ArticlePDF Available

Windle B, Draper BW, Yin YX, O'Gorman S, Wahl GM.. A central role for chromosome breakage in gene amplification, deletion formation, and amplicon integration. Genes Dev 5: 160-174

Authors:

Abstract and Figures

A CHO cell line with a single copy of the DHFR locus on chromosome Z2 was used to analyze the structure of the amplification target and products subsequent to the initial amplification event. Dramatic diversity in the number and cytogenetic characteristics of DHFR amplicons was observed as soon as eight to nine cell doublings following the initial event. Two amplicon classes were noted at this early time: Small extrachromosomal elements and closely spaced chromosomal amplicons were detected in 30-40% of metaphases in six of nine clones, whereas three of nine clones contained huge amplicons spanning greater than 50 megabases. In contrast, the incidence of metaphases containing extrachromosomal amplicons fell to 1-2% in cells analyzed at 30-35 cell doublings, and most amplicons localized to rearranged or broken derivatives of chromosome Z2 at this time. Breakage of the Z2 chromosome near the DHFR gene, and deletion of the DHFR gene and flanking DNA was also observed in cells that had undergone the amplification process. To account for these diverse cytogenetic and molecular consequences of gene amplification, we propose that chromosome breakage plays a central role in the amplification process by (1) generating intermediates that are initially acentric and lead to copy number increase primarily by unequal segregation, (2) creating atelomeric ends that are either incompletely replicated or resected by exonucleases to generate deletions, and (3) producing recombinogenic ends that provide preferred sites for amplicon relocalization.
Content may be subject to copyright.
10.1101/gad.5.2.160Access the most recent version at doi:
1991 5: 160-174Genes Dev.
B Windle, B W Draper, Y X Yin, et al.
deletion formation, and amplicon integration.
A central role for chromosome breakage in gene amplification,
References
http://genesdev.cshlp.org/content/5/2/160#related-urls
Article cited in:
http://genesdev.cshlp.org/content/5/2/160.refs.html
This article cites 61 articles, 35 of which can be accessed free at:
service
Email alerting click herethe top right corner of the article or
Receive free email alerts when new articles cite this article - sign up in the box at
http://genesdev.cshlp.org/subscriptions
go to: Genes & DevelopmentTo subscribe to
Copyright © Cold Spring Harbor Laboratory Press
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
A central role for chromosome breakage
in gene amplification, deletion
formation, and amplicon integration
Brad Windle/ Bruce W. Draper/ Yuxin Yin, Stephen O'Gorman, and Geoffrey M. Wahl
Gene Expression Laboratory, The Salk Institute for Biological Studies, La Jolla, California 92037 USA
A CHO cell line with a single copy of the
DHFR
locus on chromosome Z2 was used to analyze the structure
of the amplification target and products subsequent to the initial amplification event. Dramatic diversity in
the number and cytogenetic characteristics of
DHFR
amplicons was observed as soon as eight to nine cell
doublings following the initial event. Two amplicon classes were noted at this early time: Small
extrachromosomal elements and closely spaced chromosomal amplicons were detected in 30-40% of
metaphases in six of nine clones, whereas three of nine clones contained huge amplicons spanning >50
megabases. In contrast, the incidence of metaphases containing extrachromosomal amplicons fell to 1-2% in
cells analyzed at 30-35 cell doublings, and most amplicons localized to rearranged or broken derivatives of
chromosome Z2 at this time. Breakage of the Z2 chromosome near the DHFR gene, and deletion of the DHFR
gene and flanking DNA was also observed in cells that had undergone the amplification process. To account
for these diverse cytogenetic and molecular consequences of gene amplification, we propose that chromosome
breakage plays a central role in the amplification process by (1) generating intermediates that are initially
acentric and lead to copy number increase primarily by unequal segregation, (2) creating atelomeric ends that
are either incompletely replicated or resected by exonucleases to generate deletions, and (3) producing
recombinogenic ends that provide preferred sites for amplicon relocalization.
[Key Words: Gene amplification; chromosome breakage; deletion; gene targeting; genetic instability]
Received September 13, 1990; revised version accepted December 19, 1990.
Chromosomes in cancer cells are often rearranged rela-
tive to their counterparts in normal cells. Such changes
can alter the organization or change the copy number of
genes normally involved in the control of cell growth
and differentiation, and the ensuing alterations in gene
expression affect the proliferative capacity of the cell
(e.g., see Bishop 1987). Because the large-scale chromo-
somal alterations that arise in cancer cells occur infre-
quently in normal cells (Tlsty 1990; Wright et al. 1990),
it is likely that control mechanisms that safeguard chro-
mosomal integrity are abrogated in the development of
malignancy. Thus, understanding the molecular basis of
common chromosomal rearrangements should provide
insight into the mechanisms that maintain chromosom-
al integrity and how they may have gone awry in the
development of malignancy.
Our studies have focused on gene amplification, a pro-
cess that increases gene copy number and elevates the
expression of a variety of proto-oncogenes in vivo and a
large number of genes that confer drug resistance in vitro
(e.g., for references, see Stark et al. 1989; Wahl 1989).
'Cancer Therapy and Research Center of South Texas, University of
Texas, San Antonio, Texas 78229 USA; ^Fred Hutchinson Cancer Re-
search Institute, Seattle, Washington 98195 USA.
The amplification process can be divided into two
phases. (1) In the early phase, the amplified sequences are
typically genetically unstable. Cells with unstable am-
plification products are usually isolated by selection
with a single low concentration of selective agent and
passaged for a "minimal" time prior to analysis. In most
cases,
the instability derives from acentric extrachromo-
somal molecules ranging in size from —100 kb to >2000
kb [e.g., episomes and double minute chromosomes
(DMs);
for references, see Cowell 1982; Wahl 1989].
However, in one instance, unstable early stage amplifi-
cation products have been postulated to be chromosomal
(Saito et al. 1989). (2) In the late phase, the amplified
sequences are typically genetically stable, and they are
detected in cells subjected to multiple step selection or
extensive cell culture. Where extrachromosomal ele-
ments were detected in early stage cells, their integra-
tion into one or more chromosomal sites can generate a
stable amplification phenotype at later stages (e.g., for a
review and references, see Wahl 1989; Von Hoff et al.
1990).
Cytogenetic and molecular structures detected in late
stage cells may be generated by processes unrelated to
those that produce the initial unit of amplification on
"amplicon" (for review, see Stark et al. 1989; for a recent
160 GENES & DEVELOPMENT 5:160-174 © 1991 by Cold Spring Harbor Laboratory ISSN 0890-9369/91 $1.00
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Gene amplification mediated by chiomosome bteakage
example, see Ruiz and Wahl 1990). Thus, deductions
about mechanisms for generating the initial amplicons
should depend on analyzing amplification events as
close as possible to the occurrence of the initial event.
However, most of the proposed mechanisms for gene
amplification derive from analyses of the structures
present in late-stage cells (for recent reviews, see Stark et
al.
1989; Wahl 1989). Furthermore, unambiguous analy-
sis of the target site subsequent to amplification has
been difficult because most of the target genes studied
thus far are autosomal and present in at least two copies
in the heteroploid cell lines often employed. Studies that
circumvented this problem by using gene transfer to in-
troduce an amplifiable gene into a single chromosomal
site have indicated that deletion of the transfected gene
occurred during amplification, but it is possible that
these results reflect a special property of the insertion
sites analyzed (e.g., see Ruiz and Wahl 1988, 1990). Am-
plification of an endogenous N-myc gene in neuroblas-
toma cells has also been linked recently with deletion of
one N-myc gene copy (Hunt et al. 1990), but only one
example from a single amplification event was exam-
ined, and it is uncertain whether deletion was a primary
event in this case. In contrast, another study concluded
that deletion was not involved in the amplification of
endogenous CHO DHFR genes (Trask and Hamlin 1989).
Thus,
whether deletion of the target locus is a common
element of the amplification process at different loci, or
even within a single locus, remains an important and
controversial issue.
This paper examines the early products generated dur-
ing the amplification of an endogenous dihydrofolate re-
ductase
[DHFR]
gene in CHO cell line
UA21,
which con-
tains only a single DHFR locus on chromosome Z2
(Urlaub et al. 1986). This experimental system enabled
us to investigate directly the structures of the amplifi-
cation target and amplicons. The results implicate chro-
mosome breakage as an early or initial event in gene
amplification, as suggested by previous cytogenetic stud-
ies (e.g., see Biedler et al. 1988), and they indicate that
the initial amplicons are most frequently acentric ele-
ments. In contrast to a recent report (Trask and Hamlin
1989),
we show that cells having undergone DHFR gene
amplification are prone to deletion of the DHFR gene
and flanking sequences. A model is presented in which
chromosome breakage generates the initial acentric am-
plification intermediates, produces atelomeric ends
whose shortening creates deletions, and provides sub-
strates for the targeted insertion of amplicons.
Results
Early events generate unstable DHFR amplicons
Fluorescent in situ hybridization (FISH) was used to
characterize the number and cytogenetic features of
DHFR amplicons in methotrexate (Mtx)-resistant sub-
clones of UA21 cells approximately eight to nine cell
doublings after application of selection. The selection
protocol was designed to favor isolation of recent ampli-
fication events. Innocula of 100 UA21 cells were ex-
panded to ~2 X 10^ cells and then challenged, in a single
step,
with the lowest concentration of Mtx that would
enable isolation of cell clones with DHFR gene amplifi-
cation (see Materials and methods). Because cells with
amplification pre-exist at frequencies of~10~^tolO~'^
(Kempe et al. 1976; Tlsty et al. 1989), most of the events
that generated clones should have occurred within one
or two cell doublings of application of selection. Infor-
mative preparations were obtained from 10 clones. Of
these, nine showed evidence of DHFR gene amplifica-
tion. The exception, which probably resisted Mtx by one
of several other known mechanisms (Flintoff et al. 1978),
was not examined further.
Individual cells of each Mtx-resistant clone frequently
contained more independent DHPi?-specific FISH signals
than the parental UA2I cells, and there was substantial
heterogeneity in the number of signals displayed by in-
dividual cells in each clone (Fig. 1; Table 1). In UA21
cells,
80% of interphase nuclei showed either one or a
clustered pair of yellowish signals (Fig. la); no additional
signals were observed. In contrast, in each Mtx-resistant
single cell "clone," between 30% and 66% of interphase
nuclei showed more than two signals arrayed as a loose
cluster or dispersed throughout the nuclei (Fig. lb,c,f).
The interphase cells shown in Figure 1, b and f, effec-
tively illustrate the variation in signal number that was
routinely observed in each clone.
Examination of the metaphase spreads in these same
preparations showed that the DHFR FISH signals ob-
served in interphase nuclei of Mtx-resistant clones were
generated by a process that frequently affected the integ-
rity of the amplification target. For the parental UA21
line,
43/50 (86%) of the metaphase spreads examined
showed one or a pair of FISH signals approximately one-
third of the distance between the centromere and the
telomere on the long arm of the submetacentric chromo-
some Z2 (see Fig. 3A; Funanage and Myoda 1986; Trask
and Hamlin 1989); no signal was observed in the remain-
ing seven spreads (Table 1). In contrast, in Mtx-resistant
clones, only 15-50% of the metaphases examined con-
tained a chromosome exhibiting a wild-type hybridiza-
tion signal at the native position (e.g., the chromosome
in the center of Fig. Ih with a single pair of fluorescent
dots).
This reduction in the fraction of wild-type sites
with single-copy hybridization is statistically significant
(P < 0.005).
The nine clones analyzed could be divided into two
classes based on the amplicon size and the presence of
small extrachromosomal elements. In six clones, the
amplicons were relatively small, and DHPi^-specific hy-
bridization was detected in small extrachromosomal el-
ements (Fig. lc,d) or in clusters in chromosomes. Within
clusters, chromosomally localized amplicons displayed
little or no propidium fluorescence between adjacent
DHFR signals (e.g., see the two clusters of signals in the
dicentric chromosome in Fig. Ig). Individual cells within
a single clone also manifested differences in the loca-
tions of DHFR FISH signals. For example, the cells in
Figure 1, d and e, descended from a single progenitor, yet
GENES & DEVELOPMENT 161
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Windle et al.
162 GENES & DEVELOPMENT
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Gene amplification mediated by chromosome breakage
Table 1. DHFR amplicon distribution eight tuhne cell doublings post-Mtx selection
Clone
UA21
2A
2B
2C
2A
C3
C14
CI
C5
C16
Number
64
50
16
43
14
ND
17
22
28
11
Interphase nuclei
Percent with n FISH signals
n = 0
20
7
12
14
28
18
18
29
0
n = 1-
80
35
38
43
42
18
50
36
55
-2 n >3
0
Class 1: small
58
50
43
30
64
Class 2: large
32
35
45
ampl
Number
50
icons
4
5
4
14
6
10
amplicons
13
13
6
Metaphase spreads
Percent with
0 sites
14
0
40
0
14
33
60
0
0
0
WT
86
25
40
25
14
17
20
38
31
50
hybridization to
NWT
0
0
0
50
35
33
30
85
77
50
SEE
0
75
20
50
42
17
40
0
0
0
DHFR FISH signals were counted in interphase cells, and the locations and sizes of the amplicons were determined in metaphase
preparations such as those shown in Fig. 1. The 50 UA21 metaphases analyzed here were prepared under identical conditions to those
used for the Mtx-resistant cells. Many non-wild-type (NWT) chromosomes may be derivatives of the wild-type (WT) chromosome.
Percentages exceeding 100 exist in some cases because individual cells displayed more than one site of hybridization. Class 1 clones
contain small extrachromosomal elements (SEE) and small amplicons (see
Fig.
Ic, d, g); class 2 clones contain large amplicons (Fig. Ih).
(ND) Not determined.
one cell contained small extrachromosomal elements
while the other contained a single site of hybridization at
the end of a metacentric chromosome. The cells in Fig-
ure 1, f and g, represent another example of diverse am-
plicon locations in a single subclone; other cells in this
same subclone manifested extrachromosomal signals.
Extrachromosomal signals v^ere observed in 17-75%
of the metaphases obtained from subclones with small
amplicons (Table 1). No extrachromosomal structures
were observed in the 50 metaphases from parental UA21
cells prepared under identical conditions or the —1000
UA21 metaphases prepared from mass cultures (B. Win-
die,
Y. Yin, and G.M. Wahl, unpubl.). The extrachromo-
somal signals represent true hybridization to DNA be-
cause the yellowish fluorescence of the probe was sur-
rounded by a red halo generated by the propidium iodide
counterstain. Structures with similar characteristics to
those designated as extrachromosomal elements in these
studies were routinely observed in cell lines known, by
other criteria, to contain solely extrachromosomal am-
plicons (Y. Yin and G.M. Wahl, unpubl.).
The amplicons were large in the second class compris-
ing three clones, and there was no evidence of the small
extrachromosomal elements described above. The FISH
signals in metaphase spreads were widely spaced (Fig.
Ih).
Ring or rod structures containing widely and regu-
larly spaced amplicons and lacking obvious centromeric
constrictions were observed in several metaphases (data
not shown). In many cases, the spacing between FISH
signals was similar to the distance between the endoge-
nous DHFR gene and the adjacent telomere on the long
arm of chromosome Z2. This observation suggests that
the large amplicon in this second class of clones consists
of all sequences distal to the DHFR gene.
A different distribution of amplicons is observed
30-35 cell doublings postselection
Previous studies suggested that small extrachromosomal
elements of the type described above represent transient
intermediates in the amplification process and are prone
to integrate in CHO cells (for references, see Wahl 1989;
Ruiz and Wahl 1990). The following experiments inves-
tigate this issue and the molecular mechanisms that gen-
erate unstable amplicons. Since all of the cells in each
colony detected at eight to nine cell doublings were re-
quired for cytogenetic analysis, a second selection was
performed and six independently isolated single-cell
clones with DHFR gene amplification were obtained.
Each clone was expanded to —10'^ cells (i.e., —25 cell
doublings) to obtain sufficient material for additional
analyses.
The heterogeneity in the number of sites of DHFR
hybridization detected in the experiments described
Figure 1. DHFR gene distribution in Mtx-resistant cells eight to nine cell doublings postselection UA21 cells were selected to resist
Mtx, and metaphase spreads were prepared from 10 independently derived clones containing 300-400 cells as described in Materials
and methods. Shown are representative interphase and metaphase FISH patterns obtained from UA21
[a]
and the following Mtx-
resistant clones: C5
{b);
C3 (c); 2C (d,e); 2A
{f,g);
and CI (A). A quantitative summary of the results is shown in Table 1. The arrows
indicate bona fide FISH signals (yellowish fluorescence over red propidium iodide stained DNA). Probe: DHFR cosmid c400-30.
GENES & DEVELOPMENT 163
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Windle et al.
above predicts that most amplification events should re-
sult in substantial intercellular differences in the num-
ber of functional DHFR genes. Flow microfluorimetry
with fluoresceinated Mtx (F-Mtx), a tight binding, and
specific inhibitor DHFR (Kaufman et al. 1978; Gaudray
et al. 1986) revealed substantial heterogeneity of DHFR
expression in all six cell populations derived from the
expansion of the respective parent single-cell clones (for
one example, see Fig. 2A; P followed by the number of
the parental clone designates such populations). Fluores-
cence-activated cell sorting (FACS) revealed a 30-fold
dif-
ference in DHFR gene copy number between cells exhib-
iting the lowest and highest 10% of DHFR expression
Log Fluorescence [F-Ma]
75-20
Low
10%
Sorted
High
10%
Soncd
DHFR
1 1
Control 1 1
(rDNA) U
Relative Copy
normalized to rDNA) ^
0
0
30
Figure
2.
Heterogeneous
DHFR
gene copy number in cells —25
cell doublings postselection. (^4) Each of six clones isolated after
a single-step Mtx selection, shown to contain amplified DHFR
genes,
was analyzed for the DHFR content of individual cells
—25 cell doublings postselection using the F-Mtx staining and
FACS protocol described in Materials and methods. Shown are
the fluorescence profiles of DG44
{DHFR
", designated "back-
ground fluorescence"), UA21 (one DHFR copy), and one typical
Mtx-resistant cell population (P75-20).
[B]
P75-20 cells exhibit-
ing the highest and lowest 10% of F-Mtx binding were isolated
for quantitation of relative DHFR gene copy number (see Ma-
terials and methods). After hybridization with a DHFR-speciiic
probe, the blot was stripped of bound probe and rehybridized
with a rDNA clone (pi la-2) to provide an internal hybridization
standard. The DHFR and rDNA hybridization signals are
shown, along with the relative numbers of DHFR copies in the
two subpopulations.
(Fig. 2B). Similar results were obtained for populations
derived from the other five independently isolated clones
(data not shown).
FISH was used to determine the localization of DHFR
amplicons 30-35 cell doublings postselection. In con-
trast to the results seen at eight to nine cell doublings
postselection, only 1-2% of the cells analyzed at 30-35
cell doublings contained extrachromosomal elements (4/
390 and 2/100 metaphases of P75-20 and P75-45, respec-
tively; Fig. 3B,C). Approximately 98% of metaphases dis-
played intrachromosomal hybridization in clustered
amplicons (Fig. 3D,E), The low incidence of extrachro-
mosomal amplicons was not due to the detection sensi-
tivity since the single copy signal in UA21 cells was
detected in 99% of metaphases (Fig. 3A). None of the
clones analyzed from the second selection contained the
large amplicons detected eight to nine cell doublings
postselection. This probably derives from the small sam-
ple size since identical selection and propagation condi-
tions were utilized in both selections.
Chromosomally amplified sequences were often ob-
served (-20% of the cells in P75-20) at the end of a small
metacentric chromosome (Fig. 3D). Quinacrine (Q)-
banding revealed that this chromosome is a Z2 chromo-
some broken at the approximate position of the DHFR
locus with a cluster of DHFR genes attached to the end
(see chromosomes and nomograms in Fig. 3F). We also
detected cells in P75-20 and P75-45, in which the ampli-
fied sequences localized to a dicentric chromosome sim-
ilar to that shown in Figure Ig and to multiple chromo-
somes within a single cell, some of which were clearly
not related to Z2 (see Fig. 3E). Since other studies have
described chromosomes with amplified DHFR genes
bearing striking similarity to those described here (e.g.,
Biedler 1982; Trask and Hamlin 1989), such structures
must represent the most common types of cytogenetic
products generated by amplification of the CHO DHFR
gene in both hemizygous and diploid CHO cells.
Previous studies have suggested that chromosomal
amplicons might be unstable and capable of generating
copy number heterogeneity (e.g., see Saito et al. 1989).
We tested this possibility by isolating six subclones from
P75-20 (referred to as P75-20S5 to P75-20S10), which
contained the major classes of chromosomally amplified
DHFR genes. The amplified DHFR genes in three sub-
clones were contained in a structure similar to that
shown in Figure 3D, one subclone contained a dicentric
chromosome similar to that shown in Figure Ig, and two
subclones had derivative Z2 chromosomes not yet char-
acterized completely but clearly different from each
other (see Table 1; data not shown). Each subclone was
cultured for 25 cell doublings, and the resulting popula-
tions were analyzed for DHFR copy number heterogene-
ity. For any of the subclones, there was a maximum 1.2-
fold difference in DHFR copy number between cells ex-
pressing the highest and lowest amounts of DHFR
enzyme activity. It is significant that cells containing
amplified genes within a dicentric chromosome did not
generate copy number heterogeneity even though such
chromosomes could generate cytogenetic variants, pre-
164 GENES & DEVELOPMENT
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Gene ampUKcation mediated by chromosome breakage
Figure 3. Localization of DHFR amplicons
to extrachromosomal and chromosomal
structures at 30-35 cell doublings postselec-
tion. Shown are FISH patterns obtained
with DHFR cosmid probe c400-30 and met-
aphase chromosomes from UA21 or Mtx-re-
sistant cell populations (—300 metaphases
per Mtx-resistant cell population: see Mate-
rials and methods). The efficiency of detect-
ing the single DHFR gene in UA21 cells was
99%
in metaphase spreads prepared under
standard conditions. [A] UA21. (B) A cell
from P75-20; (arrow) a large extrachromo-
somal structure containing multiple DHFR
amplicons. (C) A cell from P75-45: (arrows)
extrachromosomal DHFR structures of var-
ious sizes. (D, E,
F]
Metaphases from cells of
P75-20. (D) The DHFR amphcons are at the
end of a broken Z2 chromosome.) DHFR
amplicons localize to both a modified chro-
mosome 1 (arrow) and the end of a broken
Z2 chromosome. (F) The Z2 chromosomes
from UA21 [left] and from a cell of P75-20
[right] that contained amplified DHFR se-
quences at the end of a metacentric chromo-
some as shown in D were stained with
quinacrine to obtain Q bands. The nomo-
gram of each chromosome is shown, along
with the site or region of DHFR hybridiza-
tion on each chromosome.
sumably by bridge-breakage fusion cycles (data not
shown; see Kaufman et al. 1983; McClintock 1984; Ruiz
and Wahl 1990). The minimal difference in copy number
observed in this experiment is within the precision of
the assay and is insignificant by comparison with the
30-fold copy number difference observed in the original
populations propagated under identical conditions.
These results demonstrate that copy number heteroge-
neity and diversity of amplicon localization were gener-
ated by intermediates that predated the formation of the
stable chromosomal amplicons detected at 30-35 cell
doublings.
Cells with deletion of DHFR sequences from the
amplification target can he isolated from each
Mtx-resistant population
Since the small chromosomal amplicons are stable in
this system; it is likely that, as in other examples, the
small extrachromosomal elements prevalent in such
clones at early times engender both copy number and
amplicon location heterogeneity (for references, see
Wahl 1989). The following studies investigate whether
deletion of chromosomal sequences is involved in the
generation of extrachromosomal elements as proposed
previously (Carroll et al. 1988; Hunt et al. 1990; Ruiz and
Wahl 1990). To obtain sufficient material for analysis,
the studies were performed 30-35 cell doublings postse-
lection.
If deletion produced the extrachromosomal elements
shown in Figures 1 and 3, then growth of cells under
nonselective conditions should produce a subpopulation
lacking all DHFR genes as a consequence of the random
loss of acentric amplicons at mitosis. Since very few
cells had solely extrachromosomal elements 30-35 cell
doublings postselection, the frequency of obtaining cells
devoid of DHFR genes after nonselective growth was ex-
pected to be low. Consequently, to make a sensitive as-
GENES & DEVELOPMENT 165
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Windle et al.
sessment of whether cells with DHFR deletion exist at
30-35 cell doublings postselection, we implemented the
following enrichment and detection strategy. Each of the
six independently derived populations with amplified
DHFR genes was grown without selection for 10 cell
doublings to allow for random segregation and loss of
unstable DHFR amplicons. The cells were then stained
with F-Mtx, and cells with little or no DHFR expression
were isolated using the FACS. This procedure enriched
for putative null cells by ~ 50-fold. Individual cells con-
taining DHFR deletions were obtained in either of two
ways.
First, the sorted cells were subdivided further by
single-cell cloning. Alternatively, each sorted subpopu-
lation was exposed to a suicide selection procedure that
kills cells expressing as iew as one DHFR gene through
the DHPi?-dependent conversion of ["^Hluridine to
[^Hjthymidine (see Materials and methods). The sub-
clones isolated using both protocols were then analyzed
by polymerase chain reaction (PCR) and Southern blot-
ting to determine whether the absence of DHFR expres-
sion derived from DHFR gene deletion.
Figure 4A shows a typical PCR analysis of a single-cell
subclone (20A1), isolated from P75-20 by cell sorting
only, which contained no detectable DHFR activity as
measured by F-Mtx binding. Each PCR reaction included
primers specific for the 5' region of the DHFR gene and
for the endogenous thymidine kinase (TK) gene to pro-
vide an internal positive control for reaction efficiency.
This subclone generated no DHPi^-specific PCR product
while substantial PCR amplification of the TK sequence
was obtained. A parallel reaction that employed UA21
cells generated PCR products of the sizes expected to be
generated from the DHFR and TK genes. No DHFR sig-
nal was obtained from the controls that lacked DNA or
from DG44 cells that contain a double deletion of the
DHFR gene (Urlaub et al. 1986).
The DHFR deletion isolation strategy was employed
for each of the six populations containing amplified
DHFR genes, and each proved to contain cells with a
deletion of the DHFR locus. To confirm the PCR results,
the clones shown by PCR to have deleted the DHFR gene
were expanded for analysis by Southern blot hybridiza-
tion. Hybridization with a DHfi?-specific probe pro-
duced intense bands with DNA isolated from UA21
cells,
but no hybridization above background was de-
tected in any of the deletion clones (Fig. 4B).
The FACS-suicide selection procedure was used in
conjunction with PCR and Southern blot analysis to
quantitate the frequency at which cells with DHFR de-
letion could be obtained from the parental UA21 cells
and from each of the six Mtx-resistant cell populations
described above. Table 2 shows that the frequency of
cells with DHFR deletion ranged from 1.3 x 10""* to
1.8 X 10"^. In contrast, no deletions were detected
among 4x10^ UA21 cells grown under the same condi-
tions.
Thus, the frequency of obtaining DHFR deletion
mutants is at least 500- to 10,000-fold higher in cells
known to resist Mtx by DHFR gene amplification than
in the parental UA21 cells. These data demonstrate that
some cells with DHFR gene deletion exist in each of
TK-
DHFR-
*** list ••'•'»
is*
<
APRT
Figure 4. DHFR deletion mutants detected by PCR and South-
ern blot analyses after nonselective growth of Mtx-resistant cell
populations. Cells from Mtx-resistant cell line 75-20 were
grown for 10 cell doublings in medium that does not require
DHFR for survival. Single cells exhibiting the lowest 2% of
fluorescence were obtained by preparative FACS (see Materials
and methods) and expanded into mass culture, and the presence
of DHFR genes was assessed by both PCR and Southern blot-
ting.
[A]
PCR of subclone 20A1 (derived from 75-20 cells). Each
reaction contained primers for the DHFR gene and for the TK
gene as a positive control. Negative controls included a reaction
with primers, but without added cellular DNA, and cellular
DNA from a mutant lacking both copies of the DHFR gene (cell
line DG44).
[B]
Southern blot analysis of DNA isolated from
putative deletion mutants identified by PCR. DNA (—40 |xg)
from putative deletion mutants obtained from each of six inde-
pendently derived Mtx-resistant mutants was digested with
£coRI, fractionated by agarose gel electrophoresis, and Southern
blotted. DHFR sequences (arrows) were detected with a probe
derived from the 3' end of the
DHFR
gene (which contains some
repetitive sequences. The hybridized probe was removed, and
the blot was rehybridized with a probe specific for the adenine
phosphoribosyl transferase [APRT] gene as a control for the
amount of DNA present in each lane.
166 GENES & DEVELOPMENT
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Gene amplification mediated by chromosome breakage
Table 2. DHFR gene deletion frequency in cells with DHFR
gene amplification
Cell
line
DG44
UA21
P75-20
P75-45
P75-69
P75-70
P75-74
PI 50-5
75-20S5
75-20S6
75-20S7
75-20S8
75-20S9
75-20S10
Amplicon
location! s)
(extrachromosomal/
chromosomal)
NA
NA
E/C
E/C
nd
E/C
ND
ND
broken Z2 ter
interstitial Z2
dicentric Z2
broken Z2 ter
broken Z2 ter
interstitial Z2
Deletion
frequency
1.00
<2.5 X 10-^
4.2 X 10"*
1.7 X 10-^
9.7 X 10"^
1.6 X 10-'*
1.3 X 10"^
1.8 X lO"'^
< 10-5
< 10-5
<io-5
< 10-5
< 10-5
< 10-5
The negative selection method developed by Urlaub and Chasin
(1980;
see Materials and Methods) was used to determine the
frequency of cells with DHFR deletion in the indicated cell
lines.
DG44 is a CHO cell line lacking both DHFR genes, which
provides a plating efficiency control. No cells with DHFR dele-
tion were obtained from 4 x 10^ UA21 cells. Cell populations
derived from the expansion of independently isolated single-cell
clones with DHFR gene amplification are designated by the
prefix P, followed by the clone number. Cell lines designated
75-20S5,
etc., are single-cell clones obtained from population
P75-20 and were shown by in situ hybridization to contain chro-
mosomally amplified DHFR genes at the indicated locations.
The deletion frequencies were determined by plating 2.5 x 10'*
to 2.5 x 105 cells per population or subclone. (NA) Not appli-
cable; (E) extrachromosomal elements detected; (C) chromo-
somally amplified regions detected; (ND) not determined.
these six cell populations descended from independently
arising parental cells with DHFR gene amplification.
The detection of cells vyith deletion of the DHFR gene
in a heterogeneous population in v^hich the majority of
cells contained chromosomally localized amplified se-
quences could be explained as a consequence of second-
ary rearrangements within unstable chromosomal struc-
tures.
For example, a dicentric chromosome such as that
shown in Figure Ig could theoretically generate dele-
tions through breakage-fusion-bridge cycles (Kaufman
et al. 1983; McClintock 1984; Ruiz and Wahl 1990). We
therefore determined whether any of the subclones with
chromosomal amplicons produce cells devoid of DHFR
sequences. If a particular chromosomal amplicon can
generate deletions, the frequency of deletion should be
substantially higher in the appropriate subclone than in
the mixed population However, Table 2 shows that no
subclone generated descendants with DHFR deletion un-
der growth conditions that readily gave rise to cells with
deletions in the initial population. These data are con-
sistent with the stability of gene copy number in cells
with chromosomally amplified DHFR genes (see above),
and they indicate that the event giving rise to the dele-
tions must have predated the formation of these stable
secondary amplification products.
Chromosome breakage during gene amplification
Since the deletions observed in each cell line may be the
products of the initial DNA rearrangements involved in
DHFR gene amplification, some of these deletions were
characterized further. We isolated two deletion sub-
clones, 20A1 and 20A2, from P75-20 and identified the
Z2 chromosomes by Q-banding. The Z2 chromosomes in
these two clones are quite different from one another
(Fig. 5A, columns 3 and 5). The Z2 chromosome of 20A1
has a terminal deletion with a break near the DHFR lo-
cus.
In contrast, comparison of the wild-type Q-banding
pattem of chromosome Z2 with that found in 20A2 re-
vealed that the latter has a large interstitial deletion that
removes a band encompassing the DHFR locus (Fig. 5).
Since 20A1 and 20A2 both descended from the same pa-
rental cell that generated clone 75-20, these results dem-
onstrate that at least two strikingly different manifesta-
tions of a DHFR deletion can be derived from a single
initiating event.
The deletions in nine subclones isolated from P75-20
were analyzed by Southern blotting of genomic DNA
using cosmid probes containing sequences that flank the
DHFR gene. This analysis revealed that although all of
the nine subclones lacked the DHFR gene, three con-
tained sequences within 40 kb of the 5' end of the DHFR
gene (encoded by the cosmid c400-13). One of these three
subclones was 20A1. The remaining six subclones, in-
cluding 20A2, contain no sequences that hybridize with
C400-13.
FISH analysis of metaphase chromosomes of 20A1 us-
ing a C400-13 probe revealed that a single copy of these
sequences remains at the terminus of the broken Z2
chromosome (Fig. 5A, column 4; 100/100 metaphases).
The position of the hybridization relative to either the
centromere or the telomere of the other arm is approxi-
mately the same as that of the DHFR gene in the native
Z2 chromosome (see Fig. 5A, column 2; cf. with Fig. le).
This confirms the Q-banding analyses of Figures 3F and
5A in which some Z2 chromosomes in these Mtx-resis-
tant cell populations are broken near the DHFR gene.
The cytogenetic and molecular analyses indicate that
sequences between the DHFR gene and cosmid c400-13
are missing in deletion mutant 20A1. We investigated
whether the deleted region extends into cosmid c400-13
by determining whether any restriction fragments were
missing from c400-13 in clone 20A1. Figure 5B shows
that only 5 of the 12 Xbal fragments detected in UA21
DNA are present in 20A1 DNA. Thus, a substantial frac-
tion of this region has been deleted from the 20 AI ge-
nome, suggesting that sequences in cosmid c400-13 lie
close to or at the end of the broken Z2 chromosome.
Further characterization of the broken chromosome is in
progress.
Although sequences within 40 kb of the DHFR gene
reside near or at the terminus of a broken Z2 chromo-
some in deletion mutant 20A1, Southern analyses re-
GENES & DEVELOPMENT 167
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Windle et al.
Figure 5. Characterization of DHFR deletions by
cytogenetics, FISH, and Southern blot analyses.
[A]
The Z2 chromosome or its derivative was identi-
fied in UA21 (column 1], 20A1 (column 3), and
20A2 (column 5) by Q-banding, and each is shown
with the corresponding nomogram and the posi-
tion of the DHFR gene. Metaphase spreads (>100
analyzed per cell line) prepared from UA21, 20A1,
and 20A2 were then hybridized with cosmid probe
c400-13,
which contains sequences —40 kb up-
stream of the 5' end of the DHFR gene (B. Draper,
B.
Windle and G. Wahl, unpubl.) or cosmid probe
C400-30, which contains the DHFR gene se-
quences (column
2).
(B) DNA (-40
M-g)
from UA21,
20AI, and the 75-20 cell population containing
amplified DHFR genes was digested with Xba\,
fractionated by agarose gel electrophoresis. South-
em-blotted, and hybridized with c400-13. The ar-
rows indicate the bands present in
UA21
and 75-20
but absent from
20A1.
The hybridization signals in
75-20 represent a sixfold amplification of DHFR.
c400-13 contains repetitive sequences that create
the diffuse hybridization smear.
2.3
2.0--
vealed the unexpected result that some of the terminal
sequences identified by cosmid c400-13 are clearly am-
plified in the 75-20 population (Fig. 5B). A mechanism
relating chromosome breakage to both DNA sequence
amplification and deletion is considered in detail below.
Discussion
Amplification mechanisms are not likely to involve
multiple, independent, low-probability recombination
events since cells with amplified genes arise spontane-
ously in tumor cell populations at rates of typically 10 "'^
to 10 ~^ per cell per cell division (Kempe et al. 1976;
Johnston et al. 1983; Capecchi 1989; Tlsty et al. 1989).
Nonetheless, our analyses show that very early events
involve substantial genetic fluidity, since the descen-
dants of each parental cell manifested diversity in the
consequences of amplification at the target site, and the
number and cytogenetic locations of the DHFR ampli-
cons.
A model is presented below that attempts to limit
the number of independent recombination events re-
quired to generate amplicon diversity by proposing that
the initial amplicons are acentric elements generated by
chromosome breakage. According to this model, cytoge-
netic diversity is created by the random segregation and
subsequent integration of such elements and the multi-
ple ways in which the broken chromosomes can be re-
paired.
It is difficult to account for all of the observations con-
cerning amplification of the CHO DHFR gene presented
here and elsewhere (e.g., see Kaufman and Schimke
1981;
Biedler 1982; Trask and Hamlin 1989) by any of
the currently proposed models. For example, unequal sis-
ter chromatid exchange requires multiple independent
recombination events to produce the copy number het-
erogeneity detected in early populations. Since recombi-
nation frequencies are typically 10~'*tol0~* (see Capec-
chi 1989) and sister strand exchanges do not occur at
higher frequencies in chromosomally amplified regions
than in unamplifled DNA (Chasin et al. 1982; Cowell
168 GENES & DEVELOPMENT
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Gene amplification mediated by chromosome breakage
1982),
it is unlikely that this mechanism could generate
the amplicon diversity detected at eight to nine cell dou-
blings. Furthermore, this mechanism cannot readily ac-
count for the generation of both terminal and interstitial
deletions in descendants of the same parental cell. A
second proposed mechanism, DNA rereplication (Mari-
ani and Schimke 1984; Schimke et al. 1986), has not
been observed in mammalian cells (Hahn et al. 1986) and
does not readily account for the frequent deletion of the
target gene or the genesis of either small or huge ampl-
icons in independent events. A third mechanism, recom-
bination within a replication intermediate (see Wahl
1989),
does not directly predict chromosome breakage,
heterogeneous deletions from a single parental cell, or
vast differences in amplicon sizes. Finally, it has been
proposed that the initial amplification intermediates are
very large, intrachromosomal, and highly unstable, and
they decrease in size over time (Saito et al. 1989). This
model requires that multiple, typically rare recombina-
tion events would have to occur every few cell divisions
to account for the observed copy number heterogeneity.
Our data do not address the possibility that some of the
small chromosomal amplicons detected in clones at 30-
35 cell doublings could have arisen by such a process. On
the other hand, it is difficult to envision how this process
could occur at a rate sufficient to generate both small
extrachromosomal elements and small intrachromo-
somal amplicons in each of six clones by eight to nine
cell doublings. We favor the interpretation that the two
differently sized amplicons detected at this early time
represent alternative products of chromosome breakage
(see Fig. 6A-C and below).
A model for gene amplification involving chromosome
breakage within replication intermediates
There is no evidence that incontrovertibly excludes any
of the existing amplification models, but the difficulties
considered above led us to develop an alternative in
which chromosome breakage initiates the amplification
process. Our data agree with previous cytogenetic stud-
ies that led to the suggestion that breakage near the
DHFR locus could have occurred as a very early event
(see Biedler 1982; Biedler et al. 1988). For example, some
cells analyzed eight to nine cell doublings postselection
contained a small metacentric chromosome with DHFR
sequences positioned at or near one terminus. This is
most likely a Z2 chromosome broken in the vicinity of
the DHFR locus since the relative lengths of the two
arms and the site of DHFR hybridization were the same
as in chromosomes frequently detected in later popula-
tions,
which were shown by Q-banding to be broken Z2s.
Importantly, chromosomes containing amplified re-
gions,
including cytogenetically unstable dicentric chro-
mosomes, did not generate similarly broken Z2 chromo-
somes at experimentally measurable frequencies.
Stalled replication intermediates might provide the
substrates for chromosome breakage. This hypothesis is
based on two observations. First, conditions that lead to
the accumulation of replication intermediates, such as
transient inhibition of DNA replication near the time
the target gene is replicated, can markedly increase the
frequency of gene amplification (Tlsty et al. 1982; Brown
et al. 1983) Second, we observed that sequences left at
single-copy level at the terminus of the broken Z2 chro-
mosome were also present in the amplicons of the Mtx-
resistant cell population. This suggests that the amplifi-
cation target contained two copies of such sequences
prior to or coincident with the initiating event and that
the breakage event was asymmetrical (see Fig. 6A-C).
A particular strength of the asymmetric breakage
model is that it provides a simple mechanism for pro-
ducing the amplicons of diverse size and structure. Fig-
ure 6A shows that two breaks introduced on opposite
sides of a replication bubble produce an amplicon con-
taining all DNA sequences between the break site and
the telomere of the same arm. Such large amplicons were
detected here and have been described in CAD gene am-
plification in Syrian hamster cells (Wahl et al. 1983; Giu-
lotto et al. 1986). Asymmetric breakage at three (Fig. 6B)
or four sites (Fig. 6C), coupled with appropriate end liga-
tion reactions, will produce circular amplicons and a bro-
ken chromosome terminated by sequences homologous
to some of those in the amplicon. Circles produced by
four breaks/ligations will contain imperfect inverted re-
peats of the type typically found in mammalian ampli-
cons (Ford and Fried 1986; Saito and Stark 1986; Looney
and Hamlin 1987; Passananti et al. 1987; Hyrien et al.
1988;
Ruiz and Wahl 1988). The sizes of such extrachro-
mosomal amplicons could easily conform to the major-
ity of amplicon sizes detected at the CHO DHFR locus
(Looney et al. 1988). However, since some amplicons are
probably larger than one replicon (Borst et al. 1987;
Looney et al. 1988; Jongsma et al. 1989), their genesis
would require simultaneous or sequential breaks in two
or more adjacent replication intermediates.
All of the breakage patterns described above produce
acentric molecules that contain the target locus. Such
molecules should segregate unequally at mitosis to gen-
erate a population in which they accumulate in some
cells and are lost from others. Unequal segregation pro-
vides an extremely simple mechanism for rapidly gener-
ating the copy number heterogeneity we observed in ev-
ery population with amplified DHFR genes. Since cells
with as few as four DHFR genes could survive the selec-
tive conditions utilized here
(B.
Windie, Y. Yin, and G.M.
Wahl, unpubl.), only two rounds of replication and un-
equal segregation would be required to generate resistant
cells by this mechanism. Larger per cycle increases in
copy number would be achieved if the replication of
acentric circular molecules were deregulated or occurred
by a rolling circle or similar mechanism (e.g., see Stark
and Wahl 1984). Continued propagation, particularly un-
der conditions where elevated gene expression confers a
selective advantage, would lead to the accumulation of
cells with higher levels of amplification. Importantly,
the only recombination events required for gene ampli-
fication to occur by this mechanism are those that gen-
erate the initial break and acentric amplicon. Copy num-
ber increases arise "passively" by unequal segregation.
GENES & DEVELOPMENT 169
Cold Spring Harbor Laboratory Press on July 10, 2011 - Published by genesdev.cshlp.orgDownloaded from
Windle et al.
4=^
Replication
i=©:
A=^
"V^
1
Chromosomal
Breakage
t-^'-r-=-
^='==^
B.
i=& Replication
4=© 'f^
%f
1
Chromosomal
Breakage
4-^-r-=-
+
>
D. i Q :
Exonucleolytic
Degradation
4=^
•=
/ \
i=© i=^ =
1 Repair of Ends
i Q :: < Q I
(20A1) (20A2)
•=© Replication
4=^ A^
1V
Chromosomal
Breakage
t-^-r
< Q :
+
Recombination
4=^=^
Figure 6. Chromosome breakage model for gene amplification. This model proposes that amplicons are acentric chromosome
fragments produced by chromosome breakage. A-C show how asymmetric breakage at two {A), three
[B],
or four sites (C) creates
amplicons of various sizes and configurations. Breakage within a single-replication intermediate is shown for simplicity, but events
comprising adjacent replicons may also occur. The large arrowheads indicate break sites. The placemen