Extrahelical (CAG)/(CTG) triplet repeat elements support proliferating cell nuclear antigen loading and MutLα endonuclease activation

Extrahelical Triplet Repeat Elements in RCC DNA Support MutLα Endonuclease Activation.

To determine whether (CAG)_n or (CTG)_n extrusions support MutLα activation, we examined a series of RCC heteroduplexes containing variable numbers of extrahelical (CAG) or (CTG) repeat units embedded within the human DMPK sequence context (Table S1). As shown in Fig. 1 (lane 6), human RFC, PCNA, MutSβ, and MutLα support endonucleolytic incision of a 3.1-kb RCC DNA that contains a (CAG)₃ extrusion, an effect that requires all four proteins (lanes 2–5), as well as presence of the lesion within the DNA. Incision in this manner depends on the endonuclease function of MutLα because production of nicked DNA was abolished when an endonuclease-dead variant of MutLα was substituted for the wild-type protein (lane 7).

Dependence of this reaction on the nature and size of the extrahelical element and the specificity of the effect for MutSβ or MutSα are summarized in Fig. 2A. RCC heteroduplexes containing two or three extrahelical (CTG) or (CAG) repeat units are the preferred substrates in this system, and activity on each of these DNAs is MutSβ-specific. A (CTG)₄ heteroduplex also supports MutSβ-dependent MutLα activation, but in this case, the effect is more modest. (CTG)₁ and (CAG)₁ heteroduplexes also trigger modest endonuclease activation, but activity levels with these heteroduplexes are similar in the presence of either MutSβ or MutSα. RCC heteroduplexes containing an extrahelical TG dinucleotide or a (CAG)₄ extrusion are poor substrates in this system.

The threshold length for (CAG)_n/(CTG)_n instability occurs as n exceeds about 30 repeat units (14). We therefore constructed model heteroduplexes in which (CTG) or (CAG) extrusions are present within the context of long repeat tracts. Fig. 2B demonstrates that heteroduplexes of the form (CTG)₅₆/(CAG)₅₄, (CTG)₅₄/(CAG)₅₆, (CTG)₈₇/(CAG)₈₄, and (CTG)₈₄/(CAG)₈₇ also support MutLα activation in a MutSβ-dependent manner. Although the nature of the unpaired segments within such heteroduplexes is not completely understood, available evidence indicates that the excess repeats on one strand are preferentially accommodated within a single extrusion that is located internal to the repeat and bounded by sequences that are duplex in nature (30). Because repeat sequences in relaxed DNAs are expected to be relatively stable and resistant to slippage because of the absence of negative superhelical strain (31), we attribute endonuclease action on these heteroduplexes to the presence of an extrahelical extrusion within the repeat tract.

Extrahelical Triplet Repeat Elements Can Serve as Sites for PCNA Loading.

Because PCNA loading is required for MutLα activation (9), the results shown in Fig. 2 suggest that extrahelical (CTG)_n or (CAG)_n elements might function as sites for clamp loading. This possibility was addressed using gel filtration to score loading of [³²P]PK-PCNA [PCNA with a N-terminal protein kinase recognition motif (32)] onto RCC DNAs. Yeast RFCΔN, which is highly efficient in loading human PCNA and is functional in reconstituted human MMR, was used in these experiments because the protein is easily prepared in large quantities and displays relatively low nonspecific affinity for DNA, which renders it useful for gel filtration assay of clamp loading (9, 33, 34).

The efficiencies of PCNA loading onto relaxed DNAs are summarized in Table 1. Production of PCNA-DNA complexes with homoduplex or an extrahelical TG dinucleotide heteroduplex is only slightly higher than that observed in background control reactions lacking ATP. These complexes may therefore correspond to PCNA that is indirectly associated with DNA via interaction with nonspecifically bound yRFCΔN. PCNA loading onto the (CAG)₄ extrusion heteroduplex is also similar to that observed with control homoduplex DNA. However, extrahelical (CTG)_n (n = 1–4) or (CAG)_n (n = 1–3) do support enhanced levels of clamp loading, with the best substrates being heteroduplexes containing two or three extrahelical repeat units.

Although extrusions of two or three repeat units are also the optimal substrates for MutLα endonuclease activation in the presence of human RFC, extrahelical (CTG)₁ and (CAG)₁ are moderately effective substrates for clamp loading by yRFCΔN but relatively poor substrates for endonuclease activation in the presence of the human clamp loader (Fig. 2 and Table 1). To address this issue, yRFCΔN was used as clamp loader to evaluate endonuclease activation in the presence of human MutSβ, MutLα, and PCNA. As shown in Fig. S1, MutLα activation is more robust in the presence of the yeast clamp loader, but the dependence of endonuclease activation on the nature of the extrusion is similar to that observed with human RFC. The exceptions to this generalization are (CTG)₁ and (CAG)₁, which support substantial endonuclease activation in the presence of the yeast clamp loader. The simplest interpretation of these findings is that yRFCΔN is more effective than human RFC in using a single repeat unit extrusion as a PCNA loading site, and that clamp loading is largely rate-limiting for MutLα activation on a relaxed heteroduplex.

Because lesion recognition by MutSβ is also required for MutLα activation in this system, affinities of the MSH2•MSH3 heterodimer for (CTG)_n and (CAG)_n extrusions were estimated by surface plasmon resonance spectroscopy. The synthetic oligonucleotide duplexes used for these experiments contained a (CTG)_n or (CAG)_n insertion within the same sequence context as that present in the RCC DNAs described earlier (Table S1). In all but a single case, MutSβ dissociation constants determined by this method are in the 10- to 20-nM range (Table 1). The single exception is (CAG)₄, for which MutSβ affinity was estimated to be 35 nM. These results are also consistent with the idea that PCNA loading is probably rate-limiting for endonuclease activation on the RCC heteroduplexes tested here.

Triplet Repeat-Provoked MutLα Incision of RCC DNA Occurs on Either Strand.

MutLα incises heteroduplex DNA containing a small bubble on either DNA strand, an effect attributed to different PCNA loading orientations at either end of the bubble (9). Lack of obvious strand directionality is also characteristic of MutLα activation on RCC DNA containing an extrahelical triplet repeat element. In the experiment shown in Fig. 3, PCNA was loaded onto a relaxed heteroduplex containing a (CTG)₃ extrusion, and the kinetics of incision were monitored after the addition of MutSβ and MutLα by hybridization with strand-specific probes. Although the rate of incision appeared to be somewhat faster on the top DNA strand, both strands were efficiently incised. Because PCNA loading orientation confers strand directionality on MutLα endonuclease (9), we infer that the (CTG)₃ extrusion supports clamp loading in either orientation relative to that of the helix.

Fine structure is evident in the product distributions on both DNA strands (Fig. 3 arrowheads), suggesting preferential sites for endonuclease action. Because previous work has failed to reveal a strong sequence preference for MutLα cleavage (7, 9), these preferential incision sites may reflect structural features of protein–DNA complexes involved in the reaction.

RCC DNA with a Triplet Repeat Extrusion Is Subject to MutSβ- and MutLα-Endonuclease–Dependent Repair in Nuclear Extracts That Can Occur on Either Strand.

To address the possibility that MutLα activation by an extrahelical triplet repeat element might trigger subsequent repair of a RCC DNA, we compared the fate of a RCC (CAG)₂ extrusion heteroduplex with otherwise identical 3′ or 5′-nicked DNAs in nuclear extracts derived from MLH1^−/− MSH3^−/− HCT116 tumor cells that are deficient in both MutSβ and MutLα (35). Because the (CAG)₂ extrusion was present within overlapping recognition sites for BglII and AlwNI, lesion rectification could be monitored by repair DNA synthesis assay and strand direction scored by restriction endonuclease cleavage.

Fig. S2 shows results of one such experiment in which extract repair in the presence of [α-³²P]dCTP was scored by agarose gel electrophoresis and associated DNA synthesis visualized by gel exposure to a phosphorimager screen. Repair on the nicked strand of the 3′-heteroduplex was evident in unsupplemented HCT116 extract (Fig. S2A, lane 1). This MMR-independent repair is likely a result of the nick-directed endonucleolytic pathway described by Hou et al. (36), as the low level of associated repair DNA synthesis is consistent with the short patch nature of hairpin removal by this system. Repair in this manner was largely abolished when the extract was supplemented with MutSβ (lane 2), suggesting that MutSβ binding to the extrusion suppresses the MMR-independent pathway. Repair in MutLα-supplemented extract did not differ from that in the extract alone, although dCMP incorporation into repair products was enhanced significantly, perhaps because of low-level MMR supported by MutSα, which is present in HCT116 cells (lane 3). However, supplementation with both MutSβ and MutLα substantially enhanced repair over that observed in the presence of MutSβ alone and dramatically increased associated repair synthesis (lane 4), consistent with the long excision repair tracts characteristic of MMR (37). Both repair and repair synthesis depend on the endonuclease function of MutLα because they were substantially reduced when extract was supplemented with MutSβ and endonuclease-dead MutLα (lane 5). Although the nature of the low level of repair and synthesis that occur under these conditions is not clear, it might be indicative of incision by MutLγ [MLH1•MLH3 (1–4)], which conceivably could be produced in extracts by subunit exchange with endonuclease-dead MutLα.

It is important to note that digestion of reaction products with AlwNI demonstrated that in all cases, both repair and repair synthesis were strongly biased to the nicked DNA strand (lanes 6–10). Results with the corresponding 5′-nicked heteroduplex were similar with the exception that no significant repair was detected in HCT116 extract in the absence of exogenous MutSβ and wild-type MutLα (Fig. S2B, lanes 1–10).

Repair products derived from the relaxed (CAG)₂ extrusion heteroduplex were not evident on ethidium-stained agarose gels, but they were readily demonstrable as repair synthesis products that were observed only when extract was supplemented with MutSβ and wild-type MutLα (Fig. S2C, lanes 4 and 9). As can be seen, repair occurred with similar efficiencies on both DNA strands. Fig. 4 summarizes repair synthesis data from two experiments such as that shown in Fig. S2. On the basis of levels of [³²P]dCMP incorporation into restriction products derived from repaired molecules, we estimate that the sum of repair on the two strands of the RCC heteroduplex corresponds to about 10% of that observed on the repaired strand of 3′- or 5′-nicked substrates.