PKD1 Unusual DNA Conformations Are Recognized by Nucleotide Excision Repair*

The 2.5-kilobase pair poly(purine·pyrimidine) (poly(R·Y)) tract present in intron 21 of the polycystic kidney disease 1 (PKD1) gene has been proposed to contribute to the high mutation frequency of the gene. To evaluate this hypothesis, we investigated the growth rates of 11Escherichia coli strains, with mutations in the nucleotide excision repair, SOS, and topoisomerase I and/or gyrase genes, harboring plasmids containing the full-length tract, six 5′-truncations of the tract, and a control plasmid (pSPL3). The full-length poly(R·Y) tract induced dramatic losses of cell viability during the first few hours of growth and lengthened the doubling times of the populations in strains with an inducible SOS response. The extent of cell loss was correlated with the length of the poly(R·Y) tract and the levels of negative supercoiling as modulated by the genotype of the strains or drugs that specifically inhibited DNA gyrase or bound to DNA directly, thereby affecting conformations at specific loci. We conclude that the unusual DNA conformations formed by the PKD1poly(R·Y) tract under the influence of negative supercoiling induced the SOS response pathway, and they were recognized as lesions by the nucleotide excision repair system and were cleaved, causing delays in cell division and loss of the plasmid. These data support a role for this sequence in the mutation of the PKD1 gene by stimulating repair and/or recombination functions.

Polycystic kidney disease (PKD) 1 encompasses a family of closely related syndromes characterized by intraparenchymal renal cysts that are lined by a single layer of epithelial cells. The several forms of PKD include autosomal dominant polycystic kidney disease (ADPKD), which is one of the most common inherited human disorders (ϳ1 per 500 worldwide). Affected individuals typically develop large cystic kidneys, but hepatic cysts, intracranial aneurysms, and cardiac valvular abnormalities are extra-renal manifestations often associated with this disorder. Approximately half of ADPKD patients develop end stage renal disease requiring renal replacement therapy and compose ϳ5% of the chronic dialysis population in the United States (reviewed in Refs. [1][2][3][4][5]. The genetic defect in Ͼ85% of ADPKD cases is mutations in PKD1, a gene that encodes a transcript of 14 kilobases from 46 exons spanning 50 kbp on chromosome 16p13.3 (6). The 5Ј portion of the gene (exons 1-34) is duplicated with more than 95% homology in at least three other copies on chromosome 16p13.1, from which transcripts of 20, 17, and 8.5 kilobases are released (7,8). However, it is unclear whether these PKD1-like mRNAs are translated into proteins. Polycystin-1, the product of PKD1, is thought to be involved in cell-cell and cell-matrix interactions (9 -11) and in calcium-permeable non-selective cation currents (12).
Genotypic analyses of PKD1 microsatellite markers revealed that cysts originate from a single cell that, in a number of cases, underwent loss of heterozygosity. These results led to the proposal that cysts form by a "two-hit" process including a germ line mutation and a subsequent somatic mutation on the functional allele leading to loss of function (13)(14)(15). The rate of somatic mutations must be high given the frequent occurrence of ADPKD and the very large number of cysts observed, suggesting the existence of a local hot spot for mutagenesis. The identification of mutations in PKD1 has been hampered by the presence of the homologous genes; several of the 92 mutations identified so far are confined to the 3Ј-unduplicated region. Therefore, it is unclear whether the frequency of mutations may vary throughout the 50 kbp of the gene (6 -8, 13, 17, 89 -106).
The PKD1 gene is intersected in intron 21 by an extraordinary 2.5-kbp poly(purine⅐pyrimidine) (poly(R⅐Y) tract (16), 1 of the 10 longest sequences of this kind. This poly(R⅐Y) tract is 66% G⅐C-rich with 95% C ϩ T in the sense strand and is partly repeated in introns 1 and 22. These poly(R⅐Y) sequences, which are also present in the PKD1-like homologues, have been proposed to contribute to the high mutation rate of PKD1 (17,18). A computer search of the 2.5-kbp R⅐Y sequence revealed 23 mirror repeat sequences, which would be expected to adopt three-stranded DNA structures (intramolecular triplexes) (16) with stems of at least 10 bp. Also, 163 direct repeat sequences were identified, which may adopt slipped, mispaired conformations (16). These structures can block transcription and/or replication and thus induce repair functions (19 -23).
Herein, the growth rates of various repair and DNA topological mutant strains of Escherichia coli that harbored plasmids containing the 2.5-kbp poly(R⅐Y) tract from intron 21 were analyzed. Unusual intramolecular DNA conformations, formed under the influence of negative supercoiling, induced the SOS system and were recognized as lesions by the NER pathway. Delays in cell division were observed as well as loss of the plasmid, suggesting that such DNA conformations were substrates for strand breaks and endonuclease activities.

Growth Curves
E. coli strains were transformed with plasmids by the CaCl 2 method (27) and selected on agar plates containing ampicillin (Ap). In all experiments that involved Ap, 100 g/ml were used. Isolated colonies (one or more depending on size) were transferred to 10 ml of LB medium with Ap and grown to an A 600 of Ϸ0.8 -1.1. The number of viable colony-forming units (CFU) from such cultures ("founder cells") was determined by plating dilutions on agar plates without Ap, during which time the 10-ml cultures were kept at 4°C. This time varied from 1 to 5 days. Control experiments indicated that during this storage time the number of CFU decreased by an average of 33 Ϯ 19%. A known number of CFU was then inoculated into triple-baffled 2-liter fermentation flasks with 1 liter of LB medium pre-warmed to 37°C, and Ap was added. These flasks were shaken at 250 rpm at 37°C, and 5-10-ml aliquots were withdrawn after 5 min (t 0 ) and subsequently at every 15-30 min for 8 -10 h. To determine the number of CFU/ml, the aliquots were diluted in LB and plated on agar plates without Ap.
The following parameters were determined for each growth curve.
The Doubling Time-The doubling time (t 2 ) during the exponential phase of growth was obtained from the relation t 2 ϭ ln 2/k, where k was the slope as shown in Equation 1, ln CFU/ml ϭ ln(CFU/ml) 0 ϩ kt (Eq. 1) The standard error of t 2 (E t2 ) was derived from the standard error of k (E k ) according to E t2 ϭ t 2 E k /k. The Maximum Fold Decrease in CFU/ml (MDC)-During the first few hours of growth, the number of CFU/ml either decreased or increased non-exponentially. These data were fit by high degree polynomials. MDC was obtained by dividing the CFU/ml at time 0 ((CFU/ml) t0 ) by the CFU/ml at the curve minimum ((CFU/ml) t(min) ). When there was no decrease in CFU/ml, (CFU/ml) t(min) ϭ (CFU/ml) t0 and MDC ϭ 1.
Duration of the Non-exponential Phase of Growth (t A )-We define t A as the time it would have taken for the bacterial population to reach (CFU/ml) t0 if it grew exponentially from the theoretical (CFU/ml) 0 given by Equation 1. Accordingly, t A ϭ ln((CFU/ml) t0 /(CFU/ml) 0 )/k.

Topology of Closed Circular DNA
The following relations were used to analyze the superhelical density of plasmids. For relaxed, circular DNA Lk 0 ϭ Tw 0 ϩ Wr 0 , where Tw 0 ϭ n/h 0 , n is the total number of bp and h 0 the helical pitch.
Addition of a ligand that binds DNA and alters h 0 (twist) gives Lk 0t ϩ Wr 0t under conditions where DNA is relaxed (i.e. by topoisomerase I). Subsequent ligand removal in the absence of strand breakage yields Lk 0t ϭ Tw 0 ϩ Wr t , where Wr t ϭ Wr 0 ϩ t , and t is the number of ligand-induced superhelical turns. For an in vivo population of negatively supercoiled topological isomers, ϽLkϾ ϭ Tw 0 ϩ ϽWrϾ (ϽWrϾ ϭ Wr 0 ϩ ϽϾ, where is the number of negative supercoils). Addition of a ligand in vivo in the presence of homeostatic control of negative supercoiling, followed by its removal in the absence of strand breaks, yields ϽLkϾ ϭ Tw 0 ϩ ϽWr t Ͼ. This population of topoisomers will have, on average, fewer negative supercoils when the ligand increases twist and more negative supercoils when the ligand decreases twist (28,29).

RESULTS
Sequence Analysis of the Poly(R⅐Y) Sequence-A previous computer search performed on the 2.5-kbp poly(R⅐Y) identified the presence of 23 perfect mirror repeats, which can form DNA triple helices (16). We find that these motifs are clustered into three distinct regions separated by ϳ600 and 450 bp, respectively ( Fig. 1). A search for perfect tandem repeats (30), which will form slipped structures (31,32), revealed more than a thousand. The dinucleotides, TC and CT (row a), are most common; however, they are excluded from the 5Ј-end where the mirror repeats predominate. These dinucleotides are found with increasing frequency toward the 3Ј-end. The trinucleotide repeats are mostly CCT and are localized within the 5Ј-half of the tract, with the exclusion of the very 5Ј-end (row b). Pentanucleotide repeats are of three types as follows: CTCCC, CTCCT, and CCCAT. The first type (row c) can be further subdivided into three clusters according to the following reading frames: 8 CTCCC tandem repeat units (TRU) at the 5Ј-end, 5 CCTCC TRU in the middle, and a cluster of 19 TCCCC TRU at the 3Ј-end. The CTCCT direct repeats are positioned exclusively in the middle of the tract, interspersed by a single CCCAT tandem repeat (row d). Longer direct repeats are common as shown in row e. The tract also contains close runs of guanines throughout its length that may form four-stranded structures (G-quartets) (33).
In summary, although many mirror and direct repeats are present, their locations are clustered, suggesting highly specific evolutionary duplication events. To the best of our knowledge, this tract contains the highest density of unorthodox simple sequence repeat features (mirror, direct repeats, and R⅐Y strand bias) of any known sequence of this length.
Previous analyses revealed that under superhelical stress the 2.5-kbp poly(R⅐Y) tract was cleaved by the single-stranded specific nuclease P1 at four locations, suggesting the formation of unusual structures (19). The location of the four P1 nucleasesensitive sites is not coincident with the position of the direct repeats with Յ5 bp (Fig. 1); however, sites 1 and 2 do superimpose with clusters of overlapping mirror repeats, as well as with two of the longest direct repeats. This suggests that the unusual DNA structures formed by the poly(R⅐Y) sequence in vitro include intramolecular triplexes and mispaired loops.
Length-dependent Delay in Cell Growth-To determine whether such structures form and influence cell physiology, we investigated the effect of the full-length tract and its truncations ( Fig. 2A) on the growth rate of wild-type E. coli KMBL1001. Fig. 2B shows the normalized growth curves obtained with each of the plasmids. The doubling time was lengthened in proportion with the length of the inserts from 21.9 min for pBS0.8 to 24.4 min for pBS4.0. Four plasmids (pBS4.0, pBS1.8, pBS1.5, and pBS0.8) caused a decrease in CFU/ml during the first few hours of growth ( Fig. 2B, inset). This loss of viable cells was largest for pBS4.0 that had the full-length poly(R⅐Y) tract, and it decreased in proportion to the length of the inserts. Finally, the duration of the non-exponential phase of growth (lag period) also lengthened in accordance with the increasing insert lengths, from 0.8 h for pBS0.8 to 1.8 h for pBS4.0.
Thus, we conclude that the poly(R⅐Y) sequence compromised the viability of a number of cells in a length-dependent manner and contributed to the lengthening of the population doubling time as well as the lag period, suggesting that the magnitude of these effects correlated with the number of unusual DNA structures that formed.
DNA Topology, Repair Functions, and SOS Response-To identify genetic factors associated with cell death, growth curves were determined for E. coli mutant strains that affected DNA supercoil density, nucleotide excision repair (NER), and DNA damage-induced SOS response (Table I). The parameters (MDC, t 2 , and t A ) obtained in E. coli strains harboring pBS4.0 were compared with those obtained for pSPL3, a control chosen because of its size (6.0 versus 7.0 kbp for pBS4.0), the absence of long repetitive sequences, and because, like pBS4.0, it was a derivative of pBluescript.
The results (Table II) reveal the following. First, the extent of cell loss mediated by pBS4.0 correlated with the steady-state level of negative supercoiling (as determined in vivo for pUC19, not shown) because the magnitude of MDC followed the order topA10 Ͼ wt Ͼ topA10gyrB226. Since the number and stability  8. Growth curves were normalized by dividing the CFU/ml at time ϭ 0 ((CFU/ml) t0 ) and at subsequent times ((CFU/ml t )) by (CFU/ml) t0 . The numbers of CFU/ml used to start the 1-liter cultures are as follows: pBS4.0, 3200 (filled circles); pBS1. 8, 3700 (open circles); pBS1.5, 3100 (squares); pBS1.4, 2800 (triangles); pBS1.3, 1600 (inverted triangles); pBS1.0, 2500 (diamonds); and pBS0.8, 2400 (hexagons). Inset, expanded version of the data at the early time points. Linear regressions of the data (in the order reported above) for MDC, t 2 , and t A gave r 2 values of 0.68, 0.46, and 0.87, respectively. of underwound unusual DNA structures increase with the extent of negative supercoiling (20 -23, 28, 34, 35), we conclude that supercoiling stabilized their formation and that this was critical for the loss of cell viability. Furthermore, the differences in doubling time between pSPL3 and pBS4.0 in the three top/gyr strains (15.3, 9.3, and 5.4 min, respectively) also depended upon negative supercoiling, suggesting that the formation of unusual DNA structures interfered with cell division.
Second, E. coli NER mutants revealed that UvrB and UvrC were necessary for the loss of cells. In fact, the MDC was negligible for both ⌬uvrB and ⌬uvrC strains compared with their isogenic wild-type KMBL1001. In addition, for ⌬uvrB there was no lengthening of the doubling time, whereas ⌬t A was shortened to ϳ15 min from Ն1.3 h observed for all the other strains. The second significant result from this analysis was the considerable loss of cells in the ⌬uvrA and ⌬uvrD strains harboring only the control plasmid pSPL3. Because deficiency in UvrD is known to lead to constitutive expression of the SOS response (36 -38), a likely explanation is that the SOS system may induce apoptosis during the first period of cell population growth.
Third, we investigated the E. coli SOS strains. During the preparation of JJC523 (⌬lexA71) "founder cells" harboring pBS4.0, we observed significant cell lysis that yielded ϳ1 ϫ 10 5 CFU/ml from overnight cultures compared with Ͼ1 ϫ 10 7 obtained normally with the same strain harboring pSPL3 or all the other strains with either pBS4.0 or pSPL3. This result clearly indicates that pBS4.0 caused extensive cell death when associated with a fully active SOS response. The growth curve started from the surviving cells did not show a large decrease in cell count. The MDC value with pBS4.0 for the wild-type JJC510 was much lower than for the other two wild-type strains, JTT1 and KMBL1001. We observed that the values of MDC (and therefore t A ) for the three wild-type strains, for topA10, and for topA10gyrB226, varied considerably between experiments. Further analyses indicated that most of this variability arose during the preparation of founder cells used to start the growth curves. At that stage a variable number of cells lost their plasmid; these cells then succumbed at the beginning of the growth culture, when MDC was measured. In the lexAind1 strain JJC123 harboring pBS4.0, where the SOS system is not inducible, the MDC value was comparable to that of its isogenic wild-type strain, JJC510.
In summary, the data indicate that the poly(R⅐Y) sequence from intron 21 of PKD1 formed unusual DNA structures as a consequence of negative superhelical density and that the detrimental effects on growth depended upon activation of the SOS system. Furthermore, both UvrB and UvrC were necessary to elicit the pBS4.0-dependent lethality.
Integrity of the poly(R⅐Y) Sequence in Surviving Cells-Previous work showed that long repetitive sequences such as (CTG⅐CAG) n and (CGG⅐CCG) n are unstable and that the average length of the repeats shortens as the cell population ages (39 -42). Therefore, we investigated whether the surviving cells represented a subpopulation that had deleted part, or all, of the poly(R⅐Y) tract. pBS4.0 was isolated from all of the strains and analyzed in two ways. First, the 4.0-kbp PKD insert was excised from the vector, and the insert and vector were separated by agarose gel electrophoresis. Most lanes showed considerable smearing throughout their length. How- ever, no discrete products were visible besides the two expected bands, except 1 sample out of 33 where a recombination event took place. Second, these inserts and vectors were radioactively labeled, and their molar ratio was calculated. The ratios ranged from 0.47 Ϯ 0.07 to 0.64 Ϯ 0.06; however, no statistical significance was found among the mean values, and no correlation was observed between these ratios and the MDC values ob-tained for the same E. coli strains. Thus, we conclude the following. First, loss of the full-length poly(R⅐Y) tract in favor of more stably transmitted deletion products was not observed. Second, both the smearing and the lower molar ratio of the PKD insert indicate that a proportion of pBS4.0 was in the process of being degraded at the time of plasmid isolation and that the starting point for such degradation was within the PKD insert. We suggest that pBS4.0 was either replicated and transmitted intact during cell division or cleaved at the poly(R⅐Y) tract and then rapidly degraded.
Influence of Negative Supercoiling-To verify further the differences observed in the topoisomerase I and gyrase E. coli mutants, as well as the ⌬uvrB strain, MDC was evaluated in identical cells following alterations in their levels of negative supercoiling. We used novobiocin to inhibit the assembly of active gyrase (43,44) and thus achieve a relaxation of the DNA in vivo. This should reduce the formation of unusual DNA structures and relieve the extent of founder cell loss. Fig. 3 shows that the population of more highly negative supercoiled topoisomers of pUC19 was progressively reduced in the presence of novobiocin, as expected. Fig. 4A shows the growth curves for the topA10 strain carrying pBS4.0 in the absence or presence of novobiocin. Addition of novobiocin during the preparation of cells maintained their complete viability in the subsequent growth, whereas Ͼ90% was lost in its absence. Superimposable growth curves were found for 2Ϫ10 M novobiocin. Cell loss was not prevented when novobiocin was added at the beginning of the growth curves rather than during the preparation of founder cells, indicating that the loss required an activity to take place at an early step.
An alternate strategy was also implemented to influence supercoiling, and hence DNA structure, by preparing cells in the presence of drugs (actinomycin D or netropsin) that bind directly to DNA and thereby influence its global topology as well as its conformations at specific sequences. Analyses of the topoisomer distributions of pUC19 in the presence of either drug confirmed their binding to the DNA (not shown).
Actinomycin D intercalates at G⅐C pairs, and it reduces the number of negative superhelical turns and stabilizes DNA in the right-handed duplex B-form (45)(46)(47)(48)(49)(50). Due to its G-C richness, the poly(R⅐Y) tract offers numerous binding sites for the drug, whose activity is expected to decrease the number of inviable cells. Fig. 4B shows the growth curves for the wild-type E. coli strain KMBL1001 harboring pBS4.0 in the absence or presence of actinomycin D. As described for novobiocin, actinomycin D was only added during the preparation of cells. The extent of cell loss within the first 2 h decreased progressively in the presence of 5Ϫ30 M actinomycin D, supporting the hypothesis that the formation of underwound non-B DNA structures was responsible for cell loss.
Netropsin binds to the minor groove of A⅐T pairs, requiring four or more such bp for optimal contacts, introduces additional negative supercoils, destabilizes triplex DNA, and increases the stiffness of the double helix (26,(51)(52)(53)(54)(55)(56)(57)(58). We reasoned that netropsin would accentuate cell loss in two ways. First, it would bind to duplex B-DNA and induce additional negative supercoiling. Second, it would bind selectively to the vector sequences (of the 131 consecutive four A-T pairs only 7% are in the poly(R⅐Y) tract) and thus increase the stiffness of the vector. Thus, the superhelical density would be preferentially partitioned into the poly(R⅐Y) tract, which is the most flexible and writhed region of the plasmid (59). Fig. 4C shows the growth curves for wild-type E. coli KMBL1001 carrying pBS4.0 in the absence and presence of FIG. 3. Negative superhelical density in vivo. Topoisomers of pUC19 were isolated from E. coli strain KMBL1001 grown to A 600 of 0.5 in the absence or in the presence of 5 M novobiocin and analyzed by agarose gel electrophoresis (1% w/v) in 90 mM Tris borate, 2 mM EDTA, pH 8.0, in the presence of chloroquine. The gel negatives were scanned with a PhosphorImager and the data were smoothed by a low pass filter that used a fast Fourier transform to remove abnormal high frequencies. The y axis is the area of each peak in pixels as given by the PhosphorImager. The negative superhelical density (x axis) was obtained as described (35). Data were fit to a three-parameter Gaussian curve. The ⌬t 2 values (t 2drug Ϫ t 2control ) for A-D were 3.2, Ϫ0.9, 0.5, and Ϫ0.2 min, whereas the ratios MDC drug /MDC control were 0.05, 0.05, 16.7, and 0.8. netropsin. As described above, the drug was only added during the preparation of cells. The data show that netropsin strongly accentuated cell selection during the first 5 h. The growth curves were not affected by the addition of netropsin to E. coli KMBL1001 cells harboring pSPL3 or when the compound was added to E. coli KMBL1001 containing pBS4.0 only at the beginning of the growth curve culture.
To verify further whether cell loss was caused by the formation of non-B DNA structures alone or whether it required their recognition by the NER system, the experiment with netropsin was conducted in E. coli ⌬uvrB carrying pBS4.0. Contrary to the wild-type strain, ⌬uvrB cells were unaffected by the drug (Fig. 4D), proving that the recognition of unusual DNA structures by NER was indispensable for eliciting cell selection.
In summary, we conclude that optimal viability required the DNA to be maintained in an orthodox right-handed B-form. We also conclude that the reactions of the NER proteins on the underwound non-B DNA structures, but not their formation alone, was indispensable for eliciting loss of cell viability.
Loss of Plasmid-The relationship between the SOS and NER pathways and cell viability became apparent when colonies derived from the exponential phase of the growth curves were analyzed. Samples of E. coli transformed with pSPL3 taken throughout the period of growth yielded colonies of essentially homogeneous size when plated without Ap. Alternatively, colonies derived from several E. coli strains transformed with pBS4.0 plated after the cultures grew for 3-5 h appeared heterogeneous in size. DNA isolated from small and large colonies revealed that pBS4.0 was only present in the small but not in the large colonies.
To determine the extent of plasmid loss for pSPL3 and pBS4.0, the number of Ap-resistant (Ap R ) and Ap-sensitive colonies for various E. coli strains was measured. Fig. 5A shows the logarithm of the ratio between the number of Ap R colonies and the total number of colonies for E. coli strains harboring either of the two plasmids. For pSPL3 (Fig. 5A, hatched bars), the greatest differences were observed among the wild-type strains, whereas only small differences were seen between the wild-types and their respective mutants. Ten and 90% of the JTT1 and JJC510 cells, respectively, contained the plasmid, whereas less than 0.1% of cells retained pSPL3 for the KMBL1001 strain. Retention of pSPL3 was slightly greater (2-5-fold) in the ⌬uvrB and ⌬uvrC mutants (the ⌬uvrA and ⌬uvrD mutants were not tested). Addition of netropsin to KMBL1001 cells caused the dramatic reduction by 4 orders of magnitude in the number of plasmid-containing cells, whereas addition of actinomycin D or novobiocin slightly increased the retention of pSPL3 5-and 10-fold, respectively.
These results indicate that plasmid stability was improved when negative supercoiling was lessened, as expected. The topA10 mutation increased pSPL3 instability by about 5-fold, whereas the stability increased 8-fold in the topA10gyrB226 strain, confirming that supercoil tension was detrimental to plasmid stability. Inactivation or constitutive expression of the SOS system in the ⌬lexA71 and lexAind1 strains, respectively, did not affect the retention of pSPL3.
In summary, these results with pSPL3 show that plasmid loss commonly took place during cell growth, that the extent of loss depended on the host genetic background, and that it correlated with negative supercoil density.
As for pSPL3, pBS4.0 was least stable in wild-type KMBL1001 as compared with JTT1 and the JJC510 strains. However, its retention was reduced relative to pSPL3 by 1000-, 10-, and 1000-fold in KMBL1001, JTT1, and JJC510, respec-  tively, indicating that the PKD1 insert was detrimental to plasmid stability. Significantly, retention of pBS4.0 improved by nearly 6 orders of magnitude in the ⌬uvrB mutant strain, strongly suggesting that loss of pBS4.0 was mediated by NER. Deletion of the uvrC gene or addition of netropsin to KMBL1001 did not further increase the number of cells without pBS4.0, whereas growth of KMBL1001 cells both in the presence of actinomycin D or novobiocin increased the retention of pBS4.0 by 4 and 2 orders of magnitude, respectively. The topA10 mutation decreased the number of cells containing pBS4.0 by 10,000-fold, whereas the topA10gyrB226 mutation increased their fraction 10-fold. Because both the genetic and pharmacological approaches agreed that loss of pBS4.0 was greatly dependent on negative supercoiling, we conclude that alternative DNA structures in the poly(R⅐Y) tract were critical for leading to plasmid loss through their interaction with the NER pathway. Inactivation of the SOS system in the lexAind1 strain increased pBS4.0 loss about 10-fold. The results with the ⌬lexA71 cells were difficult to interpret because of the substantial lysis observed in the presence of pBS4.0, as noted above.
Hence, this analysis suggests that (i) plasmid loss is a general phenomenon that depends on the host genetic background; (ii) negative supercoiling increases the propensity of losing a plasmid; (iii) the PKD1 insert in pBS4.0 greatly increases the propensity to lose the plasmid; (iv) the UvrB component of the NER system is required for the PKD1 sequence-mediated plasmid loss; and (v) constitutive induction of the SOS system leads to cell lysis in the presence of pBS4.0.
To determine whether cell loss during the growth curves (Table II) could be explained by the behavior of cells that lack plasmid, untransformed wild-type E. coli JTT1, JJC510, KMBL1001, and the ⌬uvrB mutant were prepared in the absence of Ap and then used to mock-start a growth curve in fresh medium with Ap. Fig. 5B shows that Ͼ98% of these cells lost their viability within the first 2 h of growth. This result is consistent with the hypothesis that cell death was mediated by the selection stress on cells that had lost their plasmid during the previous culture.
In summary, these data show that the poly(R⅐Y) sequence from intron 21 of the PKD1 gene contains unusual DNA structures that form under the influence of negative supercoiling, that these structures interact with components of the NER pathway, and that such interactions lead to plasmid loss.  (61,68) and triplexes was documented (63,69,70). It was hypothesized (60 -67), but not proven, that these instabilities were due to the capacity of these sequences in vivo to adopt non-B DNA conformations. Herein, we provide genetic evidence that the unusual conformations and not the sequences per se are responsible. Several cellular factors (i.e. NER, the SOS system, and gyrase and topoisomerase I functions) are involved, which interact with or influence the stability of the unorthodox DNA structures.
These data are the first to link directly DNA conformations and their polymorphic behaviors with mutagenesis in vivo. The capacity of certain types of DNA sequences to undergo facile transformations from right-handed B structures to unorthodox conformations (i.e. triplexes, slipped structures, etc.) is an integral component of these studies.
The role of DNA topology (supercoiling) was evaluated with mutant strains and with a gyrase inhibitor (novobiocin) that affected supercoil density as well as with ligands (actinomycin D and netropsin) that bind directly to DNA and thereby influence its global topology and hence its conformations at specific sequences. The composite data show that unusual DNA structures (triplexes and slipped structures) in the 2.5-kbp poly(R⅐Y) tract from the PKD1 gene exist in E. coli and are responsible for the observed genotoxicity. The PKD1 gene is a human sequence on chromosome 16p13.3, and hence the types of structural transitions formed in this plasmid system may also occur in humans. Fig. 6 outlines the main steps that may be involved in the process. The in vivo homeostatic control of negative supercoiling (71-73) (Fig. 6, A and B) generates plasmid topoisomers in which plectonemic conformations maintain the DNA under highly negative torsional stress (74,75). The poly(R⅐Y) tract undergoes structural transitions (Fig. 6C) to various non-B structures (19 -23) under the influence of both steady-state levels of negative supercoiling and waves of hyper-negative supercoiling generated by the passage of DNA helix-tracking enzymes such as during transcription, replication, or repair. Underwound unusual DNA structures should be dissipated as they are approached by an incoming polymerase due to positive supercoiling formed ahead of the complex (76 -79) and thus are predicted to not interfere with replication. E. coli strains defective in the SOS-induced genes, specifically ⌬uvrA and ⌬uvrB (80), failed to show a lengthening of the population doubling time, in contrast with the strains with a functional SOS system. Thus, unusual DNA structures may have interfered with cell division only after DNA damage (i.e. strand-break(s)) inflicted on them by NER (81) caused replication forks to collapse (82)(83)(84)(85) (Fig. 6, D and E).
Experiments performed previously with plasmids containing (CTG⅐CAG) triplet repeats (39,67,86,87) showed that the tract frequently deletes to discrete and stably transmitted new species in E. coli populations and that both the MMR and NER systems are involved (67,86,87). On the contrary, we show here that the poly(R⅐Y) tract from the PKD1 gene does not give rise to such discrete and stably transmitted deletions. A possible reason for the difference is that the presence of the triplet repetitive elements within the (CTG⅐CAG) tracts facilitates post-DNA damage repair and/or recombination and thus maintains plasmid viability, which is lost when similar damage is inflicted within the poly(R⅐Y) tract. We also show (Fig. 4) that the NER nuclease activity related to DNA stability is not directed to a particular DNA sequence but rather to DNA structures that may arise from several types of sequences. Therefore, we conclude that the NER system has the capacity of recognizing certain non-B DNA structures formed by undamaged bases, in addition to recognizing damaged bases (88). Future in vitro studies with the highly purified components of the NER system may provide further insights into the substrate requirements.