A Tandem Array of Minimal U1 Small Nuclear RNA Genes Is Sufficient To Generate a New Adenovirus Type 12Inducible Chromosome Fragile Site

Abstract

Infection of human cells with adenovirus serotype 12 (Ad12) induces metaphase fragility of four, and apparently only four, chromosomal loci. Surprisingly, each of these four loci corresponds to a cluster of genes encoding a small abundant structural RNA: the RNU1 and RNU2 loci contain tandemly repeated genes encoding U1 and U2 small nuclear RNAs (snRNAs), respectively; the PSU1 locus is a cluster of degenerate U1 genes; and the RN5S locus contains the tandemly repeated genes encoding 5S rRNA. These observations suggested that high local levels of transcription, in combination with Ad12 early functions, can interfere with metaphase chromatin packing. In support of this hypothesis, we and others found that an artificial tandem array of transcriptionally active, but not inactive, U2 snRNA genes would generate a novel Ad12-inducible fragile site. Although U1 and U2 snRNA are both transcribed by RNA polymerase II and share similar enhancer, promoter, and terminator signals, the human U1 promoter is clearly more complex than that of U2. In addition, the natural U1 tandem repeat unit exceeds 45 kb, whereas the U2 tandem repeat unit is only 6.1 kb. We therefore asked whether an artificial array of minimal U1 genes would also generate a novel Ad12-inducible fragile site. The exogenous U1 genes were marked by an innocuous U72C point mutation within the U1 coding region so that steady-state levels of U1 snRNA derived from the artificial array could be quantified by a simple primer extension assay. We found that the minimal U1 genes were efficiently expressed and were as effective as minimal U2 genes in generating a novel Ad12-inducible fragile site. Thus, despite significant differences in promoter architecture and overall gene organization, the active U1 transcription units suffice to generate a new virally inducible fragile site.

Only four loci in the human genome are known to contain tandemly repeated genes encoding small abundant structural RNAs, and these are also the only four sites of metaphase chromosome fragility induced by adenovirus type 12 (Ad12) infection of human cells at low multiplicity: the RNU1 locus encoding U1 small nuclear RNA (snRNA), the RNU2 locus (encoding U2 snRNA), the RN5S locus (encoding 5S rRNA), and the PSU1 (pseudo-U1) locus (a decaying RNU1-like locus that may still be transcribed although the RNA is no longer stably incorporated into U1 small nuclear ribonucleoprotein particles (snRNP). Apparent colocalization of four clustered multigene families with four virally induced fragile sites led us to propose that high levels of local transcriptional activity, in combination with Ad12 early functions, interfered locally with metaphase chromatin condensation (31). Some support for this proposal came from the demonstration by fluorescent in situ hybridization (FISH) that U2 genes could be found on either side of the Ad12-induced fragile site at RNU2 (11). Thus, if the RNU2 locus was not a contributing cause of fragility, it was at least the weakest link in local chromatin structure. We (4) and others (15, 28) subsequently demonstrated that an artificial tandem array of active, but not inactive, U2 snRNA genes was indeed sufficient to generate a new Ad12-inducible fragile site; moreover, a minimal 834-bp U2 snRNA gene suffices for this effect (4), ruling out essential contributions by sequences flanking the RNU2 locus or by other elements within the natural 6.1-kb U2 repeat unit.

Mutations in the Ad12 E1B 55-kDa transforming protein substantially reduce the ability of whole virus to induce fragility (47), and the E1B 55-kDa protein is known to interact with p53 (26, 55; reviewed in reference 7). We were therefore gratified to find that p53 and the Ad12 E1B 55-kDa protein alone were sufficient to induce fragility of the resident RNU2 locus in Saos-2 cells lacking endogenous p53 function (28a). Although the E4 orf6 34-kDa protein is known to interact with both p53 and the E1B 55-kDa protein (10, 16, 40), the 34-kDa protein may modulate virally induced fragility but cannot be essential for it. More recently, we found that the DNA damaging reagent actinomycin D (56a) can phenocopy p53-dependent Ad12-induced fragility of the RNU1 and RNU2 loci in the absence of virus, and similar results have also been obtained with 1-β-d-arabinofuranosylcytosine (33a).

Thus, we must explain why Ad12-induced fragility requires U2 snRNA transcription and p53 but can be induced by either Ad12 E1B 55-kDa protein or DNA damage caused by actinomycin D or 1-β-d-arabinofuranosylcytosine. The simplest unifying hypothesis is that Ad12 55K protein somehow phenocopies the effect of these DNA-damaging reagents, alerting p53 (by protein-protein interaction or by inducing phosphorylation, acetylation, or a conformational change), which then directly or indirectly causes locus-specific chromosome fragility. Activated p53 would then interact with the transcription apparatus to interfere with metaphase chromatin condensation. The ability of Ad12 to induce four specific sites of metaphase fragility, rather than generalized fragility, might reflect a specific interaction of activated p53 with the specialized U1 and U2 snRNA promoters (3, 20, 46) and termination factors (22, 23, 39). Alternatively, activated p53 may interfere mildly with chromatin condensation throughout the genome (thus causing generalized fragility at a high multiplicity), but high levels of local transcriptional activity would render the RNU1 and RNU2 loci hypersensitive to this effect.

What then of the U1 genes? Can we assume that Ad12 induces fragility of the RNU1 and RNU2 loci by the very same mechanism? The U1 snRNA genes colocalize cytologically (i.e., within 10 Mbp) with the Ad12-induced fragile site, and the many similarities between the RNU1 and RNU2 loci are impressive, but these are hardly conclusive arguments. Although the U1 and U2 transcription units are apparently very similar (9, 12, 37, 49), there are also some intriguing differences (2, 8, 19, 36; see Discussion for details). The genomic organization of the human U1 and U2 genes is also quite different. The U2 genes are organized as 5 to 25 tandem copies of a relatively small, regular 6.1-kb repeat unit that does not appear to encode any other large or small RNAs; the entire RNU2 locus therefore spans at most 300 kb and more typically 60 kb (29, 30, 41, 51, 54). In contrast, the U1 genes are organized as an irregular and highly polymorphic multigene cluster with a repeat unit that exceeds 45 kb (6, 32, 35) and contains a multitude of tRNA genes (52); the 30 true U1 genes (32) therefore span over 1.35 Mbp. In addition, there is good reason to suspect that the chromatin structures of the RNU1 and RNU2 loci are profoundly different because an Ad5-simian virus 40 hybrid virus preferentially and repeatedly integrates at various sites distributed across the RNU1 locus (44, 45) but has never been observed to integrate at the RNU2 locus.

Thus, it is not a foregone conclusion that the U1 genes are the cause of Ad12-induced fragility of the giant RNU1 locus. Indeed, virally induced fragility could be due to the many tRNA genes (transcribed by RNA polymerase III) embedded within the U1 repeat unit and not to the U1 genes themselves (transcribed by RNA polymerase II). This would help to explain how Ad12 can cause the fragility of multigene clusters transcribed by both RNA polymerase II (U1 and U2 snRNA) and RNA polymerase III (5S rRNA). We therefore thought it was prudent to ask directly whether a tandem array of human U1 genes could generate a new Ad12-inducible fragile site.

MATERIALS AND METHODS

Construction of a minimal, marked U1 snRNA gene.

A minimal human U1 snRNA gene was excised from the HSD4 clone (35) as a 681-bp HaeIII fragment, fitted with BamHI linkers, and cloned into M13mp8 (58). As shown in Fig. 1A, this minimal U1 construct (mU1), which extends from 418 nucleotides (nt) upstream of the 164-nt U1 coding region to 89 nt downstream, spans the entire transcription unit (reviewed in reference 9) from the upstream distal sequence element of the promoter (centered at −220 [2, 36, 37, 49]) to the downstream 3′-end-formation signal (+15) (1, 21, 57). The U72C mutation was introduced into this construct by the two step “megaprimer” PCR method (5). A mutagenic 20-mer (P_84–65; 5′-CAGCACATCCGGGGTGCAAT-3′, where the mutation is underscored) and the M13 reverse primer (M13RP) were annealed to the mU1 construct in M13mp8, and the megaprimer was generated by PCR amplification as described previously (5) with 4.0 mM MgCl₂ (27). After purification of the megaprimer on low-melting-temperature agarose, a second round of PCR amplification was performed with 2.5 mM MgCl₂, using the megaprimer and the M13 sequencing primer (M13SP), and a purified EcoRI/NarI fragment of the mU1 M13mp8 replicative form as template. NarI digestion severed the M13RP binding site from the template, resulting in a vast excess of mutant U1 gene compared to wild type. The PCR products were digested with BamHI, and the 684-bp U1-U72C fragment was recovered, transferred to pUC18, and the insert completely sequenced to rule out PCR artifacts.

FIG. 1.

Local BglII restriction map of resident U1 genes and recombinant constructs used to build artificial U1 arrays. (A) mU1 minigene used to generate artificial U1 arrays drawn approximately to scale. The asterisk denotes the U72C mutation. The BglII site at position −6 upstream from the coding region is shown explicitly. Also indicated are the distal sequence element (open circle) and the proximal sequence element (solid circle) of the promoter and the 3′-end-formation signal (solid box). (B) E1 selection cassette containing the neomycin resistance marker (Neo) and the dihydrofolate reductase (DHFR) minigene capable of conferring high levels of methotrexate resistance (4). (C) Consensus BglII map of resident U1 genes drawn to scale (6, 32, 35). All sites are polymorphic, but no BglII fragments more than 10 kb from the U1 gene (box) were ever detected in genomic blots with an mU1 probe.

Transcriptional activity of the marked U1-U72C gene.

To verify that the marked pU1-U72C plasmid construct was transcriptionally active, adherent HT1080 fibrosarcoma cells were cotransfected by the calcium phosphate method (Gibco/BRL) with 10 μg each of pU1-U72C and an equivalent pU2-U87C plasmid as a control (4). HT1080 was grown in minimal essential medium (Gibco/BRL) supplemented with 10% fetal bovine serum and split 24 h prior to transfection to give 10⁶ cells per 100-mm-diameter plate. After transfection, cells were grown for 24 h, harvested from the nearly confluent plates by trypsinization, and lysed with Nonidet P-40 as described previously (21); total RNA was prepared by phenol extraction. Primer extension analysis was performed with ddATP and a 5′-end-labeled 16-mer complementary to U1 positions 75 to 91 or a 20-mer complementary to U2 positions 89 to 109 (4, 59). In the presence of ddATP, the extension products derived from the endogenous and marked U1-U72C snRNAs were 3 and 8 nt longer, respectively, than the U1 primer; the products derived from endogenous and marked U2-U87C snRNAs were 2 and 13 nt longer, respectively, than the U2 primer. The primer extension products were resolved on a 15% sequencing gel, and phosphorimager analysis was performed with a GS-250 Molecular Imager (Bio-Rad Laboratories) and Molecular Analyst 2.0. A correction was made for the ability of murine leukemia virus reverse transcriptase to read through U72 on HT1080 U1 snRNA despite the presence of ddATP.

Artificial tandem arrays of minimal U1 genes.

In brief, large random arrays of the BamHI U1-U72C fragment were generated by ligation in the presence of 50 μM spermidine to reduce viscosity and to increase the average array size; the ligation was doped with a low concentration of a BamHI/BglII fragment containing the E1 neomycin resistance cassette (0.01 mass ratio) to increase the average size of the tandem arrays in Geneticin-resistant colonies (Fig. 1B). Transfection of HT1080 cells, colony isolation, characterization of the artificial U1 arrays, Ad12 infection, and FISH with the biotinylated mU1 gene as probe were essentially as described for artificial U2 arrays (4).

RESULTS

Use of phylogenetic and SELEX data to choose the U72C mutation.

To assay U1 snRNA transcribed from the artificial arrays above the massive background of endogenous U1 snRNA (10⁶ molecules per cell or 20% of the molar level of rRNA [53]), we sought a point mutation which could be detected by a simple primer extension assay but would not interfere with U1 snRNA transcription, function, or stability. Using a compilation of snRNA sequences (17), we compared the U1 snRNA sequences of human, rat Novikoff hepatoma, mouse sperm, chicken liver, Xenopus laevis, and Drosophila melanogaster to identify single-stranded phylogenetically variable regions (Fig. 2). Since no such sites could be found except in the Sm, U1A, 70K, and U1C binding regions (18, 33), we attempted to supplement the phylogenetic data with related data obtained by SELEX (systematic evolution of ligands by exponential enrichment). When the U1A protein was used to select RNA sequences from degenerate pools (50), the U1A protein binding site in loop II was shown to consist of a highly conserved 5′ region (5′-AUUGCAC-3′) and a more variable 3′ region (5′-UCC-3′). Although the underlined U (equivalent to human U72) is strongly conserved throughout metazoans, we reasoned that this might reflect functions other than U1A binding and thus that a mutant U1-U72C snRNA would probably yield a stable, marked U1 snRNP.

FIG. 2.

U1 snRNA sequence and accepted secondary structure. Differences between human U1 and U1 from rat (r), mouse (m), chicken (ch), and X. laevis (x) are indicated by callouts; for example, chC indicates that U60 is a C in chicken. The mutagenic P_84–65 oligonucleotide used to generate the “megaprimer” is denoted by a shaded overbar; the P_91–75 oligonucleotide used to assay for expression of U1-U72C snRNA by differential primer extension through the mutated U72C site is denoted by a solid overbar. 70K binds to loop I, U1A binds to loop II; and U1C binds to 70K and common snRNP proteins (38). The Sm binding site centers on ₁₂₆AUUUG₁₃₀ (25). The sequence is from reference 17.

Transcriptional activity of the marked U1-U72C gene.

The marked U1-U72C gene was constructed as described in Material and Methods and assayed for transcriptional activity by transfection into HT1080 cells by the calcium phosphate precipitation technique; cotransfection with an equivalent U2-U87C U2 minigene provided an internal control. The ratios of U1-U72C to wild-type U1 snRNA and of U2-U87C to wild-type U2 snRNA were determined by differential primer extension through the mutant sites in the presence of ddATP, resolution of the primer extension products by denaturing 15% polyacrylamide gel electrophoresis, and phosphorimager analysis. The U1-U72C and U2-U87C snRNAs were both efficiently expressed and accumulated over the course of a 24-h transfection period to 15 to 20% of the level of endogenous U1 and U2 (data not shown). Efficient expression of the marked U1-U72C genes in stably transformed cell lines is also documented in Fig. 4.

FIG. 4.

Relative expression of resident U1 and marked U1-U72C genes. Total RNA was isolated by the Nonidet P-40 method from the parental HT1080 cell line and 13 lines containing artificial tandem arrays of marked U1 genes. The relative amounts of steady-state U1-U72C snRNA and wild-type U1 were determined as described in Material and Methods and are tabulated in Table 1. Lane 5 is underloaded. Premature stops by reverse transcriptase are occasionally seen (extra bands in lanes 4, 9, and 15) because primer extension on the highly structured and modified U1 snRNA template is unusually sensitive to both annealing and extension conditions.

Cell lines containing artificial tandem arrays of the marked U1-U72C minigene.

We had previously generated large head-to-tail tandem arrays of U2 minigenes by random ligation of BamHI/BglII monomers in vitro, followed by digestion with BamHI and BglII (4); however, a BglII site just upstream from the U1 coding region ruled out this approach for the mU1 constructs. Fortunately, we had also learned that HT1080 cells invariably resolve random in vitro ligation reactions of BamHI/BglII fragments (containing head-to-head, head-to-tail, and tail-to-tail junctions) into flawless head-to-tail arrays, as evidenced by the absence of any genomic restriction fragments corresponding to head-to-head or tail-to-tail junctions (3a, 4). We therefore excised the marked U1-U72C gene as a BamHI fragment from the parental plasmid and followed the protocols previously used for the U2 minigenes (4) to generate HT1080 cell lines containing large artificial U1 tandem arrays.

Of 134 Geneticin-resistant colonies screened, 28 contained mU1 arrays and only the 13 with the largest number of mU1 repeat units were analyzed further. The genomic organization of the artificial U1 tandem arrays in these 13 cell lines was characterized by BglII digestion, and the mU1-U72C gene copy number was determined by comparison to that of the resident U1 genes (31). The marked mU1-U72C genes, like all other true U1 genes, are cut by BglII at position −6 just upstream from the promoter, generating a 684-bp monomer unit. Note that the BglII site is located asymmetrically within the mU1 repeat unit; thus, BglII digestion of a random artificial mU1 array would generate three fragments of 824 bp (head-to-head repeats), 684 bp (head-to-tail repeats), and 544 bp (tail-to-tail repeats) as can be seen for the multimer control (Fig. 3). Instead, the artificial U1 arrays generate 684-bp fragments exclusively, confirming that all mU1 minigenes in each artificial array must be tandemly repeated in head-to-tail fashion even though the input ligation reaction is random.

FIG. 3.

Genomic organization of artificial U1 tandem arrays. Genomic DNA from the parental HT1080 line and 13 lines containing mU1 arrays was digested with BglII and characterized by genomic blotting with an mU1 probe. The endogenous U1 genes are polymorphic (Fig. 1C) and generate BglII fragments of 11, 9.5, 5.0, 3.5, and 1.7 kb, whereas the artificial arrays generate exclusively 684-bp BglII fragments indicative of a perfect head-to-tail tandem repeat (see text). The multimer control is a BglII digest of a random array of mU1 BamHI fragments generated by in vitro ligation. The mU1-1 and mU1-63 lines each have only one or no marked U1 gene and therefore lack the 684-bp band diagnostic of head-to-tail U2 genes.

Although the FISH data suggest that each of these 13 cell lines contained a single mU1 artificial array (see Fig. 5), this was directly confirmed for 7 of the 13 lines (mU1-38, mU1-67, mU1-77, mU1-96, mU1-99, mU1-105, and mU1-125) by digestion of genomic DNA plugs with SpeI and NdeI. These “null cutters” excise intact artificial arrays as a single restriction fragment, so that both the number and the approximate size of the artificial arrays can be determined by field inversion gel electrophoresis followed by blotting with an mU1 probe (data not shown).

FIG. 5.

Ad12-induced fragility of the artificial tandem arrays of marked U1 genes. A gallery of representative metaphase chromosomes is shown from the indicated cell lines with and without Ad12 infection. To prevent the U1 signal from obscuring virally induced changes in chromosome morphology, each chromosome containing an artificial mU1 array is represented by a pair of images: total DNA visualized by DAPI (4′,6-diamidino-2-phenylindole) staining is shown on the left, and mU1 arrays visualized by FISH with an mU1 probe and superimposed on the DAPI image are shown on the right. See Table 2 for quantitation and details.

The genomic organization of the resident U1 genes is more complex. Unlike the perfect 6.1-kb tandem repeat unit of the human U2 genes, the repeat unit of the human U1 genes is >45 kb (6) and displays significant polymorphism (6, 31, 34). Although BglII invariably cuts at position −6 just upstream from the U1 coding region, other BglII sites flanking the U1 snRNA genes are polymorphic (Fig. 1C) and thus generate five major BglII fragments (11, 9.5, 5.0, 3.5, and 1.7 kb). The copy number of the mU1 repeats in each cell line can be determined by genomic blotting, simply by normalizing the signal derived from the 684-bp mU1 fragment to the sum of the signals derived from endogenous U1 gene fragments.

Transcriptional activity of artificial U1 arrays.

All of the mU1 cell lines expressed U1-U72C efficiently with ratios of U1-U72C to wild-type U1 ranging from 0.10 to 0.50 (Fig. 4 and Table 1). However, the relative expression per marked U1 gene, calculated by dividing the RNA ratio by the gene ratio, varied widely, from 0.15 to 0.73. Quite different results were obtained for artificial arrays of U2 genes, where the relative expression of the marked U2-U87C and wild-type U2 snRNA was nearly constant from 0.9 to 1.1 (4). These data may support the notion that U1 snRNA (33) but not U2 snRNA is subject to dosage compensation. Alternatively, U1 transcription may be more sensitive than U2 to chromosomal position effects, although this would have to be a U1-specific effect because the neomycin resistance cassette is also embedded within the artificial array and must be strongly expressed. Also, the observation that chromosomal regions surrounding newly integrated DNA seem to be activated and repressed en bloc (43) argues against position effects.

TABLE 1.

Characteristics of cell lines containing artificial tandem arrays of the marked U1-U72C minigene^a

Cell line	Copy no. of mU1 genes	Ratios
		Gene ratio	RNA ratio	RNA/gene
HT1080	0
mU1-1	5	0.18	0.13	0.73
mU1-4	16	0.53	0.15	0.29
mU1-9	24	0.82	0.50	0.61
mU1-38	15	0.48	0.35	0.71
mU1-63	7	0.22	0.12	0.56
mU1-65	14	0.48	0.20	0.43
mU1-67	35	1.17	0.25	0.21
mU1-77	21	0.71	0.15	0.21
mU1-96	45	1.50	0.42	0.28
mU1-99	30	1.01	0.30	0.30
mU1-102	12	0.39	0.10	0.26
mU1-105	24	0.80	0.37	0.47
mU1-125	37	1.21	0.18	0.15

^aThe copy number of the marked U1-U72C genes was calculated by phosphorimager analysis of the data in Fig. 3, assuming 30 true U1 genes per haploid complement (32). The gene ratio is the ratio of signal from the exogenous mU1 genes to that of the sum of the resident U1 genes, after correction both for background and for the extent of complementarity between the mU1 probe and each genomic fragment. The RNA ratio is the ratio of marked U1-U72C to wild-type U1 snRNA determined by differential primer extension as described in Materials and Methods. The RNA/gene ratio is the ratio of the RNA ratio to the gene ratio and represents the relative transcription of the marked U1-U72C genes to resident U1 genes. 

Ad12-induced metaphase fragility of the artificial U1 arrays.

We assayed Ad12-induced fragility of the artificial U1 arrays in the seven cell lines with the largest artificial arrays; Ad12-induced fragility of the resident RNU1 and RNU2 loci provided a convenient and reliable control (Fig. 5 and Table 2). FISH was performed as described previously (4) with the mU1 construct as the probe to ensure comparable signals from the artificial U1 array and the resident RNU1 locus. Chromosomal locations of the artificial U1 arrays could be deduced in several instances from chromosome length or morphology. No metaphase decondensation of the artificial arrays was observed in the absence of Ad12 infection, confirming that the integration sites and sequences are not constitutively fragile.

TABLE 2.

Ad12-induced fragility of artificial U1 tandem arrays

Cell line	mU1 copy no.a	Location of mU1 arrayb	% of loci scored as fragile
			RNU2 (17q21)	RNU1c (1p36)	mU1 array
HT1080			88	40 (71)
mU1-38	14	3q(m)	93	52 (70)	19
mU1-67	35	Cq(m)	95	59 (85)	14
mU1-77	21	Cq(t)	92	56 (84)	5
mU1-96	45	Cp(m)	77	32 (50)	21
mU1-99	30	13q(t)	87	43 (64)	16
mU1-105	24	17p(m)	95	46 (70)	27
mU1-125	36	Cq(t)	96	63 (83)	18

^aFrom Table 1. 

^bFISH signals were mapped to chromosomes 3, 13, or 17 or to a group C chromosome. The position of the signal was further defined as in the middle (m) or telomeric region (t) of the short or long arm (p or q). 

^cNumbers in parentheses indicate the percentage of cells containing at least one fragile RNU1 locus. 

Ad12-induced metaphase chromsome fragility depends on the multiplicity of infection (47, 60) At low multiplicities (<1 PFU/cell), only the RNU2 locus at 17q21 exhibits fragility; the RNU1 locus at 1p36 exhibits little or no visible damage. At higher multiplicities (10 to 20 PFU/cell), one or both RNU2 loci and about half the RNU1 loci exhibit damage. We therefore used a relatively high multiplicity (20 PFU/cell) to compare the relative fragilities of the artificial and natural U1 arrays in seven different cell lines; the cytological location of each artificial U1 array was distinct, and the mU1 gene copy number per array ranged from 14 to 45. To ensure that only fully infected cells were scored, the artificial and resident U1 arrays were examined exclusively in metaphases displaying visible damage at the RNU2 locus. At least 30, and more typically 60, metaphases were scored. Approximately 50 to 85% of the resident U1 arrays (RNU1) but only 5 to 27% of the artificial U1 arrays exhibited aberrations in infected cells, although the gene copy number per array was generally comparable (Table 2). We also calculated the relative fragility of the seven artificial mU1 loci compared to the resident RNU1 loci (Table 3). We found that relative fragility correlates reasonably well with relative expression of U1 snRNA per locus (RNA per locus) but not with gene copy number (gene ratio) or relative expression per gene (RNA per copy). The correlation is not absolute, however; the mU1-96 and mU1-105 artificial arrays are more fragile than the resident RNU1 locus but generate less U1 snRNA per locus.

TABLE 3.

Correlation of Ad12-induced fragility with transcriptional activity of resident RNU1 locus and artificial mU1 tandem arrays

Cell line	Ratios				Relative fragilitye
	Genea	RNAb	RNA/copyc	RNA/locusd
mU1-38	0.48	0.34	0.71	0.69	0.73
mU1-67	1.17	0.25	0.21	0.50	0.47
mU1-77	0.71	0.15	0.21	0.30	0.18
mU1-96	1.50	0.42	0.28	0.83	1.31
mU1-99	1.01	0.30	0.30	0.61	0.74
mU1-105	0.80	0.37	0.46	0.74	1.17
mU1-125	1.21	0.18	0.15	0.36	0.57

^aRatio of mU1 to resident U1 genes as listed in Table 1. 

^bRatio of U1-U72C to wild-type U1 snRNA as listed in Table 1. 

^cRatio of RNA ratio to gene ratio, giving expression of marked U1 genes relative to resident U1 genes. 

^dRNA ratio × 2. If we assume that the two endogenous RNU1 loci are equally expressed, this represents the relative expression per locus of the artificial and resident U1 arrays. 

^eRelative frequency of breakage at the artificial mU1 locus compared to that at the RNU1 loci (see Fig. 1 of reference 27a for scoring of breaks). 

DISCUSSION

We have demonstrated that an artificial tandem array of transcriptionally active U1 snRNA genes is sufficient to generate a new Ad12-inducible metaphase chromosome fragile site. In view of previous work showing that an artificial array of transcriptionally active, but not inactive, U2 snRNA genes is also sufficient to generate a new fragile site (4, 15, 28), these results leave little doubt that Ad12 induces fragility of the RNU1 and RNU2 loci by related, if not identical, mechanisms. Consistent with our earlier results showing that active U2 transcription is required for Ad12-induced fragility of artificial U2 arrays (4), the relative fragility of the artificial U1 arrays correlates well with the levels of U1 transcription per locus, although not with gene copy number (Table 3). Thus, U1 expression must be regulated not just by gene copy number but also by epigenetic factors such as limiting transcription factors, DNA methylation, heterochromatin structure, or perhaps association with coiled bodies (13, 14, 42).

We were surprised that the miniscule U1 gene (approximately 400 bp from the 5′ enhancer to 3′-end-formation signal) can cause fragility of a U1 repeat unit exceeding 45 kb (6); however, at least in primary human embryonic kidney cells (60), Ad12 can also induce fragility of the RN5S locus encoding 5S rRNA (an RNA polymerase III transcript). Thus, it is conceivable that the many tRNA genes which are embedded within the U1 repeat unit (52) and are transcribed by RNA polymerase III could contribute to (or even be required for) Ad12-induced fragility of the intact RNU1 locus but be unnecessary when a minimal U1 gene is multimerized.

Although it is usually safer to ask how than why, the observation that clustered U1 and U2 genes both generate artificial fragile sites tempts us to ask why Ad12 induces chromosome fragility at all. It was established very early that fragility is not a consequence of viral integration (61), and this was confirmed by the ability of coexpressed p53 and Ad12 E1B 55-kDa proteins to induce fragility of the RNU1 and RNU2 in Saos-2 cells (27a). Adenoviral infection is known to activate transcription by RNA polymerase III (24, 48, 56), and thus it would not be completely surprising if the virus induced generalized chromatin decondensation in order to activate global gene expression during lytic infection. Underlying changes in chromatin structure might manifest themselves during metaphase as generalized fragility; specific fragility would reflect the greater sensitivity of heavily transcribed multigene clusters to viral decondensation. Generalized fragility would then be useful to the virus; specific fragility would be fortuitous. This could explain why Ad2/5 and Ad12 both induce generalized fragility at a high multiplicity of infection but only Ad12 induces specific fragile sites at a low multiplicity of infection (47, 60). Chromatin decondensation might also facilitate viral integration, thus favoring transformation; however, this would not explain why an Ad5-simian virus 40 hybrid virus integrates preferentially at a variety of sites distributed across the RNU1 locus but never at the equally fragile RNU2 locus (44, 45).