Ubiquitin-conjugating enzymes (E2) enable protein ubiquitination by conjugating ubiquitin to their catalytic cysteine for subsequent transfer to a target lysine side chain. Deprotonation of the incoming lysine enables its nucleophilicity, but determinants of lysine activation remain poorly understood. We report a novel pathogenic mutation in the E2 UBE2A, identified in two brothers with mild intellectual disability. The pathogenic Q93E mutation yields UBE2A with impaired aminolysis activity but no loss of the ability to be conjugated with ubiquitin. Importantly, the low intrinsic reactivity of UBE2A Q93E was not overcome by a cognate ubiquitin E3 ligase, RAD18, with the UBE2A target PCNA. However, UBE2A Q93E was reactive at high pH or with a low-pKa amine as the nucleophile, thus providing the first evidence of reversion of a defective UBE2A mutation. We propose that Q93E substitution perturbs the UBE2A catalytic microenvironment essential for lysine deprotonation during ubiquitin transfer, thus generating an enzyme that is disabled but not dead.
Protein ubiquitination is a post-translational modification that regulates both protein function and abundance1. A trio of enzymes, ubiquitin-activating enzyme (E1), E2, and ubiquitin-ligating enzyme (E3), work in series to conjugate ubiquitin to a lysine residue on a target protein through an isopeptide bond2. The modification has myriad cellular effects, and, as a consequence, mutations in components of the ubiquitination machinery are linked to human diseases, including cancer and neurodevelopmental and neurodegenerative disorders3. Mutations in ubiquitination enzymes are documented contributors to neurodevelopmental pathophysiology in intellectual disability (ID) and autism spectrum disorders4.
A prototypical disorder involving an altered ubiquitination enzyme is X-linked intellectual disability (XLID) type Nascimento5, related to abnormalities in the UBE2A gene (gene ID 7319). The human locus for the UBE2A gene is at Xq24, and the transcript is ubiquitously expressed; the highest mRNA levels are in the heart, testis, and brain6. Individuals with this syndrome carry intragenic mutations (point mutations or small deletions) or larger Xq24 deletions encompassing the UBE2A gene5,7,8,9,10,11,12,13,14,15,16,17,18,19. Common clinical features include moderate to severe ID, prominent dysmorphic features, impaired speech, urogenital malformations, skin abnormalities, and epilepsy10. Larger deletions encompassing UBE2A and adjacent genes appear to result in severe forms of the clinical syndrome with a higher prevalence of white-matter changes, heart defects, and urogenital malformations than observed in patients with intragenic UBE2A pathogenic variants.
UBE2A shares 95% amino acid identity with UBE2B (gene ID 7320) at 5q31.1, but UBE2B variants are associated with only male infertility20. Molecular mechanisms linking mutations in UBE2A to neurodevelopmental disorders are still poorly understood. Emerging data implicate defective mobilization of the E3 Parkin as a source of neuronal vesicle trafficking and clearance of dysfunctional mitochondrial abnormalities, both of which are suspected to contribute to the neurodevelopmental phenotype in patients with UBE2A-deficiency syndrome11. Previously reported pathogenic mutations are distributed along the six exons of the UBE2A gene5,7,8,9,10,11,12,13,14,15,16,17,18,19. Limited functional and structural information on mutant forms of UBE2A limit current understanding of the contribution of specific UBE2A variants to the pathogenesis of ID.
We identified a novel missense mutation, c.277 C>G (p.Q93E) in UBE2A, in two brothers presenting with an atypical phenotype of only mild ID and impaired speech. We characterized the effects of this novel mutation on UBE2A function and structure to reveal the molecular mechanisms underlying the human disorder. We found that the Q93E mutation yields an enzyme with impaired ability to transfer ubiquitin to lysine, thereby inhibiting product formation. The defect is not rescued by the presence of an E3 ligase (RAD18) toward a specific UBE2A target, PCNA. Our results implicate Q93 of UBE2A in the deprotonation of an incoming lysine, a crucial step for ubiquitin transfer from an E2~ubiquitin conjugate to the substrate. Strikingly, the activity of the UBE2A Q93E mutant is partially recovered by increased pH, thus providing initial evidence of a potential reversion of a defective mutation in UBE2A related to XLID type Nascimento.
Identification of UBE2A Q93E mutation in patients with XLID
Two brothers from a Brazilian family (Fig. 1a) were diagnosed with idiopathic mild ID and speech impairment. Extensive clinical genetic investigation, including fragile X syndrome screening, karyotyping, and chromosome-microarray analysis, did not identify the etiology of the condition. Recent whole-exome sequencing revealed that the affected individuals carry a missense variant in UBE2A exon 5 (chromosome X: 1187165 c.277 C>G). This variant had not previously been reported. Sanger sequencing confirmed the mutation in the probands and was used to determine familial segregation (Fig. 1b). The mother and three unaffected daughters are heterozygous carriers of the UBE2A mutation (Supplementary Fig. 1). All female carriers show extreme skewed X inactivation in peripheral blood (>90%). The variant was therefore considered the cause of the phenotype. The mutation results in an amino acid change of glutamine to glutamate at position 93 near the active site of UBE2A. Position 93 is a glutamine in both UBE2A and UBE2B, but this residue is not highly conserved among E2s (Supplementary Fig. 2).
The Q93E mutation impairs UBE2A activity
Given the proximity of Q93 to the catalytic cysteine, we tested whether the mutation might alter general UBE2A activity. Purified recombinant wild-type (WT) and Q93E-mutant UBE2A were assayed for their ability to generate polyubiquitin chains21. Each E2 was incubated with E1, ATP/Mg2+, and ubiquitin, in the absence of an E3 or a substrate. WT UBE2A generated high-molecular-weight products, as detected by western blotting against ubiquitin, but there was a marked lack of products generated by the mutant E2 (Fig. 1c). To test whether the loss of activity was due to impaired conjugation of ubiquitin to the mutant E2 by E1, we compared assays in which the E2~ubiquitin (Ub) conjugate was visualized directly (Fig. 1d). Both WT and Q93E-mutant UBE2A were robustly conjugated with ubiquitin, thus suggesting that the defect was in the second (transfer) step of product formation.
In the assay shown in Fig. 1c, products are generated when a lysine on an acceptor ubiquitin (Ubacceptor) attacks the thioester of E2~Ubdonor, and an aminolysis reaction ensues. To test whether the Q93E mutation might affect the intrinsic aminolysis activity of UBE2A, we performed lysine discharge assays. Activity toward the ε-amino group of free lysine reflects the ability of E2~Ub conjugates to transfer ubiquitin without complications associated with protein substrates22,23. After the addition of free lysine to E2~Ub, WT UBE2A~Ub was depleted within ~10 min, whereas depletion of UBE2AQ93E~Ub took longer than 30 min (Fig. 1e and Supplementary Fig. 3a). Hence, UBE2A Q93E is compromised in aminolysis activity toward free lysine and lysine attached to ubiquitin.
The activity of other UBE2A mutations previously reported to be involved in XLID type Nascimento, namely R7W, R11Q, and G23R7,11, were compared. Each XLID mutant showed a detectable decrease in polyubiquitin-chain formation in vitro (Supplementary Fig. 4a), although the decreases were more modest than those observed for the Q93E mutant. Ubiquitin conjugation to the E2 active site was not impaired in these mutants (Supplementary Fig. 4b). These mutation sites are far from the E2 active site: R7 and R11 are in the putative E1/E3-binding surface, and G23 is on the ‘back side’ surface of UBE2A that binds the E3 enzyme RAD18 (ref. 21).
XLID Q93E mutation affects the UBE2A catalytic site
We determined crystal structures of WT and Q93E-mutant UBE2A at 1.85- and 2.20-Å resolution, respectively (Fig. 2a and Supplementary Table 1) and compared the structures. Superposition of the two structures revealed high similarity (r.m.s. deviation of 0.17 Å) and a typical E2 fold, featuring a four-stranded antiparallel β-sheet, four α-helices, and a short 310 helix24. The most notable difference is the loss of a hydrogen bond between the Q93 side chain and the backbone carbonyl of L89—the residue adjacent to active site C88 (Fig. 2b and Supplementary Fig. 5).
We used NMR spectroscopy to gain more insight into the basis for the lower activity of UBE2A Q93E. Assignment of backbone resonances of UBE2A Q93E was performed with standard triple-resonance approaches; the backbone assignments for WT UBE2A were based on previously reported UBE2B NMR spectra (BMRB 17443). Overlay of the (1H,15N)-HSQC spectra of WT and Q93E-mutant UBE2A indicated that the spectra are highly similar, and residues near the catalytic site (T69, V70, L89, W96, and A122) and the catalytic C88 (Fig. 2c–e) exhibit small chemical-shift perturbations (CSPs). In agreement with the loss of hydrogen-bonding between L89 and Q93 implied from the mutant crystal structure, the two largest CSPs are for L89 and C88, thus confirming that the Q93E mutation detectably alters the chemical environment of the catalytic residue.
XLID Q93E mutation preserves the UBE2A~Ub conformation
The orientation and interactions of ubiquitin within the E2 active site play a central role in its transfer to substrate25. In the structurally characterized E2~Ub conjugates, the conjugated ubiquitin is flexible and assumes a range of conformations relative to the E2 domain26. Binding of an E2~Ub to some RING-type E3s promotes a ‘closed’ state of the conjugate, in which ubiquitin is positioned against the E2 helix α2 (cross-over helix) (refs 26,27). The closed state has enhanced reactivity, presumably through positioning the thioester bond for nucleophilic attack by incoming lysine28. We investigated whether the ability of UBE2A~Ub to adopt a closed conformation might be altered in the mutant, by monitoring the formation of an oxyester conjugate of the UBE2A C88S variant, which is more stable than the thioester conjugate25. [15N]UBE2AC88S or [15N]UBE2AC88S Q93E was conjugated with unlabeled ubiquitin directly in the NMR tube, in the presence of E1 and Mg2+. After the reaction was started by the addition of ATP, spectral changes were monitored through acquisition of sequential (1H,15N)-HSQC spectra. The spectrum of oxyester-linked [15N]UBE2AC88S–O~Ub showed substantial CSPs, especially for peaks corresponding to residues surrounding the UBE2A catalytic site (Fig. 3a,b). We observed perturbations to residues on helix α2 (residues D101, S103, I105, S111, L112, and L113) (Fig. 3b), the E2 region that interacts with ubiquitin in the closed conformation26. The oxyester of the mutant [15N]UBE2AC88S Q93E (UBE2AQ93E–O~Ub) showed similar perturbations to those observed with UBE2A–O~Ub conjugate (Fig. 3c).
A reciprocal setup allowed us to observe changes to the [15N]ubiquitin spectra after oxyester conjugation (Supplementary Fig. 6a). The CSPs were highly similar for both conjugates (Fig. 3d and Supplementary Fig. 6a). A crystal structure of the yeast homolog of UBE2A, Rad6~Ub, is in an open state;29 therefore, to assess whether the UBE2A oxyester conjugates might adopt a closed state in solution, we built a molecular model of UBE2A~Ub, which was based on a Ubc13~Ub crystal structure (PDB 5AIU)30 in a closed conformation (Fig. 3e). In the model of UBE2A~Ub, the NMR-perturbed ubiquitin residues I44, K48, Q49, V70, and L71 are in the interface with E2 helix α2 (refs 31,32). Thus, the NMR results indicate that UBE2A–O~Ub visits closed conformations in solution, in agreement with UBE2A’s ability to assemble polyubiquitin chains in the absence of an E3. Furthermore, the results indicate that the UBE2AQ93E–O~Ub conjugate adopts a similar conformation. The main spectral difference between UBE2A–O~Ub and UBE2AQ93E–O~Ub conjugates is in ubiquitin Gly76, the residue that forms the oxyester bond with UBE2A (Supplementary Fig. 6a). This difference indicates that the chemical environment of the oxyester bond (the mimic of the reactive bond of E2~Ub) is altered in the UBE2AQ93E–O~Ub conjugate.
The human E2, UBE2D2 (UbcH5b), has arginine at the position homologous to UBE2A Q93. Substitution of R90 in UBE2D2 with glutamate (UBE2D2 R90E) generates a mutant with impaired polyubiquitination activity33. An interaction between the substituted glutamate residue in UBE2D2 R90E and R74 in the C-terminal tail of ubiquitin has been proposed to explain the loss of function. If a similar interaction is at play in UBE2A Q93E, we would predict a difference in the chemical shift of the ubiquitin R74 peak between UBE2AWT~Ub and UBE2AQ93E~Ub, but no such difference was observed (Supplementary Fig. 6b). Furthermore, the peak corresponding to E93 was not perturbed between the apo and conjugate form of the E2 (Supplementary Fig. 6c). Together, these observations do not support an interaction similar to the source of the activity defect in UBE2AQ93E~Ub.
Finally, given that an interaction between ubiquitin and the back sides of some E2 enzymes is critical for polyubiquitin-chain formation21,33,34, we tested whether the Q93E mutation might affect the UBE2A back-side interaction with ubiquitin. The (1H,15N)-HSQC spectra of WT and Q93E-mutant UBE2A in the presence of a tenfold molar excess of ubiquitin were highly similar (Supplementary Fig. 7), thus confirming that the weak back-side interaction was not detectably impaired.
Overall, the data demonstrate that the activity defects of UBE2A Q93E are not due to altered conformation of either the apo or conjugate form. Instead, the evidence suggests a change in the chemical microenvironment of the active site.
UBE2A residue Q93 facilitates ubiquitin transfer
The ability to observe side chain NH2 resonances for glutamine and asparagine side chains by NMR allowed us to follow the fate of the Q93 side chain during UBE2A C88S oxyester formation (Fig. 4a). Although the Q93 backbone amide resonance is unperturbed after ubiquitin conjugation (Fig. 4a), its side chain NH2 resonances are considerably perturbed (Fig. 4a). To further investigate the Q93 side chain, we generated the thioester in the NMR tube with WT UBE2A. The Q93 side chain resonances have different resonance positions in (uncharged) WT and C88S-mutant UBE2A, in agreement with the notion that Q93 senses the chemical environment of the active site (-SH versus -OH; Supplementary Fig. 8). In the first spectrum after ATP addition (collected for 80 min after ATP addition), we observed a substantial CSP in the C88 peak (Supplementary Fig. 9a) and the Q93 side chain peaks (Fig. 4b; red spectrum). At 5 h after ATP addition, a second set of Q93 side chain resonances that are further shifted appear (Fig. 4b; cyan spectrum). The observation of two sets of peaks indicates that the Q93 side chain exists in two slowly interconverting conformations at that point in the reaction. Importantly, neither of the Q93 resonances overlay with the original peak (Fig. 4b; black spectrum), so neither corresponds to discharged, unmodified UBE2A. In the same spectrum, there is a substantial loss of intensity for most peaks. A set of specific UBE2A residues exhibit an even larger decrease in peak intensity (Supplementary Fig. 9a,b), essentially the same residues that were perturbed in the UBE2A–O~Ub oxyester spectra (Supplementary Fig. 9c and Fig. 3b), thus indicating that the effect is linked to the formation of the thioester bond between UBE2A and ubiquitin. We propose that the Q93 side chain participates in a new interaction during the process. To test this hypothesis, we examined the effect of Q93 substitution with alanine on polyubiquitin-chain-assembly activity. As predicted, the Q93A substitution showed lower UBE2A activity than did WT UBE2A (Fig. 5a), in agreement with the involvement of the Q93 side chain in facilitating ubiquitin transfer. Of note, most E2 family members have residues capable of forming hydrogen bonds at the Q93-equivalent position of UBE2A (Supplementary Fig. 2).
The defective Q93E mutant is active in alkaline conditions
The Q93E mutant is more severely impaired than UBE2A Q93A in its polyubiquitination activity, thus suggesting that the presence of the carboxylate group of E93, as opposed to the loss of the amide group of Q93, is particularly detrimental to activity (Fig. 5a). In assays performed under varying pH, the ability of UBE2A Q93E to assemble polyubiquitin chains was partially restored at pH 9 or above (Fig. 5b).
An essential step in ubiquitin transfer via aminolysis is deprotonation of the incoming lysine, which acts as a nucleophile attacking the E2~Ub conjugate. Given the high pKa of lysine (approximately 10.5), residues near the E2 active site must create a microenvironment that lowers the effective pKa of the attacking lysine, thereby enabling formation of the isopeptide bond35. The calculated pKa value for a substrate lysine near the catalytic cysteine of the SUMO-specific E2 Ubc9 is almost 3 pH units lower than the pKa of free lysine35. The increase in UBE2A Q93E activity at high pH suggests that the mutant E2 lacks the ability to suppress the pKa of the incoming lysine, thereby slowing the rate of ubiquitin transfer to a lysine amino group. To test this hypothesis, we performed reactions with hydroxylamine (pKa ~ 6.0)36 as the nucleophile, because it does not require deprotonation to act as a nucleophile at neutral pH. The initial rates of discharge with hydroxylamine, in contrast to reactions with free lysine (Fig. 1e), were indistinguishable for WT and Q93E-mutant UBE2A (Fig. 5c and Supplementary Fig. 3b). Both enzymes exhibited considerable activity, thus indicating that when deprotonation is not necessary, UBE2A Q93E is active. Together, the results strongly suggest that a primary effect of the UBE2A Q93E mutation is its inability to lower the pKa of the attacking lysine side chain.
To calculate the pKa value of lysine from an Ubacceptor molecule that is proximal to the catalytic site of UBE2A, we incorporated a ubiquitin molecule into our UBE2A~Ubdonor model, which was based on the crystal structure of Ubc13~Ub presenting an acceptor ubiquitin near the thioester linkage (PDB 2GMI)37. With this model, multiconformational continuum electrostatics38 predicted an increase in the lysine pKa from 9.3 in the WT complex to 13.7 in the UBE2A Q93E complex. The calculations support the notion that substitution of glutamate at position 93 leads to impaired lysine deprotonation caused by an increase in the pKa of the attacking lysine.
To gain insight into the activation of UBE2A Q93E at high pH, we measured the rate of diubiquitin formation by WT and Q93E-mutant enzymes at pH 8 and 9 (Fig. 5d and Supplementary Fig. 10). We conducted single-turnover assays in which the charging reactions were performed with lysine-free ubiquitin to hinder polyubiquitin formation (Supplementary Fig. 10a), and the discharge reactions were performed with an equimolar concentration of WT ubiquitin (after addition of EDTA to inhibit E1 enzyme). Diubiquitin formation was measured for both enzymes at the two pH conditions (Supplementary Fig. 10b) and quantified, and the rates of product formation were determined from linear-regression analysis (Supplementary Fig. 10c). The rate of diubiquitin formation by UBE2A Q93E at pH 8 was approximately fourfold slower than that of WT UBE2A (0.013 ± 0.001 µM/min versus 0.047 ± 0.008 µM/min, respectively; Supplementary Fig. 10d). Increasing the reaction pH from 8 to 9 led to a decrease in the gap between rates (0.041 ± 0.003 µM/min for mutant versus 0.101 ± 0.006 µM/min for WT UBE2A). The rate observed for UBE2A Q93E at pH 9 was quite similar to that of WT UBE2A at pH 8, thus corroborating the idea that the impairment caused by Q93E substitution can be overcome by increased pH.
To further explore this hypothesis, we substituted Q93 in UBE2A with arginine, the residue found in numerous E2s including UBE2D2. As predicted by our model, the Q93R mutation yielded greater UBE2A activity than the WT enzyme (Fig. 5e). Thus, we conclude that the nature of the amino acid side chain of the position-93 residue in UBE2A directly contributes to enzyme activity via modulating lysine activation during ubiquitin transfer.
The UBE2A Q93E mutation diminishes monoubiquitination of PCNA
To assess the ubiquitin-transfer activity of UBE2A Q93E in the presence of an E3 and a substrate, we performed reactions containing either WT or Q93E-mutant UBE2A, the RING E3 RAD18, and the substrate PCNA. This pairing of E2 and E3 is known to transfer monoubiquitin to a specific lysine on PCNA39. Both forms of the E2 generate monoubiquitinated PCNA (Fig. 6), but both the rate and level of product formation were lower in UBE2A Q93E than the WT (half-life (t1/2) of 22.8 min for UBE2A Q93E versus 6.9 min for WT). The results indicate that the aminolysis defect in UBE2A Q93E is not compensated by the E3 ligase RAD18, thus suggesting that the presence of mutant E2 may lead to decreased levels of ubiquitinated product. To assess whether a positively charged residue might facilitate lysine activation for ubiquitin transfer to substrate, we conducted the PCNA ubiquitination experiment with UBE2A Q93R (Fig. 6a). The variant was highly active in this reaction, producing more product than WT UBE2A at a higher rate (t1/2 of 2.9 min; Fig. 6b). Together, the data show that the more active the E2 is intrinsically, the more product is generated in E3-dependent reactions.
Mutations in the UBE2A gene are known to cause XLID5,7,8,9,10,11,12,13,14,15,16,17,18,19. Uncovering the molecular mechanisms underlying UBE2A-type ID can expand understanding of ID pathophysiology and clarify the fundamental catalytic mechanisms of this important class of enzymes. As central players in protein ubiquitination, E2s carry out two reactions: transthiolation, which conjugates ubiquitin to the E2 active site, and aminolysis, which transfers ubiquitin to an amino group on a substrate23. Here, we demonstrate that a novel pathogenic missense UBE2A mutation, Q93E, impairs the enzyme’s ability to carry out aminolysis while its transthiolation activity is preserved. UBE2A Q93E is defective not only with ubiquitin serving as the nucleophile, thus inhibiting polyubiquitin-chain formation, but also with free lysine, thus indicating that the defect is in the aminolysis reaction itself. Furthermore, the mutant E2 is defective in its ability to monoubiquitinate PCNA in the presence of its cognate E3 ligase (RAD18), thereby indicating that the intrinsic defect cannot be fully overcome by an activating E3. These observations suggest that UBE2A Q93E generates its bona fide ubiquitinated products more slowly and at lower levels than WT UBE2A, even in the presence of cognate E3 ligases.
For lysine to act as a nucleophile, its ε-amino group must be deprotonated35,40. Two mechanisms are proposed for lysine deprotonation by E2s: (i) direct abstraction of a proton by a proton acceptor and (ii) a decrease in the pKa of the incoming lysine, owing to the microenvironment near the E2 active site35. In the former case, D117 of UBE2D1 (ref. 41) and H94 of UBE2G2 (ref. 42) have been proposed to serve as proton acceptors for an incoming lysine. However, neither of the corresponding residues in UBE2A (S120 and Q93, respectively) can serve as proton acceptors, and indeed none of the residues observed to be perturbed in UBE2A Q93E can receive a proton. Furthermore, the high similarity of ubiquitin resonances in UBE2AWT~Ub and UBE2AQ93E~Ub conjugates, other than for ubiquitin G76, do not support the existence of a proton acceptor in the ubiquitin moiety. Therefore, our work supports a mechanism in which lysine deprotonation is not carried out directly by a specific amino acid residue from UBE2A or Ubdonor serving as proton acceptor.
Despite the conserved architecture of the active site and the position of the catalytic cysteine among E2s, the identity and location of residues that participate in catalysis differ, thus suggesting that the positioning and deprotonation of an incoming lysine may involve distinct sets of residues43. In the SUMO-specific E2 Ubc9, residues N85, Y87, and D127 near the active site cysteine provide a microenvironment that lowers the incoming lysine pKa, thereby enhancing deprotonation and nucleophilicity at physiological pH and helping to position the incoming lysine35. Although the NMR data indicate that substitution of glutamate for glutamine at position 93 perturbs the environment around the active site, the UBE2A residues corresponding to Ubc9 N85, Y87, and D127 (N80, Y82, and S120) exhibit no or very minor perturbations in the UBE2A Q93E spectra compared with the WT UBE2A spectra.
Our study identifies a previously unappreciated contribution of the Q93 side chain to the aminolysis reaction. Previous studies on UBE2B have indicated involvement of the Q93 residue in the formation of the thioester with ubiquitin21. In contrast, the defect in UBE2A Q93E is in discharge of the ubiquitin from the thioester and not in thioester formation. Residues at the Q93-equivalent position are not strictly conserved in the E2 family, although almost half of human E2s have arginine or lysine at this position, and only UBE2A and UBE2B have glutamine (Supplementary Fig. 2). Divergent effects of mutations at this E2 position have been reported. Substitutions with alanine inhibit formation of the E2~Ub conjugate in human UBE2C (UbcH10)44, restrict polyubiquitin-chain formation by UBE2G2 (ref. 45), and have no apparent effect on the activity of UBE2D2 and UBE2K33,46. In UBE2D2, UBE2K, and UBE2A, glutamate substitution at this position results in an E2 with impaired activity33,46.
The combined observations indicate that the ‘Q93’ position, which sits at the opening to the active site of E2s, may modulate both the thioester formation and discharge activities of an E2. A positively charged residue at this position appears to enhance activity, whereas substitution with a negatively charged glutamate impairs activity. Three human E2s, UBE2J2, UBE2U, and Ubc9, contain a negatively charged residue at this position, thereby suggesting that different residues are responsible for modulating the aminolysis reaction carried out by these E2s.
Studies conducted on cells from Ube2A-knockout mice or containing patient-derived UBE2A mutations (R7W, I87MfsX14, and Q128X) have common mitochondrial defects, which are suspected to be caused by defective polyubiquitination of mitochondrial proteins11. These studies have associated XLID mutations in UBE2A with disruptions in mitophagy11. The UBE2A R7W and R11Q mutations result in a loss of interaction with UBR4, a protein linked to lysosome-mediated degradation and autophagy47. The R11Q and G23R pathogenic mutations result in defects in polyubiquitin-chain assembly21,47. Here, we show similar impairments in polyubiquitin-chain assembly by the mutants R7W, R11Q, and G23R and even more pronounced impairments by the newly described Q93E mutation in UBE2A. Therefore, defects in polyubiquitin-chain formation and disruption of the mitophagy/autophagy processes appear to be a common feature related to different XLID UBE2A mutations.
Altogether, our data support a model in which glutamate at position 93 in UBE2A markedly decreases the enzyme´s ability to transfer ubiquitin from its active site to the ε-amino group of a lysine. The defect is caused by changes in the active site microenvironment that impair the decrease in the pKa of an incoming lysine, as is needed for its nucleophilicity. The results may have important implications for strategies to specifically inhibit UBE2A, which is recognized as a potential target for melanoma, breast, and ovarian cancer48,49,50. Moreover, our finding that increased pH can partially rescue UBE2A Q93E activity suggests the possibility of modulating the activity of the mutant enzyme with compounds that restore its ability to decrease the pKa of an incoming lysine.
The two affected brothers were born to nonconsanguineous parents with no family history of neurodevelopmental disorders. Both patients were negative for fragile X syndrome and chromosome-microarray tests. The parents and the three female siblings are healthy.
Patient 1 was born at term after Caesarian section and weighed 4,160 g. A physical examination when the patient was 34 years old revealed a height of 1.81 m, a weight of 86 kg, hair whorls, thick eyebrows (without synophris), a large mouth, widely spaced nipples, normal genitalia, and shortened Achilles tendons. Temporal irritative focus was detected by routine EEG when the patient was 6 years old, but there were no convulsive or other epilepsy manifestations. He presented with mild ID, without substantially delayed motor milestones, but exhibited stuttering and difficulties in the articulation of words. He attends a special-education school, having achieved a level equivalent to the third or fourth grade of elementary schooling.
Patient 2 was born at term after Caesarian section and weighed 3,890 g. A physical examination when the patient was 32 years old revealed a height of 1.71 m, a weight of 55 kg, a moderate degree of frontal balding, hair whorls, thick eyebrows (without synophris), a large mouth, and normal genitalia. Similarly to his brother, he presented with mild ID; he did not present substantially delayed motor milestones but showed pronounced difficulties in the articulation of words. He also attends a special-education school and has achieved a level equivalent to the third or fourth grade of elementary schooling.
This study was approved by the ethics committee of the Institute of Biosciences (University of São Paulo, São Paulo, Brazil), and written informed consent was obtained from the patients’ parents. The current work is in compliance with all relevant ethical regulations.
DNA samples were prepared with the AmpliSeq Exome library, according to the manufacturer’s specifications. The AmpliSeq libraries were single-end sequenced on an Ion Proton System at Beijing Genomics Institute (BGI, Beijing, China). The raw reads were aligned to the reference genome (GRCh37/hg19), with BWA51 and preprocessed according to the GATK toolkit52. Filtering and prioritization of variants were conducted with VarSeq software (Golden Helix). The variants were filtered per population frequency (<0.01), quality (phred quality ≥20, genotype quality ≥20), and read depth (10).
Sanger sequencing was used to confirm the presence of the variants that were considered potentially pathogenic and to perform segregation studies (primer sequences available upon request).
X-inactivation, tested by determining the methylation status of the androgen receptor gene (AR), was evaluated as described previously53.
Cloning and mutagenesis
Sequences encoding the WT UBE2A gene and Q93E mutant were amplified by PCR from the patients’ mothers’ cDNA and index cDNA, respectively, and cloned into the pET28a vector (Novagen) via the NdeI and XhoI restriction sites. The human PCNA gene was also inserted into the pET28a vector through the same restriction sites. Human RAD18 was cloned between the restriction sites BamHI e XhoI of a modified version of pETSUMO (Invitrogen). Expression plasmids for human Uba1 and ubiquitin (both in pET3a) were donated by T. Sixma and C. Hill (Addgene plasmids #63571 and #61937, respectively). All point mutations in UBE2A (Q93E, Q93A, Q93R, C88S, C88S Q93E, R7W, R11Q, and G23R) were created with basic PCR-based site-directed mutagenesis and confirmed by Sanger sequencing.
Protein expression and purification
WT UBE2A (residues 1–152) and mutants (Q93E, Q93A, Q93R, C88S, C88S Q93E, R7W, R11Q, and G23R) were expressed from pET28a (Novagen) as an N-terminal hexahistidine-tag fusion. Rosetta II Escherichia coli cells (Novagen) were used to express the proteins in LB medium at 37 °C for 5 h at 200 r.p.m. The culture temperature was decreased to 18 °C for 1 h before induction with 0.2 mM IPTG, and cells were grown at 18 °C for 16 h, then harvested by centrifugation (5,000 g for 10 min at 4 °C). The cells were disrupted by sonication in lysis buffer (50 mM Tris, pH 7.5, 300 mM NaCl, 10% glycerol, 2 mM β-mercaptoethanol, and 1 mM phenylmethylsulfonyl fluoride) containing 0.2 mg/mL lysozyme and were centrifuged to remove cellular debris (40 000 g for 60 min at 4 °C). Proteins were first purified on a HisTrap column (GE Healthcare) preequilibrated with binding buffer containing 50 mM Tris, pH 7.5, 300 mM NaCl, and 2 mM β-mercaptoethanol, then were eluted with a linear gradient of the same buffer containing 500 mM imidazole, in an AKTA Purifier system (GE Healthcare). The hexahistidine tag was cleaved by thrombin protease, and the proteins were further purified with a Superdex 75 column (GE Healthcare) with buffer containing 50 mM Tris, pH 7.5, 300 mM NaCl, and 1 mM dithiothreitol. After being concentrated to 10 mg/ml, proteins were flash frozen in liquid nitrogen and stored at −80 °C. Recombinant purified UBE2A proteins are shown in Supplementary Fig. 12. Human Uba1, Ub, SUMO–RAD18, and PCNA were expressed and purified as previously described21,54,55.
Crystals were obtained by sitting-drop vapor diffusion at 18 °C by mixture of equal volumes of protein at 10 mg/mL and reservoir solution containing 50 mM disodium succinate, pH 7.0, and 12–14% PEG 3350. Harvested crystals were cryoprotected with 10% (vol/vol) ethylene glycol added to the mother liquor. Data were collected at 1.46 Å at beamline W01B-MX2 (LNLS) at 100 K, processed with XDS56, and merged and scaled with Aimless57. The structures were solved by molecular replacement with MolRep57. The human UBE2B protein (PDB 2YB6)21 was used as a search model for WT UBE2A, whose structure was used as a search model for the Q93E mutant. The models were refined with Phenix (phenix.refine)58, and the quality of the final models was assessed with MolProbity. The crystallographic parameters and final refinement statistics are summarized in Supplementary Table 1. Both final models contained 97.3% of residues in favored regions and 2.7% in allowed regions of the Ramachandran plot (no outliers).
15N-labeled WT and Q93E-mutant UBE2A proteins were produced in M9 minimal medium supplemented with 1 g/L 15NH4Cl (Cambridge Isotope Laboratories). To produce 13C/15N-labeled UBE2A Q93E, we also supplemented the M9 with 4 g/L [13C6]d(+)-glucose (Cambridge Isotope Laboratories). The expression and purification protocols were performed as described for the unlabeled proteins. The gel-filtration final buffer contained 50 mM K2HPO4-KH2PO4, pH 8.0, 200 mM NaCl, and 1 mM DTT. Labeled proteins were concentrated to 750 µM, and 10% D2O was added. NMR spectra were recorded at 25 °C on an Agilent Inova 500-MHz spectrometer. The spectra were processed with NMRPipe/NMRDraw59 and analyzed with NMRView60. The backbone assignment of UBE2A Q93E was determined with the three-dimensional experiments HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO, and the graphical interface Smartnotebook.
NMR thioester and oxyester reactions
Thioester reactions were performed in the NMR tube (in 200 µL) with 60 µM [15N]UBE2A, 180 µM ubiquitin, and 3 µM Uba1 in buffer containing 50 mM Tris, pH 8.0, 150 mm NaCl, 5 mM MgCl2, and 0.5 mM DTT. For oxyester reactions, 400 µM [15N]UBE2AC88S or [15N]UBE2AC88S Q93E was mixed with 650 µM ubiquitin and 6 µM Uba1, or conversely, 200 µM [15N]ubiquitin was combined with 325 µM UBE2A C88S or UBE2A Q93E C88S and 3 µM Uba1, in the same buffer used for thioester reactions. All reactions were conducted at 30 °C and were started by the addition of 10 mM ATP. Sequential (1H,15N)-HSQC spectra were acquired on a 600-MHz spectrometer (Agilent Inova) equipped with a cryogenic probe. We used the backbone resonance assignment of apoubiquitin available in BMRB 4769.
In vitro ubiquitination assays
Ubiquitination assays were performed at 32 °C with 1 μM human recombinant Uba1, 30 μM human recombinant ubiquitin, and 20 μM WT or Q93E-mutant UBE2A in ubiquitination buffer containing 50 mM Tris, pH 8.0, 50 mM NaCl, 50 mM KCl, 10 mM MgCl2, and 0.1 mM DTT. Reactions were started by the addition of 3 mM ATP and quenched with an equal volume of 2× SDS–PAGE buffer (reducing or nonreducing). Samples collected at specific time points were resolved by SDS–PAGE and were then western blotted with anti-ubiquitin (P4D1, Santa Cruz). In vitro ubiquitination assays at different pH values were conducted identically to those described above but with the following buffers: MES for pH 6, Tris for pH 7–9, and CAPS for pH 10. For intrinsic-reactivity assays, WT and Q93E-mutant UBE2A were charged for 15 min at 32 °C in ubiquitination buffer before being mixed with 10 mM EDTA and 25 mM lysine or 5 mM hydroxylamine, and incubated at 32 °C. Samples were collected at several time points during the reaction, quenched in nonreducing SDS–PAGE buffer, and visualized by Coomassie staining. PCNA monoubiquitination was performed by mixture of 1 μM Uba1, 30 μM ubiquitin, 20 μM WT or Q93E-mutant UBE2A, 5 μM SUMO–RAD18, and 1 μM PCNA in ubiquitination buffer supplemented with 2 μM ZnCl2, with incubation at 32 °C in the presence of 3 mM ATP. The reactions were then subjected to anti-PCNA (P8825, Sigma) western blotting.
Single-turnover diubiquitin-formation assays
Diubiquitin-formation assays were conducted in two-step reactions. First, charging of UBE2A was achieved by mixture of 20 µM of WT or Q93E-mutant UBE2A with 1 µM E1, 30 µM lysine-free ubiquitin, and 3 mM ATP in ubiquitination buffer, pH 8. After incubation for 15 min at 37 °C, the reactions were stopped by the addition of 25 mM EDTA and then diluted fourfold in ubiquitination buffer at pH 8 or 9 (both with 50 mM Tris-HCl), thus resulting in final concentrations of 0.25 µM E1, 5 µM UBE2A, and 7.5 µM lysine-free ubiquitin. Diubiquitin formation was triggered by the addition of 7.5 µM WT ubiquitin. Reactions were kept at 32 °C, and samples were collected at different time points and immediately quenched in reducing SDS–PAGE buffer. Results were evaluated by western blotting with anti-ubiquitin (P4D1, Santa Cruz), and quantification of the corresponding diubiquitin band was performed in ImageJ.
UBE2A was covalently bound to ubiquitin with the UBE2N~Ub structure (PDB 5AIU)30 as the template. UBE2A was superposed with UBE2N, and the complex formed by UBE2A and ubiquitin was subjected to energy minimization with steepest descent followed by a simulated annealing protocol (SD/SA) to remove bad contacts, with the backbone atoms of UBE2A and ubiquitin fixed. We then created the thioester bond between the carboxyl group of the terminal glycine residue of ubiquitin and the sulfhydryl group of the catalytic Cys88 of UBE2A. To correct the bond angles and lengths, a new round of energy minimization with SD/SA was performed, with the backbone atoms of Cys88 fixed. All steps were performed with YASARA software and the AMBER03 force field. To position a second free ubiquitin molecule with a lysine positioned to attack the thioester bond, we used the template of Mms2–Ubc13 covalently bound to ubiquitin and presenting the Lys63 residue from another symmetric lysine positioned close to the thioester bond (PDB 2GMI)37, with the symmetry operators of the P212121 space group applied with PyMOL. We superposed UBE2A with Ubc13 to position the second ubiquitin generated by symmetry close to the thioester bond. Then, the backbone atoms of UBE2A~Ub were fixed, and an energy minimization with SD/SA was performed to eliminate bad contacts.
Crystallographic statistics are indicated in Supplementary Table 1. All western blot experiments were repeated independently three times and yielded similar results. (1H,15N)-HSQC spectra of UBE2A proteins (WT, Q93E, C88S, and C88S Q93E) were repeated independently at least two times and yielded similar results. Each oxyester and thioester reaction, and subsequent NMR data collection, was performed once. Data from discharge assays, diubiquitin formation, and PCNA monoubiquitination assays were derived from three independent experiments and are presented as the mean ± s.d. The diubiquitin-production rates were determined from a linear regression. The PCNA monoubiquitination data were fitted according to the equation PCNA–Ub = plateau(1 – e–Kt), where PCNA–Ub is the fraction of monoubiquitinated PCNA, plateau is the PCNA–Ub value at infinite time, K is the rate constant, t is time, and t1/2 = ln(2)/K.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank LNBio/CNPEM for access to core facilities as well as for financial support. We also thank the Brazilian Synchrotron Light Laboratory (LNLS) for access to the MX2 beamline and the MX2 staff for technical assistance. We are very grateful to H. Powell for assistance in X-ray data processing. We also thank T. Sixma (Netherlands Cancer Institute) and C. Hill (University of Utah) for Addgene plasmids (#63571 and #61937, respectively). This work was supported by the Brazilian National Council for Scientific and Technological Development (CNPq, C.R., 306879/2014-0; K.G.F., 310536/2014-6 and 422790/2016-8) and grants from São Paulo Research Foundation (FAPESP, C.R., 2012/50981-5 and 2013/08028-1; M.M., 2015/06281-7) and NIH/NIGMS (R.E.K., R01 GM088055).
About this article
Nature Chemical Biology (2019)