We have previously cloned and characterized two differentdihydrofolate reductase amplicon types from a methotrexate-resistant Chinese hamster ovary cell line ...
MOLECULAR AND CELLULAR BIOLOGY, Dec. 1988, p. 5268-5279 0270-7306/88/125268-12$02.00/0 Copyright © 1988, American Society for Microbiology
Vol. 8, No. 12
The Dihydrofolate Reductase Amplicons in Different Methotrexate-Resistant Chinese Hamster Cell Lines Share at Least a 273-Kilobase Core Sequence, but the Amplicons in Some Cell Lines Are Much Larger and Are Remarkably Uniform in Structure JAMES E. LOONEY,1t CHI MA,1 TZENG-HORNG LEU,1 WAYNE F. FLINTOFF,2 W. BRAD TROUTMAN,' AND JOYCE L. HAMLIN'* Department of Biochemistry and Cell and Molecular Biology Program, University of Virginia School of Medicine, Charlottesville, Virginia 22908,1 and Department of Microbiology and Immunology, The University of Western Ontario
Health Science Center, London, Ontario, Canada N6A 5C12 Received 8 July 1988/Accepted 23 August 1988 We have previously cloned and characterized two different dihydrofolate reductase amplicon types from a methotrexate-resistant Chinese hamster ovary cell line (CHOC 400). The largest of these (the type I amplicon) is 273 kilobases (kb) in length. In the present study, we utilized clones from the type I amplicon as probes to analyze the size and variability of the amplified DNA sequences in five other independently isolated methotrexate-resistant Chinese hamster cell lines. Our data indicated that the predominant amplicon types in all but one of these cell lines are larger than the 273-kb type I sequence. In-gel renaturation experiments as well as hybridization analysis of large SfI fragments separated by pulse-field gradient gel electrophoresis showed that two highly resistant cell lines (A3 and MK42) have amplified very homogeneous core sequences that are estimated to be at least 583 and 653 kb in length, respectively. Thus, the sizes of the major amplicon types can be different in different drug-resistant Chinese hamster cell lines. However, there appears to be less heterogeneity in size and sequence arrangement within a given methotrexate-resistant Chinese hamster cell line than has been reported for several other examples of DNA sequence amplification in mammalian systems.
The phenomenon of DNA sequence amplification has been observed in a variety of drug-resistant mammalian cell lines, as well as in human neoplasms (see references 19, 37, and 39 for reviews). The units of amplification (amplicons) are invariably much larger than the gene that is thought to supply the selective advantage in each case (e.g., dihydrofolate reductase gene [DHFR] [30, 32], CAD [16, 42], oncogenes [1, 25, 38]). This raises the question why such large amounts of flanking passenger sequence are initially included in the amplicons and retained during amplification to high copy number. It is unlikely that the coamplification of a distant colinear gene is required, since a cloned selectable marker such as CAD or DHFR can be amplified at a new chromosomal location when introduced into a suitable recipient cell line by transfection (6, 29). The two general models that have been proposed to explain the mechanism of DNA sequence amplification could both account for the large size of amplicons (19, 37, 39). In the rereplication model, it is suggested that multiple initiations occur at an origin of DNA synthesis in a single S period, and that by subsequent recombination events, the resulting extra daughter duplexes become arrayed end-toend either in the body of the chromosome or as extrachromosomal, acentromeric double minutes. According to this model, replication forks would usually have to travel beyond the boundaries of one replicon during the amplification event, since the amplicons in mammalian cell lines are in the range of 240 to 10,000 kilobases (kb) in length (16, 26), whereas the largest replicons are thought to be no more than
a few hundred kilobases in length (20). However, the existence of more than one replicon per amplified unit has not been demonstrated in any system.
In the unequal sister chromatid exchange model, it is proposed that homologous elements flanking the gene in question mediate a staggered pairing event that leads to unequal recombination and the consequent reciprocal loss and gain of the intervening DNA sequence by the two respective chromatids. These recombinogenic elements would have to be situated at great distances from one another to explain the large size of most amplicons and could conceivably relate to some structural feature of higher-order chromatin folding. Another property that has been considered a hallmark of the amplicons in drug-resistant mammalian cells is the heterogeneity in amplicon size and structure within a single cell line and among cell lines of the same species that have amplified the same gene. This notion derives largely from studies on the amplified DHFR gene in methotrexate (MTX)resistant murine cell lines (4, 7) and on the CAD amplicons in N-(phosphonacetyl)-L-aspartate (PALA)-resistant Syrian hamster cells (2, 16, 43). In each of these systems, many different interamplicon junction fragments were identified in a single drug-resistant cell line, indicating variable-sized amplicons. When cloned fragments from these amplicons were utilized as hybridization probes on digests of DNA from other cell lines of the same species that were resistant to the same drug, additional complexities in amplicon structure were detected (2, 4, 7, 43). Based on these and other studies (e.g., references 5 and 40; see references 19, 37, and 39 for reviews), the following generalizations have been made. (i) Amplicons appear to be extremely variable in length within and among cell lines that are resistant to the
* Corresponding author. t Present address: Howard Hughes Medical Institute, The Johns Hopkins University, Baltimore, MD 21205.
5268
VOL. 8, 1988
same drug and can terminate quite close to the selected gene. (ii) There do not appear to be preferred endpoints (hotspots) for the joining together of amplified sequences in most systems. (iii) Initial amplified units may be quite large, but become trimmed to smaller units during the course of amplification to high copy number. (iv) Amplicon structure can change during propagation in some drug-resistant cell lines, even when cells are maintained in drug concentrations that do not require further amplification of the selected gene. (v) Sequences from distant, unrelated loci may even become joined to the amplicons and be coamplified with the selected
gene. However, there are exceptions to these generalizations. For example, Tyler-Smith and Bostock (40) have presented evidence that once a particular DHFR amplicon type is established in a MTX-resistant murine cell line, it appears to be amplified to higher copy numbers without significant further rearrangement, even though the predominant amplicon types in individual drug-resistant murine cell lines can have different structures. In addition, Hyrien et al. (23) have shown convincingly that there can be hotspots for the joining together of adenylate deaminase amplicons in coformycinresistant Chinese hamster ovary cells. In our own laboratory, we have established an MTXresistant CHO cell line (CHOC 400) that contains -1,000 copies of the DHFR gene per cell. We originally estimated that the average DHFR amplicons in CHOC 400 were at least 135 kb in length, but could not be greater than -300 kb based on the size of the involved chromosomal regions (30). When visualized in ethidium bromide-stained agarose gels, the patterns of amplified restriction fragments were quite similar among CHOC 400 and three other MTX-resistant Chinese hamster cell lines (MK42, MQ19, and A3) (3, 31, 32). In addition, we showed in hybridization studies that an identical core sequence of at least 150 kb was uniformly amplified in CHOC 400, MK42, and A3 cells (31). Thus, by these criteria, the DHFR amplicons in highly MTX-resistant Chinese hamster cell lines appeared to be relatively small and uniform in structure compared with those in most other well-studied examples of drug resistance in mammalian cell lines. We also presented evidence that there was a single replication initiation locus in the amplified DHFR domain in CHOC 400 cells (21, 22). On the basis of these observations, we suggested several years ago that the initial unit of DHFR gene amplification in Chinese hamster cells was equivalent to a parental chromosomal replicon that was overreplicated multiple times during the MTX selection process (19, 31). We have recently succeeded in isolating overlapping recombinant cosmids that represent the equivalent of two complete DHFR amplicon types from the CHOC 400 cell line (26). In the present study, we used these recombinant clones to analyze the genomes of five other independently isolated MTX-resistant Chinese hamster cell lines to determine whether the DHFR amplicons are similar in size and relative homogeneity to those that have been characterized in CHOC 400. These cell lines vary in DHFR gene copy number from -5 to more than 700. Our results indicate that the amplicons in four of five of these variants are greater in length than the prevalent types in the CHOC 400 genome. In fact, in two highly resistant cell lines, the core amplicons are at least 453 kb (and probably greater than 653 kb) in length and are quite uniform in structure within each cell line.
MATERIALS AND METHODS Cell lines and cell culture. The MTX-resistant cell line CHOC 400 was derived in our laboratory from CHO cells by
DIHYDROFOLATE REDUCTASE AMPLICONS
5269
stepwise increases in drug concentration to a final concentration of 800 ,uM; CHOC 400 contains -1,000 copies of the DHFR gene per diploid nucleus (29). MK42 was developed in the laboratory of Lawrence Chasin (Columbia University) from CHO cells (32), and the A3 cell line was developed by Biedler and Spengler (3) from DC3F, a Chinese hamster lung fibroblast. When we initially received MK42 and A3, each contained approximately 150 copies of the DHFR gene. After further increases in MTX in our laboratory, they are now resistant to 800 ,uM and 8 mM MTX and contain 300 and 700 DHFR gene copies, respectively. All these cell lines were grown in minimal essential medium supplemented with nonessential amino acids, 10% donor calf serum, and the appropriate amount of MTX. Cultures were maintained in 8% CO2 in a humidified chamber. The drug-resistant cell lines RI, RIII-1, RIII-2, RIV, and RV were derived from CHO Toronto as previously described (9-12, 17). The RI cell line was selected for resistance to 10-7 M MTX and encodes a dihydrofolate reductase enzyme with a lowered affinity for MTX; the independently isolated RIII-1 and RIII-2 derivatives of RI were selected in a single step in 10-6 M MTX and contain -17 and 5 copies of the altered DHFR gene, respectively. RIV was derived from CHO Toronto in a single selection step at 5 x 108 M MTX and contains approximately four copies of the wildtype DHFR gene; RV was then cloned from RIV in 10-6 M MTX and contains -60 copies of the gene. All tissue culture media and sera were obtained from GIBCO Laboratories (Grand Island, N.Y.), and MTX was kindly supplied by the National Cancer Institute Drug Development Branch. DNA preparation, restriction digestion, and Southern blotting. For analysis of small restriction fragments less than 20 kb in length, genomic DNA was purified by standard procedures (15) and was incubated with the appropriate restriction enzyme in the buffer recommended by the supplier (Bethesda Research Laboratories, Inc., Gaithersburg, Md.). For the parental CHO Toronto and moderately resistant RI, RIII-1, RIII-2, RIV, and RV cell lines, 20 ,ug of each digest was separated on a 0.8% agarose gel and the digests were then transferred to GeneScreen (Dupont, NEN Research Products, Boston, Mass.) and hybridized with appropriate probes as previously described (31). To eliminate background problems from the repetitive sequence elements present in cosmid probes, blots were prehybridized for 6 h prior to probe addition with 20 ,ug of CHO DNA per ml that was hybridized to a Cot of 500 in 0.6 M NaCl at 68°. For the hybridization experiments comparing CHO DNA with DNA from the highly resistant cell lines CHOC 400, MK42, and A3 (see Fig. 3 and 4), 5 ,ug of parental DNA digest and 1 ,ug from the resistant variants were separated on agarose gels and transferred to GeneScreen Plus (Dupont, NEN Research Products) as previously described (27). Hybridization probes (either whole cosmids or fragments separated on low-melting-point agarose) were labeled with [32P]dCTP (Dupont, NEN Research Products) either by nick translation (34) or by the random primer method (8). Separation of genomic DNA by pulse-field gradient gel electrophoresis was performed essentially as outlined in reference 14. Approximately 105 cells were embedded in a 20-,ul low-melting-point agarose plug (0.5%) formed in 3/32-in. (0.24-cm) Tygon tubing. Each plug (-0.5 cm in length) was removed from the tubing and processed and digested with SfiI as described previously (14). The plugs were then inserted into the wells of a 0.8% agarose gel (7.5 by 8.5 by 0.6 cm) formed in a GeneLine pulsed-field gradient
5270
s~ _
LOONEY ET AL.
MOL. CELL. BIOL.
S5
Si
*t
*dhfr
0
ISO
S3
~>
80 Si
90
40
120
160
so
S2 S3
200
S4
ori\'
4
280
240
180
13
320
360
80
13
Si
S2 S3
.
_
i
400
440
480
CHO
KB
'I
ISO
13
S2 S3
_
Type 1
13
so
Si
S2 S3
_
_.
Type I
Type I- 273 KB
80
15
S S332
S2 S3
1
160 KB
13
> Type I
280
13
80
200
§
80
13
280
13
S2SS3
IX ---Type I
, Type X- 240 KB
FIG. 1. Parental DHFR locus and type I and II amplicons from the CHOC 400 genome. (A) Five SfiI sites (Si to S5) define 453 kb of DNA sequence in the DHFR locus (27). The black box on the linear scale represents the DHFR gene, and the hatched box defines the rough location of the replication initiation locus (ori) downstream from the gene. (B) The 273-kb sequence bracketed in panel A was amplified, and the multiple copies became arranged in head-to-tail arrays to form the type I amplicons early in the course of the development of drug resistance (27). Note that the type I amplicon begins in the parental 200-kb Sfil fragment and ends in the parental 160-kb Sfil fragments (panel A) and that these two fragments became joined at the asterisk to form the 180-kb type I junction fragment. (C) The 240-kb type II sequence bracketed in panel B was amplified later in the process, and the multiple copies became organized into palindromic arrays, forming two new junctions (the head-to-head and tail-to-tail joints indicated with asterisks). A 33-kb region (dashed line, panel B) was deleted during this rearrangement. The type I junction is indicated with a squiggle line and an asterisk. Note that the formation of the type II amplicon, which represents -75% of all amplicons in the CHOC 400 genome), results in two new palindromic fragments 280 and 15 kb in length.
electrophoresis apparatus (Beckman Instruments, Inc., Fullerton, Calif.). Gels were run in 10 mM Tris acetate-0.5 mM EDTA (pH 8.2) at 15°C and were pulsed for either 15 or 30 s at 150 mA (see legends to Fig. 6 and 7). The digests were transferred to GeneScreen Plus and hybridized with appropriate probes as outlined above. The isolation of the cosmids utilized in this study was described in three previous studies (26, 29, 31). In-gel redaturation. For the experiment presented in Fig. 5, the in-gel renaturation procedure was performed exactly as described by Roninson (35). A small portion of an EcoRI digest of each genomic DNA sample was end labeled with [32P]dCTP utilizing T4 polymerase (35) and was mixed with 10 ,ug (CHO, MK42, A3) or 6 ,g (CHOC 400) of nonradioactive digest. The samples were separated on 1.0% agarose gels prior to two cycles of denaturation, renaturation, and Si nuclease treatments (35). The gels were then dried down on a glass plate at 55°C overnight and exposed to Kodak X-Omat X-ray film. For SfiI digests, DNA was labeled intrinsically by incubating cells for one cell cycle with 1 ,uCi of [3H]thymidine per ml. Approximately 1.5 x 105 cells were then embedded in an agarose plug, and the samples were processed and digested with SfiI as described above. After electrophoresis, the gel was subjected to one cycle of the modified in-gel renaturation procedure described in reference 27, and the SfiI digests were transferred to GeneScreen Plus by alkaline blotting (33). The transfer was sprayed with En3Hance (Dupont, NEN Research Products) prior to autoradiography.
RESULTS
Organization of the DHFR domain in parental CHO cells and in the MTX-resistant cell line CHOC 400. The type I amplicon arose from the parental CHO locus (or from some larger precursor amplicon) by a relatively simple rearrangement in which multiple uniform copies of a 273-kb sequence ended up in head-to-tail arrays in the genome (26, 27) (Fig. 1A and B). We have shown previously that, except for the interamplicon junctions themselves, the type I amplicon represents a faithful version of the corresponding 273-kb sequence in the parental DHFR locus (26, 27). The 240-kb type II amplicon represents a truncated version of the type I sequence that has suffered a 33-kb deletion in a region upstream from the DHFR gene (dashed line, Fig. 1B) (26). The endpoints of the type II amplicon are positioned within the body of the type I sequence. The multiple uniform copies of the type II amplicon are arranged in head-to-head and tail-to-tail arrays in the genome to form giant palindromes (Fig. 1C). The type II sequence represents -75% of all the DHFR amplicons in CHOC 400 cells, whereas the type I sequence accounts for less than 5% (26). The type II amplicon must have arisen from the type I sequence, since all of the type II amplicons include the type I junction fragment (26). Both amplicons contain the replication initiation locus that we have previously identified in the DHFR domain (indicated by the ori box in Fig. 1 and 2A) (21, 22). There are five or six other minor sequence arrangements that account for the remaining 20 to 25% of amplicons
DIHYDROFOLATE REDUCTASE AMPLICONS
VOL. 8, 1988
A
280B
80
15
1808(200)
80
_o--N 0 e
20
40
i
i
60 e
100
80
-%
- -_
_
s DHFR,
sp-
ORI 200
16KB-a
Bc B:c> u cc
1
+
-N
>
rr_
: er
c N07
cN27
cPA36
_= X
- -|
c26A31 R P ~~~~~~c32AI' ~
F
~~
I
GM2
I----
4
&
-
O
-
u Xw>a wrF XW= -
r
>,
a:wcwr=_ -
_
°
> >,
U
W
u
260273 KB
240
C
cS2J
1 r
lb
_
= = =_ r wo cJ > ccX rr o:_-
O
rI
TYPE I
c7-6
cD23 L4
.
180A(160)
~S3
220
b
-TYPE
- --
,
13
14& ljV 180
120
t
i
,
280A
13
_-~ fv-
1
5271
_
>
C4 -
urar:rr a
cPA36 --t
cH-45
_L'O
>
_
- N
°>
w ir v X oQr
r
2=
XXFr
-N-t r
*
a
!w
0;
f
0
*~. a
a'; 1 0
0
IV 11 0
cPA36
cN27
6
14
No7
cH2
c26A31
c7-6
cS21
cl1-45
CHO
RIV (4)
(ALT)
RIII-1
(17)
RIIh-2 RV (5)
(60)
FIG. 2. Two of three moderately resistant CHO cell lines amplify at least the 273-kb type I sequence. (A) Map of the type I and type II amplicons showing the position of the DHFR gene, the replication initiation locus (ori), and the SfiI fragments that characterize these amplicons. The locations of probes utilized in experiments shown in Fig. 6 are shown just below the linear scale, as is the 16-kb EcoRI fragment containing the type I junction. The cosmids used in this study are indicated below. Note that cPA36 contains the type I junction (indicated by broken line); the type II amplicon begins at the head in cHDZ23, crosses the type I junction in cPA36, and ends at the palindromic sequence in cNQ7 (Fig. 1C). Note also that the 80- and 13-kb fragments are shared by both amplicon types but that the 180-kb fragment contains the type I joint and the 280- and 15-kb fragments contain the type II joints. (B) HindIl digests of genomic DNA from five moderately resistant CHO cell lines and the wild-type parent were probed with the indicated cosmids, as outlined in Materials and Methods. The genesis of these cell lines is shown below the panel, along with the estimated number of copies of the DHFR gene per diploid nucleus. The RI cell line and its derivatives encode an altered dihydrofolate reductase enzyme (see text). Note that the digests from the CHO and Rl cell lines were exposed to X-ray film for a longer time than the other drug-resistant cell lines were to detect single-copy signals. Asterisks are defined in the text.
in the CHOC 400 genome (26, 27), but these were not characterized in detail. The entire 273-kb type I sequence is amplified in four other MTX-resistant Chinese hamster cell lines. To assess the extent of heterogeneity among the DHFR amplicons in independently isolated MTX-resistant Chinese hamster cell lines, we utilized a series of cosmids (Fig. 2A) representing the 273-kb type I amplicon as hybridization probes on digests of genomic DNA from five other MTX-resistant cell lines. These cell lines range in DHFR gene copy number from -5 to more than 700 per cell. The moderately resistant RIII-1, RIII-2, and RV cell lines
(Fig. 2B) were derived from CHO cells (which are homozyfor the DHFR locus [32]) and contain 5 to 60 copies of the DHFR gene per diploid nucleus (9, 11, 12). The RI cell line and its derivatives, RIII-1 and RIII-2, contain an altered DHFR gene that encodes an enzyme with a lowered affinity for MTX (10, 17). RIII-1 and RIII-2 amplified the same spectrum of HindIll fragments as parental CHO cells in the 273-kb region represented by the eight different cosmids (Fig. 2B). A similar result was obtained when EcoRI digests were screened with the same cosmids (W. Flintoff, unpublished observations). If either of these cell lines contained amplicons that were smaller than the 273-kb type I sequence, gous
5272
LOONEY ET AL.
variant fragments that contain junctions would have been observed (i.e., if a sequence is duplicated and then joined end-to-end, two parental-sized restriction fragments will be united at the joint to form a new restriction fragment of a different size). We conclude that the major amplicon types in the RIII-1 and RIII-2 cell lines are at least as large as the 273-kb type I amplicon cloned from CHOC 400 and represent nonrearranged versions of the parental sequence. In the RIV cell line and its more resistant derivative, RV, all the cosmids also hybridized to corresponding amplified sequences in the genome. However, with cN27, some fragments appeared to be relatively underrepresented (asterisks, Fig. 2B). Furthermore, a variant fragment that was not present in other cell lines or in the parental CHO genome was detected with the cosmid cPA36 (asterisk, Fig. 2B). While we did not map these rearrangements in detail, the data showed that some amplicons in the RIV and RV cell lines are larger than the 273-kb type I sequence (i.e., contain parental-sized fragments in the region of the type I junction represented by cosmid cPA36), while others appear to terminate at one end in the neighborhood of cosmids cPA36 and cN27. The size of these amplicons will depend on how far they extend beyond the 3' boundary of the type I sequence. The highly resistant MK42 cell line was developed from CHO cells by Chasin and colleagues (32), and the A3 cell line was derived from the Chinese hamster lung fibroblast, DC3F, by Biedler and Spengler (3). In both cell lines, the DHFR amplicons are located in chromosomal homogeneously staining regions (3, 32). DC3F is heterozygous at the DHFR locus, and many restriction site polymorphisms exist between the two alleles (28, 31). The A3 cell line was chosen for the present study because it has amplified the allele found in the homozygous CHO cell line and its derivatives, MK42 and CHOC 400 (31). MK42, A3, and CHOC 400 contain -300, 700, and 1,000 DHFR genes per diploid nucleus, respectively. In hybridization studies with the cosmid series shown in Fig. 2A, we have previously demonstrated that the sequence mapping between the ends of the 273-kb type I amplicon is uniformly amplified in both MK42 and A3; the patterns of hybridization in either EcoRI or HindIII digests were identical between parental CHO cells and these two MTXresistant cell lines (31; B. Troutman, unpublished observations). Thus, the amplicons in MK42 and A3 appear to be at least as large as the type I sequence by this criterion. However, the data presented in Fig. 3 show that if differences exist in amplicon sequence arrangement between cell lines, these differences can be detected by using cosmids as hybridization probes. For example, the cosmid cPA36, which contains the type I junction, detected the prominent 16-kb junction-containing fragment in CHOC 400 DNA, but hybridized instead to 9- and 10-kb fragments in MK42 and A3 (Fig. 3A), both of which fragments are parental sized (27). In addition, the cosmids c26A31 and c32A1, which cross the head-to-head and tail-to-tail junctions in the type II amplicon, respectively, detected multiple differences in the patterns of hybridization between CHOC 400 and the other two drug-resistant cell lines (see map in Fig. 2A and Fig. 3B and C). In particular, with both c26A31 and c32A1, some bands appeared to be underrepresented in the CHOC 400 genome relative to other fragments in the region (Fig. 3B and C, arrowheads); these fragments corresponded to the sequences in the type I (and larger) amplicons that were deleted in the formation of the type II structure (Fig. 1A and B). Also, both probes detected fragments unique to the
MOL. CELL. BIOL. 0 A °q E
