plates on a StepOne machine using Fast Sybr Green (both Life Technologies, Carlsbad, CA) and gene specific primers (Additional file 9). Quantification of expression levels was done using the ΔΔCt method with a house keeping gene (TBP) and whole embryo cDNA as a calibrator.

Histological X-ray images of representational cases of osteosarcoma were deposited to PATHBASE database (http://​eulep.​anat.​cam.​ac.​uk, acc.No. PB94713 ff). Rb 1 promoter sequence variation was deposited to the European Nucleotide Archive (http://​www.​ebi.​ac.​uk/​ena/​) and to the Database of genomic variants archive (http://​www.​ebi.​ac.​uk/​dgva/​, acc.No. pending).