INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Developmentally programmed DNA rearrangement is an integral part of the life cycle of many organisms. One of the best-known examples is the rearrangement of immunoglobulin genes that occurs during lymphocyte development, giving rise to the vast diversity of the vertebrate immune system (reviewed in reference 36). Such DNA rearrangement events must be precisely controlled to avoid deleterious effects of aberrant reorganization. For example, chromosomal translocations involving the immunoglobulin locus are frequently associated with lymphoid malignancies (reviewed in references 27 and 39). The deleterious potential of failed rearrangement underlies the importance of understanding the molecular mechanisms guiding these events.

The most dramatic examples of DNA rearrangement have been termed chromatin diminution, which refers to the developmentally programmed elimination of large portions of genetic material from all somatic progenitor cells. This phenomenon was first described a century ago by Boveri for Ascaris (7) and has since been observed in many organisms (6, 28, 30; reviewed in references 14 and 32). Chromatin diminution is ubiquitous among the ciliated protozoa studied (32). Most ciliates exhibit a nuclear duality, maintaining distinct sets of genetic material for germ line and somatic functions. The DNA of the germ line micronucleus is the full genetic complement, whereas the DNA of the somatic macronucleus is a highly rearranged subset of the germ line DNA. The form and extent of the DNA rearrangements observed in this diverse group of organisms vary greatly. For example, the sizes of eliminated regions range from tens of base pairs to tens of kilobase pairs, and the quantities of eliminated DNA range from ~10% to as much as 95% of the germ line genome. The relationships between the rearrangements that occur in different ciliate species or even within the same species are not well understood.

Among organisms that undergo large-scale DNA rearrangement, the ciliate Tetrahymena thermophila is particularly amenable to molecular genetic analysis. In Tetrahymena, conjugation initiates a developmental program that results in the formation of new germ line micronuclei and somatic macronuclei, as well as the destruction of old macronuclei. The genome of a developing macronucleus undergoes extensive reorganization. Chromosome fragmentation occurs at 50 to 200 sites (1, 13) defined by the chromosome breakage sequence (50), and specific DNA rearrangements remove 10 to 15% of the germ line genome from roughly 6,000 internal chromosomal sites (46, 47). The segments of micronucleus-limited sequences are referred to both as deletion elements and internal eliminated sequences. They consist of unique and/or moderately repetitive DNA sequences and range in size from several hundred base pairs to greater than 10 kbp. Most Tetrahymena deletion elements have been found outside of coding sequences, although one has been found within an intron (22). The eight deletion elements examined by sequence analysis share few obvious similarities other than a strong A+T nucleotide bias and the presence of short direct repeats of 1 to 8 bp at the rearrangement boundaries (4, 5, 12, 22, 25, 40).

The R and M elements were the first Tetrahymena deletion elements to be sequenced (4, 5) and remain the most extensively characterized. The R element is eliminated during macronuclear development by a 1.1-kbp deletion event (2). The M element is eliminated from the macronucleus by two alternative deletion events of 0.6 and 0.9 kbp (2). These two eliminated forms share a common right boundary but utilize different left boundaries that are 0.3 kbp apart (5). Alternate rearrangement boundaries may be used by as many as 25% of deletion elements (12). Different rearrangement events between the same boundaries of a given element usually produce the same junction sequence; even so, rearrangement of most elements exhibits some heterogeneity, producing variant junction sequences that differ by a few base pairs (3, 29, 31).

Deletion elements placed on Tetrahymena rDNA-based transformation vectors rearrange accurately when introduced into conjugating cells (20). By using this transformation assay to study M-element rearrangement, an essential cis-acting regulatory sequence, 5'-AAAAAGGGGG-3' (A₅G₅), was identified, providing the first mechanistic insight into these site-specific deletion events. This sequence is located ~45 bp outside each end of the micronucleus-limited region in a specific orientation (A₅G₅ on the left; C₅T₂AT₂ on the right) and functions to position the rearrangement boundaries a short distance away (20). Moving the location of this sequence repositions the rearrangement boundary to within 41 to 54 bp of the new location (19). This A₅G₅ sequence is not found near any of the other sequenced deletion elements, and it is the only cis-regulatory sequence that has been clearly defined.

The distance-dependent action of the M-element A₅G₅ sequence, as well as its position outside the deletion element, is unique among known rearrangement systems. The lack of any common, identifiable cis-acting sequence among the other known deletion elements, together with their size and sequence diversity, has challenged our understanding of these rearrangement events. It is still not known whether elimination of the estimated 6,000 deletion elements occurs via a common mechanism or involves several distinct rearrangement pathways. To better understand the relationship between the rearrangement of different elements in Tetrahymena, we have characterized cis-acting sequences involved in the rearrangement of the R deletion element. In this study, we have found that sequences outside the micronucleus-limited region are required for deletion. We show that these cis-acting sequences serve to position the rearrangement boundary a short distance away. The function of these flanking regulatory sequences is very similar to that determined for the M-element A₅G₅ sequence (19, 20). The finding that different flanking regulatory sequences of these two elements perform the same function provides strong evidence for a common mechanism controlling DNA rearrangement in Tetrahymena.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains. T. thermophila inbred B strains CU427 [Chx/Chx (VI, cy-s)] and CU428 [Mpr/Mpr (VII, mp-s)] (obtained from Peter Bruns, Cornell University) were used for all transformation experiments described below. Maintenance and growth of these strains were carried out under standard conditions as previously described (21).

Plasmid constructions. Recombinant DNA techniques were executed essentially as described by Sambrook et al. (33). For transformation analyses, all modified R elements were inserted into the polylinker sequence located downstream of the transcribed region of the rDNA in the Tetrahymena vector pD5H8 (20). In some cases, the polylinker of this vector had been previously modified to introduce additional cloning sites by inserting the 31-bp NSXBK/NKBXS linker sequence given in Table 1 into the unique NotI site to create pD5H8N1.

Tetrahymena transformations. Transformation of Tetrahymena with R-element-containing rDNA vectors was performed by microinjection or electroporation. Logarithmically growing cells were prepared for transformation by starvation for several hours in 10 mM Tris-Cl (pH 7.4) prior to mixing strains to initiate conjugation (20). Microinjection of mating pairs was performed as described previously (10, 38, 49). Electroporation of mating cells was performed as described by Gaertig and Gorovsky (18). Transformants generated by microinjection were used only to determine the rearrangement boundaries of some modified R elements. Transformants obtained by electroporation were used to determine rearrangement activity and boundary sites.

DNA isolation and analysis. Whole-cell DNA was isolated from transformants as previously described (4). For Southern blot analysis, DNA was digested with restriction enzymes under the conditions recommended by the suppliers. These samples were fractionated by electrophoresis in 0.8 to 1.2% agarose gels. Lambda DNA digested with either HindIII or PstI was used as a size standard. DNA was then transferred to nitrocellulose membranes (Schleicher & Schuell, Keene, N.H.) by pressure using a PosiBlot apparatus (Stratagene, La Jolla, Calif.) and then cross-linked to the membranes by UV light. Immobilized DNA was hybridized to an R-element-specific probe in 6× SSC (20× SSC is 3 M sodium chloride plus 0.3 M sodium citrate [pH 7.0]), 0.1 M Tris-HCl (pH 7.5), 0.5% sodium dodecyl sulfate, and 2× Denhardt's reagent (50× Denhardt's reagent is 1% Ficoll, 1% polyvinylpyrrolidone, and 1% bovine serum albumin) at 65°C overnight (12 to 20 h). This probe was an EcoRI/PstI restriction fragment from pDLCR5 (11) that corresponds to the rearranged form of the R element with 0.3 kbp of right and 0.9 kbp left of the rearrangement junction that was radiolabeled with [³²P]dATP (16, 17). Hybridized membranes were washed three to four times in 1× SSC-0.5%SDS at 65°C for 20 to 30 min and then exposed to X-ray film. The amounts of rearranged and unrearranged R element were quantified with a PhosphorImager (Molecular Dynamics, Sunnyvale, Calif.). To detect DNA fragments resulting from accurate rearrangement, membranes were hybridized and washed under similar conditions at 37°C to an end-labeled oligonucleotide, J1110R (Table 1), that is specific for the predominant chromosomal deletion junction (4). Autoradiograms were captured as digital images with a flatbed scanner (Epson America, Torrance, Calif.) and Photoshop version 4.0 LE (Adobe Systems) and displayed by using Canvas version 3.5.5 (Deneba Systems).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The R deletion element (Fig. 1A) is a 1,084-bp micronucleus-limited sequence (4) that is located ~2.7 kbp right of the M element (2) on micronuclear chromosome 4 (9). To aid in its description, we have divided the sequence of the element into three parts as described in the legend to Fig. 1A.

View larger version (30K): [in this window] [in a new window]   FIG. 1.   The R element and the rearrangement assay. (A) Schematic diagram of the R element. The bar above the diagram indicates the region of the element originally sequenced by Austerberry and Yao (4) and is divided into 200-bp increments. We have separated the element into three parts and assigned a numbering system to each. The micronucleus-limited region that is eliminated during rearrangement is shown as a narrow, solid box and is numbered left to right from +1 to +1084. Positions +1 and +1084 correspond to nucleotides 329 and 1413 as assigned in the original published sequence. The macronucleus-destined region on the left is represented as the wide, open box and numbered right to left from 1L to 328L. The right macronucleus-destined region is shown as the wide, shaded box and is numbered left to right from 1r to 421r. For positions beyond 328L and 421r, the distances are approximate. Positions of the predominant left and right rearrangement boundaries formed by elimination of the endogenous R element are designated by the open and shaded arrowheads, respectively. The nucleotide positions, 328 and 1414, joined by rearrangement are given under these arrowheads and correspond to the first nucleotides of the left (1L) and right (1r) flanking regions. (B) The rearrangement activity of vector-borne deletion elements are tested by transformation of conjugating Tetrahymena cells. Conjugation is initiated by mixing prestarved strains CU427 and CU428. Approximately 8 to 9 h after mixing, the transformation vectors are introduced into the developing macronuclei by microinjection or electroporation. The transformants are identified by their growth in the presence of the antibiotic paromomycin. DNA is isolated from the transformants, digested with restriction endonucleases to liberate the DNA fragment containing the deletion element from the transforming rDNA molecules, and analyzed by gel electrophoresis and Southern blot hybridization.

The rearrangement assay. To identify sequences that are required in cis to control R-element rearrangement, we used a transformation assay (Fig. 1B) that was developed previously to investigate the control of M-element rearrangement (20). In this assay, the Tetrahymena rDNA-based vector, pD5H8, that contains a modified R element is transformed into conjugating T. thermophila CU427 and CU428. Upon transformation, the rDNA including the R element inserted into the 3' nontranscribed region is cleaved from the circular plasmid at the 5' and 3' chromosomal breakage sequences. Telomeres are added near the 3' breakage site, and the entire molecule is converted to a palindromic minichromosome and amplified to ~9,000 copies per cell (48; reviewed in reference 42). The vector-carried R element undergoes rearrangement during processing of the transforming rDNA. Transformants are selected by their resistance to the antibiotic paromomycin (Pm^r phenotype), which is conferred by rRNA synthesized from the vector-borne, Pmr rDNA allele (8).

View larger version (36K): [in this window] [in a new window]   FIG. 2.   Analysis of external deletion of sequences flanking the micronucleus-limited region. Plasmid constructs containing progressively larger deletions of sequences flanking the R element were assayed for the ability to undergo precise deletion upon transformation. Plasmid DNA (P) and DNA isolated from transformants (T) were digested with NotI prior to electrophoresis and transfer to nitrocellulose membranes. Southern blot hybridization analysis with a probe specific to the macronuclear DNA from 312L to ~900r of the R element is shown at the top. A longer exposure of the right-hand lanes is shown to allow visualization of less abundant fragments. Positions of the unrearranged and rearranged elements are indicated to the left; positions of PstI-digested lambda DNA size standards are shown to the right. PhosphorImager analysis was used to quantify hybridization. To determine whether rearrangement was accurate, filters were hybridized separately with an end-labeled oligonucleotide, J1110R, that detects specifically the predominant rearrangement junction of chromosomal R elements. These hybridizing fragments are indicated by the arrowheads. The major rearranged product for the 203L/116r is ~200 bp larger than the accurately rearranged species and is denoted by the asterisk. The amount of DNA flanking the R element in each construct is indicated above each set of lanes. A diagram of the constructs is given at the bottom. The solid box represents the micronucleus-limited sequences; the wider open and shaded boxes represent the left and right macronucleus-destined regions, respectively. The rearrangement activity (ratio of hybridization to rearranged forms and the unrearranged construct) relative to an intact R element is shown on the bottom right: Normal, normal activity (>50%); , reduced activity (11 to 49%);  , greatly reduced activity (<10%); and , no detectable rearrangement.

DNA sequences flanking the micronucleus-limited region are necessary for efficient rearrangement. We used the above transformation assay to determine whether regions flanking the micronucleus-limited R element are required for efficient rearrangement. To this end, we created a series of deletion constructs that lacked progressively more of the left and right flanking regions and introduced vectors containing these modified R elements into conjugating Tetrahymena cells. Southern blot hybridization of DNA isolated from the resulting transformants by using a probe specific to the R-element flanking regions is shown in Fig. 2. For each construct, the plasmid used for transformation (lane P) is shown adjacent to the transformant DNA (lane T). The plasmid served as a marker for the size of the unrearranged R element. Both unrearranged and rearranged forms of the constructs were observed in the transformant DNA preparations. Each of the R-element constructs, 312L/900r, 203L/116r, and 100L/391r, that contained at least 100 bp of left and 116 bp of right flanking sequences exhibited normal (i.e., >50%) rearrangement activity. In contrast, the construct that contained only 70 bp right of the micronucleus-limited region, 312L/70r, displayed greatly reduced rearrangement activity. The construct that contained only 63 bp left and 70 bp right of the micronucleus-limited DNA, 63L/70r, was inactive. These data indicate that the micronucleus-limited region alone is insufficient and that the R element minimally requires ~100 bp of macronucleus-destined sequences on each side to efficiently undergo DNA rearrangement. We cannot formally exclude the alternative possibility that constructs with <100 bp of flanking DNA rearrange poorly due to the inhibitory action of vector sequences brought to within interfering distance of the boundary; however, we find this explanation unlikely and inconsistent with further analysis of the left flanking region presented below.

View larger version (41K): [in this window] [in a new window]   FIG. 3.   Analysis of small sequence deletions at the left flank of the R element. R-element constructs containing small <105-bp deletions at the left boundary of the micronucleus-limited region were transformed into conjugating Tetrahymena cells. Southern blot hybridization analysis used to assess the rearrangement of various R-element constructs is shown at the top. Plasmid DNA (P) and DNA isolated from transformants (T) were digested with AccI and NotI (312L/900r, 31L:+3, 31L::+24, 62L:1L, and 62L:+24) or NotI alone (76L::+24, 101L:+3, and 101L:61L) prior to electrophoresis and transfer to nitrocellulose membranes. The probes were the same as used for Fig. 2. A longer exposure for some lanes is given to allow visualization of DNA fragments in low abundance. Positions of PstI-digested lambda DNA size standards are shown to the right. Arrowheads indicate the fragments that hybridized to the oligonucleotide probe that detects the major chromosomal junction sequence. The region deleted from each construct is indicated above each set of lanes. A diagram showing an enlargement of the left flanking region of each construct is given at the bottom. Nucleotide positions of the deletion endpoints are indicated above the enlargement. The 6-bp ApaI site, GGGCCC, was inserted in place of the sequences removed. The efficiency of rearrangement relative to the activity of an intact R element is given to the right of each diagram: Normal, normal activity; , reduced activity; and  , greatly reduced activity.

Sequences in the left macronucleus-destined region control the position of the left deletion boundary. The M-element A₅G₅ polypurine tract has been shown to specify the deletion boundaries at a distance 41 to 54 bp 3' of the proximal G nucleotide (19). If the region flanking the R element also serves to determine the rearrangement boundary, then the small deletions (Fig. 3) effectively move this cis-acting control region closer to the micronucleus-limited region and therefore should produce a rightward shift of the junction formed by rearrangement. Alternatively, if the boundary is specified by sequences present within the micronucleus-limited region, removal of flanking sequences should not change the position of the rearrangement junction. To examine this issue, we characterized junctions created by rearrangement of four of these modified R elements (Fig. 4). We PCR amplified the rearranged elements from two to four independent transformants derived from each construct, cloned the amplified products, and determined the sequence spanning the junction. For each construct, the right side of the junction was formed at one of two positions, 1r or 4r (Fig. 4). These two positions are the same right-end boundaries that constitute the major and minor junction sites, respectively, resulting from the rearrangement of chromosomal R elements (4).

View larger version (29K): [in this window] [in a new window]   FIG. 4.   The left rearrangement boundary shifts rightward into the micronucleus-limited region a distance corresponding to the length of sequence removed from the left flanking DNA. The R element including an enlargement of the left end is shown schematically at the top. The narrow and wide boxes represent the micronucleus-limited and macronucleus-destined flanking regions, respectively, and are shaded as in prior figures. The name of each construct is indicated at the left edge of each schematic. Shaded arrowheads indicate positions of the right rearrangement boundaries; open arrowheads indicate positions of the left rearrangement boundaries. Each bar represents an independently rearranged element with its left boundary observed to be at the position indicated by the associated arrowhead. The number above each arrowhead denotes the position of the boundary relative to the start of the micronucleus-limited region as described in Fig. 1.

View larger version (31K): [in this window] [in a new window]   FIG. 5.   The rearrangement boundary and the flanking region remain in relative juxtaposition in constructs containing sequence insertions at their left ends. DNA fragments of 48, 342, and 517 bp were inserted into the ApaI site of deletion construct 31L:+3, and the rearrangement of each construct was determined by Southern blot hybridization of NotI-digested plasmid (P) or transformant (T) DNA. The probes were the same as used for Fig. 2. The positions of HindIII-digested lambda DNA size standards are shown to the right. A diagram of the constructs is given at the bottom. The solid boxes represent the micronucleus-limited sequence; the wider open and shaded boxes represent the macronucleus-destined flanking regions; the wide lines represent the inserted sequence. Positions of the rearrangement boundaries on the right are indicated by the shaded arrowheads and were found to be at the same position, 1r or 4r, for each junction analyzed for a given construct. Each shaded arrowhead indicates the position of the left rearrangement boundary; each bar represents an independently rearranged element with its left boundary observed to be at the position indicated by the associated arrowhead. The nucleotide positions given above each arrowhead denote the observed distance of each left boundary from the 32L position. Eleven of twelve left side boundaries are within a 14-bp region right of this position.

The flanking regulatory region appears to span an extended sequence. The R-element flanking region appears to control the rearrangement boundary much like the M-element A₅G₅; nevertheless, the actual sequence must be quite different. An A₅G₅-like motif does not exist in this region, nor does inspection of the sequence reveal any obvious candidate for a simple cis-acting sequence motif. Our deletion analysis above indicated that cis-acting sequences were present within the 70-bp region between 32L and 101L, but it could not determine whether the important sequences consisted of one or more simple motifs or, alternatively, a single, rather extended sequence. For instance, we could account for our data by invoking the presence of a somewhat simple motif that spans the 63L position since this region is altered or removed by all deletions that show reduced rearrangement activity. To determine whether any particular sequence within the flanking regulatory region is critical to its function, we introduced clusters of point mutations in 10-bp blocks throughout the region between 32L and 95L. The rearrangement activities of five such RPM constructs were examined. The introduced sequence changes (Fig. 6) are as follows: construct RPM1, six nucleotide changes between 33L and 42L; RPM2, six changes between 46L and 55L; RPM3, six changes between 58L and 67L; RPM4, six nucleotide changes between 70L and 79L; and RPM5, five changes between 86L and 95L. Southern blot analysis of their rearrangement activities is also shown in Fig. 6. We expected that mutation of a simple cis-acting sequence would reduce rearrangement activity similarly as was observed upon removal of these sequences (Fig. 3). However, we found that all five of these altered R elements displayed normal rearrangement activity. Hybridization with an oligonucleotide probe that is specific for the major chromosomal deletion junction showed that the rearrangement of each construct was accurate (data not shown). These results argue against the existence of a simple sequence in this region regulating rearrangement. In particular, normal rearrangement of the RPM3 construct, which contains six changes between 58L and 67L, argues against the existence of a simple regulating sequence spanning the 63L position. Because we changed only about half (29 of 63) of the nucleotides between 32L and 95L, it is possible that we failed to mutate key nucleotide positions. However, the sequences that we did not alter represent mostly blocks of A and T nucleotides that are abundant throughout this region and thus unlikely to contain such a specific signal. We believe that these data indicate that the nature of the cis-acting sequence is somewhat complex and tolerant of mutation.

View larger version (53K): [in this window] [in a new window]   FIG. 6.   Clustered mutations in the left flanking region have little effect on rearrangement efficiency or accuracy. Each R-element construct contains five or six changes within 10-bp blocks of the left flanking region between 32L and 95L. Southern blot hybridization analysis of two pools, A and B, each containing DNA from >10 transformants was used to assess the ability of each construct to undergo accurate rearrangement. The plasmids, RPM1 to RPM5, used for transformation are indicated above the pair of lanes. DNA was digested with BamHI prior to electrophoresis and transfer to a nitrocellulose membrane. Lane P is plasmid RPM1 digested with BamHI. The positions of PstI-digested lambda DNA size standards are shown to the right. A diagram of the R element and an enlargement of the left flanking sequence between 30L and 100L are shown at the bottom. The clusters of mutations that create each construct, boxed and labeled with their corresponding construct names, are shown in bold beneath this sequence.

The right-side flanking control region can substitute for the sequences flanking the left end of the micronucleus-limited region. All data collected thus far indicate that sequences in the left-end R-element flanking region control the rearrangement boundary from a distance like the M-element A₅G₅. An additional property of the A₅G₅ sequence is that two copies, one outside each boundary, are required for efficient rearrangement. Even the boundary-controlling sequence flanking the R element cannot effectively substitute for one of the A₅G₅ sequences that is located outside the M element (20). The two R-element flanking regions share no obvious sequence identity. Nevertheless, if the R-element flanking regulatory sequences are truly analogous to the A₅G₅ sequence, then the same functional sequence found flanking the left side should also be contained outside the right end. To address this issue, we created three chimeric R-element constructs each containing sequences originally from the right flanking region immediately flanking both right and left sides of the R element. Diagrams of these constructs are shown in Fig. 7. The first two constructs replaced the entire left side with either the first 135 bp or first 334 bp from the right side of the R element (LF1r:135r and LF1r:334r, respectively). The third construct, LF1r:135r:312L, contained the first 135 bp from the right side in place of the first 100 bp of the left side but retained the left side flanking sequences between 101L and 312L.

View larger version (31K): [in this window] [in a new window]   FIG. 7.   The right flanking sequence effectively substitutes for the essential left flanking region. R-element constructs containing substitutions of the sequence immediately flanking the left end of the micronucleus-limited region with sequence flanking the right end were transformed into conjugating Tetrahymena cells. Southern blot hybridization analysis used to assess the rearrangement of each construct is shown at the top. Plasmid DNA (P) and DNA isolated from transformants (T) was digested with BamHI prior to electrophoresis and transfer to the nitrocellulose membrane. The positions of HindIII-digested lambda DNA size standards are shown to the right. The sequences flanking the left end of the eliminated region are indicated above each set of lanes. A diagram of the flank substitution constructs is shown at the bottom. In each schematic, the narrow black box represents the micronucleus-limited sequences. The wide white and shaded boxes represent the sequences flanking the eliminated region to the left and right, respectively.

A wide variety of sequences at the boundaries can participate in rearrangement. In the study above, we collected a large number of novel deletion junction sequences (Table 2). As most of our constructs were altered on their left sides, most of these junctions retained the normal right rearrangement boundary but utilized a unique left boundary. We aligned these unique boundary sites and examined the sequences for nucleotide preferences at the eight positions on each side of the junction. There appears to be no sequence bias at these 16 positions. Previously, Saveliev and Cox (34, 35) found that DNA breaks that occur during conjugation at the ends of chromosomal M and R elements have a common structure. These breaks all had four-nucleotide 5' extensions, and 11 of 12 breaks terminated with a 3' adenosine residue on the recessed strand. Based on this observation, the 3' adenosine was proposed to be an important sequence feature at the rearrangement boundary. Many of the novel boundaries identified in this study have an appropriately positioned adenosine to fit this proposed consensus, but at least one-third do not. Therefore, a 3' adenosine appears not to be required at each boundary, although it may be a preferred structure that has been selected for at chromosomal boundaries. We must note that we can only infer breakage sites from our junction sequences since the mechanism of breakage and joining is still largely unknown. Nevertheless, we find that the rearrangement boundary is highly permissive of sequence variation.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The A₅G₅ polypurine tracts, which are located ~45 bp outside each end of the M element, were shown previously to specify the positions of the rearrangement boundaries (19, 20). Prior to this study, A₅G₅ was the only characterized cis-acting sequence known to control deletion element rearrangement. No identical or similar sequence had been found flanking the seven other characterized deletion elements, which raised the possibility that the M element utilizes a unique mechanism of rearrangement. Yet, it seems unlikely that the thousands of deletion elements are eliminated by different rearrangement systems. In this study, we have shown that the sequences located outside the micronucleus-limited region of the R element are required for accurate and efficient rearrangement. Upon altering the spacing between these sequences and the micronucleus-limited region, we found that the rearrangement boundary always shifted to remain in relative juxtaposition with the flanking region. This was true even when the flanking region was several hundred base pairs removed from the micronucleus-limited region. Furthermore, sequences from the right side flanking region effectively substituted for essential sequences on the left side, indicating that the two flanking regions of the R element are interchangeable despite lacking obvious sequence identity. These data demonstrate that the R-element flanking regions contain sequences that function similarly to the M-element A₅G₅ tracts in that they serve to position the deletion boundaries at a specific distance. We believe that this study suggests that many, and perhaps most, Tetrahymena deletion elements use flanking regulatory sequences to specify the location of rearrangement boundaries.

Although the R-element flanking sequences perform the same function as the A₅G₅ tract, the nature of the sequence recognition, at least superficially, seems different. Mutations within the G₅ portion of the M element are sufficient to block rearrangement (20). In contrast, we were able to localize the essential regulatory sequences on the left side of the R element only to within a 70-bp region (32L to 101L). Clusters of point mutations introduced throughout this region did not reduce rearrangement activity (Fig. 6). It has been proposed that the M and R elements both belong to a class of deletion elements that have a common rearrangement mechanism, but that they are members of different families that are distinguished by their use of particular flanking regulatory sequences (19). Our finding that the M- and R-element flanking regions contain different yet functionally equivalent sequences provides the first direct support for this view. The apparent complexity of the R-element flanking regulatory region helps us to reconcile with the difficulty of identifying common controlling sequences among the different known elements.

Aside from the A₅G₅ tract, the flanking regulatory regions of the M element may be quite similar to those of other elements. Like our modified R elements, M-element constructs that have less than 100 bp of flanking sequence displayed reduced rearrangement activity (20). In fact, the A₅G₅-containing minimal-flanks construct (which has 65 bp flanking the left boundary and 70 bp flanking the right) rearranges rather poorly (20). The rearrangement efficiency of this construct is probably not much different from that of the R element 312L/70r construct (Fig. 2). Even though the A₅G₅ tract is the primary boundary determinant, additional flanking sequences appear to greatly facilitate the use of this signal. In light of the fact that the 101L:61L construct produces a wild-type junction (Fig. 3), the R element may also contain a sufficient boundary determinant within the first 60 bp, a location very similar to that of the M-element A₅G₅. However, this accurate junction is a minor product of rearrangement for this construct, indicating that these sequences function poorly without the flanking DNA between 60L and 100L. A third deletion element, mse 2.9, also requires >66 bp of flanking DNA in order to accurately rearrange (29). These observations may imply that the flanking regulatory sequences are bipartite, containing some sequences that primarily determine accuracy and others that enhance efficiency of rearrangement.

We found that the R-element construct with only 100 bp of the left flanking sequence displays normal rearrangement efficiency. Following this observation, we were quite intrigued that a construct (101L:2L) completely lacking these sequences also displays appreciable rearrangement activity. It appears that flanking sequences beyond the first 100 bp contain some functional redundancy with the flanking region proximal to the micronucleus-limited region. Since we did not identify an obligatory role for these distal sequences in normal rearrangement, it is unclear as to whether these sequences normally contribute to efficient rearrangement activity or, alternatively, that removal of the primary regulatory sequences revealed a cryptic controlling sequence.

In this study, we did not address whether sequences within the micronucleus-limited region are required for R-element rearrangement. An M element lacking roughly half of its micronucleus-limited region displays reduced rearrangement (20). It appears that both M and R elements contain similar cis-acting sequences within their micronucleus-limited regions that function independently of orientation and position (44). The role of these internal sequences is otherwise unknown. It is clear from this study of the R element and previous studies of the M element (19, 20) that the flanking regulatory sequences are sufficient to position the boundaries. We have proposed that the internal sequences serve to target an element for elimination and rely on the flanking sequences to limit the extent of deletion (14, 43). The interaction between the functions of these two different types of cis-acting sequences remains to be determined.

We believe that using flanking regulatory sequences to position the boundary offers distinct advantages for accurate rearrangement. The pairwise use of identical, orientation-dependent regulatory sequences is likely to increase specificity and limit aberrant rearrangement. Pairing of recombination signals appears to occur prior to V(D)J recombination (23). It is compelling to speculate that Tetrahymena rearrangement requires pairing of flanking regulatory sequences, and such pairing is a common strategy used to ensure specificity of a variety of DNA deletion events. Based on the analysis of M-R-element chimeras, a boundary determined by an A₅G₅ tract does not effectively join to a boundary specified by an R-element flanking regulatory sequence (20). The use of different sequences to control the rearrangement of adjacent elements would greatly minimize aberrant recombination between elements. In addition, the use of external sequences to limit deletion appears to be an efficient way to eliminate elements of widely varying size. Increasing the size of the R element by 50% (the 31L:+3 i517 construct) did not affect the accuracy of rearrangement (Fig. 5). This feature of the mechanism may in part account for the large variability in size and sequence of the Tetrahymena deletion elements.

The finding that DNA breaks at the ends of chromosomal M and R elements have a common structure (34, 35) provides further evidence that these elements share a rearrangement mechanism. The breaks identified are consistent with a mechanism that is similar to transposition processes. At least some internal eliminated sequences of hypotrichous ciliates appear to be related to transposable elements (15, 24, 26). In contrast, the Tetrahymena deletion elements lack most of the structural characteristics normally associated with transposable elements (reviewed in reference 14). The use of sequences exclusively outside the element to position the sites of excision is not known to occur for any transposon-like process. Still, it is logical to think that developmentally programmed DNA rearrangements found in different ciliate species have a common origin, although we lack the knowledge to clearly recognize the connections. Our identification of a probable common mechanism controlling deletion element rearrangement is an important step toward understanding the regulation and origin of developmentally programmed genome reorganization.