Report

Ubiquitin-Binding Domains in Y-Family Polymerases Regulate Translesion Synthesis

See allHide authors and affiliations

Science  16 Dec 2005:
Vol. 310, Issue 5755, pp. 1821-1824
DOI: 10.1126/science.1120615

Abstract

Translesion synthesis (TLS) is the major pathway by which mammalian cells replicate across DNA lesions. Upon DNA damage, ubiquitination of proliferating cell nuclear antigen (PCNA) induces bypass of the lesion by directing the replication machinery into the TLS pathway. Yet, how this modification is recognized and interpreted in the cell remains unclear. Here we describe the identification of two ubiquitin (Ub)–binding domains (UBM and UBZ), which are evolutionarily conserved in all Y-family TLS polymerases (pols). These domains are required for binding of polη and polι to ubiquitin, their accumulation in replication factories, and their interaction with monoubiquitinated PCNA. Moreover, the UBZ domain of polη is essential to efficiently restore a normal response to ultraviolet irradiation in xeroderma pigmentosum variant (XP-V) fibroblasts. Our results indicate that Ub-binding domains of Y-family polymerases play crucial regulatory roles in TLS.

Signaling through ubiquitin (Ub) is generally thought to occur by low-affinity noncovalent interactions between Ub and a variety of specialized Ub-binding domains (UBDs) (1, 2). To analyze the Ub-interaction map, we performed yeast two-hybrid screens using wild-type Ub and Ub in which isoleucine 44 (I44) was mutated to alanine (Ub*). To date, all known characterized UBDs require the conserved I44 in the hydrophobic patch on Ub for their binding (2), and proteins interacting with Ub* might therefore contain previously unknown Ub-interacting modules. Among the clones that interacted with Ub* are two that encode the C terminus of TLS polymerase ι (polι) (fig. S1A). Moreover, full-length mouse polι expressed in human embryonic kidney (HEK) 293T cells bound to both glutathione S-transferase (GST)–Ub and GST-Ub*, but not to GST alone (fig. S1A). Thus, polι contains a Ub-binding module in the C terminus that does not require I44 for its binding to Ub. Bioinformatic analysis of the C-terminal part of polι identified two copies of a previously unknown sequence motif termed UBM (Ub-binding motif). These repeats span ∼30 residues and consist of two predicted helical segments, separated by an invariant “Leu-Pro” motif, which is conserved in all polι versions examined, as well as in Rev1, another Y-polymerase (fig. S1B). Missense mutations of the conserved residues with a presumptive crucial role in Ub binding (L508A, P509A in UBM1*, L693A, P694A in UBM2*) in either polι UBM substantially impaired polι binding to GST-Ub, whereas the inactivation of both domains by point mutations completely blocked the interaction (Fig. 1A). Similar results were obtained using polι UBM deletion (polι-Δ496-524 and polι-Δ681-709) mutants (fig. S1C). We purified isolated GST-UBM1 and GST-UBM2 of polι and analyzed their binding to Ub and the Ub-I44A mutant by nuclear magnetic resonance (NMR) spectroscopy (fig. S1D). The estimated dissociation constant (Kd) values for binding of UBM1 and UBM2 to both Ub and Ub-I44A were in the range of 180 μM. Mapping of the UBM2 binding surface on Ub revealed binding around the previously defined hydrophobic patch, but the binding surface is displaced toward L8 and away from I44 (Fig. 1B).

Fig. 1.

(A) Identification of the UBDs in Y-polymerases. Point mutations of either UBM1 (L508A,P509A in UBM1*) or UBM2 (L693A,P694A in UBM2*) of mouse polι reduce its binding to Ub as compared with wild-type polι (wt). Mutating both UBMs (UBM1*,2*) abolishes binding of polι to Ub in GST pull-down assays. (B) Surface representation of Ub interaction with UBM determined by NMR spectroscopy. The binding interface of GST-UBM2 on Ub defined by residues K6, L8, T9, G10, I13, T14, R42, K48, G53, and R72 (see supporting online material) is indicated in purple. Residue I44 (yellow) is indicated for orientation. (C) Polη UBZ mediates binding to ubiquitin. HEK293T lysates (TCL) containing FLAG-polη wild type or its UBZ mutant (D652A) (UBZ*) were subjected to Ub-agarose pull-down assays. The shift in mobility of polη visible in lane 1 represents its monoubiquitinated form. IB, immunoblot.

Apart from the UBMs, we also identified several yeast two-hybrid clones containing mononucleate Zn fingers, which were required for their binding to Ub in yeast and mammalian cells. Using profile-based sequence comparisons (3, 4), we grouped these Ub-binding Zn fingers into a separate family, which we named UBZ (Ub-binding Zn finger). These sequence profile searches showed that the UBZ-family Zn fingers can be clearly separated from the presumed DNA-binding variety and are completely unrelated to PAZ (polyUb-associated Zn finger) (5) and NZF (Npl4 Zn finger) (6). UBZ-type fingers were also found in the remaining two Y-family polymerases polη and polκ (fig. S1E). Indeed, human polη is a Ub-binding protein because it interacted specifically with monoUb coupled to agarose (Fig. 1C), and point mutation of the conserved aspartate 652 to alanine in the UBZ (UBZ*) prevented the interaction of polη with monoUb. Thus, all members of the Y-family polymerases contain UBDs in their C termini (fig. S1F).

Polι and polη colocalize with each other and with proliferating cell nuclear antigen (PCNA) in replication factories. These appear as bright foci in S-phase cells, which accumulate upon ultraviolet (UV) irradiation (7, 8). To examine whether this localization requires their UBDs, we cotransfected MRC5 fibroblasts with wild-type cyan fluorescent protein (CFP)–polι and wild-type or mutated FLAG-polι. The fraction of cells with foci increased in UV-irradiated cells to ∼60%, and as expected, both wild-type constructs colocalized in foci. However, neither ΔUBM1,2 nor UBM1*,2* mutants of polι localized in replication foci (Fig. 2A). Furthermore, the Xenopus tropicalis polι, which contains two UBMs but few other conserved amino acids in the C-terminal part (fig. S2A), bound to Ub (fig. S2B) and localized in replication factories in human cells (fig. S2C).

Fig. 2.

UBMs and UBZ are essential for the accumulation of polι and polη in replication foci. (A) MRC5 fibroblasts were cotransfected with pECFP–polι wild type (left panels) and pCMV-FLAG polι, either wild type (WT) or mutants, as indicated (middle panels). The cells were UV irradiated with 15 J m–2 and fixed 16 hours later. (B) MRC5 fibroblasts were cotransfected with YFP-polη wild type and CFP-polη, either wt or D652A, and treated as in (A).

In similar cotransfection experiments, yellow fluorescent protein (YFP)- or CFP-tagged wild-type polη formed bright foci in S-phase cells accumulated upon UV irradiation, whereas CFP-polη–D652A mutant (UBZ*) formed very faint or no foci (Fig. 2B). We conclude that foci localization of both polη and polι depends on their ability to interact with Ub, indicating a common mechanism for the accumulation of Y-polymerases in replication foci. Notably, the polymerase activity was identical in wild-type and mutant proteins used (fig. S3A).

Ubiquitination of the polymerase processivity factor PCNA controls switching of polymerases and replication of damaged DNA (913). TLS polymerases polη and polι bind directly to PCNA via their PCNA-interacting peptide (PIP box) (11, 1416). In addition, DNA damage–induced ubiquitination of PCNA on K164 increases its interaction with polη in vivo (11, 13). We therefore investigated whether the UBM domains of polι directly bind to monoubiquitinated PCNA upon DNA damage. Isolated UBMs of polι readily precipitated monoubiquitinated PCNA generated in HEK293T cells by hydroxyurea treatment (Fig. 3A). UBMs also interacted directly with monoubiquitinated His-PCNA and with a permanently ubiquitinated PCNA, engineered by fusing the cDNA for Ub in frame with that for His-PCNA-K164R mutant (PCNA*-Ub chimera) (fig. S3B). We next analyzed whether simultaneous binding of PIP and UBM domains of polι to ubiquitinated PCNA enhanced this interaction in vivo. Coprecipitation of polι and unmodified PCNA from cells was hardly detectable, whereas the PCNA*-Ub chimera strongly bound to polι in transiently transfected HEK293T cells (Fig. 3B). Coprecipitation was substantially reduced when the PIP box was mutated in polι and completely abolished if UBM domains or both PIP box and UBM domains of polι were mutated.

Fig. 3.

Dual mode of polι-monoUb-PCNA interaction. (A) The UBM domains of polι mediate interaction with monoubiquitinated PCNA (mUb PCNA) generated in HEK293T cells by hydroxyurea (HU) treatment. Total cell lysates or proteins bound to GST or GST-UBMs were analyzed by immunoblotting (IB) with antibodies to PCNA (anti-PCNA). A small amount of nonubiquitinated PCNA precipitated with GST-UBM domains, presumably owing to heterotrimerization of ubiquitinated monomers with nonubiquitinated ones. (B) Both the UBM and PIP box motifs of polι mediate its binding to PCNA*-Ub chimera. HEK293T cells were transfected with His-PCNA*-Ub and FLAG-polι (wt) or its PIP* mutant (FLAG-polι-D424A/C425A/Y426A), UBM1*,2* mutant, or PIP*,UBM1*,2* mutant. The TCL shows the expression level of the corresponding proteins that were subsequently subjected to anti-FLAG immunoprecipitation (IP). Immunoprecipitated polι and PCNA*-Ub were detected with anti-FLAG and anti-PCNA, respectively. (C) XP30RO XP-V cells were transfected with pEGFP-polη constructs, and stable clones were isolated. The survival of these clones was measured after exposure to different doses of UV irradiation and plating in the presence of caffeine (75 μg/ml). UBZ* represents the C638A mutation.

We next examined the importance of the UBDs in polη for its functions in vivo. Xeroderma pigmentosum variant (XP-V) patients are defective in polη, resulting in elevated levels of UV mutagenesis and skin cancers (17). XP30RO is an XP-V fibroblast line that is sensitive to UV irradiation in the presence of caffeine (17). Transfection of green fluorescent protein (GFP)–polη into XP30RO cells restores the UV plus caffeine sensitivity to that of normal human fibroblasts (Fig. 3C). However, when we transfected a UBZ mutant of polη (C638A) into XP30RO cells, the ability to confer resistance to UV plus caffeine treatment was substantially impaired. A reduction in this ability, though less marked, was also obtained with the PIP box mutation. Thus, both Ub- and PCNA-binding abilities of polη are required for efficient TLS. In previous work, we identified a motif in the little finger domain of polη that was required for its interaction with ubiquitinated PCNA; we thought this motif was similar to a CUE domain (11). Current bioinformatic and biochemical studies indicate that this motif is not a bona fide UBD. It is likely therefore that it is involved in the polη-PCNA interaction, irrespective of ubiquitination status.

UBDs can mediate monoubiquitination of the proteins containing them (1, 2). When FLAG-polι was transfected into HEK293T cells, a band of polι of reduced mobility was detected by immunobloting with antibodies to FLAG (Fig. 4A). Analysis of cells additionally transfected with hemagglutin (HA)–Ub showed that the protein of reduced mobility was monoubiquitinated polι (Fig. 4A). Under similar conditions, we found a species of polη with reduced mobility (Fig. 4B; see also Fig. 1C), which we showed was monoubiquitinated polη (Fig. 4B). Mutational inactivation of UBMs in polι or UBZ in polη abolished their monoubiquitination (Fig. 4, A and B). We also detected endogenous levels of monoubiquitinated forms of both polymerases (Fig. 4, C and D). The monoubiquitinated species of polι appeared to be no longer capable of binding to Ub. In GST-Ub pull-down assays, only the unmodified form of polι was detected (Fig. 4E; compare lane 3 with lane 1). Similarly, only the unmodified form of polη bound to Ub-agarose beads (Fig. 1C). In accordance with these observations, creating a permanently ubiquitinated form of polη by fusing Ub to the C terminus of polη (polη-Ub chimera) strongly reduced its ability to bind to Ub-agarose beads (fig. S3C). These results indicate that monoubiquitinated polymerases might be blocked in binding to Ub because of autoinhibitory interactions with their own UBDs (18).

Fig. 4.

Monoubiquitination of polι and polη. HEK293T cells were transfected with (+) or without (–) HA-Ub together with the indicated FLAG-polι (A) or FLAG-polη (B) constructs, and TCL or anti-FLAG immunoprecipitates (IP) were subjected to immunoblotting (IB) with anti-FLAG or anti-HA, as indicated. (C) Endogenous monoubiquitinated polι was detected with anti-polι in TCL or anti-polι immunoprecipitate. (D) Endogenous monoubiquitinated polη was detected with anti-polη in TCL from MRC5 but not in XP30RO cells (E) HEK293T lysates containing FLAG-polι were subjected to pull-down assays with GST-Ub, and bound proteins were detected by immunoblotting with anti-FLAG. The arrow indicates the monoubiquitinated form of polι.

In conclusion, we have identified two previously unknown UBDs in the Y-family TLS polymerases that enable them to interact with monoubiquitinated targets and undergo monoubiquitination in vivo. UBDs are critical for accumulation of polι and polη in replication foci in human cells and are required for efficient restoration of normal TLS in XP-V cells. Both polι (Fig. 3, A and B) and polη (11, 13) preferentially interact with monoubiquitinated PCNA, which is generated at stalled replication forks (11, 13). The PIP box provides the specificity for the interaction, and the DNA damage–induced conjugation of a Ub moiety to PCNA increases the avidity of this binding by providing an interaction surface for the UBDs. We have also shown that polι and polη are themselves monoubiquitinated in vivo. Although the precise role of monoubiquitination of the polymerases remains to be established, it is tempting to speculate that a cycling between their nonubiquitinated and monoubiquitinated forms may contribute to regulation of their compartmentalization in or out of replication factories. Taken together, our data show that Ub binding of the Y-family polymerases plays an important role in translesion DNA synthesis and provide a long-sought clue to how these polymerases can gain preferential access to the stalled replication machinery at the sites of DNA damage.

Supporting Online Material

www.sciencemag.org/cgi/content/full/310/5755/1821/DC1

Materials and Methods

Figs. S1 to S3

References

References and Notes

View Abstract

Navigate This Article