Molecular Cell, Vol. 17, 885–894, March 18, 2005, Copyright ©2005 by Elsevier Inc. DOI 10.1016/j.molcel.2005.02.028
Mechanisms at Intronic Alu
Elements in Mammalian Cells
Beth Elliott,1Christine Richardson,1,2
and Maria Jasin1,*
1Molecular Biology Program
Memorial Sloan-Kettering Cancer Center
1275 York Avenue
New York, New York 10021
tial duplication of MLL, has been demonstrated in
cases of acute myeloid leukemia (AML) in which pa-
tients lack cytogenetic defects (or have trisomy 11)
(Caligiuri et al., 1994; Schichman et al., 1994; So et al.,
1997; Strout et al., 1998). The involvement of Alu ele-
ments in MLL partial duplications is not limited to Alu-
Alu recombination, as breakpoint junctions between an
Alu element and non-Alu sequences have also been
identified (Strout et al., 1998). Alu-Alu recombination
has also been reported in the generation of a reciprocal
translocation present in tumor DNA (Onno et al., 1992).
However, in most reciprocal translocations, Alu ele-
ments are either joined to non-Alu sequences or fused
to other Alu elements or simply occur within the vicinity
of breakpoint junctions (Kolomietz et al., 2002).
Genomic rearrangements arise from breakage and
misrepair, especially of DSBs. Multiple pathways of re-
pair have been demonstrated in mammalian cells for
the repair of a single chromosomal DSB (Liang et al.,
1998). The two primary pathways are NHEJ, involving
little or no homology, and conservative homologous re-
combination (HR) (van Gent et al., 2001). A third path-
way, SSA, can also occur at homologous sequences
near a DSB. Unlike HR, which involves strand invasion,
SSA involves the annealing of DNA strands formed af-
ter resection at the DSB (Pâques and Haber, 1999).
We set out to investigate the involvement of DSB re-
pair pathways in genomic rearrangements in mamma-
lian cells, specifically chromosomal translocations. Pre-
viously, we showed that translocations could be formed
during the repair of DSBs in mouse ES cells (Richard-
son and Jasin, 2000). Translocations were a fraction of
the recovered repair products in this system, and each
translocation had the same overall structure due to
constraints in the substrate design. We devised, there-
fore, a strategy that would allow us to specifically se-
lect translocations in order to determine which mecha-
nism of DSB repair predominates as the translocation
pathway. This strategy involves an intron-based sub-
strate design that is adaptable to addressing a number
of issues regarding mechanisms of genomic rearrange-
ments in mammalian cells.
Repetitive elements comprise nearly half of the hu-
man genome. Chromosomal rearrangements involv-
ing these elements occur in somatic and germline
cells and are causative for many diseases. To begin
to understand the molecular mechanisms leading to
these rearrangements in mammalian cells, we devel-
oped an intron-based system to specifically induce
chromosomal translocations at Alu elements, the
most numerous family of repetitive elements in hu-
mans. With this system, we found that when double-
strand breaks (DSBs) were introduced adjacent to
identical Alu elements, translocations occurred at
high frequency and predominantly arose from repair
by the single-strand annealing (SSA) pathway (85%).
With diverged Alu elements, translocation frequency
was unaltered, yet pathway usage shifted such that
nonhomologous end joining (NHEJ) predominated as
the translocation pathway (93%). These results em-
phasize the fluidity of mammalian DSB repair path-
way usage. The intron-based system is highly adapt-
able to addressing a number of issues regarding
molecular mechanisms of genomic rearrangements
in mammalian cells.
Genomic rearrangements are characteristic of tumor
cells, and specific genomic rearrangements are respon-
sible for many inherited diseases, yet genetic systems
to study their etiology at the molecular level have been
limited. Repetitive elements, which comprise at least
45% the human genome (Lander et al., 2001), present
ample opportunity for genomic rearrangements (Dei-
ninger et al., 2003). Alu elements make up the largest
family of repetitive elements, numbering approximately
one million copies and comprising an estimated 11%
of the genome (Lander et al., 2001).
In germline cells, Alu-Alu intrachromosomal recombi-
nation has been implicated in the etiology of several
inherited diseases, including some cancers (Deininger
and Batzer, 1999; Kolomietz et al., 2002). In somatic
cells, recombination between Alu elements has also
been documented, with the most intensively studied
example involving the MLL gene (Hess, 2004). Recom-
bination between intronic Alu elements, leading to par-
Translocation Substrates: Breakpoint Junctions
within an Intron
Key to our translocation substrate design is an intron-
containing neomycin phosphotransferase gene (neo):
the DSBs occur within a split intron of the neo gene to
allow the selection of a variety of breakpoint junctions
in the formation of the derivative (der) chromosomes
(Figure 1A). For example, translocations involving NHEJ
and/or SSA can be selected. Another important aspect
of the design is that an Alu element is incorporated ad-
jacent to each of the DSB sites. A second set of re-
peats, derived from the puromycin resistance gene
(puro), is also incorporated adjacent to the DSB sites
but on the opposite side from the Alu elements (brack-
ets, Figure 1A).
2Present address: Institute for Cancer Genetics, Columbia Univer-
sity, 1150 St. Nicholas Ave., New York, New York 10032.
Figure 1. Translocation Substrates and Pos-
sible Reciprocal Translocation Outcomes
(A) Translocation substrates on chromo-
somes (chrs.) 17 and 14 are based on a split
neo intron design with an Alu element from
the MLL gene (blue box) inserted 3# and 5#
of the splice donor and splice acceptor por-
tions of the neo gene (neoSD and SAneo,
respectively). DSBs generated by the I-SceI
endonuclease followed by interchromosomal
repair can potentially generate translocation
chromosomes. Repair by SSA at the iden-
tical Alu elements or puro sequences (brack-
ets) will delete one of the repeats, whereas
NHEJ could fuse the Alu elements or puro
sequences, as shown. A neo+gene is formed
on der(17) by either pathway, whereas on
der(14), a puro+gene is formed by SSA and
a puro−gene is formed by NHEJ. By using
SSA and/or NHEJ, four combinations of re-
ciprocal translocations are possible. The to-
tal size of the intron after precise NHEJ is
approximately 1 kb.
(B) Gene targeting of the chrs. 17 and 14 at
the Pim1 and Rb loci to create targeted al-
leles with the translocation substrates (p5
and pF, respectively). The chr. 17 substrate
was targeted first for the creation of the p5
cell line. The chr. 14 substrate was subse-
quently targeted to the Rb locus in the p5
cell line to create the p5pF cell lines #5, #6,
and #18. Vertical black bars are exons 1–4
for the Pim1 locus and exon 20 for the Rb
locus. Abbreviations: HII, HincII and Pst,
The two translocation substrates were targeted to loci
on chromosomes 17 and 14 in mouse ES cells (Figure 1).
The chromosome 17 substrate, termed p5, consists of
the following four components: a 5# neo fragment with a
splice donor site (neoSD), intronic sequences including
an Alu element, an I-SceI endonuclease cleavage site
for DSB formation, and a 3# puro fragment (3#puro)
(Figure 1A). The Alu element, derived from intron 1 of
the MLL gene, has been demonstrated to participate in
some of the partial tandem duplications of MLL found
in patients with AML (Hess, 2004). By using a linked
hygromycin resistance gene (hyg), we targeted this
translocation substrate to the Pim1 locus on chromo-
some 17 in ES cells to create the p5 allele (Figure 1B).
The chromosome 14 substrate, termed pF, consists of
a 5# puro fragment (5#puro), an I-SceI site, the same
MLL intron 1 Alu element followed by additional intronic
sequences, and a 3# neo fragment with a splice accep-
tor site (SAneo) (Figure 1A). The two Alu elements share
290 bp of identity, whereas the 5#puro and 3#puro share
265 bp of identity (brackets, Figure 1A). By using a
linked HPRT mini-gene, the pF translocation substrate
was introduced into the p5 cell line by targeting to an
intron in the Rb locus on chromosome 14 to create the
pF allele (Figure 1B). HPRT+colonies were selected,
and three independent clones that were correctly
targeted (Figure 1B) were chosen for use in the translo-
Table 1. DSB-Induced Translocation Frequencya
Cell Lines No DNA DSB-Induced (pCBASce)
2.6 ± 1.9
5.0 ± 4.0
2.7 ± 1.4
aFor the p5pF and Het Alu cell lines, three independent clones were
examined for each. For the Hom Alu cell lines, two independent
clones were examined. Two experiments were done on each cell
line, except for one cell line of p5pF, Hom Alu, and Het Alu.
bThe frequency of neo+colonies was determined by dividing the
number of neo+colonies by the number of electroporated cells (2 ×
107), correcting for 50% viability after electroporation.
Mechanisms of Reciprocal Translocations
Figure 2. DSBs Induce Reciprocal Translocations
(A) Reciprocal translocations are observed in neo+clones derived from the p5pF cell lines after DSB induction. Parental clones have two
normal chrs. 17 (red) and 14 (green), and neo+clones have one normal chr. 17 and chr. 14 and two derivative chromosomes generated by a
reciprocal translocation, i.e., der(17) and der(14).
(B) The DSB repair pathway used in the formation of the derivative chromosomes is determined by Southern and PCR analyses. Fragment
sizes from Southern (top arrows in each panel) and PCR (bottom arrows) analyses are indicated. Whereas SSA gives a unique product, NHEJ
can occur by precise ligation or can result in deletion (del.) or insertion (ins.) of nucleotides at the breakpoint junction so as to decrease or
increase the size of the Southern or PCR fragments, respectively (see also Supplemental Data). A conservative HR event would give rise to
one derivative chromosome that is identical to that derived from SSA, i.e., with one Alu element or puro repeat, whereas the reciprocal
derivative chromosome would have the remaining repeat segments, either puro-Alu-puro or Alu-puro-Alu, respectively. Because this was not
observed in any of the neo+clones, it is not diagrammed. Abbreviations: RI, EcoRI; HII, HincII; and H3, HindIII.
(C and D) Southern (C) and PCR (D) analyses of neo+clones. Parental p5pF cell lines do not give PCR products (see [D], left), because the
primers for each pair are located on separate chromosomes. See also Supplemental Data.
Figure 3. Translocations by an SSA Mechanism Predominate at Identical Alu Elements
(A) Classes of reciprocal translocation outcomes obtained in neo+clones from the p5pF and Rev48 cell lines after I-SceI expression. These
two cell lines differ in that the p5pF cell line has the I-SceI sites in the same relative orientation on chrs. 17 and 14, whereas for the Rev48
cell line, the I-SceI sites are in opposite orientation. Thus, in the formation of translocation chromosomes, intact I-SceI overhangs at the DNA
ends have the potential to be precisely ligated in the p5pF cell line, but not in the Rev48 cell line.
(B) Sequence analysis of breakpoint junctions from der(17) and der(14) from p5pF neo+clones. The boxed nucleotides show the breakpoint
sequences on both strands after I-SceI cleavage. Breakpoint junction sequences were obtained for each of the derivative chromosomes
except in four cases, which are indicated in parenthesis. The number of nucleotides deleted (del.) from the top strand of each end are
indicated, as well as the number or sequence of nucleotides inserted (ins.). Nucleotides that could have been derived from the I-SceI overhang
on the bottom strand are in bold. The four base I-SceI overhangs are underlined with thin lines in the three junctions derived from precise
ligation (asterisks); microhomologies which occur at three other junctions are underlined with thick lines.
(C) Local derivation of the insertion on der(14) of clone 18G-3. The insertion is derived from sequences 5# and 3# of the chr. 17 DSB (thick
and thin black bars, respectively). This includes 173 bp of the Alu element, partial I-SceI sequences, and 3#puro sequence, in addition to 9
bp of a (TAn)2insertion. Spaces between thick bars represent sequences from chr. 17 that were not contiguous. The number of inserted
nucleotides is indicated in bp below each insertion. The dotted line indicates 31 bp of originally noncontiguous sequence that was repeated
in the junction.
Mechanisms of Reciprocal Translocations
To induce translocations, the three p5pF cell lines
were electroporated with the I-SceI endonuclease ex-
pression vector, and neo+colonies were selected. Each
of the p5pF cell lines gave similar numbers of neo+col-
onies after I-SceI expression, resulting in an average
frequency of 2.6 ± 1.9 × 10−5(Table 1). In the absence of
I-SceI expression, spontaneously arising neo+colonies
were not detected (<10−7), indicating that neo+colonies
arose from DSB repair. Fluorescence in situ hybridiza-
tion (FISH) demonstrated that the neo+clones carried
the two derivative chromosomes, i.e., der(14) and der(17),
expected from a reciprocal translocation (Figure 2A).
The neo+clones also had one intact chromosome 14
and chromosome 17, as expected, as the translocation
substrates are present on only one chromosome 14 and
chromosome 17 in the parental p5pF cell lines.
determine the translocation pathway. From these
clones, 18 (80%) had der(17) chromosomes generated
by SSA and all 20 had der(14) generated by SSA (Figure
3A). Thus, the bias toward SSA at the translocation
breakpoint junctions is not the result of a secondary
DSB repair event.
Translocation Breakpoint Junctions Derived
A total of 15 of the 47 translocation breakpoint junc-
tions from the p5pF neo+clones were formed by NHEJ.
These junctions were analyzed in more detail. Three of
the breakpoint junctions had a restored I-SceI site (as-
terisk, Figure 3B) from precise ligation of the I-SceI
overhangs (thin underline, Figure 3B). Small deletions
and/or insertions, i.e., involving <22 bp, were found in
seven junctions. In one junction, a larger insertion (224
bp) was found that had a complex origin (Figure 3C).
The insertion was “locally derived,” i.e., derived from
nearby sequences, as has been seen in translocation
breakpoint junctions in a number of cancers (Zucman-
Rossi et al., 1998). Larger insertions of approximately
0.9 kb, and 0.7 kb occurred at two der(17) junctions,
as determined by Southern analysis. For one der(14)
junction, no signal was obtained by Southern blot
analysis, indicating that the puro probe sequence was
deleted (data not shown). For another der(14) junction,
a more complex Southern hybridzation pattern was ob-
served, indicating that the puro gene sequences were
rearranged and/or deleted. We also sequenced five of
the SSA products (as indicated in Figure 3B) and veri-
fied that no mutations were introduced into the repeats
In summary, from 47 neo+clones, 15 of the 94 break-
point junctions were derived from NHEJ, the majority
(10 of 15 junctions) of which had little, if any, degrada-
tion or other alteration to the ends (0–21 bp); the re-
maining breakpoint junctions involved more extensive
SSA Predominates as the Translocation Pathway
with Identical Alu Elements
To identify the translocation pathway(s), 47 neo+clones
from the p5pF cell lines were analyzed by Southern
blotting and PCR (Figures 2B–2D; see Supplemental
Data available online with this article). The predominant
class of clones was found to have both derivatives
chromosomes formed by SSA (Class 1, 81%; Figure
3A). Considering the individual derivative chromo-
somes, 40 (85%) of the der(17) chromosomes were de-
rived from SSA of the identical Alu elements, with the
remaining seven (15%) derived from NHEJ (Figure 3A).
Similarly, 39 (83%) of the der(14) chromosomes were
derived from SSA at the puro repeat, whereas only
eight (17%) were derived from NHEJ. Thus, SSA is the
preferred translocation pathway for generating either
derivative chromosome in the p5pF cell lines.
Translocations associated with deletions that extend
beyond the intronic sequences into the neo coding se-
quences would preclude the formation of an intact neo
gene. We therefore took advantage of the ability to se-
lect der(14) translocations with puromycin in order to
determine if translocations selected in this way would
arise in a similar manner as when neo+clones were
selected. Of 54 puro+clones, we found that 52 clones
(96%) were also neo+(data not shown). Thus, neo+se-
lection appears to capture most of the translocations
associated with DSB induction on chromosomes 14
and 17. As before, der(17) primarily arose by SSA of the
Alu elements (data not shown).
Because the 18 bp I-SceI sites in the translocation
substrates are in the same relative orientation on both
chromosomes 14 and 17, it is formally possible that a
derivative chromosome could be formed by precise li-
gation of the I-SceI cohesive overhangs but then un-
dergo another DSB that is repaired by intrachromoso-
mal SSA. To verify that SSA is the initial repair event
leading to the translocations, we constructed another
cell line, termed Rev48, in which the I-SceI sites are in
opposite relative orientations on chromosomes 14 and
17. In this cell line, the p5 allele (Figure 1B) is present
as before on chromosome 17, but the pF allele on chro-
mosome 14 was modified to contain the I-SceI site in
reverse orientation (see Supplemental Data). I-SceI en-
donuclease was expressed in the Rev48 cell line and
20 neo+clones were analyzed at the molecular level to
Heterology between the Alu Elements Has Little
Effect on Translocation Frequency
Alu elements in human genomes are frequently quite di-
vergent from each other, with a range of divergence from
the consensus Alu element of between 2% and 30%
(Smit, 1996). We therefore asked if the substitution of a
heterologous Alu element in one of the translocation
substrates would affect the translocation frequency
and/or translocation pathway. For this, we substituted
the MLL intron 1 Alu element in the chromosome 14
translocation substrate with an Alu element from intron
6 of the MLL gene, which is within the major breakpoint
cluster region (see Figure S1). The two Alu elements,
which are both from the Alu Sxsubfamily, are 20% di-
vergent from each other, although there are regions of
up to 25 bp of complete identity. We paired these par-
ticular Alu elements because somatic recombination
events between them have been found in patients with
AML (So et al., 1997; Strout et al., 1998) (Figure S1).
To construct cell lines containing the heterologous
Alu element, termed the Het Alu cell lines, the chromo-
some 14 targeting vector (Figure 1B, Figure S1A) was
modified to contain the intron 6 Alu element and then
Figure 4. Translocations by an NHEJ Mechanism Predominate at Heterologous Alu Elements
(A) Classes of reciprocal translocation outcomes obtained in neo+clones from the Het Alu cell lines after I-SceI expression. The same class
designation is used as described in Figure 3A, except that “SSA” indicates formation of an intact Alu element that could be consistent with
an SSA mechanism. Blue and red boxes indicate MLL intron 1 and intron 6 Alu element-derived sequences, respectively.
(B) Sequence analysis of der(17) breakpoint junctions derived from NHEJ from the Het Alu neo+clones. The 69 der(17) junctions were divided
into three groups based on the amount of degradation from the DNA ends prior to joining (<100, 100–400, and >400 bp), and four junctions
from each group were sequenced. The boxed nucleotides show the breakpoint sequences on both strands after I-SceI cleavage. The number
of nucleotides deleted (del.) from the top strand of each end are indicated, as well as the number or sequence of nucleotides inserted (ins.).
Microhomologies are underlined.
(C) Structure of the breakpoint junctions containing intact or nearly intact Alu elements. In five clones, an intact Alu element of 290 bp is
found at the breakpoint junction. This hybrid Alu element consists of sequences from the MLL intron 1 Alu element (blue box) and the MLL
intron 6 Alu element (red box), as well as microhomology between the two Alu elements (white box; the length of microhomology is indicated).
A similar overall structure is found in three breakpoint junctions derived from NHEJ, although the fused Alu element is somewhat smaller or
larger than an intact element because the junction does not occur at the same position in both Alu elements.
(D) Breakpoint junctions that restore an intact or nearly intact Alu element. The positions of the five breakpoint junctions that restore an intact
Mechanisms of Reciprocal Translocations
targeted to chromosome 14 in the p5 cell line. Three
clones that were correctly targeted were used in the
subsequent analysis. Because this modification led to
some sequence changes near the I-SceI site, we re-
cloned the intron 1 Alu element into the chromosome
14 targeting vector in an identical manner to create two
“Hom Alu” cell lines in which the intron 1 Alu element
is on both chromosomes 17 and 14 (Figure S1A). In the
Hom Alu and Het Alu parental cell lines, the I-SceI site
on chromosome 14 is in the opposite orientation rela-
tive to the I-SceI site on chromosome 17, as with the
Rev48 cell line.
The I-SceI expression vector was electroporated into
the Hom Alu and Het Alu cell lines, and neo+colonies
were selected. The Hom Alu cell lines had an average
frequency of 5.0 ± 4.0 × 10−5of neo+colonies, and the
Het Alu cell lines had an average frequency of neo+col-
onies of 2.7 ± 1.4 × 10−5(p = 0.27; Table 1). By using
similar analyses as described for the p5pF cell lines,
the neo+clones from the Hom Alu and Het Alu cell lines
were found to contain reciprocal chromosomal translo-
cations (see below; data not shown). Thus, the substi-
tution of a highly diverged Alu element has little or no
effect on the overall translocation frequency.
ments. In some cases, the fusion formed a nearly unit-
length Alu element (Figure 4C). The remaining eight
clones (11%) had greater than 400 bp deleted, leading
to complete loss of one of the Alu elements (Figure 4B).
In these cases, the deletion approached the intron/
exon border of the neo gene on one side.
Junctions with various deletion lengths were se-
quenced (Figure 4B). In some junctions, the deletions
were nearly symmetrical around the DSBs, as in clone
1B-21 in which 129 and 143 bp were deleted from the
chromosome 17 and chromosome 14 ends, respec-
tively. In several other junctions, the deletions were
highly asymmetrical, as in clone 1B-15 in which 425
and 1 bp were deleted from the chromosome 17 and
chromosome 14 ends, respectively. Microhomology of
1–5 bp was present at all of the junctions that did not
contain an insertion. Positions of microhomology are
indicated for two of the fused Alu elements that formed
a nearly unit-length Alu element (clones 1B-5 and 1B-
21; Figures 4C and 4D).
In addition to the clones with bona fide NHEJ junc-
tions, five clones had events that led to the formation
of a single, intact Alu element at the breakpoint junc-
tion, which would be consistent with an SSA event
(hence, “SSA” in Figure 4A). In these intact Alu ele-
ments, the 5# portion was derived from the intron 1 Alu
element, and the 3# portion was derived from the intron
6 Alu element (Figure 4C). This was very similar to the
fused Alu elements arising from NHEJ, except that
these five junctions were “in register,” i.e, at the same
position in both Alu elements, rather than being offset
(Figure 4D). Microhomology was present at each of the
five junctions. In two clones, the microhomology was
short, i.e., 3 and 7 bp (clones 1A-1 and 1A-10, respec-
tively; Figures 4C and 4D). In the other three clones
(clones 1A-8, 2B-3, and 2B-17), the junction occurred
within the longest stretch of near identity between the
two Alu elements at the position of the breakpoint junc-
tion found in an AML patient (patient 20; Figure 4C).
Within this stretch, 32 of 33 nucleotides are identical:
the junctions of clones 2B-17 and 2B-3 occurred 5# and
3# of the single nucleotide polymorphism that exists be-
tween the two Alu elements in this stretch, respectively,
whereas the junction for clone 1A-8 contained a G/T to
A mutation at this polymorphic site.
NHEJ Predominates as the Translocation Pathway
When Heterology Is Present at the DNA Ends
We analyzed a number of neo+clones from the Hom
Alu cell lines and found that, as with the p5pF and
Rev48 cell lines, SSA predominated as the transloca-
tion pathway for both derivative chromosomes (data
not shown). We next characterized 74 neo+clones de-
rived from the Het Alu cell lines. In contrast to cell lines
with identical Alu elements in which >80% of neo+
clones were in Class 1 (SSA/SSA), 92% of neo+clones
derived from the Het Alu cell lines fell into Class 4
(NHEJ/SSA) (Figure 4A). Thus, although heterology at
the chromosome ends did not substantially affect the
translocation frequency, it dramatically shifted translo-
cation pathway usage from SSA to NHEJ in the forma-
tion of the der(17) chromosome. Formation of the recip-
rocal der(14) chromosome by SSA was not affected by
this shift in pathway usage.
To further characterize the der(17) breakpoint junc-
tions derived from NHEJ, we analyzed PCR products
from the neo+gene. The 69 clones with breakpoint
junctions derived from NHEJ were classified into three
groups depending on the estimated amount of degra-
dation from the DNA ends prior to NHEJ (Figure 4B) as
deduced from the size of the neo PCR products (data
not shown). The majority of neo+clones (55 clones;
80%) had less than 100 bp deleted at one or both
DSBs; an additional six clones (9%) had between 100
and 400 bp deleted. Thus, in most of the clones, NHEJ
led to fusion of the two heterologous Alu elements,
either in their entirety or between portions of the ele-
In this report, we investigated translocation pathway
choice in mammalian cells. In particular, we examined
the role of Alu elements, because these repetitive ele-
ments comprise a large portion of the human genome.
By using an intron-based translocation system, we
found that when identical Alu elements are present at
DNA ends, SSA predominated as the translocation
pathway, such that 85% of the derivative chromosomes
hybrid Alu element are shown in brackets below the aligned sequences. Three of these junctions occur at a position where 32 of 33 bp are
identical between the two Alu elements and at which the breakpoint junction (boxed sequence) occurs for AML patient 20 (So et al., 1997).
The breakpoint junction for patient 300 (Strout et al., 1998) is also boxed. The position of two junctions from clones with fused Alu elements
that are not in register are also indicated; the point of fusion with respect to the intron 1 Alu and intron 6 Alu is shown above and below the
had a single Alu element at the breakpoint junction.
When diverged Alu elements are present at DNA ends,
translocation frequency was not substantially altered;
however, there was a dramatic shift to the NHEJ path-
way, such that 93% of the derivative chromosomes had
breakpoint junctions involving little or no sequence
Our translocation system provides a good model for
oncogenic translocations, which frequently have break-
point junctions within intronic sequences and often in
the vicinity of Alu elements (Deininger and Batzer, 1999;
Greaves and Wiemels, 2003; Kolomietz et al., 2002). A
yeast intron-based translocation system has also been
recently developed (Yu and Gabriel, 2004), which will
be important for phylogenetic comparisons. Although
intron-based systems impose some constraints, our re-
sults suggest that the majority of translocation chromo-
somes are recovered with neo+gene selection in our
system. The intron-based system, therefore, provides
the flexibility necessary to recover a variety of break-
then anneal to each other (Pâques and Haber, 1999).
On the other hand, NHEJ efficiently utilizes intact DNA
ends and is suppressed by 5# to 3# end resection
(Frank-Vaillant and Marcand, 2002; Ira et al., 2004). As
a result, when SSA is impaired in yeast by the removal
of repeat sequences or by rad52 mutation, NHEJ is not
enhanced (Karathanasis and Wilson, 2002).
For SSA events in our cell lines containing identical
Alu elements, w290 nucleotides would need to be re-
sected from each end in order to reveal complementary
strands for annealing. By contrast, for NHEJ in our cell
lines containing heterologous Alu elements, a need for
extensive end modification would seem to be abro-
gated for most events, given the limited end modifica-
tions observed at the breakpoint junctions. A major dif-
ference between NHEJ in yeast and mammalian cells is
the ability of mammalian cells to efficiently join a
number of DNA end structures (Smith et al., 2001).
Mammalian cells may be able to either fill in resected
ends or, alternatively, make use of resected DNA ends
for NHEJ in a “micro-SSA” type of NHEJ reaction in-
volving only a few bps of sequence identity. However,
distinct repair factors are expected to be recruited for
micro-SSA events involving a few bps and for “macro-
SSA” events involving 290 bp, because in yeast, the
genetic requirement for these events differ (Kramer et
al., 1994). Therefore, these events are not expected to
Given the major class of reciprocal translocations in
the Het Alu cell lines (i.e., class 4), as well as previous
results (Richardson and Jasin, 2000), our results imply
that NHEJ and SSA are available to repair broken chro-
mosome ends at the same stage(s) of the cell cycle. In
yeast, strand resection is cell cycle regulated, being
much reduced in G1 cells (Ira et al., 2004); presumably,
therefore, SSA would also be reduced in G1 cells. In
vertebrate cells, NHEJ is considered to be the preferred
pathway in the G1 phase of the cell cycle, whereas HR,
which like SSA requires strand resection, is most active
in late S/G2 (Takata et al., 1998). Recent work, however,
has emphasized that NHEJ is not restricted to the G1
stage (Couedel et al., 2004; Mills et al., 2004; Roth-
kamm et al., 2003). Thus, both NHEJ and SSA are pre-
sumably active in late S/G2, suggesting that our trans-
locations may be occurring at this point in the cell
cycle. It should be noted that translocations are
thought to be only a small portion of total DSB repair
events in our cell lines. Translocations are recovered
at a frequency of approximately 3–5 × 10−5, whereas
intrachromosomal NHEJ and HR events are presumed
to occur at >10−2, based on results from other I-SceI
substrates in ES cells (Moynahan et al., 2001).
Translocation Pathway Choice
Our results provide an explanation for why most onco-
genic translocation junctions rarely involve recombina-
tion between repetitive elements: sequence divergence
is sufficient to shift homology-based repair events to
nonhomologous repair events. It seems remarkable
that the translocation frequency in our experiments was
not substantially altered in cell lines containing heterol-
ogous Alu elements compared with cell lines containing
identical elements, even though the DSB repair path-
way dramatically shifted. We do not expect the SSA
events on der(14) to somehow “drive” the recovery of
der(17) events by NHEJ, because we find that translo-
cations can occur in the absence of nearby homology
at a similar frequency to the events described here (D.
Weinstock, B.E., and M.J., unpublished data).
A priori it could have been predicted that cellular
factors would control repair pathway usage, as has
been seen in yeast (Frank-Vaillant and Marcand, 2002;
Karathanasis and Wilson, 2002). Our results, therefore,
highlight the potential for overlapping use of repair
pathways at a DSB in mammalian cells and emphasize
how readily cells can shift from one pathway to another,
indicating the “fluidity” of DSB repair. Competition and
collaboration of DSB repair pathways have been noted
previously in other studies in vertebrate cells; however,
these other studies have utilized DSB repair mutants
which are defective for one or more pathways of repair
(e.g., Couedel et al. , Mills et al. , Stark et
al. , and Takata et al. ).
Factors that are thought to influence the choice of
DSB repair pathway include cell cycle stage, cell type,
growth conditions, and age of cells. This control of DSB
repair pathway usage would have made it seem likely
that cells would “choose” a pathway prior to sensing
the degree of sequence homology at the DNA ends and
then begin processing the DNA ends in a manner ap-
propriate to the chosen pathway. Based on studies in
yeast, the initial processing steps at a DSB differ for
SSA and NHEJ. In SSA, DNA ends are resected in a 5#
to 3# direction to produce 3# single-stranded tails that
Alu Elements and Genomic Rearrangements
Our experiments raise the question as to the mecha-
nism of Alu-Alu recombination that gives rise to dis-
ease-causing alleles. At least in somatic cells, HR
would not appear to be a favored mechanism to gener-
ate chromosomal rearrangements, given previously
published reports (Richardson and Jasin, 2000; Rich-
ardson et al., 1998). Moreover, conservative HR events
would be dependent on RAD51, yet we find that ex-
pression of a dominant negative RAD51 protein (Stark
Mechanisms of Reciprocal Translocations
line with this article at http://www.molecule.org/cgi/content/full/17/6/
et al., 2004) does not reduce the recovery of transloca-
tions (B.E. and M.J., unpublished observations).
The five Het Alu clones in which intact, unit-length
Alu elements were generated by translocation (Figure
4C) are similar to the SSA products from cell lines con-
taining identical elements. However, it is not certain
that these junctions are derived from SSA. Annealing
of the strands from the two heterologous Alu elements
would produce multiple mismatches along the lengths
of the Alu elements and would have to escape hetero-
duplex rejection. Presumably, such an annealed pro-
duct would be a substrate for mismatch repair (Suga-
wara et al., 2004), which might be expected to produce
an Alu element containing patches from each Alu ele-
ment, rather than the hybrid element we observed with
a single crossover position. It is notable, however, that
we obtained fused Alu elements from NHEJ that were
not in register but which came close in size to restoring
an intact Alu element (Figure 4C). In each case, the
breakpoint junction for the fused elements occurred at
a region of microhomology, such that the structure of
these fused elements is strikingly similar to the intact
These observations suggest the possibility that ap-
parent Alu-Alu recombination events between heterolo-
gous elements may arise in some instances from micro-
homology-mediated NHEJ in which the two Alu
elements are in register with each other. This is espe-
cially attractive given that NHEJ is very efficient in
mammalian cells and that sequence divergence signifi-
cantly suppresses SSA and HR. Microhomology-medi-
ated NHEJ has the potential to give rise to a wide range
of events—deletions, duplications (if between sister
chromatids), and translocations. The intron-based sys-
tem we have developed in this report will allow us to
further explore the genetic requirements of these
We thank Margaret Leversha at the Molecular Cytogenetics Core
Facility (MSKCC) for performing FISH experiments and Michael
Backlund and David Weinstock for assistance and discussions.
This project was supported by the Dorothy Rodbell Cohen Founda-
tion (B.E.), National Science Foundation 0346354, and National In-
stitutes of Health GM54668 (M.J.).
Received: December 4, 2004
Revised: February 4, 2005
Accepted: February 24, 2005
Published: March 17, 2005
Caligiuri, M.A., Schichman, S.A., Strout, M.P., Mrozek, K., Baer,
M.R., Frankel, S.R., Barcos, M., Herzig, G.P., Croce, C.M., and
Bloomfield, C.D. (1994). Molecular rearrangement of the ALL-1
gene in acute myeloid leukemia without cytogenetic evidence of
11q23 chromosomal translocations. Cancer Res. 54, 370–373.
Couedel, C., Mills, K.D., Barchi, M., Shen, L., Olshen, A., Johnson,
R.D., Nussenzweig, A., Essers, J., Kanaar, R., Li, G.C., et al. (2004).
Collaboration of homologous recombination and nonhomologous
end-joining factors for the survival and integrity of mice and cells.
Genes Dev. 18, 1293–1304.
Deininger, P.L., and Batzer, M.A. (1999). Alu repeats and human dis-
ease. Mol. Genet. Metab. 67, 183–193.
Deininger, P.L., Moran, J.V., Batzer, M.A., and Kazazian, H.H., Jr.
(2003). Mobile elements and mammalian genome evolution. Curr.
Opin. Genet. Dev. 13, 651–658.
Frank-Vaillant, M., and Marcand, S. (2002). Transient stability of
DNA ends allows nonhomologous end joining to precede homolo-
gous recombination. Mol. Cell 10, 1189–1199.
Greaves, M.F., and Wiemels, J. (2003). Origins of chromosome
translocations in childhood leukaemia. Nat. Rev. Cancer 3, 639–
Hess, J.L. (2004). MLL: a histone methyltransferase disrupted in
leukemia. Trends Mol. Med. 10, 500–507.
Ira, G., Pellicioli, A., Balijja, A., Wang, X., Fiorani, S., Carotenuto,
W., Liberi, G., Bressan, D., Wan, L., Hollingsworth, N.M., et al.
(2004). DNA end resection, homologous recombination and DNA
damage checkpoint activation require CDK1. Nature 431, 1011–
Karathanasis, E., and Wilson, T.E. (2002). Enhancement of Saccha-
romyces cerevisiae end-joining efficiency by cell growth stage but
not by impairment of recombination. Genetics 161, 1015–1027.
Kolomietz, E., Meyn, M.S., Pandita, A., and Squire, J.A. (2002). The
role of Alu repeat clusters as mediators of recurrent chromosomal
aberrations in tumors. Genes Chromosomes Cancer 35, 97–112.
Kramer, K.M., Brock, J.A., Bloom, K., Moore, J.K., and Haber, J.E.
(1994). Two different types of double-strand breaks in Saccharo-
myces cerevisiae are repaired by similar RAD52-independent, non-
homologous recombination events. Mol. Cell. Biol. 14, 1293–1301.
Lander, E.S., Linton, L.M., Birren, B., Nusbaum, C., Zody, M.C.,
Baldwin, J., Devon, K., Dewar, K., Doyle, M., FitzHugh, W., et al.
(2001). Initial sequencing and analysis of the human genome. Na-
ture 409, 860–921.
Liang, F., Han, M., Romanienko, P.J., and Jasin, M. (1998). Homol-
ogy-directed repair is a major double-strand break repair pathway
in mammalian cells. Proc. Natl. Acad. Sci. USA 95, 5172–5177.
Mills, K.D., Ferguson, D.O., Essers, J., Eckersdorff, M., Kanaar, R.,
and Alt, F.W. (2004). Rad54 and DNA Ligase IV cooperate to main-
tain mammalian chromatid stability. Genes Dev. 18, 1283–1292.
Moynahan, M.E., Akgun, E., and Jasin, M. (1996). A model for test-
DNA Manipulations and Cell Line Constructions
The first intron from the adenovirus (Ad1) major late transcription
unit was previously used in a mouse lacZ recombination reporter
(Moynahan et al., 1996) and was adapted for use here in the neo
gene of pMC1neo. The resulting construct was verified to give rise
to G418Rcolonies upon transfection into mammalian cells. This
p-i-neo construct was used in subsequent steps to derive neoSD
and SAneo. The Pim1 and Rb targeting constructs have been de-
scribed previously (Richardson et al., 1998). For details on cloning,
see Supplemental Data.
For translocation experiments, 2 × 107cells in 1 ml of phosphate-
buffered saline were electroporated with 20–25 µg of the I-SceI ex-
pression vector pCBASce (Richardson et al., 1998) in 0.4 cm
electrode-gap cuvette (250 V, 960 µF). Electroporated cells were
aliquoted into 2–5 10 cm diameter dishes. Colonies were selected
in ES media with a drug 20–24 hr after electroporation and were
grown in selection media for 10–14 days before colony counts (or
after colony expansion): G418 (200 ?g/ml), hygromycin (150 ?g/ml),
puromycin (1.6 ?g/ml), or HAT (1× Sigma H0262). For analyses of
neo+clones, see Supplemental Data.
Supplemental Data include Supplemental Experimental Procedures,
Supplemental References, and one figure and are available on-
Molecular Cell Download full-text
ing recombinogenic sequences in the mouse germline. Hum. Mol.
Genet. 5, 875–886.
Moynahan, M.E., Pierce, A.J., and Jasin, M. (2001). BRCA2 is re-
quired for homology-directed repair of chromosomal breaks. Mol.
Cell 7, 263–272.
Onno, M., Nakamura, T., Hillova, J., and Hill, M. (1992). Rearrange-
ment of the human tre oncogene by homologous recombination
between Alu repeats of nucleotide sequences from two different
chromosomes. Oncogene 7, 2519–2523.
Pâques, F., and Haber, J.E. (1999). Multiple pathways of recombina-
tion induced by double-strand breaks in Saccharomyces cerevis-
iae. Microbiol. Mol. Biol. Rev. 63, 349–404.
Richardson, C., and Jasin, M. (2000). Frequent chromosomal trans-
locations induced by DNA double-strand breaks. Nature 405, 697–
Richardson, C., Moynahan, M.E., and Jasin, M. (1998). Double-
strand break repair by interchromosomal recombination: suppres-
sion of chromosomal translocations. Genes Dev. 12, 3831–3842.
Rothkamm, K., Kruger, I., Thompson, L.H., and Lobrich, M. (2003).
Pathways of DNA double-strand break repair during the mamma-
lian cell cycle. Mol. Cell. Biol. 23, 5706–5715.
Schichman, S.A., Caligiuri, M.A., Strout, M.P., Carter, S.L., Gu, Y.,
Canaani, E., Bloomfield, C.D., and Croce, C.M. (1994). ALL-1 tan-
dem duplication in acute myeloid leukemia with a normal karyotype
involves homologous recombination between Alu elements. Cancer
Res. 54, 4277–4280.
Smit, A.F. (1996). The origin of interspersed repeats in the human
genome. Curr. Opin. Genet. Dev. 6, 743–748.
Smith, J., Baldeyron, C., De Oliveira, I., Sala-Trepat, M., and Papa-
dopoulo, D. (2001). The influence of DNA double-strand break
structure on end-joining in human cells. Nucleic Acids Res. 29,
So, C.W., Ma, Z.G., Price, C.M., Dong, S., Chen, S.J., Gu, L.J., So,
C.K., Wiedemann, L.M., and Chan, L.C. (1997). MLL self fusion me-
diated by Alu repeat homologous recombination and prognosis of
AML-M4/M5 subtypes. Cancer Res. 57, 117–122.
Stark, J.M., Pierce, A.J., Oh, J., Pastink, A., and Jasin, M. (2004).
Genetic steps of mammalian homologous repair with distinct muta-
genic consequences. Mol. Cell. Biol. 24, 9305–9316.
Strout, M.P., Marcucci, G., Bloomfield, C.D., and Caligiuri, M.A.
(1998). The partial tandem duplication of ALL1 (MLL) is consistently
generated by Alu-mediated homologous recombination in acute
myeloid leukemia. Proc. Natl. Acad. Sci. USA 95, 2390–2395.
Sugawara, N., Goldfarb, T., Studamire, B., Alani, E., and Haber, J.E.
(2004). Heteroduplex rejection during single-strand annealing re-
quires Sgs1 helicase and mismatch repair proteins Msh2 and Msh6
but not Pms1. Proc. Natl. Acad. Sci. USA 101, 9315–9320.
Takata, M., Sasaki, M.S., Sonoda, E., Morrison, C., Hashimoto, M.,
Utsumi, H., Yamaguchi-Iwai, Y., Shinohara, A., and Takeda, S.
(1998). Homologous recombination and non-homologous end-join-
ing pathways of DNA double-strand break repair have overlapping
roles in the maintenance of chromosomal integrity in vertebrate
cells. EMBO J. 17, 5497–5508.
van Gent, D.C., Hoeijmakers, J.H., and Kanaar, R. (2001). Chromo-
somal stability and the DNA double-stranded break connection.
Nat. Rev. Genet. 2, 196–206.
Yu, X., and Gabriel, A. (2004). Reciprocal translocations in Sacchar-
omyces cerevisiae formed by nonhomologous end joining. Genet-
ics 166, 741–751.
Zucman-Rossi, J., Legoix, P., Victor, J.M., Lopez, B., and Thomas,
G. (1998). Chromosome translocation based on illegitimate recom-
bination in human tumors. Proc. Natl. Acad. Sci. USA 95, 11786–