Rev1, Rev3, or Rev7 siRNA Abolishes Ultraviolet Light-Induced Translesion Replication in HeLa Cells: A Comprehensive Study Using Alkaline Sucrose Density Gradient Sedimentation

When a replicative DNA polymerase stalls upon encountering a lesion on the template strand, it is relieved by other low-processivity polymerase(s), which insert nucleotide(s) opposite the lesion, extend by a few nucleotides, and dissociate from the 3′-OH. The replicative polymerase then resumes DNA synthesis. This process, termed translesion replication (TLS) or replicative bypass, may involve at least five different polymerases in mammals, although the participating polymerases and their roles have not been entirely characterized. Using siRNAs originally designed and an alkaline sucrose density gradient sedimentation technique, we verified the involvement of several polymerases in ultraviolet (UV) light-induced TLS in HeLa cells. First, siRNAs to Rev3 or Rev7 largely abolished UV-TLS, suggesting that these 2 gene products, which comprise Polζ, play a main role in mutagenic TLS. Second, Rev1-targeted siRNA also abrogated UV-TLS, indicating that Rev1 is also indispensable to mutagenic TLS. Third, Polη-targeted siRNA also prevented TLS to a greater extent than our expectations. Forth, although siRNA to Polι had no detectable effect, that to Polκ delayed UV-TLS. To our knowledge, this is the first study reporting apparent evidence for the participation of Polκ in UV-TLS.


Introduction
Multiple systems have evolved to manage the genomic photoproducts generated by harmful UV light. One such system is nucleotide excision repair (NER), which eliminates photoproducts from DNA strands by dual incision on both sides of a damaged base. The NER system cannot, however, remove all UV-damaged bases. When a replicative DNA polymerase stalls upon encountering a residual photoproduct on the template strand, it is relieved by other low-processivity polymerase(s), which incorporate nucleotide(s) opposite the lesion, extend by a few nucleotides and dissociate from the 3 -OH. The replicative polymerase then resumes DNA synthesis. This process, termed translesion replication (TLS) or replicative bypass (reviewed in [1]), is also one of the subtle systems that have evolved for the management of genomic photoproducts.
UV-C (100-290 nm wavelength) induces 2 main photoproducts [2]: the more frequent cyclobutane pyrimi-dine dimer (CPD) and the several-fold lower pyrimidinepyrimidone (6-4) photoproduct ((6-4)PP). cis-syn CPD, a predominant form of the multiple configurations, contains 2 adjacent pyrimidines that are covalently linked in parallel. Although the frequency of CPD varies with nucleotide composition, a ratio of T-T to C-T to T-C to C-C of 68 : 13 : 16 : 3 is obtained from UV-irradiated plasmid DNA. Cytosines within CPD are unstable, and are deaminated to uracil or 5methylcytosine, and further deaminated to thymine [3]. The helical distortion caused by CPD is so inconspicuous that almost half of the lesions remain unrepaired by NER, even 6 hours after UV irradiation in the case of CHO cells [2].
The PPs from T-C, C-C, and, less frequently, T-T sequences are detected in UV-irradiated DNA whereas that of C-T are not. In PP, linkage between C-6 of one pyrimidine and C-4 of the adjacent pyrimidine cause the 2 bases to be in nearly perpendicular position. Consequently, formation of this lesion causes a major distortion in the double helix. NER preferentially removes (6-4)PP more rapidly than it removes CPD from the genome in human and rodent cells [4].
At least 5 mammalian DNA polymerases are suggested to be implicated in UV-induced TLS: Pols η, ι, ζ, κ, and Rev1, all of which belong to the Y family except for Polζ (B family) (reviewed in [1,5,6]). However, the participating polymerases and their roles have not been entirely characterized.
Patients with the autosomal recessive disorder, xeroderma pigmentosum variant (XP-V), have a predisposition to skin cancer, and XP-V cells demonstrate hypermutability after UV irradiation (reviewed in [7]). The defective gene in XP-V encodes Polη, which was first purified from a HeLa cell extract as an activity that complements TLS defect in XP-V cell extract [8]. Human Polη catalysed DNA synthesis past TT-CPD very efficiently and in a relatively accurate manner, as demonstrated by the lesion-bypass assay [7,9]. When template DNA contained a (6-4)TT-PP, Polη incorporated one (random) nucleotide opposite the first thymine and another nucleotide opposite the second thymine of the lesion, but rarely continued across the lesion [7,9].
Human Polη was also identified via a search for the homolog of yeast Saccharomyces cerevisiae Rad30 gene, which encodes an error-free bypass protein [10]. Various XP-V causative mutations have been found in the Polη gene, hRAD30A, of XP-V patients [10,11].
Polι (RAD30B) is the other mammalian homolog of yeast Polη, isolated by a similar approach [12]. In contrast to Polη, Polι is less efficient and less accurate [13].
Polκ was obtained by cloning of a human homolog of the E. coli dinB gene, encoding DNA Pol IV [14]. Polκ was reported to be unable to bypass either CPD or (6-4)PP [15,16].
Originally, Rev1, 3, and 7 were cloned from S. cerevisiae isolates, in which the frequency of UV-induced reversion from cyc1 mutations was reduced [17]. Human and mouse homologs (Rev1, 3, 7 gene) were later isolated [18][19][20]. Polζ is a comprex of the Rev3 and Rev7 gene products, which act as the catalytic and regulatory subunit, respectively. Yeast Polζ is shown to be responsible for 98% of UV-induced base substitutions and 90% of frameshift mutations, in addition to spontaneous mutations [21]. Nonetheless, yeast Polζ itself was revealed to be too faithful to incorporate opposite CPD. Instead, it can efficiently extend from a matched or mismatched 3 -end (reviewed in [5,6]). Human or mouse Polζ is assumed to have similar enzymatic properties to that of yeast, because several lines of antisense RNA expression or siRNA knockdown experiments in human or mouse cells have proven that Polζ is involved in mutagenic TLS [22][23][24].
Yeast and human Rev1 encode highly specialized DNA polymerases that preferentially insert a C residue opposite an abasic site in the template. This deoxycytidyl (dC) transferase activity is, however, unlikely to be required for UV-TLS, judging by observation in a yeast rev1-1 mutant strain [25], which retains much of its dC transferase activity, but has a missense mutation (G193R) in the N-terminal BRCT domain. Rev1 protein also contains ubiquitin-binding motifs (UBMs) that interact with monoubiquitinated PCNA (a DNA polymerase sliding clamp) [26]. In the downstream Cterminal region, Rev1 interacts not only with Rev7 but also with other bypass-polymerases [27], suggesting that Rev1 acts as a mediator and physical bridges between PCNA and Polζ.
Following DNA damage, such as that caused by UV and MMS, monoubiquitins are conjugated to PCNA arrested at the lesion-site by RAD6/RAD18 and recruit bypasspolymerases [28,29]. In addition to Rev1, Pols η, ι, and κ possess UBMs and physically interact with PCNA [1,5,6]. Stalled replicative Polδ is replaced, in turn, with one of these bypass polymerases bound on the PCNA by yet unknown "polymerase switching" mechanisms (reviewed in [1,30]).
Translesion replication is typically detected with an alkaline sucrose density gradient centrifugation (ASDG) technique. Pulse-labelled replication products are smaller in UVirradiated XP-V cells than in unirradiated cells; however, on prolonged incubation, the replication products in the irradiated cells eventually attain a high molecular weight similar to that in unirradiated cells. This conversion is interpreted that DNA synthesis is temporarily retarded by UV photoproducts, and then continues beyond the lesion, leaving a gap that is subsequently sealed [31]. The initial size of the newly synthesized DNA is approximately equal to the average distance between lesions in the template strands [32]. This means the gaps in the newly synthesized DNA are opposite the photoproducts [33]. Therefore, sealing of the gaps, by translesion or other postreplication repair mechanisms, can be observed by monitoring the molecular weight of labelled DNA.
Using a modified ASDG technique [34], we precisely detected the elongation of pulse-labelled replication products in the irradiated XP-V cells, showing that UV-TLS is delayed in the cells, but not completely abolished [35]. The marginal TLS is markedly prevented by caffeine at millimolar concentrations, as Lehmann et al. pointed out [31], and by proteasome inhibitors as well (unpublished results). In contrast, these agents do not retard UV-TLS in normal diploid cells. To know more about the inefficient polymerase(s) in vivo, we added specific DNA polymerase inhibitors. Butylphenyldeoxyguanosine (BuPGdR) inhibited TLS in XP-V cells [35], suggesting that Polζ may be involved in this Polη-independent bypass.
We recently reported that caffeine or proteasome inhibitors inhibit UV-TLS also in human cancer cells [36], and that, similar to XP-V cells, UV-TLS was much slower than in normal cells. These results suggested that UV bypass in cancer cells is predominantly of the Polη-independent type. Therefore, we expected that Polζ plays a major part in UV-TLS in cancer cells. Although Polη exists in normal quantity in these cells, it was supposed to be inactivated by some reasons. We explored these hypotheses here.

HeLa Cell Culture and siRNA Treatment.
HeLa cells were maintained in monolayers in Dulbecco's modified Eagle's medium (D'MEM) supplemented with 10% fetal calf serum (FCS) ("normal" medium), trypsinized and seeded into culture dishes (2 × 10 5 cells/φ 60 mm dish). About 6-8 hours later, the cells were treated with micelles of siRNA and OligofectAMINE (Invitrogen), formed according to the manufacturer's protocol, except Opti-MEM was replaced with D'MEM. The siRNA concentration used in RT-PCR analysis and western blot analysis was 5 nM.

UV Irradiation and Translesion
Replication. Forty hours after siRNA addition, HeLa cells were exposed to UV light (10 J/m 2 ) from a germicidal lamp (Toshiba GL15) at 0.6 J/m 2 per s. After 30 minutes in culture, the medium was changed to labeling medium consisting of D'MEM supplemented with 10% FCS and 10 μCi/mL of [U-14 C]thymidine (Moravek MC267, 470 mCi/mmol). UV-irradiated cells were pulse-labelled for 1 hour, while nonirradiated cells were labelled for 30 minutes. The medium was changed to normal medium, and the cells were chased for 5 hours. These cells were harvested by trypsinisation and examined by ASDG [34].

3.1.
Effect of siRNAs to Rev3 or Rev7 on HeLa UV-TLS. We selected 4 target sites from the human Rev3 mRNA sequence (NCBI locus: NM 002912), according to the JBioS guide. All the Rev3 siRNAs (Table 1) effectively knocked down expression of Rev3 (Figure 1(a)) and, at 5 nM, abolished UV-TLS in HeLa cells (Figure 1(b)). Replication products immediately after UV irradiation were sedimented as a sharp peak, illustrated by a thin line in Figure 1(b), slightly larger in size than the T4 phage DNA marker. When only Oligofectamine-treated cells were chased in normal medium for 5 hours, the products joined to form larger DNA with lengths in the order of megabases (Mb), illustrated by a thick line. In Rev3 siRNA-transfected cells, the products remain in smaller size, as depicted by a thin line with open marks, demonstrating these siRNAs prevent UV-TLS (Figure 1(b)).
We assessed how many mismatched nucleotides (nt) are necessary at minimum for the negative control siRNA (Figure 1(c)) and found that siRev3cont-C, designed from siRev3-D with 2 nt mismatches, did not prevent UV-TLS, indicating that these Rev3 siRNAs degrade Rev3 mRNA with high specificity. The dose-response profile of siRev3-D, shows that only 1 nM siRNA sufficiently inhibited UV-TLS (Figure 1(d)). The siRev3-D siRNA had no effect on normal replication (Figure 1(e)).

SiRNAs to Rev1
Significantly Abrogated UV-TLS. For knockdown of Rev1 expression, we selected 4 sites from the human Rev1 mRNA sequence (NM 016316) (Table 1) (Figure 3(a)). The siRNAs (5 nM) targeted to these sites abolished UV-TLS in HeLa cells (Figure 3(b)). The siRev1cont-E, designed from siRev1-C with 4 nt mismatches, had little to no effect. The dose-response profile of siRev1-C shows that 1 nM of the siRNA was enough to inhibit UV-TLS (Figure 3(c)); siRev1-C or siRev1-D siRNA had no effect on normal replication (Figure 3(d)).

Polκ siRNAs Delayed UV-TLS, While siRNAs to Polι
Did Not. Although Polι siRNAs efficiently prevented Polι expression ( Figure 5(a)), we could not detect these effects on ASDG profiles (Figure 5(b)). The siPolι-5 was reported by Journal of Nucleic Acids Choi et al. [38], and the target sequence of siPolι-A was 4 nt downstream from the one reported by Machida et al. [39] (human Polι mRNA sequence: NM 007195).

Discussion
We verified the involvement of multiple bypass polymerases in UV-TLS in HeLa cells using original siRNAs and ASDG technique, which is consistent with the recent model of 2 polymerase mechanisms [40,41]. Rev3 and Rev7, which comprise Polζ, were confirmed to participate in mutagenic UV-TLS. Also, Rev1 was suggested to play an important role Journal of Nucleic Acids in human TLS, although in avian DT40 cells, Rev1 may have a distinct role [42]. We were surprised to find that siRNAs against Polη prevented TLS to a great extent. TLS was delayed in Polκ siRNA-transfected cells, but not in siPolι-transfected cells.
We anticipated a limited participation of Polη, because UV-TLS in HeLa cells is very slow (i.e., inefficient) and caffeine-sensitive [35]. However, siRNAs against Polη, particularly siPolη-A and siPolη-B, prevented TLS to a great extent. Since both Rev3 and Rev7 siRNAs also significantly abolished UV-TLS, these results suggest that the Polζ-dependent TLS pathway and the Polη-dependent process are not mutually exclusive but overlapped.
Enzymology of yeast Polζ revealed that this polymerase is too faithful to insert nucleotides opposite a CPD, although it efficiently extends from a matched or mismatched 3 end [5,6]. Therefore, we assumed that mutagenic (errorprone) TLS proceeded through the insertion by Polι or Polκ of mismatched nucleotides opposite UV photoproducts, followed by extension by Polζ. Our data showed, however, that siPolι had no effect, and siPolκ partially prevented TLS. These results suggest that in some cases, Polη, and to a lesser extent Polκ, may insert nucleotide(s) opposite UV photoproducts, followed by extension by Polζ.
Polη is capable of bypassing a CPD without aid of other TLS polymerases. Both yeast and human Polη, however, incorporate wrong nucleotide at a fairly high rate and can extend these mismatched primer termini with only a frequency of ∼10 −2 to 10 −3 relative to extension from matched primer termini [6,43]. Plausibly, Polη dissociates from there and the proof-reading exonuclease of Polδ removes the wrong nucleotide [44]. To the primer termini, Polη is recruited again and incorporates a new nucleotide. This cycle is repeated until Polη incorporates a correct nucleotide. We suppose that disruption or malfunction of this cooperation renders mismatched primer termini accessible to Polζ.
Recently, Yoon et al. published 2 papers describing the effects of siRNA knockdown on the efficiency of TLS at TT-CPD [45] or (6-4)TT PP [46] on duplex plasmids in human cells. They also reported the effects of siRNA knockdown on mutation frequencies in the λ phage cII gene lysogenized in mouse cells expressing a (6-4)PP photolyase [45] or CPD photolyase [46]. The results of this tremendous and detailed study demonstrated that Pols η, κ, and ζ contribute to CPD bypass, wherein Pols κ and ζ promote mutagenic TLS and Polη executes error-free bypasses (Polι siRNA had no effect) [45]. As for (6-4)PP bypass, Pols η and ι provide alternate pathways for mutagenic TLS, and Polζ acts in a predominantly error-free manner (Polκ siRNA had no effect) [46].
The participation of Pols κ or ι in CPD bypass was similarly demonstrated by our results and those of Yoon et al. [45]. Because (6-4)PP is a minor photoproduct, which is removed predominantly by NER, and because HeLa cells   possess high NER activity (unpublished observation), it is reasonable to conclude that our phenomena observed in HeLa cells by ASDG are largely attributable to CPD, although we have not yet determined the extent of remaining (6-4)PP.
We may also conclude that Rev1 is indispensable for TLS across CPD. Thus far, it is unknown if Rev1 is equally involved in TLS across CPD and (6-4)PP, or if it exhibits some preference. Nelson et al. [25] demonstrated that Rev1p participates in UV-TLS across (6-4)PP, based on yeast transfected with a (6-4)PP-carrying plasmid; only slight differences were observed with a CPD-carrying plasmid.
presented complex results showing involvement of multiple bypass polymerases. They used SV-untransformed XP-A and XP-V cells, but did not include SV-untransformed normal fibroblasts, wherein we detected quick and caffeineinsensitive UV-TLS [35,36]. It is possible that the kind of damage, as well as cell status (normal, transformed, or cancerous) may determine the participation of bypass polymerase(s).
We have presented the first apparent evidence that Polκ participates in UV-TLS. Polκ knockout mouse embryonic cells are known to be UV sensitive [47], but the mechanism had not yet been determined. Polκ is also thought to play a part in the repair-synthesis step of NER [48,49]. From the results of lesion-bypass assays, human Polκ was suggested to be unable to bypass CPD or (6-4)PP. Because the outcomes of such in vitro assays depend on the assay conditions [12], these results must be validated in vivo, such as by ASDG analysis.

Conclusions
Using siRNAs originally designed and ASDG technique, we verified the participation of multiple bypass polymerases in UV-induced TLS in HeLa cells, which is the consistent with recent model of 2 polymerase mechanisms. UV-TLS was largely abolished by siRNAs to Rev3 or Rev7, suggesting that these 2 proteins, which constitute Polζ, play a primary role in mutagenic TLS. Rev1-targeted siRNAs also significantly abolished UV-TLS, consistent with prior suggestions that Rev1 is indispensable in mammalian mutagenic TLS. Unexpectedly, siRNAs to Polη prevented TLS to a great extent, implying that the Polη-and Polζ-dependent processes do not alternate but overlap. Polκ siRNAs, but not siRNAs to Polι, delayed TLS; this is the first apparent evidence for the participation of Polκ in UV-TLS.