The Dihydrofolate Reductase Origin of Replication Does Not Contain Any Nonredundant Genetic Elements Required for Origin Activity

The Chinese hamster dihydrofolate reductase (DHFR) origin ofreplication consists of a broad zone of potential initiationsites scattered throughout a 55-kb intergenic spacer, with atleast three sites being preferred (ori-ß, ori-ß',and ori- ${gamma}$ ). We previously showed that deletion of the most activesite or region (ori-ß) has no demonstrable effecton initiation in the remainder of the intergenic spacer noron the time of replication of the DHFR locus as a whole. Inthe present study, we have now deleted ori-ß', bothori-ß and ori-ß', an 11-kb region just downstreamfrom the DHFR gene, or the central $~$ 40-kb core of the spacer.The latter two deletions together encompass >95% of the initiationsites that are normally used in this locus. Two-dimensionalgel analysis shows that initiation still occurs in the earlyS phase in the remainder of the intergenic spacer in each ofthese deletion variants. Even removal of the 40-kb core failsto elicit a significant effect on the time of replication ofthe DHFR locus in the S period; indeed, in the truncated spacerthat remains, the efficiency of initiation actually appearsto increase relative to the corresponding region in the wild-typelocus. Thus, if replicators control the positions of nascentstrand start sites in this complex origin, either (i) theremust be a very large number of redundant elements in the spacer,each of which regulates initiation only in its immediate environment,or (ii) they must lie outside the central core in which thevast majority of nascent strand starts occur.

INTRODUCTION

Top

Introduction

In the chromosomes of bacteria, bacterial plasmids, viruses,and lower eukaryotes such as Saccharomyces cerevisiae, replicationinitiates at genetically defined sites known as replicators,which serve as recognition signals for cognate sequence-specificDNA binding protein complexes termed initiators (34, 42). Inthe chromosomes of metazoans, however, many replication originsconsist of broad zones of closely spaced initiation sites, someof which are used with greater efficiency than others (for reviews,see references 14, 19, and 30).

An example of the latter is the early-firing Chinese hamsterdihydrofolate reductase (DHFR) origin, which resides in the55-kb spacer between the convergently transcribed DHFR and 2BE2121genes (Fig. 1A) (17, 18, 22, 58). Neutral/neutral (7) and neutral/alkaline(49) two-dimensional (2-D) gel studies on several adjacent andoverlapping restriction fragments have demonstrated that thereare at least 20 nascent strand start sites (and probably manymore) scattered throughout the 55-kb spacer (Fig. 1) (23). However,more-quantitative assays showed that the central 40-kb core,and particularly two subregions within it known as ori-ßand ori- ${gamma}$ , are preferred (Fig. 1) (23, 43, 44). The results ofan in-gel renaturation analysis of early labeled restrictionfragments are shown in Fig. 1A (44), as is the frequency ofinitiation measured in a quantitative early labeled fragmenthybridization (ELFH) assay that used very small origin-containingnascent DNA as a probe (23). The results of a PCR-based smallnascent strand abundance assay that focused on a 12-kb subregionof the spacer encompassing ori-ß, which identifiedan additional, somewhat less active, site or region immediatelydownstream (termed ori-ß') (41) are also shown inFig. 1A. Interestingly, the one silent locus detected by thehigh-resolution ELFH assay lies in the approximate center ofthe spacer just upstream from a prominent matrix attachmentregion (Fig. 1A) (16) and coincides remarkably well with theleast-active region as measured in the lower resolution in-gelrenaturation study (44) and by neutral/neutral 2-D gel analysis(23).

View larger version (24K):
[in this window]
[in a new window]
FIG. 1. Engineering deletions in the DHFR initiation zone. (A) Map of a 190-kb region encompassing the DHFR locus, showing the positions of the DHFR, 2BE2121 (26), and MSH3 genes and the directions of transcription. The intergenic region constitutes the DHFR origin, which contains a centered matrix attachment region (M), indicated by the box (16). The gray curve is a tracing of the data obtained in a PCR-based small nascent strand abundance analysis of the 12-kb sequence encompassing the ori-ß and ori-ß' regions (41), while the open curves traced with solid and dashed lines represent the distribution of initiation sites as measured by high-resolution ELFH (23) and in-gel renaturation experiments (44), respectively. Below are shown EcoRI, Asp718, and (incomplete) HindIII and XhoI maps of the intergenic spacer, along with probes used in the Southern and 2-D gel analyses (see text). The vertical marks on the linear axis correspond to initiation sites identified by 2-D gel analysis (23). (B) The DHFR-deficient variant, DR-8A7, which has only one copy of the DHFR locus and has suffered a radon-induced 14-kb deletion encompassing the 3' end of the DHFR gene (35). The resulting deletion junction fragments are indicated above and below the map (compare with Fig. 2). (C) Cartoons representing the constructs that were used to introduce deletions into the DHFR locus. Homologous recombination at sites a and b results in restoration of the wild-type locus (as shown in panel D), while recombination at sites a and c or d (depending on the donor) restores the gene and at the same time deletes the indicated sequence from the endogenous locus (indicated in panel E). (D) Map and diagnostic wild-type fragments regenerated by recombination at sites a and b. (E) Maps of KO cell lines generated by recombination between sites a and either c or d (depending on the donor), showing diagnostic fragments detected in the Southern analyses pictured in Fig. 2 (see text).

The seeming complexity of the DHFR origin raises the questionof whether its activity is controlled by classic replicator-initiatorinteractions. If so, the proposed replicators could serve asrecognition sites for proteins such as the origin-recognitioncomplex (6), which was first identified in S. cerevisiae andwhich facilitates loading of other initiation proteins suchas CDC6, CDC45, cdt1, and the minichromosome maintenance proteins(reviewed in reference 5). Homologues of these proteins havenow been identified in several higher eukaryotes (5, 28), includingChinese hamsters (2; M. Alexandrow and J. L. Hamlin, unpublisheddata). In the latter system, at least minichromosome maintenanceproteins 2 and 5 generally localize to the intergenic regionin the early S period and then appear to translocate to flankingregions as cells move through the S period (2), mimicking thebehavior of their counterparts in S. cerevisiae (4).

In one extreme model that corresponds most closely to the paradigmestablished for simpler systems, initiation complexes couldload at one or a small number of master replicators residingwithin the DHFR origin itself. The proposed replicator(s) wouldpresumably correspond to one or more of the most active initiationsites or regions in the spacer (i.e., ori-ß, ori-ß',and/or ori- ${gamma}$ ). After loading, the initiation complexes couldtranslocate randomly up or down the template via a helicaseactivity before priming nascent strands, the latter activitydisplaying some preference for particular sequences in the template.This mechanism could lead to the dispersive mode of nascentstrand initiation suggested by the 2-D gel approaches, but itcould also explain why the regions around ori-ß, ori-ß',and/or ori- ${gamma}$ (the suggested loading sites) are preferred. Inan alternative multireplicator model, ori-ß, ori-ß',and ori- ${gamma}$ would represent the most active subset of at least20 redundant replicators distributed throughout the spacer thatattract initiation proteins with different efficiencies andwhich control initiation only in their immediate environments.Clearly, variations on these scenarios can be imagined. Withany model, the efficiencies of individual replicators couldbe modulated by additional factors such as chromatin architectureand disposition relative to genes, matrix attachment regions,and boundary elements, etc.

In a previous study (36), a reach-out-and-knockout (ROKO) homologousrecombination strategy was utilized to delete ori-ß,which was considered at the time to be the predominant initiationsite or region in the spacer and therefore might serve as amaster replicator (10, 12, 13, 52). However, deletion of ori-ßhad no detectable effect on initiation in the remainder of thespacer nor on the time of replication of the DHFR locus in theS period (36). Thus, we concluded that, while ori-ßis the most efficient initiation site in the spacer, it doesnot constitute a master, nonredundant replicator for the DHFRlocus. This finding raised the possibility that other preferredsites (e.g., ori-ß' or ori- ${gamma}$ ) might correspond to amaster replicator. Alternatively, ori-ß' and/or ori- ${gamma}$ might represent a redundant replicator that assumes the roleof master when ori-ß is deleted by relief of origininterference, a phenomenon that occurs between neighboring autonomouslyreplicating sequence (ARS) elements in S. cerevisiae (9).

In the present study, we have used the ROKO approach to deletevarious parts of the intergenic spacer to determine whetherany critical, nonredundant, cis-regulatory elements lie withinthe spacer itself. We find that even a large deletion encompassingthe central 40-kb core fails to inhibit initiation in the remainderof the spacer or dramatically change the time of replicationof the locus in the S period. Remarkably, this deletion encompasses>90% of the sites that support initiation in the native locus,including ori-ß, ori-ß', and ori- ${gamma}$ . Thus,if classic replicators control the positions of start sitesin this complex origin, they either control initiation onlyin their immediate environments or lie outside the central corein which the great majority of nascent strand starts occur.Our data further suggest that the efficiency of initiation atindividual sites may be influenced by the length of the spacer,which, by definition, is controlled by local transcriptionalactivity.

(This work represents a part of the requirements for the Ph.D.degree for X. Li from the Department of Biochemistry and MolecularGenetics, University of Virginia School of Medicine, Charlottesville,Va.)

MATERIALS AND METHODS

Top

Materials and Methods

Preparation of donor constructs for the ROKO strategy. BAC-KZ381, BAC-KP454, and pWe15-KZ381 were constructed by transferringthe NotI-resected insert from pWe16-KZ381 or pWe16-KP454 (46)into the respective vector. BAC-KZ/KP was engineered from thecorresponding bacterial artificial chromosomes (BACs), resultingin an insert extending from map position (mp) 10- to 85-kb (Fig.1A) (L. D. Mesner, unpublished data). The ßß'knockout (KO) donor is missing the 12-kb sequence extendingfrom mp 36 to 48, and it was derived by recircularizing an XhoIpartial digest of BAC-KZ/KP. The intergenic region KO (IRKO)donor was constructed by deleting a 12-kb XhoI fragment (Fig.1A) from pWe15-KZ381 (mp 36 to 48) and replacing it with a 6-kbKpnI-NotI fragment (mp 78 to 84) from KP454, resulting in adeletion extending from mp 36 to 77. The KZ381 ß'KO donor is missing $~$ 4 kb extending from mp 43 to 47 and wasconstructed essentially as described previously (36), exceptthat this region in the subcloned 12-kb XhoI fragment (Fig.1A) was replaced (via a HindIII complete digestion) with a 4-kbneomycin resistance (Neo^r) gene flanked by loxP sites. By introducingthe cre recombinase in trans, the Neo^r marker was subsequentlyremoved (36). The DHFR 3'-KO donor was constructed by circularizinga 37-kb BamHI fragment obtained by partial digestion of BAC-KZ381,resulting in an $~$ 12-kb deletion extending between mp 24 and 36(Fig. 1A). (All clones and maps are available upon request.)

Cell culture and synchronization protocols. The DR-8A7 cell line is a DHFR-deficient variant of the hemizygousCHO cell line UA21 (57) and has a deletion of the 14-kb regionextending downstream from approximately the center of the lastintron of the DHFR gene into the intergenic zone (mp 20 to 34)(Fig. 1A) (35). DR-8A7 cells were propagated on minimal essentialmedium (MEM) supplemented with nonessential amino acids, 100µM hypoxanthine, 16 µM thymidine (57), and 10% fetalclone II (HyClone) and were maintained in an atmosphere of 5%CO₂. Homologous recombinants that restored the missing partof the gene were propagated on MEM containing nonessential aminoacids and 10% fetal clone II. To prepare synchronized populations,cells were arrested in G₀ by starving them for isoleucine for36 h and they were released into complete MEM containing 200µM mimosine (Sigma) for 12 h to collect them at the G₁/Sboundary (17, 48). The drug was washed out and replaced withdrug-free medium to allow entry into the S phase.

Transfection, selection, and characterization of potential homologous recombinants. NotI-digested donor constructs were delivered to the DHFR-deficientDR-8A7 cell line by electroporation as previously described(36). After transfection, cells were seeded at a density of $~$ 2 x 10⁶ per 10-cm-diameter plate and were maintained in F12medium containing 10% fetal clone I for 36 to 48 h. The mediumwas then changed to F12 lacking thymidine, hypoxanthine, andglycine, and surviving (DHFR⁺) colonies were isolated $~$ 10 dayslater. Genomic DNA samples were purified by standard procedures(31), digested with the relevant restriction enzymes (see figurelegends), and separated on 0.6% agarose gels for Southern blottingonto HybondN+ (Amersham) and hybridization with the appropriateprobes (see figure legends) as previously described (21). Theprobes utilized in this study were as follows: probe 100, a1.2-kb XbaI/KpnI fragment; probe 122, a 0.4-kb BamHI/EcoRI fragment;probe 35, a 1.0-kb KpnI/EcoRI fragment; probe 3, a 2.0-kb EcoRI/BamHIfragment; probe 12, a 0.3-kb BamHI/PvuII fragment; probe 38,a 0.4-kb PvuII/XmnI fragment; probe 19, a 1.1-kb HindIII/KpnIfragment; probe 67, a 0.7-kb EcoRI/XbaI fragment; probe 15,a 0.8-kb HindIII/BamHI fragment; and probe 203, a 0.25-kb XbaI/BstEIIfragment. Based on the patterns of diagnostic fragments, colonieswere selected that had undergone clean double homologous recombinationevents at a site in the DHFR gene and at the correct generallocation in the intergenic region downstream (Fig. 1 and text).The restored cell line represents a control that underwent recombinationexchanges near sites a and b, thus restoring the wild-type arrangement.

2-D gel analysis (7). After 12 h in mimosine, cell lines were released into the Speriod by washing them with and returning them to prewarmedMEM containing 10% fetal clone II (supplemented with thymidineand hypoxanthine for DR-8A7 cells). At selected intervals thereafter(see figure legends), cells were harvested and replication intermediateswere prepared by digesting matrix-attached DNA with EcoRI asdescribed previously (21). Replication intermediates were analyzedby neutral/neutral 2-D gel electrophoresis exactly as describedpreviously (21).

FISH-based replication timing analysis (36, 40). The determination of replication timing by fluorescence in situhybridization (FISH) was performed as described previously (36,56), except that prior to fixation, cells were swollen in 40mM KCl and 5 mM glycine (pH 9.5) instead of the typical neutralswelling regimen (61; L. D. Mesner, P. A. Dijkwel, and J. L.Hamlin, unpublished data). Only those alkaline FISH samplesthat showed no appreciable loss in the number of cells comparedto neutral FISH samples were analyzed.

RESULTS

Top

Results

Construction of the intergenic KO cell lines by the ROKO approach (36). To engineer deletions into the intergenic region in loco, the3' end of the DHFR gene in the DR-8A7 variant is restored tothe wild type by homologous recombination with a donor constructthat provides (i) the missing part of the gene, (ii) a smallregion of homologous overlap, and (iii) part of the spacer fromwhich the relevant downstream target has been removed (Fig.1B and C). Since DR-8A7 has no functional copy of the DHFR gene,it cannot survive on minimal medium lacking thymidine, hypoxanthine,and glycine (57) unless the missing 3' sequences are restored.Homologous exchanges near sites a and b yield a wild-type restoredderivative (Fig. 1B and D) ( $~$ 1/10⁶ transfected cells) while exchangesnear a and a position downstream from the desired deletion (e.g.,near site c or d) restore the gene and also effect the deletion(Fig. 1E) (1 to 3% of DHFR⁺ cells).

In the present study, donor constructs were either BACs or cosmidsfrom which sequences encompassing the desired region had beenremoved (see Materials and Methods). Donors with the followingdeletions were constructed (Fig. 1C): (i) a 4.4-kb sequenceencompassing the ori-ß' locus; (ii) a 12-kb sequenceencompassing both ori-ß and ori-ß'; (iii)the 40-kb central core of the intergenic spacer, in which >90%of initiations occur in the native locus (compare the IRKO deletionin Fig. 1C to the ELFH data in Fig. 1A); and (iv) a 12-kb sequencecontaining the segment of the intergenic region deleted in theDR-8A7 cell line. After transfecting each donor BAC or cosmidinto DR-8A7 by electroporation, potential intergenic deletionvariants were selected on MEM and characterized by Southernblotting and hybridization with the relevant probes. Controlcells were UA21, which contains a single wild-type locus, andthe starting dhfr-deficient DR-8A7 recipient. Usually, two differentdigests were run on the same gel and hybridized with a mixtureof two probes, at least one of which illuminates a diagnosticfragment in each digest of the individual cell lines. Examplesof Southern analyses are shown in Fig. 2.

View larger version (49K):
[in this window]
[in a new window]
FIG. 2. Southern analysis confirms successful restoration of the DHFR gene and the relevant deletions in DHFR-proficient survivors. Relevant restriction digests of DNA from the indicated cell lines were separated on a 0.6% agarose gel along with a 1-kb ladder and high-molecular-weight standard (Invitrogen), transferred to a HybondN+ membrane (Amersham), and hybridized with the indicated probes to detect fragments expected for the desired deletions. Diagnostic fragments for each successful recombinant are indicated in Fig. 1B, D, and E and are discussed in the text. The probes utilized are pictured in Fig. 1A and specified in Materials and Methods.

In an Asp718 digest (Fig. 2A), a mixture of probes 3 and 100detects the following fragments (Fig. 1A and E): (i) 15.4- and4.0-kb wild-type fragments in UA21 and the wild-type restoredrecombinant; (ii) 15.5-kb junction and 4.0-kb wild-type fragmentsin DR-8A7; (iii) 12.0-kb junction and 4.0-kb wild-type fragmentsin the ori-ß' KO (ß' KO); (iv) 6.5-kb junctionand 4.0-kb wild-type fragments in the ß and ß'KO; and (v) 5.7-kb junction and 4.0-kb wild-type fragments inthe IRKO. In an EcoRI digest (Fig. 2A), a mixture of probes3 and 100 detects the following: (i) wild-type 3.2- and 2.6-kbfragments in UA21 and the restored cell line; (ii) 2.6-kb wild-typeand 2.6-kb junction fragments in DR-8A7; (iii) 3.2- and 2.6-kbwild-type fragments in the ß' KO; (iv) 3.4-kb junctionand 3.2-kb wild-type fragments in the ß and ß'KO; and (v) 7.2-kb junction and 3.2-kb wild-type fragments inthe IRKO cell line. For the 3'-end deletion (Fig. 2B), a mixtureof probes 100 and 12 and 38 on a HindIII digest detects 20.5-and 4.3-kb wild-type fragments in UA21, a 12.6-kb HindIII junctionfragment in DR-8A7, and 9.5-kb junction and 4.3-kb wild-typefragments in the 3'-end KO.

To demonstrate that the desired regions were actually deletedby successful double homologous recombination events, genomicDNAs were digested with either EcoRI or Asp718 and were hybridizedwith small probes complementary to the deleted region in eachcase (maps and probe positions are given in Fig. 1A). For example,in the results shown in Fig. 2C, a mixture of probes 19 and15 recognizes 6.1-kb (probe 19) and 11.0- and 2.6-kb (probe15) EcoRI fragments in wild-type UA21 and DR-8A7 cells. As predicted,probe 19 fails to illuminate the 6.1-kb fragment in the ß'and ß/ß' KOs, and probes 19 and 15 do notdetect either the 11.0-, 6.1-, or 2.6-kb fragments in the IRKOdeletion variant. For analysis of the 3' KO, an Asp718 digestwas hybridized with a mixture of probes 35, 67, and 122. Asseen in the results shown in Fig. 2D, UA21 DNA displays thepredicted 15.4-, 10.4-, and 8.1-kb Asp718 fragments, DR-8A7lacks the 8.1-kb fragment, and the 3' KO lacks both the 8.1-and 15.4-kb Asp718 fragments. Southern analyses with severalother restriction enzymes and probes confirmed that clean restorationsof the gene and deletion of the relevant sequences had occurred,resulting in the KO cell lines diagramed in Fig. 1E (Mesner,unpublished).

Deletion of ori-ß' alone or in combination with ori-ß has no demonstrable effect on initiation in the remainder of the intergenic zone. We first asked whether ori-ß' might correspond tothe master replicator in the region or to a redundant masterreplicator that takes over in the absence of ori-ßby relief of origin interference (9). To test these possibilities,the first recombinant cell lines constructed were the ori-ß'and ori-ß and ori-ß' KO cell lines (ß'KO and ßß' KO) (Fig. 1E). Any effects oninitiation in the remainder of the spacer resulting from thesedeletions were assessed by neutral/neutral 2-D gel analysisof replication intermediates isolated from synchronized cells90, 180, and 360 min after entry into the S phase (the S phaselasts $~$ 8 h in these cell lines). EcoRI digests of replicationintermediates were separated on a 2-D gel, transferred to amembrane, and analyzed with the relevant hybridization probes.As shown in Fig. 3A and B, different arcs are traced in thesegels depending upon whether a fragment is replicated passivelyby single forks entering from either side or if it containsan initiation site (7).

View larger version (57K):
[in this window]
[in a new window]
FIG. 3. Neutral/neutral 2-D gel patterns of origin-positive and -negative cell lines. (A and B) Principle of the neutral/neutral 2-D gel method. A restriction digest containing replication intermediates is separated in the first dimension gel (left to right) largely according to molecular mass and in the second dimension according to both mass and shape (top to bottom). After transfer to a membrane, the digest is hybridized with a probe specific for the fragment of interest. The patterns traced by a fragment replicated either passively by a single fork or actively from an internal, centered origin are shown in panels A and B, respectively. (C to E) DR-8A7 and the restored derivative were synchronized at the G₁/S boundary with mimosine, and samples were collected 90, 180, and 360 min after drug removal. After purification of EcoRI-digested replication intermediates, separation on a 2-D gel, and transfer to a membrane, the digests were hybridized with probes 12 and 38 (rows C and E) or a probe specific for a 6.5-kb fragment in the rhodopsin origin (row D) (20) (see Fig. 1A and Materials and Methods for probe positions). The average film exposure time with an intensifying screen for all of the hemizygous cell lines studied here was 10 to 21 days at -70°C.

Figure 3C to F shows replication intermediates detected in theintergenic region of the starting DR-8A7 cell line and a restoredwild-type derivative of DR-8A7 (Fig. 1B and D). In DR-8A7, virtuallyno replication intermediates can be detected in a 6.2-kb fragmentencompassing the ori-ß region at 90 and 180 min intothe S phase when hybridized with probes 12 and 38 (Fig. 3C and1A). These data are in agreement with earlier studies showingthat early-firing origin activity is inhibited in this locusas a consequence of the deletion encompassing the 3' end ofthe gene and part of the intergenic region (36). By 360 min,a relatively intense single fork arc and the absence of a detectablebubble arc indicate that this region is replicated late in theS phase, probably by forks from distant active origins (Fig.3C, compare to Fig. 3A) (see Discussion).

In contrast, when the transfer shown in Fig. 3C was strippedand rehybridized with a probe for the control, early-replicatingrhodopsin origin (27), a composite pattern is detected at 90min consisting of a complete bubble arc and a single fork arc(Fig. 3D) (20). This pattern is characteristic of fragmentsresiding in broad initiation zones that fire in the early Sphase (including the wild-type DHFR locus), since such fragmentssometimes sustain internal initiation events but more frequentlyare replicated passively by forks from neighboring initiationsites in the same zone (17, 18, 20, 22, 58). Very few bubblescan be detected in the rhodopsin origin at 180 min, indicatingthat initiation is largely complete by this time, but a singlefork arc persists in the region until $~$ 360 min. This is becausethe rhodopsin origin, like the wild-type DHFR origin, is inefficientoverall, firing only 30 to 40% of the time in any given cellcycle (17, 20, 22); thus, in many cells, the locus is replicatedpassively well into the S period by forks from active originslying upstream or downstream from the locus.

When DR-8A7 is converted to the wild-type arrangement in therestored derivative (Fig. 1B and D), the DHFR origin revertsto the early-firing phenotype characteristic of the wild-typelocus and the rhodopsin early-firing control, as indicated bythe patterns obtained with probes 12 and 38 for the ori-ßlocus and probe 203 for a region lying downstream from the 2BE2121gene (Fig. 3E and F) (20). (Note that the transfers shown inFig. 3E and F are overexposed relative to that shown in Fig.3D, but the overall patterns are similar.) Thus, in studiesdiscussed below on the effects of intergenic deletions, therhodopsin locus will be utilized as the early firing internalcontrol. Importantly, faithful restoration of the truncatedgene in DR-8A7 with a concomitant deletion of a downstream target(as in Fig. 1E) should restore the origin to the early-firingphenotype only if deletion of the target has no effect on initiation.If a deletion renders the DHFR origin either inactive or latefiring, the cell line should display the late-replicating phenotypethat characterizes DR-8A7.

We next examined replication intermediates from the ß'KO cell line by the 2-D gel method (Fig. 4). The pattern detectedwith a mixture of probes 12 and 38, which illuminates the ori-ß-containingfragment, is remarkably similar to that detected when the sametransfer was stripped and rehybridized with a probe specificfor the rhodopsin locus (compare Fig. 4A and C; also compareto the ori-ß locus in the wild-type restored controlin Fig. 3E): initiation is seen at 90 min and is largely completeby 180 min while single forks indicative of passive replicationpersist until about 360 min into the S phase. Probe 35, whichis specific for a fragment in the 5' end of the initiation zone(Fig. 1A), also detects an early-firing pattern in this regionof the ß' KO cell line, although the intensity ofthe bubble arc relative to the single fork arc is lower thanat ori-ß (compare Fig. 4A and B), just as it is inthe wild-type locus (23). Even more interesting is the observationthat when ori-ß and ori-ß' are coordinatelydeleted in the ßß' KO cell line (Fig. 4D),probe 35 again displays an early replicating pattern similarto the rhodopsin origin in the same cells (Fig. 4E) and to thatof the ß' KO cell line (Fig. 4B).

View larger version (74K):
[in this window]
[in a new window]
FIG. 4. Deletion of ori-ß' alone or ori-ß and ori-ß' has no apparent effect on initiation in the remainder of the spacer region. The ori-ß' and ori-ß/ori-ß' KO cell lines were synchronized and processed as described in the legend to Fig. 3 and Materials and Methods. The images shown in panels A to C were derived from successive hybridizations and stripping of the same blot, as were the images shown in panels D and E (22).

Critical, nonredundant, genetic elements required for early-firing origin activity do not reside in the 50-kb region extending between the 3' ends of the DHFR and 2BE2121 genes. Thus, neither ori-ß nor ori-ß' correspondsto an essential replicator responsible for controlling initiationin the DHFR origin as a whole. Furthermore, these two sitesor regions do not appear to correspond to redundant master replicatorsthat interfere with each other. However, the possibility remainedthat these two regions are redundant with yet another masterelement elsewhere in the spacer (possibly ori- ${gamma}$ ). We addressedthis issue directly by constructing the IRKO cell line, whichhas a deletion of the central 40-kb core of the intergenic spacer,including ori-ß, ori-ß', and ori- ${gamma}$ , and inwhich >90% of initiations normally occur as assessed fromthe quantitative ELFH data (Fig. 1A) (23). The positions ofthis deletion and the diagnostic fragments are shown in Fig.1C and E.

The IRKO cell line was synchronized and sampled as describedpreviously, and EcoRI digests were separated on a 2-D gel. Thedigests were blotted and hybridized successively with probes35 and 203, which are specific for 4.1- and 5.5-kb fragmentsnear the ends of the intergenic spacer in the wild-type configurationbut recognize 4.1-kb wild-type and 7.2-kb junction fragmentsin the truncated spacer in the IRKO variant. In the IRKO variant,probes 35 and 203 each detect bubble arcs in their cognate fragmentsin the 90-min sample (compare to the bubble arc in the rhodopsincontrol on the same blot) (Fig. 5C). Furthermore, replicationof the locus is largely completed by 360 min into the S phase,which would not be predicted if early-firing origin activityhad been severely compromised by the IRKO deletion (i.e., asin DR-8A7 cells) (Fig. 3C). Thus, we can conclude that critical,nonredundant sequences required to fire this origin in the earlyS phase do not reside within the 40-kb central core.

View larger version (72K):
[in this window]
[in a new window]
FIG. 5. Deletion of the central 40-kb core of the DHFR origin or sequences lying immediately downstream from the DHFR gene does not noticeably change the pattern of replication intermediates in the remainder of the intergenic spacer. The IRKO (panels A to C) and 3' end (panels D and E) deletion variant cell lines were synchronized and sampled as described in the legend to Fig. 3. Replication intermediates were prepared by using EcoRI to digest the DNA, and the resulting digests were hybridized with the indicated probes. In the IRKO variant, probe 35 detects a wild-type 4.1-kb fragment, but probe 203 illuminates a 7.2-kb deletion junction fragment. In the 3' end deletion variant, probe 19 recognizes a 6.1-kb wild-type EcoRI fragment encompassing ori-ß'.

We then asked whether any critical genetic element(s) requiredto activate this origin resides in the 12-kb region of the spacerlying just downstream from the DHFR gene. This region correspondsto the part of the spacer that is deleted in DR-8A7, whose originis inactive, but is retained in the IRKO deletion variant, whoseorigin is active (Fig. 1A, B, and E). However, as shown in Fig.5D and E, removing the region between the poly(A) sites in theDHFR gene and the 5' boundary of the IRKO deletion does notevoke a demonstrable change in the early-firing origin activity:probe 19 detects the classic early replicating pattern in anEcoRI fragment containing ori-ß', which displays thesame high level of replication bubbles as does ori-ßin the wild-type locus (17, 18, 22, 58).

The efficiency of initiation in the spacer appears to increase when the 40-kb core is deleted. Thus, deletion of a region containing the majority of startsites from the intergenic region in the IRKO cell line did notnoticeably reduce origin activity in the remainder of the spaceraccording to the criterion of 2-D gel analysis. Additional probingsdid not detect initiation within the body of the DHFR gene itself,suggesting that the initiation zone in the IRKO variant didnot spread outward (data not shown). This raised the questionwhether the efficiency of initiation in the truncated spacerwas affected by the loss of the 40-kb core.

This was first assessed by directly comparing the intensitiesof the bubble arcs in the spacers of the IRKO and wild-typecell lines in a 2-D gel-mixing experiment. IRKO cells and therestored derivative were synchronized with mimosine by the regimendescribed above and were sampled 90 min after release into theS period. Fluorescence-activated cell sorter analysis indicatedthat the two cell lines were synchronized almost identicallyby this regimen (data not shown). Equal numbers of each celltype were combined, and replication intermediates were preparedfrom the mixture by using EcoRI to digest the DNA. After separationon a 2-D gel and transfer to a membrane, the digest was hybridizedwith probe 203, which recognizes a 5.5-kb fragment in the wild-typelocus and a 7.2-kb deletion junction fragment in the IRKO cellline formed by the fusion of the 5.5- and 2.6-kb wild-type versions(Fig. 1A). Because of the difference in size, the two fragmentscould be separated enough to distinguish the intermediates arisingfrom each. Importantly, the two wild-type fragments that werefused in the IRKO deletion reside near the ends of the spacerin the wild-type locus, which normally support fewer initiationevents than the 40-kb core (Fig. 1A).

The results of this analysis with probe 203 are shown in Fig.6A. The intensities of both the bubble and fork arcs in the7.2-kb deletion junction fragment in the IRKO cell line areclearly higher than they are in the 5.5-kb wild-type fragmentfrom the restored wild-type derivative, even though the nonreplicated1n spots are comparable. To estimate the level of replicationintermediates in the IRKO junction fragment compared to a fragmentin the active central core in the restored wild-type locus,the same blot was stripped and rehybridized with probe 19; thisprobe recognizes a 6.1-kb EcoRI fragment encompassing ori-ß'in the wild-type locus of the restored cell line. This samefragment is deleted in the IRKO cell line (Fig. 1C). As shownin Fig. 6B, the intensities of the bubble and fork arcs detectedwith probe 19 in the ori-ß' locus in the restoredderivative are approximately equal to those detected with probe203 in the deletion junction fragment in the IRKO variant, eventhough the intensities of the 1n spots are approximately equal.In other words, the junction fragment in the truncated spacernow appears as active as does a fragment in the most-activeregion of the wild-type locus. Although some of the differencesin the intensities of the bubble arcs in the 7.2- and 5.5-kbfragments shown in Fig. 6A can be attributed to a size difference(45), the magnitude of this difference suggests that the frequencyof initiation in the truncated spacer may have actually increasedover that of the corresponding sequences in the wild-type locus.

View larger version (81K):
[in this window]
[in a new window]
FIG. 6. The efficiency of initiation in the truncated spacer in the IRKO cell line appears to have increased relative to the same sequences in the wild-type locus. The IRKO and restored wild-type control cell lines were synchronized as described previously, and samples were taken in the early S phase (90 min). Equal numbers of cells from each cell line were mixed, and replication intermediates were prepared by using EcoRI to digest the DNA. The transfer of the resulting 2-D gel was first hybridized with probe 203, which detects a wild-type 5.5-kb EcoRI fragment lying downstream from the 2BE2121 gene in the restored cell line (Fig. 1D) and a 7.2-kb deletion junction fragment in the IRKO variant (Fig. 1). The blot was then stripped and rehybridized with probe 19, which detects a 6.1-kb EcoRI fragment containing ori-ß' in the wild-type restored control that is deleted in the IRKO variant.

A quantitative replication timing assay suggests that deletion of the central 40-kb core of the DHFR origin has little effect on the time of replication of the DHFR locus. To address the question of efficiency of initiation more quantitatively,the IRKO variant was examined by using a replication timingassay in which the number of copies of a locus of interest ina given cell (and thus its replication status) is compared tothe copy number of an internal, early replicating standard (Fig.7A) (36). In this assay, cells were immobilized on microscopeslides and hybridized with a mixture of probes specific forthe relevant loci. In the present case, a cosmid from the early-firingrhodopsin control locus was labeled with biotinylated dATP andwas mixed with digoxigenin-labeled cosmids spanning varioussections of the DHFR locus (Fig. 7C). The probe for rhodopsinwas detected with streptavidin-Texas Red, while the DHFR-specificprobes were detected with sheep anti-digoxigenin and fluoresceinisothiocyanate-conjugated donkey anti-sheep immunoglobulin G(36, 56). By counting the numbers of each of the fluorescentpatterns shown in Fig. 7A (panels A to F), one arrives at areplication index, which is the ratio of cells that have replicatedthe DHFR locus versus the early-firing rhodopsin control (Fig.7A, panel J). If both markers replicate at the same time ina given cell line, a replication index near unity will result,while late-replicating loci will display lower replication indices(36). Although this method does not measure effects on initiationper se, if a region normally replicates early because it containsan early-firing origin, the locus as a whole will remain earlyreplicating only if its origin still fires in the early S phase.Thus, the assay will be sensitive to changes that cause an early-firingorigin to fire later or not at all, and the magnitude of thedelay can be quantified.

View larger version (21K):
[in this window]
[in a new window]
FIG. 7. The time of replication of the DHFR locus in the IRKO deletion variant does not differ significantly from that observed in wild-type cells. (A) Principle of the FISH-based replication timing assay (see text). (B) Synchronized cells were sampled at 90, 180, 360, and 540 min after release from mimosine, fixed, and spread on microscope slides. The spreads were hybridized with a mixture of a rhodopsin (RHO) probe and a second probe from the DHFR locus (see figure and text), and the numbers of each of the patterns shown in panel A were determined by fluorescence microscopy for each cell line. A replication index for each sample was calculated as shown in panel A (J) and discussed in the text. Data were expressed either as a replication index as a function of time in the S period (B) or as percent nuclei with the locus replicated (C). The cosmids used to detect various parts of the DHFR domain are indicated below panel C.

As an example, the DR-8A7 and restored cell lines were subjectedto this analysis in synchronized cell populations with cosmidKD504 (which is centered in the intergenic region) as a hybridizationprobe. The late-replicating phenotype of DR-8A7 detected by2-D gel analysis is clearly recapitulated here, with the replicationindex of DR-8A7 approaching that of rhodopsin (i.e., unity)only after $~$ 540 min in the S period (Fig. 7B, compare to therestored cell line). When the cosmids cH1, KZ381, and KC393were used to probe the IRKO cell line, which lacks >90% ofthe start sites utilized in wild-type cells, each of the regionsrepresented by these cosmids can be seen to replicate at approximatelythe same time as the early replicating rhodopsin control. (Notethat the DHFR origin appears to fire somewhat later than rhodopsin,even in the wild-type restored control, because we have scoredrhodopsin as having replicated even if only one of the two locihas doubled [as in Fig. 7A, panels B and E]. If only those cellsthat had doubled both rhodopsin loci had been scored as havingreplicated, the single DHFR locus would appear to replicateearlier than rhodopsin.)

In Fig. 7C, the data for the IRKO and restored cell lines areexpressed as percentages of cells that have duplicated a givenmarker as a function of time. This plot shows that all regionsof the DHFR locus examined here replicated at approximatelythe same time in the IRKO cell line as in the wild-type restoredcontrol. (Note that there is no data point for the cosmid KD504in the IRKO cell line, since the cognate sequences have beendeleted.) This plot also shows that, although the DHFR originis early firing in wild-type cells, it takes somewhere between180 and 360 min to replicate all copies of the locus (as shownin Fig. 3E and F). The FISH data are consistent with the observationthat the DHFR locus is often replicated passively by forks fromdistant origins at later times in the S phase. We have alsoexamined each of the deletion variants in unsynchronized exponentiallygrowing populations by the FISH-based timing assay. With theexception of DR-8A7, every deletion variant examined by thiscriterion replicates the DHFR locus at approximately the sametime as in the wild-type restored control (Mesner, unpublished).

DISCUSSION

Top

Discussion

In most bacteria, plasmids, viruses, and yeast, well-definedgenetic elements (replicators) serve as loading sites for thereplication initiation complex, thereby directing nascent strandpriming to positions very near the replicator itself. In caseswhere there is a single replicator for the entire genome (asin bacteria, plasmids, and most viruses), the genome is equivalentto a replicon, and with very rare exceptions, deletion of thereplicator results in a failure to initiate DNA synthesis (reviewedin reference 42). In S. cerevisiae and Schizosaccharomyces pombe,the situation is somewhat more complex, because the genomescontain multiple linear chromosomes, each consisting of manyreplicons and each of which contains its own genetic replicator(defined originally as ARS elements) (11, 53). Replicators inS. cerevisiae are $~$ 100 bp in length and consist of a very highlyconserved AT-rich ARS consensus sequence flanked by necessarybut ill-defined auxiliary sequences (37, 51). S. cerevisiaeorigins are distributed at 30- to 40-kb intervals in the genome(50) but rarely in the bodies of genes (51, 60). In S. pombe,replicators are somewhat larger ( $~$ 500 bp long) and multipartite,with no obvious common sequence elements other than adeninetracts (38, 39).

Unlike bacteria, yeast can usually tolerate the deletion ofone or more of these replicators because the region will eventuallybe synthesized passively by forks from active origins lyingupstream or downstream (15). A similar situation characterizesARS elements that are less than 100% active (which appears tobe most of them) (51) and which must also be replicated passivelyin some cell cycles. One is led to the conclusion that thereare usually no fixed termini between origins so that the boundariesof replicons are remarkably fluid (a notable exception is areplication fork barrier at the 3' end of ribosomal DNA [rDNA]genes) (8).

The multiple origins in mammalian genomes are distributed at $~$ 15- to 300-kb intervals (32). With a few exceptions (for examples,see references 1, 29, and 55), most mammalian origins consistof broad zones of closely spaced, potential initiation sitesrather than narrow zones closely flanking a replicator (seereference 19 for a review). Interestingly, homologues of mostof the proteins that facilitate initiation at yeast ARS elementshave been identified in mammalian cells (5, 28), but there isvery little data on the nature of the sequences with which theyinteract. Within the DHFR initiation zone, ori-ß,ori-ß', and ori- ${gamma}$ are clearly preferred initiationsites (Fig. 1A) and thus would be likely candidates to serveas replicators. However, they share no easily recognizable sequencemotifs that could correspond to initiator recognition elements(P. J. Mosca, L. D. Mesner, and H.-B. Lin, unpublished data).Furthermore, the data presented here and in a previous study(36) show clearly that none of these sites is uniquely requiredfor controlling initiation in this locus. Even when all threeare deleted in the IRKO variant, the locus still replicatesat approximately the same time in the S period. The latter observationrules out the possibility that ori-ß, ori-ß',and ori- ${gamma}$ correspond to redundant master replicators that cantake over in each other's absence. The lack of a significanteffect of the 3' deletion (Fig. 5D and E) on initiation alsoruled out the possibility that a uniquely required element residesin the part of the spacer deleted in the origin-negative DR-8A7cell line. The IRKO deletion also retains $~$ 5 kb downstream fromthe 2BE2121 gene that are normally used infrequently for initiation(Fig. 1A and E) (23) but which could theoretically contain acritical genetic element. Although we have not yet deleted thisregion, any critical element within it cannot be sufficientfor initiation. This follows because this region is retainedin the DR-8A7 and DG22 deletion variants, which lack the 3'and 5' ends of the DHFR gene, respectively, and whose originsare inactive (36; S. Saha, Y. Shan, L. D. Mesner, and J. L.Hamlin, unpublished data). Indeed, a distal ancillary elementsuch as the locus control region identified in the ß-globindomain (1a) could also reside either in the 5' or 3' regionsdeleted in DG22 and DR-8A7, respectively.

Because we have not removed the entire spacer in a single deletionvariant, we have not formally ruled out the possibility thata small number of critical but redundant genetic elements residewithin the origin itself. The minimum number would be three(i.e., one residing in the 40-kb core, one in the 11 kb immediatelydownstream from the DHFR gene, and a third in the 5-kb regiondownstream from 2BE2121). However, we believe that most of theobservations made about this locus so far conform to a somewhatdifferent model that relies in part on the yeast paradigm. Inthese simple eukaryotes, deletion of an active replicator froma chromosome can have two consequences: (i) the activity ofa nearby less-active or silent replicator can increase becauseof relief of origin interference (9, 59), resulting in an insignificantchange in replication timing of the locus, or (ii) the regionhas to wait to be replicated passively by forks from a distantactive origin, which would result in a measurable change inreplication timing (the magnitude depending upon the distancebetween the deletion and the nearest origin) (15, 51). Sincewe have not detected a significant change in replication timingas a consequence of any of the deletions in the DHFR origin,it appears that the region is being replicated from start sitesin the immediate vicinity of each intergenic deletion. Coupledwith the observation that initiation occurs at virtually everyposition tested in the spacer (whole or truncated), it is likelythat potential replicators are distributed at very frequentintervals in the spacer, some of which are more active thanothers (at least 20 sites, but probably many more) (23).

It follows from our data that each of these proposed replicatorsmust control initiation only in its immediate environment, sincedeletion of the prominent start sites has no apparent negativeeffect on initiation at other sites in the spacer. This againraises the question of whether there are bona fide replicatorsin mammalian genomes or whether initiation proteins can bindto large numbers of degenerate sequences almost at random, withepigenetic factors largely determining the efficiency of utilization.In this regard, mutagenesis studies on ectopically integratedcopies of ori-ß have shown that its activity as anorigin can be partially suppressed by certain deletions nearthe peak of initiation shown in Fig. 1A (3). In this study,initiation activity was assessed by the PCR-based nascent strandabundance assay, which allowed relatively high resolution analysisof the region surrounding ori-ß. In this experimentalsituation, there appear to be at least some sequence requirementsfor binding initiation complexes, in agreement with studieson the ß-globin locus in human cells (1). However,sequence analysis of $~$ 30 kb of the intergenic region has uncoveredno obvious simple repeated elements analogous to the ARS consensussequence that could correspond to the 20 or more other potentialreplicators in the DHFR intergenic spacer (L. D. Mesner, A.Pemov, and P. J. Mosca, unpublished data) (also see reference24). A fascinating question for the future is whether (and ifso, how) the mammalian counterparts of replication initiationproteins that were originally identified in yeast recognizethe large spectrum of initiation sites that characterize thisand many other mammalian origins. It is conceivable that someor all of these proteins have lost much of their specificityfor defined sequence elements during evolution.

Another surprising finding is that deletion of the 40-kb coreof the spacer, in which the majority of initiation normallyoccurs, does not dramatically affect the time of replicationof the remainder of the spacer nor the flanking genes (Fig.7). The usual time of replication appears to be approximatelymaintained by an increase in the efficiency of initiation inthe truncated spacer. Thus, regions that were not very activein the native locus (the outer edges of the spacer) assume amore-prominent role as substrates for initiation in the absenceof the core. Somehow, loss of the central core is transmittedto the remaining sequences, probably by epigenetic factors.One obvious possibility is the closer juxtaposition of the twogenes in the IRKO variant (i.e., reduction in the size of theintergenic spacer). In this regard, it is interesting that originswhose start sites or regions are thought to be very circumscribed(e.g., lamin B2 [29] and globin [1]) are situated in very narrowintergenic regions. Perhaps transcription itself or chromatinremodeling specific to active transcription units affects theactivity of neighboring origins in the intergenic spacers. Infact, there is a precedent for the effects of transcriptionon replication initiation in the rDNA loci in developing Xenopuslaevis (33). Prior to the mid-blastula transition when transcriptioncommences, initiation sites are distributed throughout the rDNArepeats (including the inchoate genes). Once transcription begins,initiation is confined to the intergenic spacers. It is notclear whether transcription actually increases the efficiencyof initiation in the spacer or merely prevents it in the bodyof active genes.

Another explanation for the apparent increase in the efficiencyof initiation in the truncated IRKO spacer might be that thereis a defined number of initiators for each permissive initiationzone, whether large or small, with the consequence that theconcentration in a small zone would be higher. A third possibilityis that initiation complexes take advantage of the open configurationsof promoters to load onto the template, and are then deliveredto intergenic origins either by looping, as has been suggestedto explain the activities of enhancers (25, 47, 54), or by deliveryto the spacer via the transcription machinery itself.

ACKNOWLEDGMENTS

We thank Lawrence Chasin and Adelaide Carothers for providingus with the UA21 and DR-8A7 cell lines, respectively. We arealso appreciative of the cheerful and expert technical assistanceprovided by Carlton White and Kevin Cox. We thank the othermembers of the Hamlin laboratory for excellent input duringthe course of this project.

This work was supported by NIH grant RO1 GM26108.

FOOTNOTES

* Corresponding author. Mailing address: Department of Biochemistry and Molecular Genetics, University of Virginia School of Medicine, Charlottesville, VA 22908. Phone: (434) 924-5858. Fax: (434) 924-1789. E-mail: jlh2d{at}virginia.edu.

REFERENCES

Top