Linkage reprogramming by tailor-made E3s reveals polyubiquitin chain requirements in DNA-damage bypass.

A polyubiquitin chain can adopt a variety of shapes, depending on how the ubiquitin monomers are joined. However, the relevance of linkage for the signaling functions of polyubiquitin chains is often poorly understood because of our inability to control or manipulate this parameter in vivo. Here, we present a strategy for reprogramming polyubiquitin chain linkage by means of tailor-made, linkage- and substrate-selective ubiquitin ligases. Using the polyubiquitylation of the budding yeast replication factor PCNA in response to DNA damage as a model case, we show that altering the features of a polyubiquitin chain in vivo can change the fate of the modified substrate. We also provide evidence for redundancy between distinct but structurally similar linkages, and we demonstrate by proof-of-principle experiments that the method can be generalized to targets beyond PCNA. Our study illustrates a promising approach toward the in vivo analysis of polyubiquitin signaling.

Introduction

Ubiquitin as a post-translational protein modifier affects most aspects of cellular metabolism by modulating the stability, localization, or molecular interactions of its targets. Its versatility as a signaling molecule derives in part from its ability to form polymeric chains of varying topologies (Komander and Rape, 2012; Kwon and Ciechanover, 2017). Ubiquitin is generally coupled to its substrates via an isopeptide bond between its carboxy (C) terminus and the ε-amino group of a substrate lysine (K) residue, and each of ubiquitin’s 7 lysines as well as its amino (N) terminus can be used for chain formation (Kerscher et al., 2006; Tokunaga et al., 2009). This results in a variety of possible chain topologies that are controlled by the combination of ubiquitin-conjugating enzymes (E2s) and ubiquitin protein ligases (E3s) contributing to their assembly. The biological effects of ubiquitylation are usually mediated by the so-called ubiquitin receptors (Husnjak and Dikic, 2012). These exhibit a basal affinity for the substrate and harbor one or several ubiquitin-binding domains (UBDs) that recognize the modification in a more or less linkage-selective manner. Whereas K48-, K29-, and K11-linked chains have been implicated in degradation by the 26S proteasome, K63 linkage is mainly involved in proteasome-independent pathways ranging from the DNA-damage response to inflammatory signaling, endocytosis, and intracellular membrane trafficking (Erpapazoglou et al., 2014; Schwertman et al., 2016; Ulrich and Walden, 2010; Wu and Karin, 2015). A contribution to proteasomal degradation has also been reported for K63-linked chains (Ohtake et al., 2018; Saeki et al., 2009). Linear, i.e., methionine (M)1-linked chains are assembled by a single E3, linear ubiquitin chain assembly complex (LUBAC), and they are best known for their cooperation with K63 chains in the inflammatory response (Rieser et al., 2013; Rittinger and Ikeda, 2017; Tokunaga et al., 2009), but they have also been used to target proteins for proteasomal degradation (Bragança and Kraut, 2020).

Structural biology has provided insight into the principles of ubiquitin recognition by dedicated UBDs, and a wide range of analytical tools including mass spectrometry, ubiquitin mutants, linkage-selective deubiquitylating enzymes (DUBs), antibodies, and affinity probes have facilitated the characterization of ubiquitin conjugates (Heap et al., 2017; Leestemaker and Ovaa, 2017). However, in many cases the relevance of a particular linkage for the biological function of a given polyubiquitylation event is still unclear. Our poor understanding of this factor results to a large extent from a lack of suitable tools to directly manipulate ubiquitin chain linkage. To gain insight into the functional consequences of the linkage of a particular polyubiquitin chain, it would be necessary to replace it with an alternative linkage in a physiological setting. Due to the inalterable linkage and substrate selectivity of most E2-E3 pairs, however, this has never been accomplished.

Here, we present an enzyme-based strategy of in vivo ubiquitin linkage reprogramming, designed to explore the functional relevance of polyubiquitin chain linkage in a defined biological context. Our approach entails the replacement of an endogenous ubiquitin ligase by a set of tailor-made, substrate-selective E3s of defined alternative linkage specificities.

The experimental system underlying our study is based on a conserved and well-characterized K63-linked polyubiquitin chain assembled on the replication factor PCNA in response to replication stress. Modification of PCNA at a single invariant lysine, K164, facilitates DNA replication in the presence of lesions and thereby contributes to genome maintenance (Cipolla et al., 2016; Ulrich and Walden, 2010). Monoubiquitylation by the E2-E3 pair Rad6-Rad18 promotes a mutagenic pathway of damage bypass called translesion synthesis (TLS), mediated by a set of specialized damage-tolerant DNA polymerases (Stelter and Ulrich, 2003). By means of a PCNA-interacting peptide (PIP) in combination with a UBD, these enzymes preferentially interact with monoubiquitylated PCNA, enabling them to directly copy-damaged DNA (Bienko et al., 2005; Plosky et al., 2006). Extension of the modification to a K63 polyubiquitin chain by a heterodimeric E2, Ubc13-Mms2, with a cognate E3, Rad5 in budding yeast, initiates an alternative, error-free pathway known as template switching (TS) (Hoege et al., 2002). The function of the polyubiquitin chain in the TS pathway is still a matter of debate. An involvement of proteasomal degradation has been ruled out by our previous work (Zhao and Ulrich, 2010). Factors interacting preferentially with polyubiquitylated PCNA have been reported, such as yeast Mgs1, its human homolog WRNIP1 (Crosetto et al., 2008; Saugar et al., 2012), and human ZRANB3 (Ciccia et al., 2012; Vujanovic et al., 2017; Weston et al., 2012). However, their relevance as major effectors of TS is still uncertain. Further complication arises from the multifunctional nature of Rad5, which—in addition to its E3 function—exhibits DNA-dependent ATPase activity (Johnson et al., 1994) and contributes to the recruitment of TLS polymerases (Gallo et al., 2019). Although these activities are separable and independent of Rad5’s involvement in TS (Ball et al., 2014; Chen et al., 2005; Choi et al., 2015; Gallo et al., 2019), a contribution of the protein to events downstream of PCNA polyubiquitylation has not been excluded. Using a set of non-cleavable tandem fusions of ubiquitin to PCNA as genetically engineered polyubiquitin mimics, we previously showed that a key parameter for functionality appears to be the geometry of the ubiquitin-ubiquitin junction itself, as the linear chain mimics were incapable of activating TS (Takahashi et al., 2020; Zhao and Ulrich, 2010).

Surprisingly, by applying our strategy of linkage reprogramming via replacement of Rad5, we now found that—in contrast to the pre-formed chain mimics—an enzymatically assembled M1-linked (linear) polyubiquitin chain on PCNA can substitute for the K63 modification and support TS in budding yeast. A K48-linked chain is not only inactive in TS but inhibits TLS by targeting the modified PCNA for proteasomal degradation. These findings provide insight into the relevance of linkage for ubiquitin signaling in DNA-damage bypass that would have been impossible to obtain by any other approach. In addition, we demonstrate that our method of linkage reprogramming is generalizable to targets beyond PCNA. Our study thus illustrates a novel and generalizable approach to analyzing polyubiquitin signaling in a physiological context.

Design

Our rationale for attempting a reprogramming of polyubiquitylation stems from the conundrum that pre-formed linear chain mimics on PCNA do not function in TS, even though potential effector proteins of polyubiquitylated PCNA, such as Mgs1 and ZRANB3, do not discriminate between linear and K63-linked chains (Saugar et al., 2012; Takahashi et al., 2020; Zhao and Ulrich, 2010). Three alternative, non-exclusive models can be envisioned to explain this discrepancy: (1) the existence of a yet unidentified, highly K63-selective effector, (2) an additional contribution of the E3, Rad5, to TS that necessitates its action during or after chain assembly, or (3) a requirement of polyubiquitin chain assembly or disassembly in situ, i.e., at the sites of replication stress. Replacing Rad5 with unrelated E3s of defined linkage selectivity should allow us to differentiate between these scenarios: (1) if TS strictly required the K63 linkage due to a highly linkage-selective effector, only a K63-specific E3 could act as a functional Rad5 substitute. Alternatively, (2) if Rad5 itself were directly involved in the TS process, no tailor-made E3 should be able to replace Rad5. Finally, (3) if chain assembly in situ were critical for TS, any E3—irrespective of its linkage characteristics—might be expected to compensate for the loss of Rad5. Similarly, chain disassembly could play a role, depending on the linkage selectivity of the involved DUBs.

We thus designed three tailor-made enzymes from well-characterized and highly linkage-specific E3 domains. To implement K63 specificity, we chose a domain from the budding yeast RING-finger E3, Pib1 (aa 91–286), that exclusively cooperates with the K63-specific Ubc13-Mms2 complex (Hofmann and Pickart, 1999; Renz et al., 2020). K48 specificity was realized by means of a fragment of budding yeast Cue1 (aa 24–203), harboring an interaction domain for the K48-specific E2, Ubc7 (Weber et al., 2016), and a ubiquitin-binding CUE domain (Bagola et al., 2013). The latter preferentially stimulates the growth of chains already containing two or more ubiquitin moieties (Bagola et al., 2013; von Delbrück et al., 2016). An M1-specific E3 was designed by using a part of human HOIP, the catalytic subunit of LUBAC (aa 699–1,072). The fragment includes the catalytic RING-in-between-RING (RBR) domain, known to cooperate with E2s of the UBCH5 family, and the linear ubiquitin chain-determining domain (LDD) (Smit et al., 2012; Stieglitz et al., 2013). Finally, substrate selectivity was programmed by fusing each E3 domain to a relevant targeting module. For PCNA as a substrate, a PIP motif derived from budding yeast Msh6 (aa 23–42) was chosen (Clark et al., 2000). To generalize the approach, we used GFP as a substrate and a GFP-specific nanobody, LaG-16, for substrate targeting (Fridy et al., 2014). After optimization of the PIP fusion constructs, catalytic activities as well as substrate and linkage specificities were verified in a series of in vitro ubiquitylation assays with recombinant proteins before analyzing their ability to support ubiquitylation in vivo.

Results

Tailor-made E3s polyubiquitylate PCNA in a linkage-specific manner

Standardized in vitro polyubiquitylation assays with recombinant proteins served to assess the activities and linkage specificities of the designed E3s. A linear fusion of monoubiquitin to the N terminus of PCNA was used as an established surrogate substrate to overcome the need for prior monoubiquitylation of PCNA by Rad6-Rad18 (Parker and Ulrich, 2009). Substrate-independent E3 activity was assayed by monitoring the stimulation of unanchored polyubiquitin chain synthesis, a common property of many E2s (Hochstrasser, 2006).

In the course of its optimization, the M1-specific E3 was expanded by a head-to-tail di-ubiquitin unit, tethered by flexible linkers, because it was reported that full HOIP activity requires allosteric activation by binding to a ubiquitin chain (Lechtenberg et al., 2016). Several arrangements of the individual domains were explored to maximize the intramolecular auto-activation effect and were found to increase activity toward Ub-PCNA to variable extents (Figures S1A–S1D). The final, optimized construct was named PIP-E3(1) and used alongside the respective K48- and K63-selective constructs, PIP-E3(48) and PIP-E3(63), for all further assays (Figure 1A).

Figure 1

Tailor-made E3s exhibit linkage-specific activity toward PCNA

(A) Domain structures of designed PIP-E3s. Constructs contain GST and/or His6 (H) tags for purification after expression in E. coli. Numbers in parentheses specify the range of amino acids used from the respective proteins of origin. PIP, PCNA-interacting peptide (KKMKQSSLLSFFSKQVPSGT); N, nuclear localization signal (PKKKRKV); Ub∗, ubiquitin mutant K29R K48R K63R G76L; L, hydrophilic linker (ASGAGGSEGGGSEGGTSGAT).

(B) PIP-E3s catalyze polyubiquitylation of Ub-PCNA. Reactions were set up at 30°C for 40 min under standard conditions, containing 2 μM Ub-PCNA, 5 μM ubiquitin, 0.1 μM E1, and 1 μM of the indicated E3. Ubc13-Mms2 was used at 0.2 μM, other E2s at 2 μM. Products were analyzed by western blotting against PCNA. Simultaneous detection of PCNA and ubiquitin is shown in Figure S1E. Analogous reactions monitoring the synthesis of unanchored chains in the absence of Ub-PCNA are shown in Figure S1F.

(C) PIP-E3s act in a linkage-specific manner. Standard Ub-PCNA polyubiquitylation reactions were set up as in (B) and analyzed by western blotting with linkage-specific antibodies as indicated. Analogous assays for free-chain synthesis are shown in Figure S1G.

(D) PIP-E3s act in a linkage-specific manner. Standard Ub-PCNA polyubiquitylation reactions were set up and analyzed as in (B) but using the indicated ubiquitin variants. Simultaneous detection of PCNA and ubiquitin is shown in Figure S1H.

(E) Kinetic analysis of PIP-E3 reveals substrate-binding properties similar to Rad5, but a range of catalytic rate constants. Michaelis-Menten parameters and standard deviations were determined by measurements of initial velocities as described in the STAR Methods section and in Figure S2, using ubiquitin variants that limit conjugation to a single junction.

In combination with the appropriate E2s, all PIP-E3s generated both free and PCNA-linked polyubiquitin chains (Figures 1B, S1E, and S1F). The distinct electrophoretic mobilities of the conjugates assembled by the individual PIP-E3s suggested differences in their topologies. The intended linkages were confirmed by western blotting with linkage-specific antibodies for both PCNA-bound and unanchored chains (Figures 1C and S1G) and by using ubiquitin variants UbK63R and UbK48R as well as HisUb, which carries an N-terminal His6-tag that blocks chain formation via M1 (Figures 1D and S1H). While the pattern of free chains produced by Rad5, PIP-E3(63), and PIP-E3(1) was suggestive of a distributive assembly mode, PIP-E3(48) generated predominantly high-molecular-weight species with few intermediates, indicating a high processivity. This behavior is consistent with the mechanism of Cue1 (Bagola et al., 2013; von Delbrück et al., 2016).

To compare the catalytic properties of the designed PIP-E3s to those of Rad5, we determined Michaelis-Menten parameters reflecting substrate affinity (KM) and relative rate constants (kcat) under conditions of rate-limiting E3 concentration (Figures 1E and S2). To simplify the measurements, reactions were limited to conjugation of a single additional ubiquitin moiety to Ub-PCNA by employing appropriate ubiquitin variants (UbK63R, UbK48R, or HisUb, respectively). Consistent with the use of an identical PIP motif for substrate binding, all PIP-E3s exhibited similar KM values between 1.4 and 2.7 μM, which coincidentally resembled that of Rad5 (1.7 μM). Moreover, we measured a kcat value of PIP-E3(63) only slightly lower than that of Rad5. PIP-E3(1) exhibited a ca. 10-fold lower rate constant, and PIP-E3(48) was found to be the least active among the E3s, with a kcat of more than 100-fold lower than that of PIP-E3(63). However, given Cue1’s preference for substrates already carrying at least two ubiquitin moieties (von Delbrück et al., 2016), this value is likely a significant underestimation of the overall activity of PIP-E3(48) in PCNA polyubiquitylation with wild-type (WT) ubiquitin.

Tailor-made E3s are selective for their cognate E2s and for monoubiquitylated PCNA

As linkage specificity is dictated by the combination of E2 and E3, it was important to rule out unintended cooperation of the PIP-E3s with non-cognate E2s. In systematic two-hybrid assays with all budding yeast E2s, PIP-E3(63) and PIP-E3(48) exclusively interacted with their intended partners, Ubc13 and Ubc7, respectively, while PIP-E3(1) did not show any significant interactions (Figure S3A). Similarly, catalytic assays (Figures 2A and S3B) showed that like Rad5, PIP-E3(63) was active only with Ubc13-Mms2, but not with Ubc7 or two non-selective E2s, Ubc4 and Ubc5, whereas PIP-E3(48) cooperated exclusively with Ubc7. PIP-E3(1) produced chains with Ubc4, Ubc5, and human UBCH5A, but this generic E2 interaction was expected based on the properties of LUBAC. Moreover, it did not interfere with linkage specificity, as the observed chain pattern was identical with all E2s, consistent with the fact that HOIP itself controls linkage via its LDD (Stieglitz et al., 2013).

Figure 2

Tailor-made PIP-E3s are E2 and substrate selective

(A) PIP-E3s are selective for their cognate E2s. Standard Ub-PCNA polyubiquitylation reactions were set up with the indicated E2-E3 combinations and analyzed by anti-PCNA western blotting. An anti-ubiquitin blot and simultaneous detection of PCNA and ubiquitin are shown in Figure S3B.

(B) PIP-E3s require prior monoubiquitylation of PCNA. Standard polyubiquitylation reactions were set up using either Ub-PCNA or native PCNA as a substrate and analyzed by anti-PCNA and anti-GFP western blotting. Simultaneous detection of the two substrates is shown in Figure S3C.

(C) PIP-E3s are substrate selective. Activities toward Ub-PCNA and Ub-GFP were compared in standard polyubiquitylation assays, analyzed by western blotting against PCNA and GFP.

As genuine Rad5 replacements, the PIP-E3s should be selective for the monoubiquitylated form of PCNA as a target. As intended, none of the enzymes promote de novo ubiquitylation of native PCNA despite the availability of K164, indicating that—like Rad5—they can only extend, but not initiate a ubiquitin chain on PCNA (Figure 2B). In addition, while Ub-PCNA was readily used as a substrate, an analogous Ub-GFP construct was not efficiently modified by PIP-E3(63) or PIP-E3(48) (Figures 2C and S3C). PIP-E3(1) showed some activity toward Ub-GFP, but it strongly preferred Ub-PCNA in direct competition (Figure 2C). Moreover, the presence of the intended substrate, Ub-PCNA, suppressed free-chain polymerization by PIP-E3(1) (Figure 2B), while Ub-GFP had no such effect (Figure S3D). This indicates that all PIP-E3s—like Rad5—are highly selective for monoubiquitylated PCNA as a substrate.

In summary, our biochemical analysis confirms the expected properties of the PIP-E3s with respect to linkage, E2, and substrate selectivity. Unexpectedly, our design strategy has yielded an artificial enzyme, PIP-E3(63), with catalytic properties almost indistinguishable from native Rad5. Importantly, we have succeeded in generating two additional E3s that—albeit less active—share Rad5’s affinity and selectivity for PCNA but generate chains of alternative linkages on the substrate.

Assessment of unspecific activities of the tailor-made PIP-E3s in vivo

To characterize the selectivity of the tailor-made PIP-E3s in vivo, we explored their effects in undamaged cells where PCNA ubiquitylation is minimal. Under these conditions, expression of PIP-E3(63) or PIP-E3(48) did not affect the overall pattern of polyubiquitin conjugates as detected by western blotting with linkage-selective anti-ubiquitin antibodies in the absence or presence of a proteasome inhibitor, MG132 (Figure S4A). In contrast, expression of PIP-E3(1) caused a detectable accumulation of high-molecular-weight linear polyubiquitin conjugates, possibly indicating unspecific activity or synthesis of unanchored chains by the enzyme. To obtain more accurate information, we used diGly proteomics, involving the proteome-wide detection of ubiquitylation sites by isolation of branched Kϵ-GG remnant peptides from tryptic digests of total cell lysates by SILAC-based mass spectrometry (Kim et al., 2011; Wagner et al., 2011). For the analysis of PIP-E3(48), cells were treated with MG132 prior to harvesting. Comparison of native cells with cells expressing one of the PIP-E3s revealed less than 1.2-fold change in the relative abundance of peptides corresponding to the K63 and K48 linkages of ubiquitin (Figure S4B) and little (2 ubiquitylation sites) or no significant (FDR < 0.01) enrichment of ubiquitylation site abundance for PIP-E3(63) and PIP-E3(48), respectively (Figures S4C and S4D). Upon expression of PIP-E3(1), however, more than 50 ubiquitylation sites were identified as significantly enriched, derived mainly from ribosomal proteins, but none of them from a known PCNA interactor or replication fork component (Figure S4E). Although we cannot exclude indirect effects of PIP-E3(1) expression, these factors may represent off-target substrates of this enzyme that should be considered when performing substrate-specific functional analyses.

Tailor-made PIP-E3s modulate in vivo DNA damage resistance via PCNA polyubiquitylation

To assess substrate-specific activities, PIP-E3s were constitutively expressed in a rad5Δ strain carrying a His6-tagged allele of PCNA, POL30 (Figure S5A), and their in vivo activities toward PCNA were examined by anti-ubiquitin and anti-PCNA western blotting of HisPCNA conjugates isolated under denaturing conditions. Due to reported differences in the sensitivities of various ubiquitin-specific antibodies toward individual linkages (Emmerich and Cohen, 2015), we used two monoclonal antibodies, P4D1 and FK2 (Figure S5B). Whereas P4D1 detected K63-modified PCNA upon expression of Rad5 or PIP-E3(63), FK2 was less sensitive toward K63 chains but showed ubiquitin conjugates with a distinct electrophoretic mobility dependent on expression of PIP-E3(1), consistent with linear chains (Figures 3A and S5C). No polyubiquitylated PCNA was observed upon expression of PIP-E3(48). Importantly, both K63 and M1 modification of PCNA were strongly dependent on prior treatment of cells with the DNA-alkylating agent methyl methanesulfonate (MMS), indicating that PCNA monoubiquitylation (by Rad18) triggered by DNA damage is a prerequisite for in vivo activity of the PIP-E3s toward PCNA.

Figure 3

PIP-E3s mediate PCNA ubiquitylation and affect DNA damage resistance in vivo

(A) PIP-E3s promote damage-dependent PCNA polyubiquitylation in vivo. PCNA modification was analyzed via denaturing pull-down of HisPCNA from WT cells or rad5Δ mutants expressing the indicated VSV-tagged E3s under control of the strong constitutive ADH1 promoter, either untreated or exposed to 0.02% MMS for 90 min. S: SUMO; e.v.: empty vector.

(B) Survival of WT and rad5Δ cells expressing the indicated constructs after irradiation with UV light (254 nm). Error bars indicate standard deviations from three independent experiments with three technical replicates each.

(C) MMS sensitivities of WT and rad5Δ cells expressing the indicated constructs, determined by growth on the indicated concentrations of MMS. PIP-E3(1)∗: catalytically inactive mutant (C885A of HOIP).

(D) The effects of PIP-E3s on MMS sensitivity are abolished in a ubiquitylation-deficient PCNA mutant, pol30. Note that the pol30 mutation partially suppresses the sensitivity of rad5Δ due to interfering with PCNA sumoylation (Hoege et al., 2002). Dashed lines indicate irrelevant parts removed from the image of the plates.

(E) The ubi mutant selectively abolishes the positive effect of PIP-E3(63) on MMS resistance. Dashed lines indicate irrelevant parts removed from the image of the plates.

(F) Deletion of UBC7 selectively abolishes the negative effect of PIP-E3(48) on MMS resistance. Dashed lines indicate irrelevant parts removed from the image of the plates.

(G) Expression of an N-terminally His6-tagged allele (UBI) as the only source of ubiquitin selectively abolishes the positive effect of PIP-E3(1) on MMS resistance. Dashed lines indicate irrelevant parts removed from the image of the plates.

DNA-damage sensitivity assays revealed distinct in vivo activities of all three PIP-E3s (Figures 3B and 3C): while PIP-E3(63) suppressed the sensitivity of rad5Δ mutants toward UV and MMS to near-WT levels, PIP-E3(1) afforded an intermediate sensitivity between that of WT and rad5Δ, and PIP-E3(48) conferred a dominant-negative effect, causing a further sensitization beyond the level of rad5Δ. Out of concern that the di-ubiquitin unit inherent in PIP-E3(1) might exert a non-catalytic effect when close to PCNA, we also examined a catalytically inactive mutant of PIP-E3(1) but found this to have no influence on the damage sensitivity of rad5Δ (Figures 3C and S5A).

To rule out potential effects of non-specific activity toward substrates other than PCNA or of unanchored chains formed by the E3s, we repeated the assay in a pol30 mutant, defective in PCNA ubiquitylation. In this background, the effects of the constructs were largely abolished (Figure 3D), strongly suggesting that all PIP-E3s act via polyubiquitylation of PCNA as intended. Expression of the PIP-E3s in a rad18Δ mutant yielded an overall similar pattern with no effects by PIP-E3(63) and PIP-E3(48) (Figure S5D). Interestingly, PIP-E3(1) rescued the extreme MMS sensitivity of rad18Δ to a minor extent at low MMS concentrations (0.00025%). However, this effect was independent of the E3’s catalytic activity, suggesting that it might be a consequence of recruiting the E3’s di-ubiquitin motif to the proximity of PCNA. Overall, these data confirm that PCNA monoubiquitylation by Rad18 is a prerequisite for efficient effects of the E3s on DNA damage resistance in cells.

Correct linkage and E2 specificities in vivo were verified by a set of additional mutant backgrounds: in ubi, ubc13Δ, or mms2Δ, only PIP-E3(63) lost its effect on the rad5Δ mutant (Figures 3E and S5E), deletion of UBC7 exclusively abolished the dominant-negative effect of PIP-E3(48) (Figure 3F), and use of a strain expressing a His6-tagged allele as the only source of ubiquitin (UBI) selectively blocked the effect of PIP-E3(1) (Figure 3G). Minor effects of UBI on PIP-E3(63) could result from the proximity of ubiquitin’s N terminus to K63. Collectively, these data indicate that the tailor-made E3s are active toward PCNA in vivo and convey distinct effects on damage resistance via their designated substrate and linkage specificities as intended by their design. This raised the question of how the three different linkages impinge on PCNA-dependent DNA damage bypass.

A tailor-made K63-specific E3 can replace Rad5 function in the TS pathway

Rescue of the rad5Δ phenotype by PIP-E3(63) suggested that this enzyme genuinely contributes to error-free TS by K63 polyubiquitylation of PCNA. To confirm this, we took advantage of the notion that a TS defect suppresses the cold sensitivity of a deletion mutant of a non-essential DNA polymerase δ subunit, pol32Δ (Karras and Jentsch, 2010). Reactivation of TS by expression of PIP-E3(63) in pol32Δ rad5Δ should therefore render cells cold sensitive. Figure 4A shows that PIP-E3(63) indeed prevented growth at 15°C, confirming our hypothesis that the E3 supports TS. However, our initial damage sensitivity assays had indicated an incomplete rescue of rad5Δ at high damage loads (Figure 3B), indicating that PIP-E3(63) cannot fully substitute for Rad5, despite a comparable activity toward PCNA in vivo (Figure 3A). We reasoned that this residual sensitivity might be caused by a defect in additional, TS-independent pathways of genome maintenance, mediated by Rad5’s ATPase activity and its interaction with the damage-tolerant polymerase Rev1 (Choi et al., 2015; Gallo et al., 2019). We therefore asked whether full complementation of rad5Δ could be achieved by co-expression of PIP-E3(63) with a ligase-deficient RING-finger mutant, Rad5I916A, that would supply the other activities (Ulrich, 2003). Co-expression of both factors fully complemented the damage sensitivity of rad5Δ (Figure 4B), demonstrating that while the presence of Rad5 is not required for TS, full complementation of the rad5Δ mutant can be achieved by providing ligase and ATPase/Rev1 interaction activities separately.

Figure 4

PIP-E3(63) and PIP-E3(1) activate the TS pathway in vivo, while PIP-E3(48) inhibits TLS

(A) Expression of PIP-E3(63) or PIP-E3(1) confers cold sensitivity to a rad5Δ pol32Δ mutant, but PIP-E3(63) does this only in the presence of Ubc13, whereas the effect of PIP-E3(1) is independent of Ubc13.

(B) Co-expression of PIP-E3(63) and a ligase-defective allele of RAD5 (rad5) fully restores the damage resistance of a rad5Δ mutant.

(C) The dominant-negative effect of PIP-E3(48) expression on the MMS resistance of rad5Δ is abolished by the proteasome defect of the ump1Δ mutant.

(D) The dominant-negative effect of PIP-E3(48) expression on the MMS resistance of rad5Δ is abolished by deletion of the TLS polymerases, whereas the positive effects of PIP-E3(63) and PIP-E3(1) are independent of TLS. TLSΔ: rev1Δ rev3Δ rad30Δ.

(E) Expression of PIP-E3(48) in a rad5Δ pol32Δ mutant causes a spontaneous growth defect.

K48 polyubiquitylation of PCNA induces degradation and inhibits TLS

Given the prominent role of K48 polyubiquitylation in proteasomal degradation, we considered whether the dominant-negative effect of PIP-E3(48) was due to proteolysis of PCNA upon DNA damage, which would explain our failure to detect K48 polyubiquitin chains on HisPCNA (Figure 3A). Compromising proteasome activity by means of deleting UMP1, important for maturation of the 20S proteasome (Ramos et al., 1998), indeed reduced the detrimental effect of PIP-E3(48), despite a growth defect and a slight increase in overall damage sensitivity conferred by ump1Δ (Figure 4C). Moreover, K48 polyubiquitylation and degradation of PCNA upon damage-induced monoubiquitylation by Rad18 should not only prevent TS but also negatively affect TLS. Consistent with this model, expression of PIP-E3(48) in a TLS-deficient background did not further sensitize the rad5Δ mutant, while rescue by PIP-E3(63) was TLS independent (Figure 4D). Finally, PIP-E3(48) conferred a growth defect on a pol32Δ rad5Δ mutant even in the absence of exogenous DNA damage (Figure 4E). Considering the fact that deletion of POL32 results in constitutive PCNA monoubiquitylation likely due to spontaneous replication problems (Karras and Jentsch, 2010), this finding is consistent with a scenario where PIP-E3(48) inhibits proliferation by inducing selective degradation of the replication-engaged PCNA pool. Thus, our findings imply that the dominant-negative effect of PIP-E3(48) is attributable to an inhibition of TLS due to K48 polyubiquitylation and thus proteasomal targeting of the monoubiquitylated pool of PCNA.

Linear polyubiquitin chains on PCNA can substitute for the K63 linkage in TS

Although yeast is not naturally capable of assembling M1-linked polyubiquitin chains, suppression of the damage sensitivity of rad5Δ by PIP-E3(1) (Figure 3C) suggested that this chain type could at least partially substitute for K63 polyubiquitylation. As expected, this rescue did not require the TLS polymerases (Figure 4D). To verify that the activity was due to a functional TS pathway, we again exploited the effect of PIP-E3 expression on the cold sensitivity of a pol32Δ rad5Δ mutant. Figure 4A shows that PIP-E3(1) sensitized pol32Δ rad5Δ to a similar extent as PIP-E3(63), indicating successful reconstitution of TS. Moreover, in contrast to PIP-E3(63), the effect of PIP-E3(1) was independent of UBC13, as sensitization was even observed in a pol32Δ rad5Δ ubc13Δ background. This shows that PIP-E3(1) does not rely on cooperation with the K63-specific Ubc13-Mms2 complex but mediates TS independently.

Given the lower catalytic activity of PIP-E3(1) compared with PIP-E3(63) or Rad5, it is not surprising that suppression of the damage sensitivity of rad5Δ was incomplete. While we were able to rule out a mislocalization of PIP-E3(1) (Figure S6A), we found that E3 protein levels also contributed to the efficiency of TS, as damage resistance of individual clones correlated with PIP-E3(1) levels (Figure S6B). However, when PIP-E3(63) expression was titrated down by means of a regulable TET-promoter to levels below the detection limits of western blotting (Figure S6C), its performance in damage sensitivity assays was still comparable with that of PIP-E3(1) expressed at higher levels (Figure S6D), thus emphasizing that intrinsic catalytic activity also appears to be important for overall function.

To further address the efficacy of linear as compared with K63-linked polyubiquitin chains on PCNA, we also considered their length. We had previously shown that a single K63 junction linked to a ubiquitin moiety at the N terminus of PCNA conferred significant rescue of TS activity (Takahashi et al., 2020). In an analogous experiment using a fusion of monoubiquitin to PCNA we found that PIP-E3(1) conferred damage resistance, provided that the N terminus of the substrate-bound ubiquitin moiety was available (Figures S6E and S6F, compare strains #7 and #8). This showed that a linear chain at the N terminus of PCNA activates damage bypass. Note that the slight non-catalytic effect of PIP-E3(1) described above in the rad18Δ background provided some rescue of MMS sensitivity even in those strains where the E3 is unable to modify PCNA (strains #3, #5, and #8). However, allowing extension of the N-terminal ubiquitin by a single M1-junction, using a strain harboring a His6-tagged allele as the sole source of free ubiquitin (strain #10), had marginal effects beyond the non-catalytic action of PIP-E3(1).

While this finding could indicate a requirement for longer linear chains, the lack of activity could also be due to an insufficient degree of modification due to the lower catalytic activity of PIP-E3(1). We therefore asked whether extending chain length in the context of engineered head-to-tail fusions (Takahashi et al., 2020; Zhao and Ulrich, 2010) would provide damage resistance. However, a fusion of eight ubiquitin moieties to the N terminus of PCNA did not render cells more damage-resistant than a construct carrying a single or four ubiquitin moieties (Figures S6G and S6H), indicating that chain length alone is unlikely a critical factor for TS.

Overall, our findings demonstrate that in the context of DNA-damage bypass enzymatically generated linear polyubiquitin chains can substitute for K63-linked chains on PCNA, thus calling into question the relevance of the K63 linkage as an exclusive mediator of TS.

Linkage-specific E3s can be redirected to other substrates

To demonstrate the potential of the tailor-made E3s for application to substrates beyond PCNA, we turned to the ubiquitin fusion degradation (UFD) pathway (Johnson et al., 1995). Here, a combination of a HECT-type E3, Ufd4, and a chain elongation factor (E4), Ufd2, assembles a branched ubiquitin chain on a monoubiquitin unit permanently fused to a substrate protein (Koegl et al., 1999; Liu et al., 2017). Truncation of the U-box motif of Ufd2 abolishes E4 activity (Richly et al., 2005). Use of a ufd4Δ ufd2 background should thus allow us to examine whether our tailor-made E3s would operate on monoubiquitylated proteins other than PCNA.

As a simple, well-characterized UFD model substrate, we chose a fusion of ubiquitin to GFP, Ub-GFP (Dantuma et al., 2000). Substrate specificity of the E3s was adapted by replacing their PIP motif with a nanobody directed against GFP, LaG-16 (Fridy et al., 2014; Figure S7A). For the K63-specific E3, this involved C-terminal fusion of the nanobody. As shown in Figure 5, all E3s, expressed under control of the constitutive ADH1 promoter, afforded polyubiquitylation of the substrate in budding yeast. The extent of substrate polyubiquitylation by LaG16-E3(48) and E3(63)-LaG16 was significantly improved by overexpression of their cognate E2s (Figures S7B and S7C). Moreover, as we noticed residual modification of Ub-GFP at K48 even in the absence of the tailor-made E3s, we used a K48R mutant of Ub-GFP with E3(63)-LaG16 to reduce this background (Figure S7D). Importantly, mutation of K48 or K63 of the substrate-linked ubiquitin moiety abolished chain extension by LaG16-E3(48) and E3(63)-LaG16, respectively, while modification of the N terminus by a His6-tag prevented action of LaG16-E3(1) (Figure 5). This indicates that the activity of each of the E3s depended on the availability of its correct linkage-specific acceptor site.

Figure 5

Tailor-made E3s can be targeted toward a GFP-tagged model substrate in vivo

Modification of Ub-GFP was analyzed in total lysates of ufd4Δ ufd2 cells expressing the indicated VSV-tagged LaG16-E3s. In the strains harboring E3(63)-LaG16, UBC13 and MMS2 were overexpressed via insertion of the strong, constitutive TDH3 promoter upstream of the two open reading frames. Strains harboring LaG16-E3(48) carried the TDH3 promoter upstream of the UBC7 coding region. The substrate and its derivatives were detected via western blotting with GFP-specific antibodies. Expression of the E3s was confirmed by means of anti-VSV blotting.

Taken together, these experiments show that despite variations in the relative activities of the three E3s, all of them can be redirected to other substrates, thus allowing a generalization of our approach.

Discussion

Linkage reprogramming as a versatile tool to investigate ubiquitin signaling

The notion that the in vivo fate of a polyubiquitylated protein is controlled by the linkage of its modification is intuitive. However, direct evidence for this concept is lacking because it has not been possible until now to deliberately change the destiny of a ubiquitylation target by “mutating” the linkage of the attached polyubiquitin chain. This lack of progress is to a large extent due to the absence of suitable tools for such manipulation. Using substrate-selective, tailor-made E3s of distinct linkage specificities, we have now succeeded in reprogramming the linkage of polyubiquitin chains in two different settings. This approach has demonstrated that polyubiquitylation triggered by a common stimulus on one and the same attachment site of a substrate can direct this substrate to distinct fates, depending on the linkage of the chain. In the context of DNA damage bypass, we found that the K63 linkage can be replaced by a non-physiological M1 linkage, indicating that linkage identity is not necessarily a requirement for functionality.

Our strategy of linkage reprogramming has the potential to greatly expand the ways in which polyubiquitylation can be studied. It should be generalizable to a wide range of substrates in vivo and possibly even to linkages beyond K63, K48, and M1. Further applications of linkage reprogramming are envisioned wherever polyubiquitylation is carried out by multiple, sequentially acting sets of conjugation factors. Examples include RNA polymerase II, where a monoubiquitin unit attached by Rsp5 is extended to a K48-linked chain by a cullin-based E3 (Harreman et al., 2009), substrates of APC/C-mediated K11 polyubiquitylation, which requires the sequential action of two distinct E2s in both yeast and human cells (Rodrigo-Brenni and Morgan, 2007; Wu et al., 2010), as well as other proteins subject to polyubiquitin chain extension by E4 activity (Hoppe, 2005). In these systems, linkage reprogramming would serve as an ideal tool to explore the effect of chain structure on proteasomal recognition and probe the ubiquitin threshold model, which postulates that multiple short or even branched chains are better degradation signals than a uniform K48-linked chain (Swatek and Komander, 2016). In the context of endocytosis, the system would be applicable to two-stage ubiquitylation, for example of MHC class I molecules, induced by the Kaposi’s-sarcoma-associated herpesvirus (KSHV) E3s, K3 and K5, and involving two distinct E2s, UBCH5B/C and UBC13, for mono- and polyubiquitylation, respectively (Boname et al., 2010; Duncan et al., 2006).

In cases where a single E2-E3 pair mediates chain initiation and extension, the tailor-made E3s could also be used to probe the effects of chain branching. Thus, the tool described here has the potential to provide insight into ubiquitin signaling both in situations where a uniform linkage is expected to be required for physiological activity, as well as in less clearly defined cases where more complex rules may apply (Kravtsova-Ivantsiv and Ciechanover, 2012).

Limitations of the study

Our approach resembles and expands the degron technology, where a substrate is targeted to an endogenous or deliberately introduced E3 for polyubiquitylation and subsequent proteasomal degradation (Natsume and Kanemaki, 2017). However, in contrast to the E2s and E3s involved in degron systems, the enzymes underlying our strategy are highly specialized in chain extension by using a single site on ubiquitin as a substrate (Parker and Ulrich, 2009; Petroski et al., 2007; Renz et al., 2020; Smit et al., 2013). This intrinsic preference endows them with exquisite linkage specificity, but at the same time prevents them from efficient chain initiation. Thus, the system relies on prior monoubiquitylation, in case of PCNA triggered by damage-induced activation of Rad6-Rad18. Generalization of the method is therefore currently limited to situations that merely require the extension of a pre-existing ubiquitin unit. Simply fusing a monoubiquitin unit to the desired target protein to supply this signal would in most cases not be advisable due to potential undesired effects of monoubiquitylation by itself, for example, in the context of the UFD pathway.

A two-step strategy could potentially provide a way around this limitation even in cases where mono- and polyubiquitylation are carried out by the same E2-E3 pair. This would involve an inhibition of the physiological linkage, for example, by replacement of endogenous ubiquitin with or overexpression of a ubiquitin mutant deficient in assembling this chain type, which would restrict the signal-induced modification by the endogenous enzymes to monoubiquitylation. An engineered E3 with an alternative linkage specificity could then be used to produce the desired conjugate by means of chain extension.

Whereas PIP-E3(63) and PIP-E3(48) did not exhibit significant effects on overall ubiquitylation in vivo, PIP-E3(1) appears to have off-target effects in yeast, as evidenced by the appearance of linear polyubiquitin chains and the enhanced ubiquitylation of several cellular proteins unrelated to PCNA. Although such unintended activity could potentially be problematic, it does not invalidate the use of the tailor-made E3s for functional analyses because it can be controlled for in a straightforward manner by mutation of the ubiquitin acceptor site on the substrate of interest. If such mutation abolishes the phenotypic readout, the activity of the E3 can be attributed to the relevant substrate despite the modification of other proteins beyond the desired target.

Another limitation pertains to the variable activities of the tailor-made E3s that we observed toward the different substrates. These might not only reflect intrinsic differences in their catalytic properties when combined with their specific substrate targeting motifs, but also result from linkage-selective in vivo deubiquitylation activities that may vary depending on the subcellular location. As we have shown, this variability among the E3s can to some extent be overcome by means of altering their domain arrangement, adjusting their expression levels or co-overexpression of the cognate E2s. A feature that is more difficult to control and somewhat limits the direct comparability of the different linkages is the length of the polyubiquitin chain assembled by the respective E3s. This factor not only depends on their overall activity but also on their processivity and may well affect the selectivity of downstream effector proteins (Piotrowski et al., 1997). Finally, expansion of the system to other linkages beyond K48, K63, and M1 will have to await the development of suitable linkage-specific conjugation factors.

Insight into the consequences of PCNA polyubiquitylation

Polyubiquitylation of PCNA was identified as a hallmark of the DNA damage bypass pathway and has since served as a paradigm of how the ubiquitin system contributes to genome maintenance (Hoege et al., 2002; Ulrich and Walden, 2010). Yet, its mode of action remains enigmatic because the consequences of the modification are still poorly understood. Our approach has revealed key mechanistic information by defining the features of PCNA polyubiquitylation required for its action in vivo and has allowed us to evaluate the three possible models presented earlier: first, by demonstrating that both PIP-E3(63) and PIP-E3(1) support TS, we have shown that any downstream effector of this pathway involved in recognition of polyubiquitylated PCNA is likely not absolutely specific for the K63 linkage. The structural similarity of the M1 to the K63 linkage makes it conceivable that such a reader would interact with both, even though many UBDs exhibit varying degrees of selectivity for either K63 (Rap80-UIM or TAB2-NZF) or M1 chains (OTULIN-OTU or NEMO-UBAN) (Erpapazoglou et al., 2014; Fennell et al., 2018). Indeed, the two most relevant interactors of polyubiquitylated PCNA, ZRANB3 and Mgs1, do not discriminate between the K63 and the M1 linkage and also recognize linear chain mimics with non-native junctions (Saugar et al., 2012; Takahashi et al., 2020; Zhao and Ulrich, 2010). However, ZRANB3 is not conserved in yeast, and the phenotype of mgs1 mutants is inconsistent with a direct participation in TS (Hishida et al., 2001, 2006; Saugar et al., 2012). Thus, an alternative yet unidentified reader of the modified clamp responsible for mediating TS cannot be ruled out.

Second, we have defined the role of Rad5 in the TS pathway. While it is established that its ATPase activity and its participation in TLS are separable from its ligase function (Ball et al., 2014; Choi et al., 2015; Gallo et al., 2019), we have now shown that engineered, PCNA-specific E3s are able to compensate for loss of Rad5 function in TS. This indicates that the ligase activity, directed toward PCNA as a substrate, is necessary and sufficient for its contribution to this pathway.

Finally, our observation that enzymatically assembled, but not pre-formed polyubiquitin conjugates support TS possibly suggests a relevance of PCNA ubiquitylation as a dynamic process. From the activity of both K63- and M1-linked chains in TS, it could be concluded that the function of the polyubiquitin chain on PCNA were to simply provide a high concentration of ubiquitin that does not trigger proteasomal degradation. However, this model does not explain the failure of the genetically encoded chain mimics fused to PCNA to support TS. A dominant-negative effect, for example, by sequestering an important effector away from the relevant sites, has been ruled out (Takahashi et al., 2020). Given that these mimics harbor mutations in ubiquitin’s C-terminal residue to prevent their disassembly, the underlying cause could thus be a downstream effector that interacts with both K63- and M1-linked chains but does not tolerate any alterations of ubiquitin’s C terminus or steric bulk at the junction. However, an alternative explanation would be a requirement for PCNA ubiquitylation or deubiquitylation during damage bypass. In support of this argument, the persistence of monoubiquitylated PCNA was found to cause replication fork destabilization and cell death in BRCA1-deficient tumor cells (Lim et al., 2018). Since inactivation of two relevant DUBs, Ubp10 and Ubp12, does not sensitize budding yeast to DNA damage (Álvarez et al., 2019; Gallego-Sánchez et al., 2012), PCNA deubiquitylation is unlikely to be critical for successful completion of TS in this organism. It is therefore attractive to speculate that in situ assembly of the polyubiquitin chain on PCNA, i.e., at the sites of damage bypass, is important for TS. For example, the growth of the polyubiquitin chain itself could potentially constitute an essential signal for TS that cannot be supplied by the chain mimics. Understanding the molecular basis for this phenomenon may provide further insight into the mechanism of the TS pathway and the relevance of chain geometry. When compared with the chain mimics, our strategy of linkage reprogramming highlights the advantages of using enzymatic approaches to direct linkage- and substrate-specific polyubiquitylation toward a physiologically relevant location within the cell.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Helle D. Ulrich (h.ulrich@imb-mainz.de).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact upon request.

Experimental model and subject details

All Saccharomyces cerevisiae strains used in this study are haploid and are derived from DF5 (Finley et al., 1987), with the exception of the strains used for protein production and two-hybrid assays (see Table S1). They were cultured at 30°C unless otherwise noted, either in YPD or – if noted – in synthetic complete (SC) medium supplemented with the relevant amino acids.

Method details

Plasmid construction

Relevant oligonucleotides are listed in Table S2. Plasmids, including those for protein production, are listed in Table S3. E3 constructs for in vitro activity assays were generated with a His6- or GST-tag for purification. If not stated otherwise, E3 constructs for expression in yeast were cloned without affinity tag in integrative yeast vectors under control of the strong, constitutive ADH1 promoter. A VSV-tag was used for detection. In addition, regulable expression of PIP-E3(63) was achieved by a doxycycline-inducible TET-ON promoter (Bellí et al., 1998). Vectors expressing Ub∗-PCNA∗ fusions under control of the native POL30 promotor (Takahashi et al., 2020; Zhao and Ulrich, 2010) or a His6-tagged version were integrated into the URA3 locus. Analogous head-to-tail mimics of polyubiquitylated PCNA were integrated into the LEU2 locus. Expression vectors for LaG16-E3(48), LaG16-E3(63) and LaG16-E3(1) were generated by replacing the PIP motif of the PIP-E3s with the sequence of a GFP-specific nanobody, LaG-16 (Fridy et al., 2014) that had been codon-optimized for expression in yeast, followed by a linker peptide. The K63-specific E3 construct was further optimized by swapping the E3 domain with the nanobody domain to yield E3(63)-LaG16. The GFP substrates (Ub-GFP and its derivatives) bear a G76V mutation at the C terminus of the ubiquitin moiety, followed by a sequence derived from the E. coli lac repressor, eΔK (Johnson et al., 1995). They were encoded by episomal plasmids under control of the CUP1 promotor, using URA3 as selection marker. For two-hybrid assays, fusions with the GAL4 activation and DNA-binding domains (GAD, GBD) were generated in the pGAD-C and pGBD-C vector series (James et al., 1996). UBI as the sole source of ubiquitin was expressed from a 2μ vector under control of the CUP1 promoter as described for ubiquitin lysine mutants (Spence et al., 1995).

Protein production and purification

Recombinant, E. coli-derived human HisUb was purchased from BioTechne, and WT ubiquitin was from Sigma. All other recombinant proteins were produced in E. coli, except for HisRad5, which was overexpressed in S. cerevisiae strain BJ5460. Untagged UbK48R and UbK63R were purified by acid precipitation (Pickart and Raasi, 2005). Murine HisUba1 (E1) and budding yeast HisRad5, HisUbc13, Mms2, and PCNA were purified as described (Parker and Ulrich, 2009). HisUbc4, HisUbc5, GSTUBCH5A, MBPUbc7His, HisPIP-E3(1) and its variants, GSTPIP-E3(48) and HisUlp1 were purified by standard affinity chromatography (Ni-NTA or glutathione Sepharose for His- or GST-tagged proteins, respectively), followed by size exclusion chromatography in buffer A (40 mM Tris-HCl, pH 7.4, 200 mM NaCl, 10% glycerol, 1 mM EDTA and 1 mM DTT) on a Superdex 200 or Superdex 75 Increase 10/300 GL column. Ub-GFP was used without further purification as a diluted, cleared E. coli lysate. Ub-PCNA was prepared as follows: HisSmt3-Ub-PCNA expression in E. coli was induced with 0.2 mM IPTG for 20 h at 18°C. Cells were resuspended in buffer B (50 mM Tris-HCl pH 7.4, 250 mM NaCl, 10% glycerol, 1 mM DTT, 20 mM imidazole) and lysed by high-pressure cell disruption, followed by addition of 0.05% Tween 20 and SIGMAFAST protease inhibitor and clearance of the lysate by centrifugation at 18,500g for 30 min. The supernatant was subjected to affinity chromatography on Ni-NTA resin, and eluted protein was rebuffered in buffer C (50 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM DTT) and subjected to cleavage with recombinant HisUlp1 at 16°C overnight. After renewed passage through Ni-NTA, untagged Ub-PCNA was purified by gel filtration in buffer A. PIP-E3(63) was expressed as a GST fusion and purified as follows: cells were resuspended in buffer D (40 mM HEPES pH 7.5, 150 mM NaCl, 1 mM DTT and SIGMAFAST protease inhibitor), and the cleared lysate was applied to glutathione Sepharose for affinity purification. Bound protein was eluted in buffer E (40 mM Tris-HCl pH 8, 150 mM NaCl, 15 mM glutathione, 1 mM DTT), rebuffered into Merck Millipore’s Thrombin cleavage buffer, and the GST moiety was cleaved with Thrombin (Merck Millipore) at 4°C overnight. After renewed passage over glutathione Sepharose, untagged PIP-E3(63) was subjected to gel filtration in buffer A.

In vitro ubiquitylation assays

Standard reactions were performed in 10 μL aliquots in a buffer containing 40 mM HEPES, pH 7.4, 8 mM magnesium acetate, 50 mM NaCl, 100 μM ATP, 0.1 μM E1 and 1 μM E3 as indicated. Ubc13-Mms2 were used at 0.2 μM, other E2s at 2 μM. For analysis of unanchored chain synthesis, ubiquitin was used at 50 μM and reactions were incubated at 30°C for 7 min. For ubiquitylation of substrates, the ubiquitin concentration was 5 μM and Ub-PCNA, PCNA or Ub-GFP was present at 2 μM. Reactions were incubated at 30°C for 40 min, quenched by addition of NuPAGE LDS sample buffer supplemented with 25 mM DTT and analyzed by SDS-PAGE, followed by western blotting against ubiquitin and PCNA. Blots were imaged with an Odyssey CLx system (LI-COR) using secondary antibodies tagged with near-infrared fluorophores (LI-COR Biosciences IRDye 680RD and 800CW). Kinetic parameters were measured with ubiquitin variants that would limit chain formation to a single extension step and thus yielded a unique, quantifiable product (UbK63R, UbK48R or HisUb for PIP-E3(63), PIP-E3(48) and PIP-E3(1), respectively). Product formation was quantified by SDS-PAGE, followed by SYPRO Orange staining of the gels and imaging on a Typhoon FLA9500 reader. In a first step, reaction conditions were identified where E3 concentration was rate-limiting. Final reactions contained 0.15 μM E1, 2 μM E2, 5 μM ubiquitin mutant and 0.05 μM Rad5 or PIP-E3(63), 0.2 μM PIP-E3(1) or 2 μM PIP-E3(48) in the standard buffer described above. Next, initial reaction velocities (V0) were determined by measuring product formation at different substrate concentrations at time points of 1, 2 and 3 min (for Rad5: 0.5, 1 and 1.5 min).

Construction of yeast strains

Gene deletions were created via PCR-based methods or combined by mating and tetrad dissection. The ufd2 allele was generated by replacing the U-box sequence with a PCR cassette encoding a 9myc-tag. Promoter replacements or marker exchanges were generated analogously by transformation with PCR cassettes. E3 constructs were introduced into the relevant strains via integration into the LEU2 or URA3 locus and GFP substrates were expressed from episomal vectors with the URA3 marker.

DNA damage and temperature sensitivity assays

Quantitative survival assays were performed by plating a defined number of cells from exponential cultures in triplicate and UV irradiation (254 nm) with the indicated doses. Colonies were counted after incubating the plates in the dark at 30°C for 3 days. For comparative spot assays, 1.6ˑ106 cells from exponential cultures were spotted in 5-fold serial dilutions onto freshly prepared plates that had been supplemented with the indicated concentrations of MMS or were exposed to the indicated doses of UV radiation (254 nm) after spotting. Plates were incubated at 30°C or – for temperature sensitivity assays – at the indicated temperatures for two days before scoring.

Yeast two-hybrid assays

Interactions of PIP-E3s with all budding yeast E2s were analyzed in the two-hybrid system as described previously (Renz et al., 2020), using strain backgrounds PJ69-4A and PJ69-4α (James et al., 1996). E2-E3 interaction assays were spotted by robotic pinning in quadruplicate.

Detection of E3s and PCNA modifications in yeast

Total cell extracts were prepared as described (Morawska and Ulrich, 2013). To induce replication stress, exponentially growing cultures of relevant yeast strains were treated with 0.02% MMS for 90 min. To detect PCNA modifications, HisPCNA or its derivatives were isolated by Ni-NTA affinity purification under completely denaturing conditions as described previously (Davies and Ulrich, 2012; Hoege et al., 2002), and ubiquitylated forms were detected by western blotting for ubiquitin and PCNA. E3s were detected in total extracts by means of anti-VSV antibodies. To detect PIP-E3(1) by fluorescence microscopy, co-localization of a GFP-tagged version with mCherry-tagged Hta2 (histone H2A) was analyzed in live cells on a Deltavision TM Elite widefield microscope equipped with a GFP/mCherry wide filter set.

In vivo polyubiquitylation of Ub-GFP

Modification of Ub-GFP was detected by western blotting of total lysates from cells growing exponentially in SC-URA medium. Cell extracts were prepared as described (Morawska and Ulrich, 2013) and substrates were detected by anti-GFP western blotting and imaging with an Odyssey CLx system (LI-COR). Expression of the relevant E3s was confirmed by anti-VSV blots. Equal loading of the gel was confirmed by Ponceau staining.

Analysis of total ubiquitin conjugates by western blotting

Yeast strains (lys1Δ pdr5Δ) expressing PIP-E3s or harboring an empty vector were cultured in SC-Complete medium to an OD600 of 0.5 and treated with 75 μM MG132 or DMSO for 90 min. Total lysates were analyzed by western blotting. For probing membranes with linkage-specific anti-ubiquitin antibodies, wet blots were performed at 4°C for 2 h at 30 V. Blots were imaged with an Odyssey CLx system (LI-COR).

SILAC labeling and yeast lysis for diGly proteomics

For SILAC experiments, lys1Δ pdr5Δ cells expressing PIP-E3s or harboring an empty vector were cultured twice to saturation in SC complete medium containing 15 mg·L-1 L-arginine and 30 mg·L-1 L-lysine0 (light) or L-lysine8 (heavy), respectively. All experiments were performed in triplicates, including one label switch. PIP-E3(48)-expressing strains and their corresponding empty vector controls were treated with 75 μM MG132 for 90 min before harvesting. At an OD600 of approximately 3, equal numbers of cells (total: 700 OD600) were mixed and harvested by centrifugation. Cell lysis was performed with a PreCellys instrument in 4 mL of RIPA buffer (50 mM HEPES, pH 7.5, 1 mM EDTA, 140 mM NaCl, 0.1% (v/v) Triton, freshly supplemented with SIGMAFAST Protease Inhibitor Cocktail and 10 mM N-ethylmaleimide). Aliquots of 800 μL cell suspension were treated with 200 μL of zirconia beads each for 10 cycles at 6800 rpm at 2°C. Triton concentration was increased to 1% (v/v) and after 15 min incubation at 4°C the supernatant was collected through centrifugation (2,000g for 2 min and again at 20,000g for 20 min at 4°C). Protein concentration of cell lysates was determined using a BCA assay.

Proteome-wide analysis of ubiquitylation sites by diGly proteomics

Proteins were precipitated in a fourfold excess of ice-cold acetone and re-solubilized in denaturation buffer (6 M urea, 2 M thiourea in 10 mM HEPES, pH 8.0). Cysteines were reduced with 1 mM dithiothreitol and alkylated with 5.5 mM chloroacetamide. Proteins were digested with Lys-C (Wako) and sequencing-grade modified trypsin (Serva). Digestion was stopped by adding trifluoroacetic acid to 1%, and precipitates were removed by centrifugation. Peptides were purified using reversed-phase Sep-Pak C18 cartridges (Waters) and eluted in 50% acetonitrile. Immunoprecipitation buffer (1× IAP: 10 mM Na2HPO4, 50 mM NaCl in 50 mM MOPS, pH 7.2) was added to 0.2× and acetonitrile was removed by vacuum centrifugation. DiGly-remnant peptides were enriched using 80 μL of diGly-Lys antibody resin (Cell Signaling Technology). Peptides were incubated with the antibody for 4 h at 4°C on a rotation wheel. Beads were washed four times in ice-cold 1×IAP, followed by two washes in water. Enriched peptides were eluted with 0.15% trifluoroacetic acid in water and desalted on reversed-phase C18 StageTips (Rappsilber et al., 2007). Enriched peptides were analyzed on a quadrupole Orbitrap mass spectrometer (Exploris 480, Thermo Scientific) equipped with a UHPLC system (EASY-nLC 1200, Thermo Scientific). Peptide samples were loaded onto a C18 reversed-phase column (50 cm length, 75 μm inner diameter, packed in-house with ReproSil-Pur 120 C18-AQ 1.9-μm beads, Dr. Maisch GmbH) and eluted with a gradient from 2.4 to 32% acetonitrile containing 0.1% formic acid in 90 min. The mass spectrometer was operated in data-dependent mode, automatically switching between MS and MS2 acquisition. Survey full scan MS spectra (m/z 300–1,650, resolution: 60,000, target value: 3 × 106, maximum injection time: 40 ms) were acquired in the Orbitrap. The 15 most intense precursor ions were sequentially isolated, fragmented by higher energy C-trap dissociation (HCD) and scanned in the Orbitrap mass analyzer (normalized collision energy: 30%, resolution: 15,000, target value: 1 × 105, maximum injection time: 40 ms, isolation window: 1.4 m/z). Precursor ions with unassigned charge states, as well as with charge states of +1 or higher than +8, were excluded from fragmentation. Precursor ions already selected for fragmentation were dynamically excluded for 25 s.

Mass spectrometry data analysis

Raw data files were analyzed using MaxQuant (version 1.5.2.8) (Cox and Mann, 2008). Parent ion and MS2 spectra were searched against a reference proteome database containing S. cerevisiae protein sequences obtained from UniProtKB (version 2017_02) using Andromeda search engine (Cox et al., 2011). Protein sequences of the engineered E3s (PIP-E3(1), PIP-E3(48) and PIP-E3(63)) were appended to the reference genome. Spectra were searched with a mass tolerance of 4.5 ppm in MS mode, 20 ppm in HCD MS2 mode, strict trypsin specificity and allowing up to two miscleavages. Cysteine carbamidomethylation was searched as a fixed modification, whereas protein N-terminal acetylation, methionine oxidation, N-ethylmaleimide modification of cysteines (mass difference to cysteine carbamidomethylation), and diGly-Lys were searched as variable modifications. Site localization probabilities were determined by MaxQuant using the PTM scoring algorithm (Olsen et al., 2006). The dataset was filtered based on posterior error probability (PEP) to arrive at a false discovery rate of below 1% estimated using a target-decoy approach (Elias and Gygi, 2007). DiGly-Lys-modified peptides with a minimum score of 40, delta score of 6 and localization probability higher than 90% are reported and used for the analyses. Statistical analysis and MS data visualization were performed using the R software environment. P-values and false discovery rates were calculated using a moderated t test (Limma algorithm) (Ritchie et al., 2015).

Quantification and statistical analysis

To determine kinetic parameters of the PIP-E3s, initial rates were plotted as a function of substrate concentration and fitted with GraphPad Prism 7 to the standard Michaelis-Menten equation (V0 = Vmax ˑ [S]/(KM+S); kcat = Vmax/[E]T). Averages and standard deviations were calculated from three technical replicates.

UV sensitivity assays were quantified by determining survival after irradiation in at least three independent experiments, each performed with three technical replicates, and generating averages and standard deviations from all values combined.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-PCNA	In-house	N/A
Mouse monoclonal anti-ubiquitin (clone P4D1)	Cell Signaling Technologies	Cat# 3936; RRID: AB_331292
Mouse monoclonal anti-ubiquitin (clone FK2)	Merck KgaA	Cat# 04-263; RRID: AB_612093
Rabbit polyclonal anti-ubiquitin	Dako	Cat# Z045801-2; RRID: AB_2315524
Rabbit monoclonal anti-ubiquitin-K48 (clone Apu2)	Merck KgaA	Cat# 05-1307; RRID: AB_1587578
Rabbit monoclonal anti-ubiquitin-K63 (clone Apu3)	Merck KgaA	Cat# 05-1308; RRID: AB_1587580
Rabbit monoclonal anti-ubiquitin-M1 (clone 1E3)	Merck KgaA	Cat# MABS199; RRID: AB_2576212
Rabbit monoclonal anti-α-tubulin (clone EPR13799)	Abcam	Cat# ab184970
Mouse monoclonal anti-VSV (clone P5D4)	Merck KgaA	Cat# A5977; RRID: AB_439710
Mouse monoclonal anti-GFP (clone 7.1/13.1)	Roche	Cat# 11814460001; RRID: AB_390913
IRDye 680LT donkey anti-rabbit IgG secondary antibody	LI-COR	Cat# 926-68023; RRID: AB_10706167
IRDye 680RD donkey anti-mouse IgG secondary antibody	LI-COR	Cat# 926-68072; RRID: AB_10953628
IRDye 800CW goat anti-rabbit IgG secondary antibody	LI-COR	Cat# 926-32211; RRID: AB_621843
IRDye 800CW donkey anti-mouse IgG secondary antibody	LI-COR	Cat# 926-32212; RRID: AB_621847
Bacterial strains
Escherichia coli BL21-CodonPlus(DE3)-RIL	Agilent Technologies	CatST230240
Chemicals, peptides, and recombinant proteins
Recombinant human His6-ubiquitin	Bio-Techne	Cat# U-530
Ubiquitin from bovine erythrocytes	Merck KgaA	Cat# U6253
SIGMAFAST protease inhibitor cocktail	Merck KgaA	Cat# S8830
Ni-NTA agarose	Qiagen	Cat# 30250
Glutathione Sepharose	Cytiva	Cat# 17-5132-02
Chitin beads	New England Biolabs	Cat# S6651
Heparin column	Cytiva	Cat# 17-0406-01
Methyl methanesulfonate	Merck KgaA	Cat# 129925
Imidazole	Merck KgaA	Cat# I2399
Glutathione	Merck KgaA	Cat# G4251
IPTG	Generon	Cat# GEN-S-02122
ATP	New England Biolabs	Cat# P0756
N-ethylmaleimide	Merck KgaA	Cat# E3876
SIGMAFAST Protease Inhibitor Cocktail Tablets, EDTA-free	Merck KgaA	Cat# S8830
MG132	Enzo Life Sciences	Cat# BML-PI102
Trypsin, MS-approved	Serva	Cat# 37286
Lys-C	Wako	Cat# 129-02541
L-Lysine monohydrochloride	Merck KgaA	Cat# L8662
L-Arginine monohydrochloride	Merck KgaA	Cat# A6969
L-Lysine-8 dihydrochloride (13C6, 99%; 15N2, 99%)	Euroisotop	Cat# CNLM-291-H-1
Critical commercial assays
Thrombin cleavage capture kit	Merck KgaA	Cat# 69022-3
Vivaspin concentrators, 10,000 & 30,000 MWCO	Thermo Fisher Scientific	Cat# 10738231, Cat# 10029460
HisTrap column (5 mL)	Cytiva	Cat# 17-5248-02
GSTrap HP column (5 mL)	Cytiva	Cat# 28-4017-48
HiTrap Q column	Cytiva	Cat# 17-1153-01
HiTrap SP column	Cytiva	Cat# 17-1151-01
MonoQ 5/50 GL column	Cytiva	Cat# 17-5166-01
Superdex 75 Increase 10/300 GL column	Cytiva	Cat# 29-1487-21
Superdex 200 Increase 10/300 GL column	Cytiva	Cat# 28-9909-44
SYPRO Orange protein gel stain	Thermo Fisher Scientific	Cat# S6651
Pierce BCA Protein Assay Kit	Thermo Fisher Scientific	Cat# 23227
PTMScan Ubiquitin Remnant Motif (K-ε-GG) Kit	Cell Signaling Technology	Cat# 5562
Deposited data
DiGly proteomics data	This study	PRIDE: PXD030735
Experimental models: Organisms/strains
Saccharomyces cerevisiae: strain background BJ5460	ATCC	Cat# 208285
Saccharomyces cerevisiae: strain background PJ69-4A	James et al., 1996	N/A
Saccharomyces cerevisiae: strain background DF5	Finley et al., 1987	N/A
Other yeast strains: see supplemental information	This study	Table S1
Oligonucleotides
Oligonucleotides for cloning: see supplemental information	Sigma-Aldrich Chemie GmbH	Table S2
Recombinant DNA
Plasmids used in this study: see supplemental information	This study	Table S3
Software and algorithms
Prism 7	GraphPad	N/A
MaxQuant	Development version 1.5.2.8	N/A
RStudio	Development version 1.3.1093	N/A
Original code: see supplemental information	This study	Methods S1