ABSTRACT
Junctions at class switch recombination sites in the genome are characterized by
a unique sequence feature. Nucleotide substitutions and small deletions are
common on either of the two sides of the switch junction, but not on both
together. We have previously reported an extrachromosomal substrate assay
system for analyzing the recombination of class switch sequences. Here we have
sequenced nine junctions on each side of the break point and compared these to
17 recombination junctions of control substrates from the same cells. Five of
the nine switch recombination junctions have nucleotide substitutions and
deletions, with multiple nucleotide changes being more common. Furthermore,
mutations were found only on a single side of the junction, just as for the
recombination of switch sequences in the genome. In contrast, only one of 17
control substrate junctions had a mutation, and this was a single nucleotide
insertion. This difference is highly significant (
P
< 0.00007) and indicates that the fundamental recombination mechanism is likely
to be similar for switch sequences in the chromosome and on minichromosome
substrates.
There are two types of DNA recombination that are universal in the production of
immunoglobulins in mammals. The first involves assembly of the variable domain
exon by V(D)J recombination (
1
). The second involves appending the heavy chain variable domain exon to a
variety of different constant domains. This change from IgM and IgD to IgG, IgA
or IgE is called class switch recombination and occurs at switch regions
located just upstream of each of the constant domains (
2
-
5
). An understanding of the physiological mechanism of switch recombination is
important because we would like to understand how this process fails in human
disease. The switch regions are involved in translocations in various lymphoid
malignancies. The c-
myc
gene translocates to the Ig switch regions in all sporadic Burkitt's lymphomas;
in ~20% of diffuse, large cell non-Hodgkin's lymphomas; in all L
3
-type acute lymphoblastic lymphomas and in a large fraction of HIV-associated non-Hodgkin's lymphomas (
6
).
In the classification of DNA recombination reactions, class switch is unusual
because the target zone is distinctive for its extensive length and for its
repetitive nature. Because of these features, the term regionally specific
recombination better describes switch recombination than the term site-specific recombination (
7
). The switch sequences are 1-10 kb long and the recombination crossover points can be anywhere within
them (
2
). The switch regions are highly repetitive and G-rich on the non-template strand. The repeat lengths vary from 20 to 80 nt. The
upstream or donor switch region is S[mu]. The downstream or acceptor switch region can be any of S[gamma]3, [gamma]1, [gamma]2b, [gamma]2a, [epsilon] or [alpha] in mouse and any of S[gamma]3, [gamma]1, [alpha]1, [gamma]2, [gamma]4, [epsilon]
or [alpha]2 in human, in that physical order along the chromosome. The downstream
break points are usually within the switch regions. In contrast, only ~60% of the S[mu] switch region break points are within it; 30% are upstream and 10% are
downstream by best estimates (
3
).
A distinguishing feature or signature of switch recombination is difficult to
identify, i.e. it is difficult to find characteristics that unambiguously
identify a recombination event as having been mediated by switch recombination
activity rather than by non-homologous DNA end joining at a random break site. Approximately 60% of
the donor and acceptor switch break point sequences show one or more
nucleotides of homology between the two ends at the recombination junction such
that these few nucleotides could be assigned to either participating DNA end (
3
), but this feature does not distinguish switch recombination from non-homologous end joining (
8
). An undetermined fraction of the recombination events at switch regions
involve nucleotide mutations in the vicinity of the recombination crossover (
3
,
9
,
10
). Interestingly, in the cases examined to date, the mutations are always on one
side or the other of the crossover point, but not on both sides. This feature,
the temporal association with transcription (
11
-
14
) and the switch sequence dependence are the most distinguishing features of
switch recombination that have been noted thus far (
3
).
Explanation of switch recombination within a chromosomal context is the ultimate
goal, however, the study of genomic switch recombination events is not suitable
for all purposes of analysis because cloning each recombination junction is
time consuming, determination of the rate of recombination is often impossible
and manipulation of processes such as transcription and replication must be
indirect. Recently, we (
15
) and others (
16
-
18
) have described an extrachromosomal switch sequence substrate assay that
recapitulates many of the features of switch recombination in the genome. These
switch sequence-bearing minichromosomes recombined at high frequency in murine mature B
cell lines, in some but not all pre-B cell lines and in two plasmacytoma lines. In the B lineage cell lines
that scored positive, a control substrate with non-switch sequences in place of switch sequences recombined at levels that
were typically 50-fold lower. The recombination was transcription- and switch region orientation-dependent; transcription in the physiological orientation
stimulated recombination, but transcription in the reverse direction did not.
Here we have analyzed the primary structural features of recombinants generated
in the minichromosome substrate assay. We have examined the degree of targeting
of the recombination to the switch regions and the sequences of recombinants.
We find that truncated versions of the switch regions show some evidence of
targeting, and they show the same propensity for nucleotide mutation on one
side of the recombination crossover point as has been described for genomic
switch recombination events. This adds an important parallel between the
recombination of switch sequences extrachromosomally and within the context of
the chromosome. Because the extrachromosomal assay can permit determination of
the recombination rate and the mutation frequencies, it may permit distinction
among models for how switch recombination occurs.
The cell lines used in this study (1-8 and Bal17) and their growth conditions have been described previously (
15
).
pGD244 and pGD209 are the switch region and control plasmids, which have also
been described previously (
15
). In brief, we have bypassed tissue-specific transcription from the promoters by the use of constitutive
promoters, as outlined in Figure
1
. These promoters are positioned immediately upstream of positions A and B, the
sites where test sequences [switch ([mu] or [gamma]3) or eukaryotic control (E1 or E2) DNA segments] are inserted. In
the control substrates, positions A and B contain portions of the
RAG
-1 and -2 genes (oriented in an antisense direction) (pGD209). The SV/
tk
and hCMV promoters function well in a wide variety of cells and are active in
each of the cell lines used in this study. In order to facilitate the analysis
of recombinants, a second prokaryotic selection marker was inserted upstream of
the promoter for region A. By selecting for recombinants which preserve both
the [beta]-lactamase (
bla
) and kanamycin (
kan
) resistance genes, we can readily observe a more restricted group of
recombinants by selecting against recombination events extending upstream of
region A of the plasmid. The prokaryotic origin of replication begins ~400 bp downstream of site B, limiting the deletions from extending a large
distance downstream of this. The majority of these recombinants will have
junctions lying within or close to regions A and B. Because class switch
recombination in the genome also includes larger deletions which extend outside
the switch regions, we were interested in using this assay system to examine a
broader target region. Positive selection of recombinants with ampicillin alone
allows identification of recombinants with junctions further upstream of S
[mu]
. Transformation of switch sequence-bearing substrates, such as pGD244, directly into
Escherichia coli
without passing the plasmid through eukaryotic cells results in a 0.01% (10
-4
) or lower background of white colonies.
Recombination substrates were transfected into eukaryotic cells using either
hypotonic DEAE-Dextran (
19
) or electroporation (
15
). The transfected cells were plated into 100 mm dishes at a density of 3 * 10
5
cells/ml in complete medium. Covalently closed, circular DNA molecules were
recovered by alkaline lysis at the indicated times and digested with an amount
of
Dpn
I that was in 10-fold excess over that specified as needed to cleave at the appropriate
sites in DNA (New England Biolabs). Therefore, the plasmid DNA that is
recovered includes only those molecules which have entered the nucleus and
undergone at least one round of replication (
20
). The digested DNA was electroporated into strain MLB7070 containing an amber
mutation in the [beta]-galactosidase gene. The transformed bacteria were plated onto X-gal/IPTG (US Biochemical) LB plates containing either
ampicillin (100 [mu]g/ml) or ampicillin and kanamycin (25 [mu]g/ml). The substrate molecules give rise to blue colonies and any
recombinant molecules which have deleted
supF
will give rise to white colonies (
15
). The ratio of white colonies to the total colonies therefore represents the
percentage of recombination for that substrate. Recombination values (
R
) are the percentage of recombination per hour from 24 to 48 h post-transfection.
R
values are given for recombinants selected on ampicillin (
R
Ap
)- or ampicillin-kanamycin (
R
ApKn
)-containing plates. The number of blue colonies arising from transformation
of
E.coli
by the substrate plasmids is the same using either ampicillin or ampicillin-kanamycin plates. Each plasmid was transfected in at least three
independent experiments with multiple replicates for each time point in each
experiment and the average
R
value +- SE is given.
We used manual Saenger sequencing and automated fluorescent sequencing with an
Applied Biosystems Model 373 to analyze recombinant plasmid molecules on both
DNA strands around the recombination point.
The switch- and control sequence-bearing substrate structures have been described previously (
15
). Truncated segments of the S[mu] and S[gamma]3 regions are positioned downstream of constitutive promoters in the
switch substrate (pGD244) (Fig.
1
). In the control substrates, these same positions contain segments of
eukaryotic (pGD209) DNA in place of the switch regions. Recombinant plasmids
can be cloned by transformation into
E.coli
and can be selected on ampicillin-kanamycin plates as detailed in Materials and Methods.
We have sequenced the junctions of recombinants selected in this way. In order
to be certain of the sequence of the starting substrates, we fully resequenced
both the S[mu] and S[gamma]3 regions. They were nearly identical to the published sequences (
21
,
22
), except for the site of deletion within S[mu]. As the work of others has shown (
3
), this indicated that the propagation of the switch regions in plasmids did not
result in a high spontaneous mutation rate.
We then randomly selected members of the AmpKan pool of recombination products
for detailed analysis. The ratio of recombinants in the AmpKan pool divided by
that in the broader Amp pool was ~0.45. This ratio primarily reflects the targeting efficiency of region A in
the substrates (Fig.
1
), which for pGD244 is the truncated S[mu] segment. This compares to a targeting efficiency of ~0.6 in the chromosome for full-length S[mu] (
3
). Sequence analysis of the recombinants was preceded by first estimating the
approximate location of the break point by restriction digestion. For all nine
junctions sequenced, we examined at least 72 nt on each side (typically >200
nt), except for recombinant GAD8313, which had >200 nt on one side but only 35
nt on the other.
The location of the S[mu] and S[gamma]3 break points in each of the nine recombinants is shown in Figure
2
. As is seen in the genome, there does not appear to be any targeting to a
specific portion of the switch sequences, i.e. the process is apparently only
regionally specific. The distribution of recombinants is relatively well
constrained to the switch sequences. However, one of the break points on the S[mu] side occurs 423 bp upstream of the S[mu] region and one occurs 24 bp downstream. It should be noted that the S[mu] segment used here is only 1.2 kb long, whereas the actual switch
region within the genome is more than twice this length. In the course of
cloning this segment (
23
), it appears that it underwent an internal deletion. Hence, it may no longer be
as active in targeting as the full-length version in the genome. In addition, there are protein binding sites
upsteam and downstream of the switch regions which may be important in
constraining the process more fully (
24
-
26
) and these are not included in this study.
The sequences surrounding the break point in the nine switch recombinants are
shown in Figure
3
. Five showed mutations. All are point mutations except one, which is a 4 bp
deletion. Point mutations and deletions are both seen at genomic switch
recombination junctions (
3
). In parallel with genomic switch recombination events, all five recombinants
with mutations had the mutations only on one side of the junction. In some, the
mutations were on the S[mu] side, while in others, on the S[gamma]3 side. The most distant mutation in this collection was 96 bp away.
The control substrates did not recombine at nearly as high a level as the switch
substates in the same cell lines. Nevertheless, a low level of recombination
was detectable. We were interested in determining the frequency of mutations
among the recombinants that do arise in order to see whether they might have
the same high mutation frequency found for switch sequence recombinants.
We sequenced at least 72 bp on each side (typically several hundred) of the
recombination junction for 16 control substrate recombinants (Fig.
6
). In contrast to the switch recombination events, none of which were
inversions, five of 16 control recombination events were inversions. In one
case, both recombination junctions of the inversion were sequenced. Hence,
there are actually 17 different control recombination junctions.
Dunnick
et al
. (
9
) have previously proposed a model based on error-prone DNA synthesis (Fig.
7
). The most error-prone polymerase known in eukaryotes is pol[beta], which makes ~1 nt incorporation error every 3000 nt (
29
). It is not known how much new DNA synthesis occurs around the break points.
For comparison purposes, if we estimate that there are no more than 200 bp of
new synthesis on each side of each junction, then the minimum mutation rate is
11/(400 * 9) = 0.00300 mutations/bp; for non-switch sequence substrates it is 1/(400 * 17) = 0.00015 mutations/bp. The mutation rate at switch
junctions analyzed here (1 in 333) is notably higher than the pol[beta] error rate (1 in 3000).
One can devise models that do not rely on error-prone DNA synthesis. In one that we call the error-prone repair model, error-prone repair is stimulated by non-B DNA recombination intermediates (Fig.
8
). Recent work by others indicates that non-B DNA configurations are targeted for transcription-coupled nucleotide excision repair, but in an error-prone manner (
30
). These non-B DNA configurations may represent pause sites at which reiterative rounds
of excision repair might introduce mutations. It is interesting to note that
for uncertain reasons, these mutations have also been noted only on one side of
the non-B DNA configuation (
30
). Reaban and Griffin (
31
) have reported for S[alpha] and we have reported for S[gamma]3, S[gamma]2a and S[mu] (
32
) that these switch regions acquire a non-B DNA configuration upon transcription in the physiological direction.
This unusual configuration consists of a stable RNA-DNA hybrid. We believe that this configuration may be the active
substrate for synapsis of the two sites and for cleavage. We have incorporated
these concepts into the error-prone repair model (Fig.
8
).
It is important to note that the mutations that occur are not restricted to the
repetitive region. In three of the five switch recombinants with mutations, the
mutations are located outside the repetitive region [400 bp upstream of the S[mu] segment (GAD8271); 24 bp downstream of S[mu] (GAD8313); 900 bp downstream of S[gamma]3 segment (GAD8306)]. These mutations are in regions of the
substrate that represent unique sequences. In genomic recombination of Ig class
switch sequences, asymmetric mutations are also found in cases where the break
point is outside the repetitive region and in the unique sequence region (
3
). Because of these mutations in the unique sequence regions, error-prone synthesis models cannot invoke the error-prone nature of the polymerase as being induced by the repetitive
nature of the sequences. Mutations attributable to error-prone repair induced by non-B DNA configurations have been noted at distances of >60 bp away
from the site of the non-B configuration; deletions in such substrates can be >100 bp away (
30
). The mutations in non-repetitive DNA are another similarity between minichromosomal and
chromosomal switch sequence recombination.
Regardless of the mechanism by which the mutations arise, in the events here
they can occur on either the upstream (S[mu]) side or on the downstream (S[gamma]3) side. This indicates that there is no discrimination between the
two sides. If the mutations always occurred on the S[mu] segment side, then there would have to be some way in which these DNA ends
(after the initial cleavages) were distinguished. This is not the case.
Therefore, the two DNA ends, S[mu] and S[gamma]3, have the potential to be handled equivalently. However, as part of
the recombination mechanism, in any one recombination event, one side may be
distinguished from the other such that the mutations predominate on one side.
The results presented in this study indicate that the extrachromosomal switch
recombination assay is characterized by many features thought to be a hallmark
of chromosomal switch recombination and, therefore, appears to be a suitable
system to study mechanistic features relevant to chromosomal class switch
recombination. The advantages of this approach are that one can quantitate the
rate of recombination over specific time intervals. One can readily alter the
sequences and determine the effect on recombination. One can also metabolically
manipulate transcription and replication. The cloning of specific recombination
junctions is achieved with fewer manipulations than when cloning from a phage
library and with fewer problems than are encountered when using PCR through
repetitive DNA. Therefore, this additional approach may be very helpful in
assessing mechanistic issues in class switch recombination (
33
).
G.A.D. was supported by PHS grant 5T32CA09302. This work was supported by NIH
grants to M.R.L., who is a Leukemia Society of America Scholar. G.A.D. and J.L.
contributed equally to this work and either can be designated as initial
author.
REFERENCES
Return