Rb1 Haploinsufficiency Promotes Telomere Attrition and Radiation-Induced Genomic Instability

The Fukushima nuclear accident has reawakened concerns about the long-term health consequences of radiation exposure, especially the risk of cancer and the contribution of individual genetic susceptibility to that risk. However, little is known about the cellular events that follow radiation exposure and lead to the development of cancer in a susceptible individual (1).

Osteosarcoma is a tumor that is readily induced by exposure to ionizing radiation, in particular through the deposition of α-particle emitting radionuclides in the mineralizing skeleton. This is reflected by the high osteosarcoma incidences in luminescent dial painters who ingested large quantities of radium salts (2), and in children treated with preparations of thorium (3). Exposure of the skeleton to external photon irradiation is also osteosarcomageneic, with this tumor being a frequent secondary cancer arising in radiation therapy fields (4).

Germline mutation of the RB1 tumor suppressor gene increases sensitivity to both sporadic and radiation-induced osteosarcoma (5–7). A study of the genetic determinants that predispose inbred mouse strains to radiation-induced osteosarcoma after injection of bone-seeking ²²⁷Th revealed a major modifying gene mapping to the Rb1 locus (8). Analysis of the increased susceptibility to radiation that is associated with Rb1 is hindered by the high sporadic rate of cancer in animals engineered to lack one copy of Rb1 in the germline (9). However, the increased sensitivity to radiation could be reiterated in mice that had been rendered Rb1 haploinsufficient through the conditional deletion of one Rb1 allele in the osteoblastic lineage (6).

Osteosarcomas are characterized by the frequent activation of the alternative lengthening of telomeres pathway of telomere maintenance (10) and a high degree of genomic instability. As the loss of telomeric DNA may lead to karyotypic instability (11), and as RB1 haploinsufficiency has been associated with increased genomic instability in premalignant retinoma cells (12), we postulate that the Rb1 gene may be the common denominator-linking susceptibility, genomic instability, and telomeric integrity. We report here that the loss of a single copy of the Rb1 gene is sufficient to cause sustained telomeric attrition and spontaneous genomic instability in cells of the osteoblastic lineage. The extent of the genomic instability is increased after exposure to radiation.

FVB/N Rb1^loxP/loxP mice (13) were obtained from the NIH Mouse Models of Human cancer Consortium, whereas the FVB/N COL1AI-Cre-Tg mice (14) were obtained from the NIH Mutant Mouse Regional Resource Centre at UC Davis, California. The ROSA26R reporter mice B6: 129-Gtrosa26^tm1Sor (15) were purchased from the Jackson Laboratory, Maine. All animals were housed in facilities approved by the animal welfare committee of the State of Bavaria. The correct targeting of the Cre expression to the osteoblastic lineage of in vivo was confirmed using (FVB/N COL1AI-Cre-Tg × B6:129-Gtrosa26^tm1Sor) mice (Supplementary Fig. S1). Inactivation of 1 allele of the Rb1 gene in the osteoblastic lineage was achieved by mating FVB-COL1A1-Cre-Tg and FVB-Rb1^loxP/loxP mice.

Long bones of 2- to 4-week-old offspring of FVB-COL1A1-Cre-Tg × FVB-Rb1^loxP/loxP mice were isolated as described previously (16). In brief, long bones were excised under sterile conditions and scraped free of adherent tissue. After removal of the epiphyses, the bone marrow was flushed using 1 mL PBS. Fragments of the cleaned bone were placed into 6-well dishes and cultured in 1 mL Dulbecco's Modified Eagle Medium supplemented with 10% fetal calf serum. After 1 week, the emergent osteoblastic cells were replated to give passage 1 (P1).

All cells were genotyped for Cre, Rb1, and the excised Rb1^+/Δ19 locus (Supplemental Fig. S2A). For all studies, the wild-type Rb^+/+ and Rb1^+/Δ19 cells used were Cre positive, unless specifically stated otherwise. As expected, no Cre mRNA transcripts were detectable in cultured osteoblasts (Supplemental Fig. S3) due to the repression of the collagen gene expression in the absence of differentiation-promoting medium supplements (16).

The assay was conducted using the β-gal staining kit (Sigma-Aldrich) according to the manufacturer's instructions.

The presence of DNA micronuclei fragments was quantified using a standard protocol (17). A total of 250,000 cells were seeded onto glass microscope slides and 24 hours later were irradiated with either 2 or 4 Gy using an Isovolt 200 X-ray device operating at 230 kVp and 20 mA, with a 1 mm Al and 0.5 mm Cu filter. Sham-irradiated cells were handled in the identical manner without exposure to the X-irradiation. Cytochalasin B (0.5 μg/mL) was added to the cultures for 18 hours at time 0 after irradiation to block the next cytokinesis. Cells were subsequently fixed in 80% ethanol for 15 minutes. DNA was visualized using 4′,6-diamidino-2-phenylindole (DAPI) staining (150 ng/mL) for 5 minutes. Slides were washed in distilled H₂O, air-dried and coverslip mounted using 1 drop of Vectashield. Scoring of micronuclei in binucleate cells was done according to the HUMN project protocols.

The fidelity of chromosomal segregation was determined by visualization of the chromatin in postmitotic cells. A total of 250,000 cells per glass slide were irradiated 24 hours after seeding at 0, 2, and 4 Gy. Cells were grown for a further 48 hours to allow at least 4 cell divisions (cycle time of approximately 12 hours). After fixation in 80% methanol for 15 minutes, the slides were stained with Hoechst 33342 for 5 minutes, washed twice with PBS, and coverslip mounted with 1 drop of Vectashield. Scoring was done visually by counting the number of bridges per 250 cells. Each cell revealing a spontaneous or radiation induced anaphase bridge was scored once, even if multiple bridges existed within that cell. Chromatin bridges between 2 postmitotic daughter cells were also scored and were considered to belong to 1 parental cell.

Cells were trypsinized, washed with 5 mL PBS, and spun down at 1,400 rpm for 5 minutes 5 mL of 70% ethanol (cold) was added while gently vortexing and left in ethanol at 4°C for 4 hours. The fixed cells were spun down at 1,400 rpm for 5 minutes and resuspended 475 μL PBS, plus 3.3 μL RNAse (30 mg/mL) and 25 μL propidium iodide (1 mg/mL). Cell-cycle analysis was conducted using a FACSCoulter (Becton Dickinson).

Telomere length was determined by in situ FISH labeling of genomic DNA (18). Chromosome spreads or interphase nuclei were prepared by incubating cells with 0.05 μg/mL Colcemid for 6 hours. Cells were harvested, incubated in 0.056 M KCl at 37°C for 10 minutes and fixed in methanol/acetic acid (3:1). Metaphase chromosomes were spread on glass slides, air dried overnight, and fixed in 4% formaldehyde in 1 × PBS for 4 minutes. To conduct the hybridization, the samples were treated with 100 μg/mL RNase A in 2 × single-strand conformational polymorphism at 37°C for 1 hour, dehydrated through an ethanol series (70%, 85%, and 100% ethanol for 4 minutes each) at room temperature and air dried. A hybridization mixture (70% formamide, 0.5% bovine serum albumin (BSA), 10 mmol/L Tris-HCl (pH 7.2) and 0.5 μg/mL fluorescein isothiocyanate (FITC) conjugated (C₃TA₂)₃ peptide nucleic acid (PNA) probe (PANAGENE) was applied onto the slides. DNA was denatured at 75°C for 5 minutes. Hybridizations were carried out at room temperature for 4 hours. Slides were washed 2 × 15 minutes in 70% formamide, 10 mmol/L Tris-HCl (pH 7.2) and 3 × 5 minutes in 0.1 M Tris-HCl, 0.15 M NaCl, 0.08% Tween-20 (pH 7.5) at room temperature. Slides were covered with one drop of Vectashield (containing 100 ng/mL DAPI for counterstaining) and mounted with a coverslip.

FISH-labeled telomeres were quantified as previously described (19). Primary cells in culture were washed in PBS and centrifuged, the supernatant was discarded and 500.000 cells resuspended in 500 μL of hybridization mixture containing 70% deionized formamide, 20 mmol/L Tris buffer, pH 7, 1% BSA, and 0.3 μg/mL of the FITC-conjugated PNA probe. Samples were denatured for 10 minutes at 80°C and left to hybridize for 3 hours in the dark at room temperature. Samples without PNA probe were used as negative control. Excess probe was removed in a wash solution containing 70% formamide, 10 mmol/L Tris buffer (pH 7), 0.1% BSA, and 0.1% Tween-20 followed by 2 washes in PBS, 0.1% BSA, and 0.1% Tween-20. Cells were then incubated in 500 μL PBS, 0.1% BSA, 10 μg/mL RNAse A, and 0.1 μg/mL propidium iodide for 1 hour at 4°C before analysis. All samples were stored on ice until analysis by flow cytometry.

Analysis was done on fresh samples using a FACSCoulter (Becton Dickinson). The FITC signal was detected in the green channel and the propidium iodide fluorescence in the red channel with no compensation set on the instrument.

RB1-expressing lentivirus (LV-RB1) was produced by polymerase chain reaction (PCR) amplification of the RB1 cDNA insert of cDNA clone MGC 29887 (Life Technologies). The insert was cloned into the lentiviral vector (pCDH-EF1-MCS-T2A-copGFP; Biocat). The Rb1 shRNA-expressing lentivirus (LV-shRb1) was generated by insertion a short hairpin sequence targeting Rb1 (GCTTAAATCAGAAGAAGAA) into the pFUGW plasmid (20). Control shRNA lentivirus (LV-scr) was generated using a scrambled sequence (Dharmacon Research). Constructs were sequenced to confirm error-free amplification before transduction. Cre-expressing lentivirus under the control of the cytomegalovirus (CMV) promoter with a green fluorescent protein (GFP) expression cassette (LV526) for transduction control was purchased from GenTarget. In this case, the control lentivirus expressed only the GFP mRNA (LV-GFP).

Production of lentivirus was done as recently described (21). For all lentiviral transductions, we used a multiplicity of infection of 2 (2 × 10⁵ cells per well and 4 × 10⁵ TU/mL). Transduction efficiency was determined 72 hours after infection by flow-cytometric analysis of GFP-positive cells. Three or 4 days after infection cells were analyzed by reverse-transcriptase (RT)-PCR and Western blotting to ensure efficacy of the expression.

The coding sequences of Rb1, Cdkn2a/Arf, and p53 were amplified from cDNA and subjected to cycle sequencing using the BigDye 3.1 kit (Applied Biosystems). Mdm2 gene copy number was determined by real-time quantitative PCR of genomic DNA (SybrGreen; Applied Biosystems). Relative copy numbers of Mdm2 were calculated by the ΔΔC_t method, using normal mouse tissue as the diploid reference. The primers used for amplification and sequencing are listed in the Supplementary Material.

To confirm the conditional deletion of 1 Rb1 allele and the expected haploinsufficiency, we used exon-specific PCR to amplify both wild-type and exon 19–deleted copies of the Rb1 gene. The content of the wild-type Rb1 allele was almost halved in cells where the exon 19–deleted allele was present (Supplementary Fig. S2A). The same change was evident at both the transcriptional (Supplementary Fig. S2C) and translational levels (Fig. 1A and B) confirming haploinsufficiency. No reduction of the remaining wild-type Rb1 allele was evident, even after 22 continual passages (Supplementary Fig. S2B).

As expected from the known tumor suppressor role of Rb1 in cell-cycle regulation, the reduced level of Rb1 altered the cell-cycle kinetics, with a reduced number of cells present in both the G₁ and S phases assayed by flow cytometry (Fig. 1C). The presence of a large fraction of polyploid cells may account for this shift as the population doubling time remained unaffected. A functional deficit in the cell-cycle checkpoints was evident following a challenge with 2 Gy acute irradiation. In wild-type cells, the expected rapid G₂–M arrest after irradiation was evident, but was severely abrogated in haploinsufficient Rb1^+/Δ19 cells (Fig. 1C). This was associated with a further increase in the polyploid fraction but not the appearance of a sub-G₀ apoptotic fraction. Cell stress induced by the exposure to ionizing radiation effectively induced senescence of wild-type cells in a dose-dependent manner. However, the induction of senescence was diminished in haploinsufficient cells, even following a 4 Gy exposure (Fig. 2A).

Possible alterations in the stability of the genome in Rb1 haploinsufficient cells were studied by examining micronuclei and polyploidy. The number of spontaneous micronuclei (acentric fragments of genomic DNA; Fig. 2B) was elevated in Rb1^+/Δ19 cells when compared to Rb1 proficient wild-type cells. As with the polyploid fraction, the number of cells showing micronuclei rose with increasing passage number. Polyploid cells were also present in wild-type osteoblasts, where they also showed an increase in numbers at higher passages (Fig. 2C). In contrast, the size of the polyploid cell fraction in haploinsufficient Rb1^+/Δ19 cells increased dramatically with increasing passage numbers, comprising almost 25% of the population by passage 12 (Fig. 2C). Even though we are unable to definitively determine the origin of the polyploid cells a dot plot of cell size versus DNA content of the haploinsufficient cells indicates that a small percentage of the measured G₂–M population are possibly an uncertain overlap of tetraploid G₁ cells.

The stability of telomeric regions is associated with loss of genome stability. We therefore quantified telomeric integrity after loss of 1 Rb1 allele. Labeling of telomeric DNA sequences with the fluorescent Cy3-PNA telomeric DNA probe revealed the presence of telomeric DNA within the micronuclei in both wild-type and Rb1^+/Δ19 cells at passage 11. The Rb1 haploinsufficient cells exhibited more than 2 times the number of telomere-positive micronuclei than the wild-type cells (Fig. 2D). An examination of interphase cells indicated an overall reduction in the level of PNA-labeled telomeric sequences in Rb1^+/Δ19 cells, compared to their wild-type counterparts. Labeling of metaphase spreads revealed that the decrease in telomeric signal strength was due to a general reduction in signal intensity and to the development of considerable heterogeneity in the labeling intensity of different chromosomes in Rb1^+/Δ19 primary osteoblastic cells (Fig. 3A).

Assay of telomere length using quantitative genomic PCR amplification (Fig. 3B and C) established that there was a global reduction in telomere repeats in the Rb1^+/Δ19 cells that coincided with the reduced PNA labeling. As with the parameters describing genomic instability, the reduction in the average telomere lengths of Rb1^+/Δ19 cells grew larger with subsequent passages compared to wild-type cells. The mean number of 102 telomeres detected in wild-type cell at passage 8 approached the theoretical mean value of 108 telomeres that would be expected, taking into account the relative numbers of diploid and tetraploid cells we observed in asynchronous cultures. The number of distinct PNA-labeled telomeres in Rb1^+/Δ19 cells was reduced by more than half (Fig. 3D).

To confirm that the engineered Rb1 haploinsufficiency is indeed responsible for the observed changes, we studied the effect of a direct knockdown of Rb1. Wild-type and haploinsufficient Rb1^+/Δ19 cells were transduced with the LV-shRb1 lentivirus, This lead to the expected reduction in Rb1 expression (Supplementary Fig. S4) and was associated with a reduction in telomere length as early as the third passage posttransduction in both wild-type and haploinsufficient cells, compared to the same cells transduced instead with the LV-scr control lentivirus (Fig. 3E). The decrease in telomere length was more pronounced when LV-shRb1 was used to further deplete the already reduced Rb1 expression of the Rb1^+/Δ19 cells (Fig. 3E).

Transduction of Cre negative wild-type (Rb1^+/+) cells and of Cre negative cells containing a nondeleted floxed Rb1 allele (Rb1^+/flox) with the Cre-expressing lentivirus resulted in in vitro Cre expression (Supplementary Fig. S5). Telomere shortening was observed in the Cre-expressing primary osteoblasts containing the nondeleted floxed Rb1 allele (Rb1^+/flox) but not in the Cre-expressing wild-type (Rb1^+/+) cells 3 passages after transduction (Supplementary Fig. S6).

As a final proof of the role of Rb1 in telomere maintenance, we followed telomeric status after restoration of Rb1 expression. The accelerated rate of telomere loss in Rb1^+/Δ19 cells could be prevented by transduction with LV-RB1 lentivirus whereas transduction with control LV-GFP lentivirus had no effect, with telomere shortening continuing unabated (Fig. 4A). The expression of RB1 in the haploinsufficient Rb1^+/Δ19 cells rescued the increased level of sporadic genomic instability, as measured by the number of anaphase bridges per metaphase plate (Fig. 4B). The expression of RB1 following transduction was equivalent to that of endogenous Rb1 in wild-type cells (Supplementary Fig. S7) and did not cause cell-cycle arrest. Thus, the proportion of cells in G₁, S, and G₂–M phase were comparable to those of wild-type cells at the same passage (38.5%, 21.5%, and 40% vs. 39.6%, 22.8%, and 37.6%).

The direct sequencing of the p53, Cdkn2a/Arf, and the remaining wild-type Rb1 allele in Rb1^+/Δ19 cells at the 12th passage did not reveal any alterations in the coding regions of these genes. In addition, the copy number of the Mdm2 oncogene was shown to be the same as that found in wild-type tissue (Supplementary Fig. S8). P53 expression was undetectable in wild-type and haploinsufficient Rb1^+/Δ19 cells, but in a subclone of apparently spontaneously transformed cells detected in long-term cultures of Rb1^+/Δ19 cells, we were able to detect a point mutation in exon 8 of the p53 gene, leading to accumulation of immunoreactive p53 protein (Supplementary Fig. S9).

The double-strand breaks induced by ionizing radiation may promote genomic instability in cells already rendered susceptible by lack of telomeres. Consequently, we exposed haploinsufficient Rb1^+/Δ19 cells to an acute dose of radiation. The exposure to ionizing radiation caused cytogenetic abnormalities (micronuclei) in equal numbers of wild-type and haploinsufficient Rb1^+/Δ19 cells 24 hours after exposure, attesting to the uniformity of the radiation field (Supplementary Fig. S10; Fig. 5A). However, the number of micronuclei induced in 2 Gy irradiated haploinsufficient Rb1^+/Δ19 cells greatly exceeded that induced by the same exposure in wild-type cells (Fig. 5B) from 1.45 ± 0.13 in the wild-type to 2.2 ± 0.26 in the haploinsufficient cultures. The number of radiation-induced anaphase bridges in Rb1^+/Δ19 cells was more than 4-fold higher than that in irradiated wild-type cells (Fig. 5C). The size of the polyploid cell fraction was also increased by radiation exposure in both wild-type and Rb1^+/Δ19 cells. The number of polyploid cells arising in the irradiated haploinsufficient Rb1^+/Δ19 cells was considerably greater (29.6%) than that seen in the irradiated wild-type cells (2.8%; Fig. 5D). Importantly, the radiation exposure did not influence the telomere length in either the wild-type or haploinsufficient cells (Fig. 5E). Telomeric DNA was present in both acentric fragments (micronuclei) and anaphase bridges (Fig. 5F), indicating involvement of the truncated telomeres in the radiation-induced increase in genomic instability.

Loss of one copy of the retinoblastoma gene (RB1) increases predisposition to the development of both sporadic and radiation-induced osteosarcoma (4, 22). The increase in sensitivity to radiation-induced osteosarcoma is reiterated in a mouse model upon conditional inactivation of the Rb1 gene in the osteoblastic lineage (6). We now show that loss of only one Rb1 allele in nontransformed primary osteoblast cell cultures is accompanied by a number of hallmarks of genomic instability. Thus, in Rb1^+/Δ19 cells, as well as in cells where the endogenous Rb1 expression was suppressed by shRNA, we observe the appearance of a polyploid cell fraction, the presence of sporadic micronuclei (acentric DNA fragments), and extreme shortening of the telomeric DNA. Restoration of the Rb1 status by lentiviral expression of the human RB1 gene was sufficient to prevent the continued increase in instability. These observations are consistent with a role for Rb1 in maintaining telomere integrity and ultimately genomic stability.

Reduced telomere length is closely associated with increasing numbers of completed cell cycles. This raises the formal possibility that the shortened telomeres in haploinsufficient Rb1^+/Δ19 cells may have arisen through accelerated cell division. However, no proliferation-promoting mutations were detected in the tumor suppressor genes Cdkn2a/Arf and p53 nor alterations of copy number changes in the oncogene Mdm2 or in the remaining wild-type Rb1 allele. Thus, there is only a small likelihood of expansion of a rapidly proliferating clone of mutant cells. Moreover, the in vitro deletion of the floxed Rb1 allele by lentiviral-mediated expression of the Cre-recombinase resulted in telomere loss within 3 passages. Wild-type cells would require more than 12 passages to reduce telomere length to that induced 3 passages after the loss of the floxed Rb1 allele, making it unlikely that telomeric losses can be explained simply by increased proliferation.

A direct action of the Cre recombinase on telomere length can also be excluded as the collagen transgene–driving Cre expression is not active in the undifferentiated osteoblast cells studied in vitro, leading to undetectable levels of Cre in these cells (16). The presence of lentiviral expressed Cre recombinase also had no effect on the telomere length of wild-type cells, indicating that the recombinase plays no part in the telomere shortening. The telomere reduction showed a consistent relationship to the copy number of the Rb1 gene, with attrition increasing further when the remaining wild-type allele was targeted by shRNA knockdown, and reduced attrition when the deficiency in retinoblastoma was reversed by expression of RB1.

Exposure of haploinsufficient Rb1^+/Δ19 cells to ionizing radiation raised the level of instability, as indicated by increases in the polyploid fraction, in failed chromosomal segregation, and the appearance of acentric DNA fragments. All 3 parameters reveal levels of instability that are far above those that are induced by the exposure of wild-type cells to radiation. The presence of telomeric DNA within both acrocentric fragments and anaphase bridges in irradiated cells suggests involvement of the shortened telomeres. This is consistent with the formation and breakage of end-to-end fusions between chromosomes capped by truncated telomeres or between truncated telomeres and radiation-induced double-strand breaks (23, 24). Such fused chromosomes would be subject to physical tension at the next cytokinesis, causing random breakage with the potential to form chromatin bridges and micronuclear fragments during the next cell cycle (25). Such repeated cycles of breakage-fusion-breakage (BFB) would continually increase the level of instability, as we have observed in the haploinsufficient Rb1^+/Δ19 cells. In support of this, we observed that telomere length was unaffected by radiation exposure. This indicates that the increased instability arises due to the combination of the induced DNA damage and the preexisting truncated telomeres.

The loss of one copy of the Rb1 gene, and the resultant haploinsufficiency, impaired the ability of the cells to handle radiation-induced stress by either G₂–M arrest or senescence. Impairment of both of these Rb1-dependent DNA damage-restoring functions that follow cell stress (23, 26, 27) may explain the persistence polyploid cells generated by the increased genomic instability. The concomitant loss of the canonical cell-cycle regulatory function of Rb1 would prevent cells with chromosomal damage from entering G₁–S arrest and subsequently entering the apoptosis or senescence pathways. This sequence of events is consistent with the existence of a mutator phenotype originally proposed by Loeb (28) to explain the appearance of mutations in tumor cells at a rate above that predicted from the sporadic mutation rate alone. The duality of action (cell cycle and telomere length) that we ascribe to Rb1 can explain the increases susceptibility to both sporadic and radiation-induced osteosarcoma that is associated with retinoblastoma gene mutations (4, 6). This may also give an explanation for the high levels of chromosomal rearrangements that is frequently observed in osteosarcoma, and more recently in retinoma before the loss of the second RB1 allele (12).

Another major component of the Rb1 pathway, P16INK4A, has been shown to promote telomere erosion, formation of anaphase bridges, and genomic instability in tumors of epithelial origin (29). The function of the retinoblastoma gene product in maintaining genomic instability has been suggested to involve chromosome condensation (30), centromere function, and mitosis-related genomic instability (31, 32). However, these previous studies indicate an apparently paradoxical situation, where loss of the retinoblastoma family members p107 and Rb2/p130, but not of Rb1 itself, were seen to promote telomere lengthening (33, 34). This prompts us to speculate that there is a balance between mutually antagonistic interactions of the retinoblastoma family of proteins that serve to regulate telomeric length.

No potential conflicts of interest were disclosed.

Conception and design: M.J. Atkinson, I. Gonzalez-Vasconcellos, M. Rosemann

Development of methodology: I. Gonzalez-Vasconcellos

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Gonzalez-Vasconcellos, M. Rosemann

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.J. Atkinson, N. Anastasov, I. Gonzalez-Vasconcellos, M. Rosemann

Writing, review, and/or revision of the manuscript: M.J. Atkinson, I. Gonzalez-Vasconcellos

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): O. Klymenko, B. Sanli-Bonazzi

Study supervision: M.J. Atkinson, M. Rosemann

The studies are partly funded by grant 02NUK002E of the Bundesministerium für Bildung und Forschung. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

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.