Proximity of Chromosomal Loci That Participate in Radiation-Induced Rearrangements in Human Cells

See allHide authors and affiliations

Science  06 Oct 2000:
Vol. 290, Issue 5489, pp. 138-141
DOI: 10.1126/science.290.5489.138


Rearrangements involving the RET gene are common in radiation-associated papillary thyroid cancer (PTC). TheRET/PTC1 type of rearrangement is an inversion of chromosome 10 mediated by illegitimate recombination between the RETand the H4 genes, which are 30 megabases apart. Here we ask whether despite the great linear distance between them, RETand H4 recombination might be promoted by their proximity in the nucleus. We used two-color fluorescence in situ hybridization and three-dimensional microscopy to map the positions of the RETand H4 loci within interphase nuclei. At least one pair ofRET and H4 was juxtaposed in 35% of normal human thyroid cells and in 21% of peripheral blood lymphocytes, but only in 6% of normal mammary epithelial cells. Spatial contiguity ofRET and H4 may provide a structural basis for generation of RET/PTC1 rearrangement by allowing a single radiation track to produce a double-strand break in each gene at the same site in the nucleus.

Chromosomal rearrangements involving the RET gene are highly prevalent in radiation-induced thyroid tumors from children exposed to environmental radiation after the Chernobyl accident (1–3) and in thyroid cancers from patients with a history of medical external irradiation (4). They can also be detected 48 hours after exposing either human fetal thyroid explants or undifferentiated thyroid carcinoma cells to x-rays (5, 6). Two common types of RET rearrangement in PTCs are intrachromosomal inversions that fuse the DNA sequence encoding the tyrosine kinase domain of RET with a portion of either theH4 gene (RET/PTC1) or the ELE1 gene (RET/PTC3) (7, 8). In mostRET/PTC3 tumors, the ELE1-RET fusion gene is formed by joining ELE1 intron 5 to RET intron 11 by illegitimate recombination events, most of which conserve all of the DNA in both genes (9). Previous analysis of theRET/PTC3 breakpoint sites in 12 post-Chernobyl tumors showed that the tumors were all different with respect to the exact intronic sites at which ELE1-RET fusion occurred, suggesting that the breaks that mediated illegitimate recombination could occur anywhere within the two introns (9). It was, therefore, surprising that the two breakpoints associated with the formation of theELE1-RET fusion in a given tumor tended to be located directly across from each other when the ELE1 intron 5 was aligned with the inverse orientation of the RET intron 11 (9). This pattern of breakpoint positions suggested that these genes might be aligned and adjacent to each other in the nucleus at the time that double-strand breaks (dsb) were induced by radiation. To assess the spatial proximity of the RET locus to one of its recombination partners, we chose to study the H4 locus, because H4 is 30 Mb away from RET (whereasRET and ELE1 are about 0.5 Mb apart), thus allowing accurate measurements of interphase distances by fluorescence in situ hybridization (FISH). In addition, a probe (D10S539) for a locus between RET and H4 was available (10).

Normal thyroid cells were obtained from four unrelated adult individuals. The primary cultured cells were hybridized withRET and H4 or RET and D10S539 probes (11). The three-dimensional (3D) microscopy showed that in 35% (93/263) of interphase nuclei of thyroid cells, at least one pair of RET and H4 signals were juxtaposed (Fig. 1, A to D). To exclude the possibility that the high frequency of RET and H4juxtaposition was due to the presence of a RET/PTC1 rearrangement in a subpopulation of these presumably normal thyroid cells, 10 to 20 metaphase spreads per individual were analyzed on the same slides (Fig. 1E). The green (RET) and red (H4) signals were separated in all metaphases studied, and the orientation of RET and H4signals relative to the centromere was identical on both chromosome 10 homologs and corresponded to the expected positions of these genes along the linear map of chromosome 10. Additional confirmation of the absence of a RET/PTC1 rearrangement in these cells was obtained by reverse transcriptase polymerase chain reaction (RT-PCR) (12). The sensitivity of the reaction was sufficient to detect cells with a RET/PTC1 rearrangement mixed with cells that lack this rearrangement in a ratio of 1:1000. Thyroid cells from each donor were studied and showed no evidence ofRET/PTC1 transcripts (13).

Figure 1

Two-color FISH of normal thyroid cells with the RET probe (RMC10P013, green spots) and the H4probe (29F6, red spots). First, cells were viewed by 2D microscopy to identify cells where both pairs of RET and H4 signals were separated (A) and cells with at least one pair of adjacent RETand H4 signals (B and C). Cell in the latter group were then visualized in 3D by viewing 30 optical sections of each nucleus. Signals were scored as juxtaposed if they were seen touching or overlapping each other in at least one optical section. To illustrate, the juxtaposed signals in (C) were seen in optical sections 16 to 18 (D). A metaphase spread was analyzed on the same slide showing separated RET andH4 signals on both sister chromatids of both homologs of chromosome 10 (E). The RET signals on sister chromatids are closer together than the H4 signals are as expected because the RET gene is closer to the centromere. Scale bar, 3 μm.

The prevalence of cells with at least one juxtaposed pair ofRET and H4 signals ranged from 29 to 40% in the four individuals (29, 30, 36, and 40%). This variation was not statistically significant (P = 0.76, Fisher's exact test). Of 93 nuclei that had at least one pair of juxtaposedRET and H4 loci, five nuclei had two pairs of juxtaposed signals. The number of cells with both pairs of signals juxtaposed was not different from that expected by chance (P = 0.11) (14), indicating that juxtaposition of one pair of genes neither prevented nor stimulated juxtapositioning of the other pair.

By contrast with the data for H4, only 6% (14/256) of thyroid cells had RET juxtaposed to the D10S539 locus in three individuals (3, 5, and 7%, a nonsignificant interpersonal variation, P = 0.53). The difference between the frequency of association of RET with H4 andRET with DS10S539 was highly significant (P< 0.0001).

In the second stage of analysis, the two-dimensional (2D)RET to H4 and RET to D10S539 distances were measured in interphase nuclei (15). The RETto D10S539 distances ranged from 0 to 3.32 μm and the RETto H4 distances from 0 to 4.34 μm, which was consistent with the known positions of these three loci on chromosome 10; D10S539 is in between RET and H4. The distributions of these distances were then compared with a theoretical Rayleigh distribution, as calculated from a polymer model (16,17). Linear polymers can be considered as long, flexible molecular chains that fold in a random manner. Previous studies have shown that 2D interphase distances between chromosomal loci that are greater than 10 Mb apart do, in fact, conform to the Rayleigh distribution (18). The distribution of distances between RET and D10S539 conformed to the Rayleigh distribution, indicating that the spatial relation between these two loci was dominated by random influences (Fig. 2A). As can be seen in Fig. 2A, the Rayleigh distribution predicts that random factors will causeRET and another locus on the same chromosome to be close to each other in a small fraction of cells, hence the fact that RETjuxtaposed to D10S539 in 6% of thyroid cells may have been due to chance, because there was no evidence of a specific association between these loci. By contrast, the distribution of RET andH4 distances showed marked deviation from the Rayleigh model (Fig. 2B). This was primarily due to the signals that were either juxtaposed or closer than expected. Thus, out of 526 measurements, a statistically significant deviation was found predominantly in the 0- to <0.2-μm range (125 actual measurements compared with 13 expected by chance) and in the 0.2- to <0.4-μm range (43 compared with 36 expected) (19). The remainder of the distribution conformed to the Rayleigh curve, except for a slight excess in the number of cells in which the signals were at least 3.6 μm apart (four compared with one expected). These results provide further evidence that theRET and H4 loci are nonrandomly associated in the nuclei of thyroid cells.

Figure 2

Distribution of interphase distances betweenRET and D10S539 (A) and RET andH4 (B) in thyroid cells as compared with the theoretical Rayleigh distribution (solid line). Dark bars indicate the measurements that were in excess of the number expected on the basis of the Rayleigh distribution (19).

The juxtapositioning of RET and H4 was commonly observed in thyroid cells, raising the question of whether interphase proximity of these loci is a general feature of human cells. To address this question, we studied the interphase distances betweenRET, H4, and D10S539 in two additional cell types, peripheral blood lymphocytes (PBLs) and normal mammary epithelial (NME) cells (20). PBLs showedRET-H4 juxtaposition in 21% (39/182) of cells (Fig. 3) and RET-D10S539 juxtaposition in 4% (8/180) of cells. In NME cells, RET andH4 were juxtaposed in 6% (9/153) of cells andRET and D10S539 in 5% (6/114) of cells. The difference between the frequency of RET-H4 and RET-D10S539 juxtaposition was highly significant in PBLs (P < 0.0001), but not in NME cells. The frequency of RET-D10S539 juxtaposition was similar in all cell types studied.

Figure 3

Two-color FISH of PBLs with theRET probe (RMC10P013, green spots) and the H4probe (29F6, red spots). 2D image of nuclei showing two pairs of signals at a distance (A) and one pair of adjacent RET and H4 signals (B). The latter cell was scored as positive for juxtaposition because RET and H4 signals were seen touching each other in optical sections 7 to 9 (C). Scale bar, 3 μm.

Analysis of 2D interphase distances betweenRET-H4 and RET-D10S539 showed that theRET-H4 distances were not random in PBLs, whereas all other sets of measurements largely conformed to the Rayleigh distribution (Fig. 4). Hence, the association betweenRET and H4 is not a universal feature of human cells, although it is definitely present in cells other than the thyroid.

Figure 4

Distribution of interphase distances betweenRET and D10S539 and RET and H4 in PBLs (A and B) and NME cells (C andD) and the theoretical Rayleigh distribution (solid line). The dark bar indicates the measurements that were in excess of the number expected on the basis of the Rayleigh distribution (19).

Despite more than 50 years of study, the mechanism by which radiation produces chromosomal rearrangements remains unclear. RegardingRET/PTC1 rearrangements specifically, the problem can be reduced to understanding how the RET and H4genes, which share little sequence identity and are 30 Mb apart, become joined. The high frequency with which the RET andH4 loci are juxtaposed offers an explanation by placing both potential recombination partners at the same place in the nucleus. Colocalization would serve to make these genes vulnerable to simultaneous incision by a single radiation track. The breaks produced will be very near each other, creating the opportunity for ends to join cross-wise to produce a rearrangement. It is relevant to point out here that we recently found that in post-Chernobyl thyroid tumors withRET/PTC3 rearrangement, the breakpoints in theRET and ELE1 genes tended to be located across from each other, which would be expected if they were a result of concerted dsb produced by a single radiation track in two adjacent chromosomal loci (9).

If RET-H4 proximity facilitates formation ofRET/PTC1 in irradiated thyroid cells, then gene proximity would be implicated in susceptibility to radiation-induced cancer. TheRET/PTC1 chimeric gene has been shown to be able to cause thyroid cancer and mammary cancer in transgenic mice (21,22). Yet, RET/PTC1 rearrangements are not found in breast cancers in humans (23). In light of our findings, it is reasonable to postulate that RET/PTC1 rearrangements in human mammary cells are rare because the RET gene is not usually near its translocation partner in these cells.

Our data also show that a high frequency ofRET-H4 juxtaposition can occur in human cells that are not known to suffer oncogenic RET/PTC1 rearrangements, i.e., lymphocytes. Data from transgenic mice expressingRET/PTC1 suggest an explanation. Although they have theRET/PTC oncogene, these animals do not develop lymphomas, suggesting that lymphocytes may be resistant toRET/PTC-mediated transformation. Perhaps signaling through the RET tyrosine kinase is not sufficient for transformation of mouse lymphocytes. The same may be true in human lymphocytes.

The concept that two loci are able to participate in a radiation-induced chromosomal rearrangement because they are very close to each other in the nucleus is not new [reviewed in (24)]. However, little direct evidence of juxtapositioning of loci is available. The BCR andABL genes, which are on different chromosomes, were reported to be located less than 0.3 μm apart in 2 to 8% of normal human lymphocytes (25). However, this frequency of proximity could be considered to be marginally higher than background taking into account that the data were collected with 2D analysis.

The frequency of RET and H4 contiguity we report here is far and away the most striking example of spatial association between heterologous genes yet described. The high rate of these gene interactions in interphase nuclei would provide a structural basis for occurrence of RET/PTC rearrangements in human thyroid cells.

  • * To whom correspondence should be addressed. E-mail: Yuri.Nikiforov{at}


View Abstract

Navigate This Article