The Microsatellite Sequence (CT)_n · (GA)_n Promotes Stable Chromosomal Integration of Large Tandem Arrays of Functional Human U2 Small Nuclear RNA Genes

Abstract

The multigene family encoding human U2 small nuclear RNA (snRNA) is organized as a single large tandem array containing 5 to 25 copies of a 6.1-kb repeat unit (the RNU2 locus). Remarkably, each of the repeat units within an individual U2 tandem array appears to be identical except for an irregular dinucleotide tract, known as the CT microsatellite, which exhibits minor length and sequence polymorphism. Using a somatic cell genetic assay, we previously noticed that the CT microsatellite appeared to stabilize artificial tandem arrays of U2 snRNA genes. We now demonstrate that the CT microsatellite is required to establish large tandem arrays of transcriptionally active U2 genes, increasing both the average and maximum size of the resulting arrays. In contrast, the CT microsatellite has no effect on the average or maximal size of artificial arrays containing transcriptionally inactive U2 genes that lack key promoter elements. Our data reinforce the connection between recombination and transcription. Active U2 transcription interferes with establishment or maintenance of the U2 tandem array, and the CT microsatellite opposes these effects, perhaps by binding GAGA or GAGA-related factors which alter local chromatin structure. We speculate that the mechanisms responsible for maintenance of tandem arrays containing active promoters may differ from those that maintain tandem arrays of transcriptionally inactive sequences.

In primates, the genes encoding U2 small nuclear RNA (U2 snRNA) are organized as a nearly perfect tandem array at a single chromosomal locus designated RNU2 (40, 49, 58, 64). The number of repeat units within an array can range from 5 to 25, but this number is heritable and somatically stable (33, 49). In humans, the 6.1-kb U2 repeat unit contains the 188-bp U2 snRNA coding region, a solo retroviral long terminal repeat, two Alu elements, a 3′ truncated L1 LINE element, and an imperfect d(CT)_n · d(GA)_n microsatellite sequence (n ≈ 70) located about 300 bp downstream of the U2 coding region (49). Of these identified sequence elements, only the U2 snRNA gene is known to have a function. Yet every copy of the 6.1-kb repeat unit within a single tandem array is essentially identical. Thus, the repeat units within an array are subject to concerted evolution; all the repeats evolve in concert, most probably as the result of gene conversion (34;; also see references 19 and 48). The only exception is the CT microsatellite, which exhibits minor length and sequence polymorphism, perhaps because microsatellite variants arise faster than gene conversion can purge or fix them throughout the array (33).

Maintenance of sequence homogeneity could be an intrinsic property of all tandem arrays, or it could reflect the presence of recombinogenic elements within the repeat unit which enhance gene conversion or stabilize the arrays. One candidate recombinogenic element is the CT microsatellite. The CT microsatellite (n ≈ 70) in the tandemly repeated U2 genes is located 0.3 kb downstream from the U2 coding region (28, 49); a similar CT microsatellite (n ≈ 15 to 25) is also located about 1.8 kb downstream from the U1 snRNA coding region in the tandemly repeated human U1 genes (27, 38); a GT microsatellite (n = 36) is located 1.4 kb downstream (0.3 kb upstream) from the 5S rRNA transcription unit (35, 51); and a GT microsatellite (n = 29) is found downstream from the putative transcription terminator for 45S rRNA (13). The CT microsatellite in the U2 snRNA repeat units is well conserved from baboons to humans, despite some length polymorphism (33). The CT microsatellites in the human U1 and U2 genes are sensitive to cleavage by S1 nuclease in relaxed DNA at acidic pH (27, 28) and in negatively supercoiled DNA over a pH range from 4 to 7 (28). This suggests that similar structures, formed under physiological conditions, might be susceptible to single-strand invasion. CT microsatellites can also form triple helical structures known as H-DNA, at least at pH 4.5, providing further support for a possible role in recombination (17, 29, 41). Finally, many simple repeats appear to be free of nucleosomes, preferentially exposing them to the recombination machinery (15, 17, 37, 57). Taken together, these data suggest that the CT microsatellite might be a cis-acting element involved in the formation and/or maintenance of tandem arrays.

In previous work on the mechanism of adenovirus type 12-induced fragility of the human RNU2 locus (5), we generated cell lines containing artificial tandem arrays of either the intact 6.1-kb U2 repeat unit (iU2), a minimal 0.8-kb U2 snRNA gene (U2 minigene or mU2), or a U2 minigene from which one or both key promoter elements had been deleted (mU2ΔDSE and mU2ΔDSEΔPSE). Surprisingly, although we readily isolated cell lines containing large artificial arrays (>20 repeat units) of the intact repeat unit (iU2) and the transcriptionally inactive U2 minigene (mU2ΔDSEΔPSE), we were unable to isolate large arrays of the transcriptionally active U2 minigene (mU2). These observations suggested (albeit only anecdotally) that some element which is present in the intact U2 repeat unit, but missing from the U2 minigene, can overcome a bias or selection against introduction or maintenance of a large tandem array of active U2 genes. We now show that the presence of the CT microsatellite within each repeat unit increases both the average and maximum size of the artificial U2 tandem arrays, but only when the repeat unit contains an active U2 promoter. Our data suggest that specialized functional elements are required for the stability or maintenance of tandem arrays containing active transcription units.

MATERIALS AND METHODS

DNA constructs.

The E1 selection cassette contains the mouse dihydrofolate reductase gene driven by the simian virus 40 early promoter and the neomycin resistance gene driven by the herpes simplex virus thymidine kinase gene promoter. The E1, mU2, and mU2ΔDSEΔPSE constructs have been described previously (5). A 700-bp AccI fragment containing the entire CT microsatellite (33) was excised from the iU2 construct, subcloned into the SmaI site of pUC18Bgl, and excised with BglII and BamHI. This BglII/BamHI fragment spans the entire AccI fragment and was used in the sixth array ligation described below; the same BglII/BamHI fragment was also cloned into the BamHI site of mU2 to generate the mU2+CT construct and into the BamHI site of mU2ΔDSEΔPSE to generate the mU2ΔDSEΔPSE+CT construct. In all constructs, the U2 snRNA genes were marked with an innocuous U87C point mutation. This mutation in a phylogenetically variable nucleotide allowed us to use a primer extension assay to monitor steady-state levels of U2 snRNA derived from the transgenes relative to the background of resident U2 snRNA (5).

Ligation of artificial arrays.

Artificial tandem arrays were generated by ligation of BamHI/BglII fragments excised from the plasmid constructs described above (Fig. 1). After digestion of the appropriate plasmid with BamHI and BglII, fragments were resolved by preparative, low-melting-temperature agarose gel electrophoresis. Gel slices containing the fragments were excised, melted at 65°C, and then incubated at 40°C for 1 h with 1 U of β-agarase for each 100 μl of agarose gel in buffer provided by the supplier (New England Biolabs). The DNA was phenol extracted, and agarose oligomers were removed by the LiCl precipitation method (16). To ensure one E1 selection cassette per artificial array, BamHI/BglII fragments bearing the U2 gene and the E1 selection cassette were ligated at a mass ratio of 100 to 1 (5). Six ligations were performed: (i) 6.0 μg of mU2 plus 0.06 μg of E1, (ii) 6.0 μg of mU2+CT plus 0.06 μg of E1, (iii) 6.0 μg of mU2ΔDSEΔPSE plus 0.06 μg of E1, (iv) 6.0 μg of mU2ΔDSEΔPSE+CT plus 0.06 μg of E1, (v) 6.0 μg of mU2 plus 1.1 μg of mU2+CT plus 0.06 μg of E1, and (vi) 6.0 μg of mU2 plus 0.5 μg of CT plus 0.06 μg of E1. All ligations were carried out in two steps: DNA at 1 mg/ml was incubated for 4 h at room temperature with 100 U of T4 DNA ligase (Boehringer Mannheim) per ml in buffer provided by the manufacturer, to which 50 mM spermidine had been added to promote DNA condensation (5, 8); the volume was then doubled, 100 U of additional ligase per ml was added, and the reaction was continued overnight. Dilution was necessary to ensure that the amount of glycerol, introduced with the enzyme, did not exceed 10%; dilution also reduced the viscosity, which increased perceptibly during the first ligation step.

FIG. 1.

Constructs and ligation conditions. (A) An intact U2 repeat unit and the five fragments used to construct artificial tandem arrays. Key restriction sites are shown (Bg, BglII; K, KpnI; Ea, EagI; A, AseI; N, NdeI; Hc, HincII; Sm, SmaI; D, DraI; RI, EcoRI; Bm, BamHI; Sf, SfuI; F, FokI). The 5′ and 3′ ends of each construct are BglII and BamHI, respectively. Sites in parentheses were destroyed by ligation. Open arrow, U2 snRNA gene; asterisk, U87C mutation; open circle and square, DSE and PSE, respectively; hatched rectangle, d(CT)_n · d(GA)_n or CT microsatellite, where n ≈ 75; shaded rectangle, truncated L1 repeat. LTR, long terminal repeat. (B) Representative ligation reactions. The mU2 and mU2+CT constructs were subjected to the two-step ligation as described in Materials and Methods, and the products were separated by agarose FIGE under conditions that resolve both large and small fragments. FIGE gels were then blotted and probed with the labeled mU2+CT construct. Either 1.0 (left) or 0.02 (right) μg of each ligation reaction was loaded per lane. As an internal control, 100 ng of the mU2 fragment was loaded to the right of the mU2 lane in each of these panels. The residual monomer and dimer are indicated; these species account for <30% of the input fragments compared to DNA mass markers (data not shown). Size markers are shown in kilobases (center).

Cell culture.

The human HT1080 fibrosarcoma line was grown in minimum essential medium (Gibco/BRL) with 10% fetal bovine serum under 5% CO₂. One day prior to transfection, 10⁶ cells were plated per 100-mm dish. The medium was completely replaced 3 h before transfection, and the cells were then transfected with complete ligation reaction mixtures but without carrier by the calcium phosphate precipitation method using a commercial kit (Gibco/BRL). We used the calcium phosphate method throughout, because electroporation was found in pilot experiments to select against large artificial arrays (5). The calcium phosphate-containing medium was replaced with fresh medium 18 h later. After 24 h of growth, the cells were split into 10 150-mm dishes and challenged with 600 μg of G418 (Geneticin; Gibco/BRL) per ml. Selection continued for 17 days, with fresh medium and G418 every 48 h. For all transfections, the number of colonies per dish ranged from 36 to 68. Well-separated colonies were collared with sterile glass cylinders, trypsinized, and transferred to 48-well microtiter trays. As each colony approached confluency in the microtiter well, it was trypsinized, transferred to a 60-mm dish, and grown until confluent (about 3 × 10⁶ cells). Approximately 80 to 90% of isolated colonies survived transfer to microtiter trays. For each cell line, half the cells were used to make genomic DNA agarose plugs, each containing about 10 μg of DNA, and half were slowly frozen in 50 μl of complete medium containing 10% dimethyl sulfoxide as a stock for future use. No significant difference in the yield of neomycin-resistant colonies was observed for transfection with any of the six ligation reactions (see above).

Characterization of cell lines.

Genomic plugs were prepared and digested as described elsewhere (5). The number of artificial arrays and the average number of repeats per artificial array were determined by quantitative genomic blotting. For each cell line, one-third of a genomic plug (about 3 μg of DNA) was digested with a restriction endonuclease that cut both the resident and artificial tandem arrays at least once per repeat unit but produced fragments of different lengths. In this way, the signal derived from artificial repeats was easily and reproducibly normalized to that for the resident U2 snRNA genes. SfuI was used to digest the mU2 (ligation 1), mU2+CT (ligation 2), mU2 and mU2+CT (ligation 5), and mU2 and CT (ligation 6) cell lines; SmaI was used for the mU2ΔDSEΔPSE (ligation 3) line; and DraI was used for the mU2ΔDSEΔPSE+CT (ligation 4) line. Genomic digests were resolved by 1% agarose gel electrophoresis, quantitatively blotted onto nylon membrane (Zetabind; CUNO) by depurination followed by alkaline transfer (59), probed with the labeled mU2 repeat unit, and quantitated by phosphorimager analysis (5). The mU2 fragment was radiolabeled to a high specific activity with [α-³²P]dATP, random hexamer primers, and the Klenow fragment of DNA polymerase I (Boehringer Mannheim). For cell lines containing >20 artificial repeats, the number of artificial arrays was determined by digesting another third of the genomic plug with BamHI, which does not cut either the resident or the artificial arrays and therefore excises both resident and artificial U2 arrays intact. As discussed in Results, failure of BamHI to cut artificial arrays confirms the absence of tail-to-tail junctions. The BamHI digests were resolved by field inversion gel electrophoresis (FIGE), and the dried agarose gel was probed with the mU2 repeat unit as described elsewhere (49).

Expression of marked U2 genes.

RNA was prepared by Nonidet P-40 lysis, followed by phenol extraction and ethanol precipitation (24). Quantitation of the marked U2-U87C snRNA by primer extension was performed essentially as described elsewhere (66), except that the 5′-end-labeled primer was complementary to U2 nucleotides 89 to 109, and ddATP was substituted for dATP. Primer extension products were resolved by 15% denaturing polyacrylamide gel electrophoresis, and the relative intensities of the signals were determined with a GS-250 Molecular Imager (Bio-Rad Laboratories) and the program Molecular Analyst 2.0.

Analysis of CT microsatellites.

Resident and marked U2 repeats were resolved as described for FIGE above, and the dried gel was hybridized with the labeled mU2 probe. After autoradiography, regions of the dried agarose gel corresponding to resident and marked U2 fragments were excised and then melted for 15 min at 100°C in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0). Melted agarose was used directly for PCR as previously described (33). Primers were U2CT1 (5′-TAGGATCTCAGCTTGGCAGT-3′) and U2CT2 (5′-TAGACGACTGGTGGATAGGT-3′). The amount of template RNU2 DNA in the melted gel slice was estimated by determining the volume needed to give the same amount of cold PCR product as 250 ng of total genomic DNA. This volume was then used as template for a labeled PCR. Reactions were carried out in 25 μl of buffer (10 mM Tris-HCl [pH 8.3]; 50 mM KCl; 1.5 mM MgCl₂; 200 μM [each] dATP, dCTP, dGTP, and dTTP) containing 0.75 μM (each) primer, 1.2 U of Taq DNA polymerase (Perkin-Elmer), and 2 μCi of [α-³²P]dCTP (3,000 Ci/mmol). The PCR protocol was denaturation at 94°C for 3 min, followed by 30 cycles of amplification (30 s at 94°C, 30 s at 55°C, and 1 min at 72°C), and 7 min at 72°C. Samples were then transferred to a 65°C water bath, diluted with 75 μl of TE buffer, and extracted twice with phenol. The aqueous supernatant was made 0.4 M in LiCl by addition of a 4 M solution and then incubated for 5 min on ice. Residual phenol and agarose oligomers were removed by pelleting at 15,000 rpm in a microcentrifuge. DNA was precipitated with ethanol, washed with 70% ethanol, and dried. After resuspension in TE buffer, the DNA was digested with MspI and MnlI and then extracted and precipitated. The resulting DNA digests were resolved by denaturing 6% polyacrylamide gel electrophoresis alongside a sequencing ladder generated with an M13 template. As controls, the entire procedure was also performed on total genomic DNA and appropriate plasmid clones, except that the DNA was not prepared by agarose gel electrophoresis.

RESULTS

As described previously (5) and detailed in Materials and Methods, all artificial arrays were generated by in vitro ligation of BamHI/BglII fragments, followed by transfection into HT1080 cells by the calcium phosphate precipitation technique. The calcium phosphate method was used because electroporation apparently imposed an upper limit on the sizes of the arrays obtained, perhaps reflecting the limited size or negative surface charge of transient plasma membrane pores generated by capacitance discharge (25). The original purpose of using BamHI/BglII fragments was to digest the ligated arrays with BamHI and BglII before transfection, thus eliminating all head-to-head and tail-to-tail junctions between repeat units. However, we discovered that double digestion was unnecessary and possibly counterproductive; the cells themselves efficiently convert mixed arrays (containing head-to-head, head-to-tail, and tail-to-tail repeat units) into pure head-to-tail arrays (5) (also see Fig. 4 below).

FIG. 4.

Artificial tandem arrays are integrated at a single site. Genomic DNA from representative mU2+CT cell lines was digested with BamHI, and the fragments were resolved by agarose FIGE. All tandem arrays are excised intact; BamHI does not cut the resident U2 repeat unit, nor the artificial arrays which are entirely free of tail-to-tail junctions (see Fig. 1 and text). The two resident U2 arrays are indicated; the artificial arrays (marked by asterisks) vary in size, but only one is seen in each cell line. Note that the average size of a genomic BamHI fragment is roughly 8 to 10 kb; thus, the intensity of an artificial array correlates only with the number of exogenous U2 repeats (as determined in Fig. 3) and not with the size of the artificial array excised by BamHI. Fragment sizes (kilobases) were estimated relative to Midrange Markers I (Gibco/BRL). The intense signal below 30 kb reflects a weak cross-reaction of the probe with bulk genomic DNA, possibly with abundant U2 retropseudogenes (58).

We were acutely aware that the distribution of tandem arrays might be influenced by the precise concentration of DNA in the in vitro ligation, by the activity of the DNA ligase, or by damage to the ends of the input DNA fragments from contaminating phosphatase or exonuclease activity. Strenuous efforts were therefore made to perform all key steps of the protocols in parallel and to ensure that the results obtained with different constructs would be comparable. For example, in vitro ligations of the mU2 and mU2+CT constructs are shown in Fig. 1B; a low level of monomer and oligomers can be seen, together with a range of multimers from approximately 15 to 90 repeats. However, the transfected cells themselves are capable of efficient recombination and ligation; not only are mixed arrays efficiently converted into head-to-tail arrays in vivo (see above), but empty sites (a neomycin resistance marker without accompanying U2 sequences) are rarely found for the mU2+CT construct, although significant levels of monomer are present in the corresponding ligation products (Fig. 1B). Thus, the ability of the transfected cells to ligate and recombine the input DNA may compensate for unavoidable minor differences in the efficiency of in vitro ligation.

The CT microsatellite affects recovery of tandem arrays containing an active U2 gene.

Two experiments were performed in parallel to ask whether the intact 6.1-kb U2 repeat unit contained elements that facilitate recovery of large artificial U2 tandem arrays.

(i) Tandem arrays of the mU2 construct. Ninety cell lines were isolated from HT1080 cells transfected with the ligated mU2 minigene construct. The copy number of the exogenous mU2 genes in each cell line was determined by genomic blotting with SfuI (see Fig. 2A for a representative blot) followed by phosphorimager analysis to compare the intensity of the exogenous 0.8-kb mU2 fragment with that of the resident 6.1-kb U2 fragment (Fig. 3A). More than half of the cell lines (49 of 90, or 54%) had no detectable mU2 repeats; another third (30 of 90, or 33%) had one or two mU2 repeats; a tenth (9 of 90, or 10%) had three, four, or five mU2 repeats; and two cell lines had 14 and 17 mU2 repeats. Overall, the average length of the mU2 arrays was 1.2 repeat units. These data tally well with our impressions from earlier, nonqualitative experiments (5). The results also raise the question of why it is difficult to obtain cell lines with tandem arrays containing more than six repeats of the U2 minigene. We know that U2 genes are subject to dosage compensation (5a, 39), so one possibility is that epigenetic or transcriptional regulatory mechanisms might inhibit insertion or maintenance of large arrays of active U2 genes.

FIG. 2.

Characterization of artificial arrays by genomic blotting. Only representative blots are shown. The probe was the labeled mU2 repeat unit in every case; signal strength was corrected for the absence of PSE and DSE sequences in mU2ΔDSEΔPSE lines. Exogenous and resident U2 gene fragments are indicated (in kilobases). (A) mU2 cell lines, SfuI digest; (B) mU2+CT cell lines, SfuI digest; (C) mU2ΔDSEΔPSE cell lines, SmaI digest; (D) mU2ΔDSEΔPSE+CT cell lines, DraI digest; (E) mixed mU2 and mU2+CT cell lines, SfuI digest; (F) mixed mU2 and CT cell lines, SfuI digest.

FIG. 3.

Distribution of cell lines by number of exogenous marked U2 genes. For each cell line, the ratio of marked U2 genes to resident U2 genes was determined by phosphorimager analysis of genomic blots, and the number of marked U2 genes was calculated by assuming 22 resident U2 genes in pseudodiploid HT1080 cells (49). In representative cases, pseudodiploidy for chromosome 17 was confirmed by fluorescent in situ hybridization. For each transfection, a tally was made of the number of cell lines having n marked U2 genes where n ranged from 0 to 105. The tally data were plotted as the number of cell lines with n marked U2 genes versus n. Insets in each panel indicate the average n and the percentages of empty sites containing the neomycin resistance marker but no detectable U2 genes. (A) mU2 transfection; (B) mU2+CT transfection; (C) mU2ΔDSEΔPSE transfection; (D) mU2ΔDSEΔPSE+CT transfection; (E) mixed mU2 and mU2+CT transfection; (F) mixed mU2 and CT transfection.

(ii) Tandem arrays of the mU2+CT construct.

Unlike the mU2 minigene, the intact U2 repeat can form exogenous arrays of at least 80 repeats (5). Since we suspected that the presence of a CT microsatellite in the intact U2 repeat might allow the formation of such large arrays, we transfected HT1080 cells with the ligated mU2+CT construct, generating cell lines in which each repeat unit contained a U2 minigene followed by the natural CT microsatellite. Although the CT microsatellite-containing fragment also contained an internal fragment of L1 sequence (Fig. 1A), this seemed unlikely to be biologically active; the fragment is short (483 bp compared to the active 6-kbp L1 element) and atypically truncated at both the 5′ and 3′ ends and has undergone many secondary mutations including loss of flanking direct repeats and the poly(A) tract (49). Seventy-four cell lines were analyzed as described for the mU2 minigene lines above, except that the mU2+CT repeat unit yields a 1.5-kb SfuI fragment (see Fig. 2B for a representative blot). Unlike the mU2 cell lines, a broad distribution of mU2+CT array sizes was observed, ranging from 0 to 51 with an average of 8.2 repeats (Fig. 3B). Moreover, only 3% of the cell lines (2 of 74) failed to have any exogenous U2 genes, compared to 54% (49 of 90) for the mU2 construct. This distribution resembles that found for artificial arrays of the intact 6.1-kb U2 repeat unit (5). Clearly, the presence of the CT microsatellite has a profound effect on both the average and maximum size of the U2 tandem arrays that can be recovered in stable cell lines.

The CT microsatellite does not affect recovery of tandem arrays containing inactive U2 genes.

Two additional experiments were performed in parallel to ask whether the CT microsatellite works together with the U2 promoter to facilitate recovery of large artificial U2 tandem arrays.

(i) Tandem arrays of the mU2ΔDSEΔPSE construct. We were unable in previous experiments to obtain cell lines containing large tandem arrays (>8 repeats) of the active mU2 minigene, but we had no difficulty obtaining very large arrays (20 to 98 repeats) of the transcriptionally inactive mU2ΔDSEΔPSE promoter deletion construct (5). To generate quantitative data, we transfected HT1080 cells with the ligated mU2ΔDSEΔPSE construct, an mU2 minigene lacking all U2 promoter elements. Sixty-four cell lines were analyzed as described above, except that digestion with SmaI reduced the exogenous mU2ΔDSEΔPSE repeat units to 0.6 kb and the resident U2 genes to 1.5 kb (Fig. 2C). Over 42% of the resulting cell lines (27 of 64) failed to have a single mU2ΔDSEΔPSE repeat unit. In the remaining cell lines, a clear and significant shift toward arrays containing >2 repeats was apparent when the profiles of the mU2 and mU2ΔDSEΔPSE cell lines were compared (Fig. 3C). The average array size also increased to 5.4 repeats for mU2ΔDSEΔPSE arrays, compared to 1.3 repeats for mU2 arrays. As in previous experiments, we also obtained several cell lines with very large arrays ranging from 20 to 80 repeats. Thus, the average size of the mU2ΔDSEΔPSE arrays approaches that of mU2+CT (5.4 versus 8.2 repeats), but the number of empty sites containing only the Neo marker (42%) resembles that for mU2 (54%) more than that for mU2+CT (3%). We conclude that larger arrays can be obtained when the U2 promoter is deleted from the mU2 minigene, but the transcriptional activity of the mU2 minigene accounts for only some of the difficulty in obtaining large tandem arrays.

(ii) Tandem arrays of the mU2ΔDSEΔPSE+CT construct.

Addition of the CT microsatellite (mU2+CT [Fig. 2B]) had a more dramatic effect on the behavior of the mU2 minigene than deletion of the U2 promoter (mU2ΔDSEΔPSE [Fig. 2C]). To test whether the CT microsatellite might function simply as a recombination hot spot that enhances the ability of any repeat unit to form tandem arrays, we asked whether addition of the CT microsatellite could affect the distribution of tandem arrays formed by the promoter-deleted mU2ΔDSEΔPSE minigene construct. Ninety-two cell lines were isolated after transfection of HT1080 cells with the ligated mU2ΔDSEΔPSE+CT construct and were analyzed as described above; DraI reduced the exogenous U2 repeats to 1.3 kb and the resident U2 genes to 6.1 kb (see Fig. 2D for a representative genomic blot and Fig. 3D for phosphorimager analysis). Although the addition of the CT microsatellite enabled us to isolate larger arrays of the active mU2 minigene (compare Fig. 3A and B), addition of the CT microsatellite actually decreased the average size of arrays of the inactive mU2ΔDSEΔPSE minigene from 5.4 to 2.3 repeats (compare Fig. 3C and D). One view of these data would be that deletion of the U2 promoter reduces the ability of the CT microsatellite to favor larger arrays (compare Fig. 3B and D); another view would be that the CT microsatellite inhibits formation of tandem arrays of the inactive mU2 minigene (compare Fig. 3C and D). In any event, it is clear that the CT microsatellite can favor formation of larger tandem arrays only when combined with an active U2 gene.

Must the CT microsatellite be present in every repeat unit?

The preceding experiments demonstrated that the CT microsatellite does not function as a genetic recombination hot spot but works together with an active U2 snRNA gene to enhance formation, integration, or maintenance of large tandem arrays. We next asked whether the CT microsatellite must be present in every repeat unit of the tandem array or whether it can function when it is present in only a few interspersed repeats. For this purpose, arrays were made by in vitro ligation of a 10/1 molar ratio of the mU2 and mU2+CT constructs or the mU2 and the 700-bp CT microsatellite fragments.

(i) Mixed tandem arrays of the mU2 and mU2+CT constructs.

One hundred four cell lines were isolated from the mixed mU2 and mU2+CT transfection and analyzed. To distinguish mU2 repeat units from mU2+CT repeat units, genomic plugs were digested with SfuI, which cuts once in the U2 promoter but not in the CT microsatellite; the resident U2 genes are 6.1 kb, the mU2+CT minigene (mU2+CT) is 1.5 kb, and the U2 minigene (mU2) is 0.8 kb (see Fig. 2E for a representative blot and Fig. 3E for phosphorimager data). The number of exogenous U2 genes in each cell line was determined by summing the signal from the 0.8- and 1.5-kb fragments. Only 28 of the 104 cell lines had two or more exogenous U2 genes, and none had more than six, giving an average array size of 1.1 repeats. This does not differ significantly from the results with the mU2 cell lines (compare Fig. 3A and E).

(ii) Mixed tandem arrays of the mU2 and CT constructs.

One hundred six cell lines were isolated from the mixed mU2 and CT transfection and analyzed. Genomic DNA plugs were digested with SfuI and analyzed as described for the mixed mU2 and mU2+CT transfection (see Fig. 2F for a representative blot and Fig. 3F for the phosphorimager data). Only 24 of the 106 cell lines analyzed had two or more repeats, the largest number of repeats observed was nine, and the average array size was only 1.1 repeats. Thus, the CT microsatellite, when present as approximately 1 in 10 of the input repeat units, had no significant effect on the distribution or average size of the arrays recovered in either the mixed mU2 and mU2+CT cell lines or the mixed mU2 and CT cell lines.

Intriguingly, the mU2+CT repeat unit was highly overrepresented in the mixed mU2 and mU2+CT arrays (Table 1). Although CT repeats are not required to build a tandem array (see column 5 in Table 1; also see Fig. 3A and C), the mU2+CT repeats account for much more than the 10% of total repeats expected for random incorporation (see column 6 in Table 1). This may indicate that one function of the CT microsatellite is to enhance recombination between input arrays (see Discussion).

TABLE 1.

Molar representation of CT repeats in artificial tandem arrays recovered from mixed transfection with mU2 and mU2+CT^a

Total repeats in array	No. of cell lines with n CT repeats:				Arrays containing no CT repeats (%)	CT repeats/total repeats for all cell lines (%)
	0	1	2	3
1	41	24			63	37
2	9	8	3		45	35
3	3	4	3	2	25	44
4	3	2	3	0	38	25
5	0	0	1	0		40
6	2	2	1	0	40	13
7	0	0	0	1		43
8	0	1	0	0		13
9	0	0	1	0		22

^aThe mU2 and mU2+CT fragments were ligated together at a molar ratio of 10 to 1 and introduced into HT1080 cells by transfection. The relative numbers of mU2 genes and CT repeats in each array were determined by digestion with SfuI, genomic blotting, and phosphorimager analysis as described in Materials and Methods.  

Additional controls.

The validity of our conclusions requires that all of the artificial U2 arrays were stable, integrated at a single site, and efficiently transcribed only when the U2 promoter was intact.

(i) Single integration site. Our analysis of the size of the tandem arrays (Fig. 3) requires that all U2 repeats be integrated at a single site in each cell line. We digested one-third of each genomic plug with BamHI, an enzyme which does not cut within any of the constructs or within the resident 6.1-kb U2 repeat unit (Fig. 1) and therefore excises the natural and artificial tandem arrays intact. Intact arrays were resolved by FIGE, and the dried agarose gels were hybridized with a U2 probe as described elsewhere (49). Representative data from the mU2+CT transfection are shown in Fig. 4. The two resident U2 arrays are indicated; the artificial arrays vary in size, but only one is seen in each cell line. The failure of BamHI to digest artificial arrays of BamHI/BglII fragments also confirms that recombination in vivo eliminates all tail-to-tail junctions within the artificial arrays (5); similar digestions with BglII demonstrate the complete absence of head-to-head junctions (data not shown).

(ii) Stability of the CT microsatellite in artificial arrays.

The CT microsatellite dramatically increased recovery of large arrays of the active mU2 minigene (compare Fig. 3A and B). To ask whether the CT microsatellite was well behaved, we used PCR to amplify the CT microsatellite from the resident and artificial U2 repeats of the U2+CT and the U2ΔDSEΔPSE+CT cell lines, from the parental U2+CT and U2ΔDSEΔPSE+CT plasmid constructs, and from a cloned 6.1-kb U2 repeat plasmid (iU2). As expected (Fig. 5), the plasmid CT tracts were less heterogeneous than those in the resident U2 repeats, but the CT microsatellites in the artificial arrays were very similar to those of the parental plasmid constructs. Thus, the CT microsatellites were substantially unchanged by transfection, integration, and propagation of the artificial U2 arrays.

FIG. 5.

The CT microsatellite was unaltered during generation of cell lines containing artificial tandem arrays of mU2+CT and mU2ΔDSEΔPSE+CT. Marked and resident arrays were isolated from the indicated cell lines by preparative FIGE. CT microsatellites were amplified by PCR in the presence of [α-³²P]dCTP, digested with MspI and MnlI, and separated by electrophoresis on a 6% denaturing acrylamide gel as described previously (33) and in Materials and Methods. Three cell lines were examined from each transfection. Control reactions with mU2ΔDSEΔPSE+CT, iU2, CT, and mU2+CT plasmid DNA or with undigested HT1080 genomic DNA are shown in the rightmost panel. An M13 sequencing ladder provided size markers (base pairs).

(iii) Transcriptional activity of the artificial arrays.

To argue that the CT microsatellite works together with an active U2 snRNA gene to enhance formation, integration, or maintenance of large tandem arrays, we had to verify that the integrated mU2 constructs were transcribed but that the integrated U2ΔDSEΔPSE constructs were not. As described in Materials and Methods (also see reference 5), all U2 genes are marked with the innocuous U87C mutation (a U-to-C transition at position 87), enabling us to use primer extension to distinguish steady-state levels of U2 snRNA derived from the resident and exogenous U2 genes. Representative data are shown for the parental HT1080 cell line, an mU2 cell line, an mU2+CT cell line, and a mU2ΔDSEΔPSE cell line (Fig. 6). All of the mU2+CT cell lines expressed the marked U2-U87C snRNA, but the mU2ΔDSEΔPSE cell line did not. As observed previously (5), U2 expression per gene can vary from one resident or artificial array to another, but even the largest artificial arrays (which constituted 50 to 70% of the total U2 genes) contributed only 15 to 36% of the total U2 snRNA.

FIG. 6.

The marked U2 genes are efficiently expressed in stable mU2 and mU2+CT lines. RNA from representative mU2, mU2+CT, and mU2ΔDSEΔPSE lines and from the parental HT1080 line was analyzed by primer extension with ddATP and a ³²P-labeled oligonucleotide complementary to positions 89 to 109 of U2 snRNA as described in Materials and Methods. Extension products corresponding to resident U2 snRNA and marked U2 snRNA are indicated on the left.

DISCUSSION

In humans, the U1 snRNA, U2 snRNA, and 5S and 45S rRNA genes are all tandemly repeated (6, 12, 13, 27, 28, 35, 51, 58, 64). The U1 and U2 repeat units each contain a substantial CT microsatellite sequence [d(GT)_n · d(CA)_n] located downstream of the snRNA transcription unit (n = 15 to 25 for U1; n ≈ 70 for U2), whereas the 5S and 45S repeat units contain substantial GT microsatellites in comparable positions (n = 36 for 5S; n = 29 for 45S). The presence of dinucleotide repeats downstream of the transcription unit in the only four tandemly repeated multigene families in the human genome is unlikely to be coincidental; either tandem repetition of the large repeat unit fosters embedded dinucleotide repeats, or the embedded dinucleotide repeats foster maintenance of the larger tandem array. (We distinguish here between the essentially perfect tandem repeats that encode homogeneous structural RNAs and the multitude of imperfect tandem repeats resulting from gene duplication such as the α and β globin genes; the immunoglobulin V, D, J, and C regions; and the major histocompatibility complex genes. In these imperfect repeats, the problem is to prevent homogenization of the duplicated gene copies, not to promote it [14, 26, 45, 46].)

To explore possible functions of the CT microsatellite, we examined the effect of the CT microsatellite on integration and maintenance of artificial tandem arrays of a minimal U2 transcription unit (U2 minigene). As summarized in Fig. 3, our data show that the transcriptionally active U2 minigene does not readily form large tandem arrays when introduced into HT1080 cells by the calcium phosphate transfection method (an average of 1.3 repeats per array). Although deletion of U2 promoter elements significantly increased the size of the tandem arrays (average of 5.4 repeats), addition of the CT microsatellite to the active U2 minigene construct had a more dramatic effect (average of 8.2 repeats). Moreover, the CT microsatellite appeared to inhibit recovery of tandem arrays of inactive U2 genes and had no effect when present in only a fraction of the repeat units of an array. Thus, the CT microsatellite enhances integration or maintenance of transcriptionally active tandem arrays of U2 genes but does not affect (and may even inhibit) integration or maintenance of transcriptionally inactive tandem arrays of U2 genes.

The CT microsatellite does not enhance array size simply by repressing transcription of nearby U2 genes. The marked U2 snRNA genes within the artificial tandem arrays are efficiently transcribed whether the array consists of intact 6.1-kb U2 repeats (iU2), a U2 minigene with the CT microsatellite (mU2+CT), or the U2 minigene lacking the CT microsatellite (mU2) (Fig. 6) (5a; see also reference 5). Moreover, if lack of U2 transcriptional activity were the only factor favoring recovery of large arrays, the promoter-deleted mU2ΔDSEΔPSE construct should have generated larger arrays on average than mU2+CT.

There is some evidence that simple dinucleotide repeats can function as recombination hot spots in a variety of systems. Repeats of d(GT)_n · d(CA)_n enhance recombination between plasmids in vivo (10, 43, 53, 56, 60). In addition, d(GT)_n · d(CA)_n repeats frequently define the boundaries of recombination events in introns of the ɛ immunoglobulin heavy chain constant region gene (22) and in the VDJ region of the immunoglobulin gene cluster in certain lymphoid tumors (9). Perhaps more to the point, the distribution of both d(GT)_n · d(CA)_n and d(CT)_n · d(GA)_n repeats in Drosophila correlates with transcriptional activity, the ability to undergo meiotic recombination, and dosage compensation (36). Conservation of microsatellite location suggests that the same could conceivably be true in mammalian genomes (52).

The CT microsatellite also does not appear to function as a simple hot spot for recombination. A hot spot could facilitate (i) recombination between input repeat units before integration, (ii) chromosomal integration of tandem arrays, or (iii) recombination within the integrated chromosomal array. Arguments can be made against all three scenarios. First, the CT microsatellite seems unlikely to significantly enhance recombination between input repeat units because the level of recombination is sufficient to convert random ligation reactions (a mixture of head-to-head, head-to-tail, and tail-to-tail repeats) into perfect head-to-tail tandem arrays whether or not the input repeats contain the CT microsatellite. (Note that transfection with the random mU2ΔDSEΔPSE ligation reaction yields perfect tandem arrays containing as many as 80 repeats [Fig. 3C].) Second, the CT microsatellite seems unlikely to facilitate chromosomal integration of tandem arrays because there were no significant differences in the recovery of neomycin-resistant colonies from any of the six transfections (mU2, mU2+CT, mU2ΔDSEΔPSE, mU2ΔDSEΔPSE+CT, mU2 and mU2+CT, and mU2 and CT). Third, the CT microsatellite seems unlikely to facilitate recombination within the integrated tandem array because the integrated arrays are almost invariably stable. With the sole exception of the iU2A37 line (5), which apparently undergoes breakage-fusion-bridge cycles (65a), we have never observed spontaneous gain or loss of repeat units or heterogeneity of intact artificial arrays resolved by FIGE, even when cell lines with doubling times of about 24 h were propagated for many months or recovered from storage (Fig. 4) (5a).

The stability of artificial U2 tandem arrays, once integrated into the chromosome, suggests that the CT microsatellite must affect an event prior to integration. Intriguingly, we found that the mU2+CT repeat unit was greatly overrepresented in the mixed mU2 and mU2+CT transfections (Table 1). On the surface, this suggests that the CT microsatellite may favor recovery of larger arrays by enhancing recombination between input DNA molecules. This leads to a paradox, however, because recombination between input DNA molecules is invariably sufficient to resolve random ligation reactions into perfect head-to-tail arrays regardless of whether the CT microsatellite is present. How then could the CT microsatellite favor recovery of larger arrays? We suggest that the paradox can be resolved by recognizing that recombination within a tandem array is always a double-edged sword: homologous recombination between repeats is as likely, a priori, to excise repeats from a tandem array as to increase the number of repeats by recombination between two arrays. Thus, the CT microsatellite may bias recombination toward events which generate larger arrays, without significantly affecting events that resolve mixed ligations into perfect head-to-tail arrays. Indeed, the CT satellite might even function as a negative element to retard integration of extrachromosomal tandem arrays, thereby allowing more time for assembly of larger extrachromosomal arrays.

How might the CT microsatellite act synergistically with transcriptional machinery to enhance recombination? The GAGA transcription factor binds to short CT repeats in the promoter or first intron of a number of genes (7, 15, 17, 20, 37, 57) and appears to activate transcription by clearing nucleosomes from the promoter in an ATP-dependent manner (37, 57) or perhaps by insulating the gene from position effects (47). Although the CT microsatellite is clearly not required for U2 snRNA expression (Fig. 6) (see also references 1 and 2), GAGA family factors might nonetheless be bound to the CT microsatellite. Since transcription factors bound to promoter (55, 65) and enhancer (32) elements can stimulate recombination, even in the absence of actual transcription, a GAGA family factor bound to the CT microsatellite might collaborate with an active U2 promoter to enhance recombination.

Alternatively, as certain dinucleotide repeats are relatively free of nucleosomes, the CT microsatellite might favor recovery of large tandem arrays by providing preferential access sites for the recombination machinery (15, 17, 29, 37, 41, 57). In this scenario, an active promoter alone (mU2) would occlude the recombination machinery, but when the promoter is combined with the CT microsatellite (mU2+CT), the active transcription unit would ensure that the CT tract remains truly nucleosome free. When the promoter is deleted (U2ΔDSEΔPSE), the DNA would assume an inactive chromatin configuration that would dominate the effect of nearby CT repeats (U2ΔDSEΔPSE+CT). Consistent with this scenario, the chromatin structure of transiently expressed DNA is known to markedly affect transcription (31, 50).

Although the CT repeat clearly favors recovery of large tandem arrays of transcriptionally active U2 genes, large arrays of a transcriptionally active minimal U1 gene were readily obtained by very similar protocols (32a). This paradox may reflect the well-known but poorly understood differences between the apparently similar U1 and U2 promoters (12). The U1 and U2 promoters can both be decomposed into an upstream enhancer-like element known as the distal sequence element (DSE) (centered around −200) and a downstream TATA-like element known as the proximal sequence element (PSE) (centered around −55). The DSE consists of adjacent Sp1 and Oct-1 sites, and the two proteins bind synergistically (54). The PSE has no TATA homology but functions similarly by binding a large complex known as SNAPc (23) or PTF (4) which contains or can associate with TATA binding protein. Oct-1 bound to the DSE and SNAPc bound to the PSE also bind synergistically (18, 44). The U2 promoter may be strictly modular; deletion of sequences between the DSE and PSE does not appear to impair transcriptional activity (2). In contrast, the U1 promoter contains at least one additional element between the DSE and PSE which can stimulate transcription by three- to fivefold (42). The existence of developmentally regulated U1a and U1b variants in mice (11) and the ability of simian virus 40 T antigen to enhance transcription of human U2 snRNA but not U1 snRNA (21) also suggest that U1 and U2 may be regulated differently. The U1 and U2 transcription units might therefore interact differently with whatever function the CT repeat provides.

We do not yet understand the mechanism by which the CT microsatellite cooperates with a transcriptionally active U2 minigene to favor recovery of large tandem arrays; however, we have established that the CT microsatellite could potentially affect the generation, structure, or maintenance of a transcriptionally active tandem repeat. Our data strongly suggest that the mechanisms responsible for maintenance of tandem arrays containing active promoters may differ from those that maintain tandem arrays of transcriptionally inactive sequences like alphoid satellite (62), minisatellites (3, 30), and oligonucleotide repeats (61, 63). Many questions remain. Does the ability of the U2 transcription unit and the CT microsatellite to cooperate functionally depend on the distance between them? Can other transcription units such as those for U1 snRNA or mRNA (RNA polymerase II), 5S rRNA (RNA polymerase III), 45S rRNA (RNA polymerase I), or even a prokaryotic gene (T7 bacteriophage RNA polymerase) substitute for the U2 minigene? Do short transcription units cooperate more effectively with the CT microsatellite than long units do? Does the chromatin structure of artificial U2 repeats, and of the CT microsatellite in particular, depend on U2 transcriptional activity? The answers to these questions could begin to explain the remarkable fact that CT or GT microsatellites are found downstream of the only four tandemly repeated multigene families known in the human genome. The genome projects can reveal more sequences, but functional studies are required to understand why these sequences look the way they do.