Premature condensation induces breaks at the interface of early and late replicating chromosome bands bearing common fragile sites

Materials and Methods

Culture and Treatment of Human Lymphocytes. Whole peripheral blood was obtained by venipuncture from volunteer donors and grown for 96 h as described in ref. 43. Common fragile site activation was performed by treating the cells with 60 ng/ml Cal (Sigma) for 50 min or with 0.1 μg/ml Aph (Sigma) for 20 h before cell harvesting. High-resolution prometaphase banding was achieved according to ref. 43. Briefly, cells were blocked by overnight thymidine (350 μg/ml) treatment. Cells were then rinsed twice with Hanks solution (GIBCO) and grown for 4.5 or 7 h (for optimum banding of chromosomes 3 and X, and 16, respectively) in complete medium containing 10 μg/ml BrdUrd (Sigma). Twenty minutes before harvesting, cells were treated with 50 ng/ml of colchicine (Sigma). Slides were prepared according to standard cytogenetic procedures (43).

Preparation of R- or G-Banded Chromosomes. BGCs were observed on preparations stained by Giemsa (Prolabo) without pretreatments to obtain an homogeneous staining of the chromosomes. Preparations were then destained and treated as described in ref. 43 to reveal R- or G-banding. Preparations were restained by Giemsa, and the chromosome alterations previously detected were then located relative to the bands. Breaks were generally scored from 50 cells from each donor and each experiment.

Genomic Localization and Preparation of the Bacterial Artificial Chromosomes (BACs). BACs were selected from the human genome project RP11 library. FRA3B has been assigned to band 3p14.2 (44). This site was probed with BAC 641C17, which bears the STS markers D3S2757 and D3S1481 close to its telomeric and centromeric ends, respectively. FRA16D has been assigned to band 16q23.2 (45, 46). We selected BAC 571O6, covering markers D16S3138, D16S3125, and D16S3101 in the region of fragility. FRAXB has been assigned to band Xp22.3 (47). We chose BAC 483M24 that covers the GS1 gene and the STS marker DXS1130 at the near distal end of FRAXB. Bacterial strains containing the BACs were spread on LB agar supplemented with 12.5 mg/ml chloramphenicol and grown at 37°C overnight. Clone identity was verified by PCR with probes corresponding to known STS markers. BACs were then extracted according to the manufacturer's instructions (Qiagen). Probes were biotinylated by Nick Translation (Invitrogen) and recovered by filtration through Quick Spin columns (Roche).

Fluorescent in Situ Hybridization (FISH) and BrdUrd Detection. FISH was performed essentially as described in ref. 48 without the proteinase K step. Biotin was revealed with alternating layers of fluoresceinated avidin (FITC) (Vector Laboratories) and biotin-conjugated goat anti-avidin (Vector Laboratories). Two layers of avidin/antiavidin were applied. For BrdUrd detection, chromosomes were denatured as for FISH, and the analog was revealed with three layers of mouse anti-BrdUrd (Becton Dickinson), rabbit anti-mouse Alexa 350, and goat anti-rabbit Alexa 350 (Molecular Probes). Slides were washed three times with PBS solution (Sigma) after each layer. Chromosomes were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) or with propidium iodide (PI, Sigma) and mounted in Vectashield (Vector Laboratories).

Results

Cal Induces Chromosome Damage. PCC was obtained by treating human lymphocyte cultures for 50 min with Cal. Chromosomes condensed in G₁ phase have a spiral appearance. In S phase, they present interspersed blocks of condensed and noncondensed chromatin. In late S and G₂ phases, they resemble mitotic chromosomes and display classical G- and R-banding patterns (Fig. 4, which is published as supporting information on the PNAS web site). Because they are not easily distinguishable from each other, chromosomes in late S, G₂, and mitosis will be collectively referred to below as mitotic-like chromosomes. Cytogenetic analysis of mitotic-like chromosomes from lymphocyte populations of four donors revealed the presence of BGCs. Their frequency was compared to those found in metaphase chromosomes of the same donor in the absence of treatment or after Aph treatment. The mean number of BGCs per genome ranged from 0.36 to 0.88 in untreated cells (mean 0.6), from 2.5 to 11.93 in Cal-treated cells (mean 5.55), and from 1.17 to 7.48 after Aph treatments (mean 4.34). The percentage of BGCs involving a single chromatid of the chromosomes was 70% in untreated cells and reached 94% and 91%, respectively, after Cal or Aph treatment (Table 1). Hence, the chromosomal consequences of both drugs appeared quantitatively very similar.

Cal Induces BGCs at Specific Loci. To determine whether Cal-dependant damage is targeted to specific chromosomal sites, we analyzed BGCs with a two-step procedure that helps minimize observation biases. First, BGCs were detected on homogeneously stained mitotic-like chromosomes of lymphocytes from eight donors. After destaining, chromosomes were R-banded and 839 of these BGCs were localized with respect to the bands. Thirty-five loci displayed a minimum of eight breaks (at least 1% of recurrence) (Table 2). Almost 70% of the BGCs observed (565 of 839) were found among these 35 loci. Thirty-four of these loci had BGCs in at least two donors, and 19 loci in four or more donors, which indicates that Cal induces common fragile sites. We found that the vast majority of these sites lie outside R bands, and, thus, they were assumed to lie in G bands (data not shown).

Cal-Sensitive and Aph-Sensitive Sites Comap on R-Banded Chromosomes. Surprisingly, the localization of Cal-sensitive sites was reminiscent of that of AphS, with a systematic shift toward the closest G band as compared to classical maps (4). To clarify this point, we used the two-step procedure described above to localize 1,231 BGCs on R-banded chromosomes of Aph-treated lymphocytes from 10 donors. Six of these donors were among the eight involved in the study of Cal-induced damage. Twenty-three loci displayed at least 12 BGCs (1% of recurrence), all 23 were observed in at least two donors, and 16 in five or more donors (Table 2). Damage targeted to these sites represented ≈80% of the total number of the mapped BGCs (1,006/1,231). Strikingly, of a total of 39 sites identified with the two drugs, 19 appeared to be induced by both Cal or Aph treatments. If less-stringent analysis was performed, i.e., taking into account sites that displayed at least 1% of BGCs with one agent but had a lower frequency with the second agent, 36 of 39 sites responded to both drugs.

Localization of Aph-Sensitive Sites on G- or R-Banded Chromosomes. Our localization of AphS differs significantly from the map established by Yunis et al. (6), who localized ≈95% of the aphidicolin-induced BGCs in R bands by studying G-banded chromosomes. With R banding, we found that ≈90% of AphS lie in G bands. To address the question of whether R versus G banding is responsible for this discrepancy, we localized 423 Aph-induced BGCs on G-banded chromosomes of one donor by using slides from the same series of spreading as those used for R banding. Our localization of AphS on G-banded chromosomes coincided very well with the classical map of AphS. In striking contrast, the coincidence with our own results after R banding was <5%, which indicates that the staining used for mapping introduces major observation biases. The comparison of the two maps (Table 2) shows that only 2 of 34 AphS studied here appeared to be colocalized by the two techniques (1q31 and 14q23). Fifteen were localized on each side of coalescent G and R bands, in the G band when the chromosomes were R-banded, and in the R band when they were G-banded, suggesting that they actually lie at the interface of these G and R bands. We localized 17 sites in the core of a G band, which were almost systematically assigned to a small R band nested in this G band by Yunis et al. (6). Hence, the bias affecting classical AphS maps also results from an a priori interpretation of the results.

FRA3B, FRA16D, and FRAXB Comap with Cal-Sensitive Sites at the Molecular Level. To establish, at the molecular level, whether Cal and Aph induce damage within the same sequences, we performed FISH experiments with BACs 641C17, 571O6, and 483M24, which are believed to map within FRA3B, FRA16D, and FRAXB, respectively (Fig. 5, which is published as supporting information on the PNAS web site). We verified this point by studying the relative position of the FISH signals of the BACs to the BGCs induced in Aph-treated lymphocytes (Fig. 1). Table 3 shows that BGCs were observed close to and on both sides of the FISH signals of all three BACs. Furthermore, in all of the cases, 25-28% of these BGCs directly involved the sequence identified by the BACs (signals “split” and “in”), which confirms that the three BACs probe for the expected sites. Very similar proportions of BGCs were found proximal to, distal to, or within the FISH signal of any of the BACs in mitotic-like chromosomes of cells treated with Cal (Table 3 and Fig. 1). These results confirm, at the molecular level, the colocalization of BGCs induced on Aph or Cal treatment.

FRA3B, FRA16D, and FRAXB Lie at the Interface of R and G Bands. The difficulties encountered in mapping BGCs on banded chromosomes and the availability of BAC probes prompted us to verify the localization of FRA3B, FRA16D, and FRAXB by FISH approaches. We focused on prometaphase chromosomes displaying high resolution R-banding revealed by detection of BrdUrd incorporated during early S phase (49, 50). Because FISH signals may travel along relatively large regions at this banding resolution, we performed a statistical analysis of the position of the spots relative to the bands. We studied 60 to 74 chromosomes hybridized with each BAC and obtained results suggesting strongly that FRA3B maps at the interface of 3p14.1 and 3p14.2 (Fig. 2A) and FRA16D at the interface of bands 16q23.1 and 16q23.2 (Fig. 2B). Even though bands were poorly resolved in the chromosomal region bearing FRAXB, our results suggest that the site maps close to the interface of bands Xp22.32 and Xp22.31 (Fig. 2C). Note that Yunis et al. (6) have assigned FRA3B and FRA16D to small R bands within G bands and FRAXB to the terminal R band of Xp arm. Hence, our high-resolution mapping supports the conclusion that most AphS lie at the junction of chromosome bands, i.e., at the boundary of domains replicating at different times through S phase.

Cal Does Not Markedly Impair S Phase Progression. Cal has been routinely used to trigger PCC, but because it perturbs the phosphorylation status of various proteins involved in cell cycle control, the drug may also impact on S phase progression (51). To clarify this point, we treated lymphocyte populations with both Cal and BrdUrd for 1 h, or with Cal alone for 30 min followed by 30 min of incubation with both drugs. Cytogenetic analysis showed that BrdUrd was incorporated in the DNA of all S phase cells we observed, whatever the labeling protocol (Fig. 3). Moreover, the mean length of the BrdUrd fluorescent tracks roughly paralleled the length of the labeling period (Fig. 3A). These results indicate that elongation occurs in the presence of Cal and does not notably slow down during Cal treatment. As indicated by the normal patterns and intensities of labeling observed in mitotic-like, in comparison with normal mitotic chromosomes (Fig. 3 B and C), completion of S phase is not affected in the presence of Cal. This observation strongly suggests that Cal-induced damage at AphS results from PCC rather than from replication perturbation. It is generally admitted that the interspersed blocks of condensed and noncondensed chromatin observed in cells undergoing PCC reflects the inability of unreplicated sequences to condense properly in S phase. In good agreement with this view, we observed that BrdUrd labeling is associated with at least partially condensed chromosome domains (Fig. 3 A and B). Why unreplicated DNA is incompetent for condensation in S phase, and possibly in G₂, while it condenses efficiently in G₁, is not known.

Postreplicative Sensitivity of FRA3B to Cal. Because Cal treatment allowed us to observe condensed chromosomes in G₁ and G₂ phases, we compared the frequencies of damage induced by PCC before and after replication. The recognition of the different chromosomes by their banding pattern in G₁ phase being difficult, we used a probe for band 3p14 to study FRA3B in Cal-treated lymphocytes from one donor. We examined 140 cells displaying chromosome morphology typical of G₁ phase. Eight of them had BGCs in or close to the FISH signal on one homologue (5.7%), a percentage roughly similar to that found in metaphase chromosomes of untreated lymphocytes of this donor. In contrast, we found that among 193 cells displaying mitotic-like chromosomes, 76 (≈40%) had damage within or close to the signal: 58 on one homologue and 18 on both homologues (Fig. 6, which is published as supporting information on the PNAS web site). Altogether, these results reveal that Cal sensitivity is cell cycle-dependent.

Fragile Sites Are Committed to Break in G₂ Cells Exiting an Unperturbed S Phase. To determine the lifespan of the DNA features responsible for postreplicative sensitivity to Cal, cells were grown for 1 h in the presence of BrdUrd, then Cal was added for 50 min without BrdUrd removal (Fig. 7, which is published as supporting information on the PNAS web site). Because we have shown that replication proceeds normally in the presence of Cal, the labeling period lasted ≈2 h in this experiment. As expected, labeled cells displayed various patterns of BrdUrd incorporation, depending at which stage of S phase the cells were supplied with BrdUrd (Fig. 7 B and C). We also observed cells with unlabeled mitotic-like chromosomes (Fig. 7D). These cells exited S phase before BrdUrd addition, i.e., at least 1 h before the beginning of Cal treatment. Because the length of G₂ phase ranges from 2 to 2.5 h in cultured lymphocytes, most of these unlabeled cells were treated with Cal during the second half of G₂ and all went through a completely unperturbed S phase. As shown in Table 4, BGCs were observed in both labeled and unlabeled cells. However, we observed an increase in the number of cells devoid of BGCs and a decrease in the number of chromosome alterations per cell in unlabeled as compared to labeled cells. We conclude that specific DNA features appear at fragile sites on replication, even in cells that went through an unperturbed S phase. These features remain frequent in early G₂ but are progressively repaired before M phase entry.

Discussion

Yunis et al. (6) mapped 110 common fragile sites, all within R bands, on G-banded chromosomes of human lymphocytes. However, 10 years before, recurrent breakpoints were mapped on irradiated lymphocytes of normal donors and on untreated lymphocytes of patients affected by Fanconi anaemia. Almost systematically, when chromosomes were R-banded, breakpoints were localized within G bands, and when chromosomes were G-banded, they were localized in R bands. This result implies that the mapping of chromosome breaks is strongly biased by the type of chromosome staining used (52). Such approximation of a few megabases, which is about the DNA content of a prometaphase chromosome band, makes it impossible to correlate the position of BGCs to the replication timing of chromosome domains. Therefore, we reassessed the localization of BGCs induced by Aph by using both R- and G-banding. Our results confirm the existence of the bias described above, show that the currently accepted location of these sites must be reconsidered, and strongly suggest that the vast majority of AphS actually lie at the interface of bands with different replication timing (Table 2). This conclusion was supported by FISH localization of BACs probing for FRA3B, FRA16D, and FRAXB on prometaphase chromosomes (Fig. 2).

The mechanisms responsible for the transition between chromosomal domains replicated at different times during S phase are still unknown. It is tempting, however, to postulate that replication forks originating from early replicating bands are blocked or slowed down at specific pause sites, the replication of regions located downstream of these sites being carried out later by forks emanating from origins of the late replicating bands. This view is supported by the fact that ATR, the activation of which likely results from the accumulation of RPA-bound single-stranded DNA at stalled forks (53), controls FRA3B stability (35, 36). When activated, ATR triggers a phosphorylation cascade that, in turn, prevents collapse of stalled forks or helps collapsed forks to restart. Activation of ATR also blocks the onset of mitosis until the completion of replication. As suggested by Casper et al. (35), by delaying replication of AphS, Aph may increase fork instability or impair fork restart, a phenomenon reinforced in ATR-deficient cells. Altogether, these and our results suggest that AphS are genetically programmed pause sites that ensure the transition between some early and late replicating bands.

We observed that mitotic-like chromosomes resulting from Cal treatment display BGCs, the majority of which colocalize with known AphS at the cytogenetic level (Table 2). This finding was verified at the molecular level in the case of FRA3B, FRA16D, and FRAXB (Table 3). We also showed that Cal does not substantially impair S progression (Fig. 3), which strongly suggests that Cal-induced destabilization of AphS results from PCC. Moreover, we established that breaks do not occur at FRA3B in cells condensed in G₁ (Fig. 6), showing that Cal sensitivity is a postreplicative characteristic of the site. Thus, damage does not result from an intrinsic sensitivity to condensation of the FRA3B nucleotide sequence but is rather triggered by treatments perturbing the relative timing of replication and condensation along the site. We observed that mitotic-like chromosomes of cells treated with Cal at different stages of G₂ phase display BGCs at AphS (Table 4 and Fig. 7), indicating that specific features are present at this level in cells that enter G₂ after an unperturbed S phase. We also showed that the mean number of BGCs per cell decreases when these cells progress through G₂, a result in good agreement with the fact that damage is infrequently observed in timely condensed mitotic chromosomes of untreated cells.

The molecular nature of these features is not fully elucidated. PCC may simply allow observation of preexisting breaks. Indeed, in prokaryotes and lower eukaryotes, several reports have established that pause sites are prone to fork collapse and breakage (54, 55). However, the demonstration that loss of ataxia telangiectasia-mutated function does not impact on AphS stability (35) strongly argues against this hypothesis. The specific role of ATR is better explained whether replication and/or chromatin reorganization after replication remain incomplete in G₂, rendering these loci incompetent for condensation, even in untreated normal cells. In contrast to the S phase exit, the G₂/M transition being stringently controlled, ATR would delay mitotic onset until their complete duplication via its downstream target BRCA1 (37). Hence, these sites might be integral components of the G₂/M checkpoint.

Supplementary Material