Ubiquitin RING E3 ligases (E3s) catalyze ubiquitin (Ub) transfer to their substrates by engaging E2∼Ub intermediates with the help of their RING domains. Different E3s have been found to contain a conserved tryptophan residue in their RING that plays an essential role in E2 binding and, hence, enzymatic activity. Many active E3s, however, lack this specific residue. We mined through the existing data to observe that the conservation of the tryptophan and quaternary organization of the RING domains are remarkably correlated. Monomeric RINGs possess the tryptophan while all well-characterized dimeric RINGs, except RNF8, contain other amino acid residues. Biochemical analyses on representative E3s and their mutants reveal that the tryptophan is essential for optimal enzymatic activity of monomeric RINGs whereas dimeric E3s with tryptophan display hyperactivity. Most critically, the introduction of the tryptophan restores the activity of inactive monomeric RNF4 mutants, an obligatory dimeric E3. Binding studies indicate that monomeric RINGs retained the tryptophan for their optimal functionality to compensate for weak Ub binding. On the other hand, tryptophan was omitted from dimeric RINGs during the course of evolution to prevent unwanted modifications and allow regulation of their activity through oligomerization.
Ubiquitination has emerged as a major post-translational modification (PTM) that virtually regulates almost all cellular processes across eukaryotes . The molecular mechanism of ubiquitination is remarkably conserved in all organisms that begin with the ATP-dependent activation of the ubiquitin (Ub) C-terminus by ubiquitin E1 enzymes followed by the transfer of the activated Ub to a conjugating E2 and culminate with the E3-catalyzed transfer of the activated Ub onto a substrate lysine. E3 ligases ascertain specificity to the process by modifying desired substrates. They also selectively bind E2∼Ub thioester conjugates from the available cellular repertoire to warrant the necessary topology of modification. Thus, appropriate regulation of E3 activity is crucial for the normal cellular physiology and their malfunctioning has been associated with diverse pathophysiological conditions such as neurodegeneration and cancer [2,3].
E3 ligases can be broadly classified into either the RING/U-box or the HECT/RBR groups based on their Ub transfer mechanism. HECT or RBR E3s react with E2∼Ubs to form thioester-linked E3∼Ub intermediates before transferring the Ub to its final destination [4,5]. On the other hand, RING and U-box containing proteins act as scaffolds to promote direct transfer of the Ub from E2∼Ub conjugates to the substrate. Research in recent years have shown that RING E3s do not merely bring E2∼Ub and the substrate in close proximity, rather they engage E2∼Ub intermediates in a precise orientation dubbed as the ‘closed conformation’ congenial for nucleophilic attack on the thioester bond by acceptor lysines [6,7]. To stabilize the E2∼Ub thioester in closed conformation, E3 ligases interact with both the E2 and the Ub moieties simultaneously. Different E3s employ different mechanisms and structural elements to bind the Ub, whereas E3:E2 interaction interfaces are mostly conserved [6–10]. Structures of all the E3:E2 complexes solved to date have revealed that the canonical E2-binding interface of RING E3 ligases is quite small (500–750 Å2). And, consistent with this smaller interface, the dissociation constant (Kd) lies in the 1–150 µM range for most RING:E2 complexes, with a few exceptions like ZNRF1  and FANCL .
Nonetheless, E3–E2 interaction acts as a crucial determinant for the ubiquitination activity of E3 ligases. The activation of RNF146 by the ADP-ribosyl moiety is a prime example where binding of the latter allows RNF146 to interact with Ube2D2 . NMR titrations and ligase activity assays substantiated by mutational analyses revealed the importance of various E3 residues that contribute towards their E2 affinity and specificity . In multiple such studies conducted on E3s such as c-Cbl, EL5, and Kaposi's Sarcoma-associated Herpes Virus K3 protein (K3) led to the identification of a conserved tryptophan residue in their RING domains that critically governs their E2 affinity [15–17]. Consequently, Trp to Ala mutation in these E3 ligases essentially abrogated their ligase activity. Interestingly, the necessity of this tryptophan does not appear universal as E3 ligases, such as BRCA1, CNOT4, and Ciap2, having hydrophobic side chains also display ubiquitination activity. Furthermore, robustly active E3 ligases, such as RNF4 and Traf6, contain serine in the corresponding position suggesting against the universal requirement of the tryptophan or other hydrophobic residues in the E2-binding surface of the E3s.
We systematically looked into the extent of conservation of this tryptophan in all known RINGs across all organisms available in the INTERPRO  and searched through the literature to correlate its presence with other characteristics. Surprisingly, we find most monomeric RINGs characterized to date contain the tryptophan while dimeric ones almost invariably contain other amino acid residues. This unusual correlation suggested towards the functional importance of tryptophan in E3s containing RING monomers. Experimental investigations reveal that mutating the tryptophan severely impairs the ubiquitination activity of monomeric E3s but not that of the dimeric RNF8. Strikingly enough, the introduction of the tryptophan reinstates ubiquitination activity of inactive monomeric mutants of RNF4, a dimeric E3 ligase. Binding studies show that conserved tryptophan compensates for weaker Ub binding wherein a single RING domain interacts with the E2∼Ub and removal of this amino acid allows regulation of activity in dimeric ones.
Database preparation and sequence analysis
Full-length protein sequences containing RING domains obtained from the INTERPRO database were trimmed to retain only the sequences corresponding to RING with two additional residues in each side, if available. This database was then curated to remove identical sequences to avoid redundancy along with incomplete sequences with missing amino acids. We only considered 100% identical sequences as redundant as we looked into single amino acid variations within the RING domain. In other words, even single residue substitutions were considered non-redundant for our intended purpose. For example, despite being nearly identical, the IR6-4 residue of XtRNF4 and HsRNF4 RING domains is different (Ala versus Ser). This variability in the IR6-4 suggests that a mutation of Ala to Ser in the course of evolution was not detrimental for RNF4 activity. The final database containing ∼20,000 RING sequences was subsequently analyzed using simple character search method utilizing a home-brewed PERL script that gave an output containing the number of RINGs having a specific amino acid at a specific position. The output file containing the numbers was then imported in Microsoft EXCEL for presentation purposes. For sub-classification, RING sequences were first segregated in different classes based on their chelating residues  followed by residue counting as described above.
Cloning, expression, and purification of proteins
Clones for all the E2s, Ub and its variants, ZNRF1CTD, full-length RNF8, HsRNF4NΔ22 (referred to as RNF4 for convenience), and human Uba1 were available in the laboratory and have already been described [11,20]. ORF corresponding to the RING domain of human RNF13 (residues 216–290) was PCR amplified from human cDNA sample with Kod DNA polymerase (Toyobo Life Sciences, Japan) and cloned into pETSUMO2 vector digested with NsiI and XhoI using the Infusion HD cloning kit (Takara Clontech Inc.). RNF8RING and RNF8RC were generated by PCR amplifying the desired regions using suitable primers with Kod DNA polymerase (Toyobo Life Sciences, Japan) and were cloned into the pETSUMO2.1 expression vector generated in our laboratory having an in-frame BamHI site that allowed the expression of the chimeric protein having N-terminal 6xHis-SUMO2 tag. RNF4RING was generated by PCR-based site-directed mutagenesis deleting the entire N-terminal region from our existing RNF4 construct in pET32a vector. All the point mutants used during the course of study such as RNF13W270A, RNF13W270S, RNF13W270L, RNF13W270D, RNF13W270F, RNF13W270Y, RNF8W430A, RNF8RC,W430A, and RNF4 mutants were generated by PCR-based site-directed mutagenesis using Kod DNA polymerase (Toyobo Life Sciences, Japan) and DpnI (New England Biolabs, U.S.A.). All clones were confirmed by DNA sequencing using Applied Biosystems sequencers (3130XL and 3500XL).
Expressions of all the proteins were induced with IPTG in Escherichia coli Rossetta2 (DE3) cells containing the necessary plasmid (Novagen Inc., U.S.A.) in M9ZB media with 0.5% glycerol at 16°C. Expression of RNF13 in the soluble form also required the addition of 25 µM ZnSO4 along with IPTG. All recombinant proteins were purified as described before . Labeling of UbS20C with fluorescein-5-maleimide has also been described . UbY59C is labeled identically to that of UbS20C. Ube2D2C85K∼Ub conjugate was prepared following the published protocol . The fluorescently labeled Ube2D2 used for anisotropy was labeled at the active site cysteine using fluorescein-5-maleimide identically as UbS20C.
All multi-turnover assays containing 25 µM fluorescent-labeled Ub (UbFL) were performed and visualized as described earlier . E3 concentrations were between 50 nM and 10 µM as indicated. For the Ube2D2-based assays, the E2 concentration was 500 nM (for RNF13) or 250 nM (for RNF4), whereas Ube2N/Ube2V1 was 500 nM in all the assays.
For E2∼Ub thioester release assays, 20 µM of Ube2D2 was incubated with 20 µM of UbFL and 500 nM of Uba1 in 1× assay buffer for ∼30 min at 37°C to generate Ube2D2∼UbFL thioesters. Ubiquitin release was subsequently carried out at 25°C by mixing the above mixture with a release mix containing lysine and the E3 variant. Release mix also contained a large excess of unlabeled Ub (final concentration 300 µM) to prevent recycling of the UbFL upon hydrolysis. We opted for this ‘chase’ assay instead of inhibiting the E1 for better reproducibility. Lysine and E3 concentrations were maintained at 50 mM and 250 nM for the RNF13-containing assays while for RNF4, either 25 mM lysine and 50 nM E3 (Figure 4H) or 50 mM lysine and 125 nM E3 (Figures 6C and 7E) were used. Samples withdrawn from the reaction mixture after the desired time interval were mixed with non-reducing SDS–PAGE sample buffer, resolved and imaged with a Typhoon RGB Imager using the 526(SP) filter. Bands were quantitated with Image Studio Lite Ver.5.2 (Licor Biosciences, U.S.A.). Normalized band intensities were plotted against time.
For FRET-based di-Ub formation assays, two Ub mutants, namely UbK63C and UbG75C, were purified and labeled with the maleimide derivatives of Dylight 488/Dylight 594 FRET pair and purified by desalting. UbG75C also contained a non-cleavable 6xHis tag to alter its mobility in SDS–PAGE compared to the untagged Ub. For di-Ub assays, reaction mixes were assembled with 5 µM UbK63C-dylight 594, 6 µM His6-UbG75C-Dylight 488 along with 500 nM Ube2N/Ube2V1 in 1× assay buffer. Reaction mixes also contained 500 nM of the designated E3, if required. Reactions were initiated at 37°C by a 100 nM Uba1 in a fluorescence cuvette. Di-Ub formations were monitored by measuring the change in fluorescence intensity at 518 nm (Dylight 488) and 614 nm (Dylight 594) as a function of time. All the reactions were carried out in triplicate and the normalized mean intensities were plotted as a function of time along with the standard deviations.
For visualizing the di-Ub formation in a gel, an identical reaction mix was incubated at 37°C as described above. Aliquots were withdrawn at designated time points, mixed with SDS–PAGE loading buffer and resolved and imaged in a Typhoon RGB using 488 nm (for Dylight 488) and 532 nm (for Dylight 594) lasers. Images are false-colored for presentation purposes.
Analytical size exclusion chromatography and binding studies
Analytical size exclusion chromatographies were performed in an AKTA Avant 25 chromatography system (GE Healthcare Lifesciences, Sweden) using either a Superdex 75 10/300 GL column (GE Healthcare Lifesciences, Sweden) at 8°C as described earlier or a Superdex 75 increase 5/150 GL column (GE Healthcare Lifesciences, Sweden) at 25°C. For probing complex formation between RNF4 mutants and the E2∼Ub conjugate, E3 proteins at 70 µM were incubated with 85 µM of the conjugate for ∼10 min at room temperature. 20 µl of the mix was loaded onto the column and developed using 20 mM sodium phosphate buffer (pH 8.0) containing 150 mM NaCl. For the free proteins, an identical amount was loaded to allow comparison of the UV absorbance at 214 nm. Isothermal titration calorimetry was performed in a semi-automated Affinity ITC LV microcalorimeter (TA Instruments, U.S.A.) as described previously . Anisotropy-based binding assays were carried out in a Jasco FP-8500 spectrofluorimeter (Jasco Inc., Japan) equipped with automated polarizers and Peltier-based temperature control. The data were collected at 25°C using the manufacturer-provided software module with an excitation wavelength of 490 nm and emission was monitored at 518 nm using the same buffer as the SEC. The labeled protein (100 nM) was taken in the cuvette and titrated with the unlabeled E3 samples.
Results and discussion
Conservation of a tryptophan residue in RING E3 ligases
To determine the extent of conservation of the tryptophan across RING domains, we prepared and analyzed a database containing ∼20,000 unique RING sequences encompassing all known RING domain-containing proteins present in the INTERPRO (see Methods for details). We chose to generate such a database as we were looking into a single amino acid change rather than the overall sequence conservation. Moreover, as RING domains from diverse E3 ligases vary not only in their amino acid sequence but also in the lengths, we devised an alternative numbering strategy for the ease of comparison and representation. In this scheme, all the eight chelating residues were numbered C1 through C8 and the residues in-between are designated with IRX–Y (Intervening Regions), where X indicates the number of the immediately preceding chelating residue and Y denotes the distance of a particular amino acid from X (see Figure 1A). The amino acid preceding C1 was referred to as ‘−1’ while the one following C8 was referred to as ‘+1’. Following this scheme, the conserved tryptophan in all the known E3s could be located at IR6-4, whereas the ‘linchpin’ arginine is found at the ‘+1’ position (Figure 1B).
Conservation of the tryptophan in RING domains.
We first scanned through our database to segregate and count RINGs based on the residue at IR6-4. The result shows that nearly half of the RINGs contain tryptophan at IR6-4 while all other amino acids together made up for the rest (Figure 1C). The second most preferred amino acid at this position is leucine (10%) followed by tyrosine (7%). Additionally, we observe an obvious preference for hydrophobic residues at IR6-4 with >75% RINGs having one of the aromatic or aliphatic side chains. Negatively charged residues and glycine, on the other hand, are least preferred with the latter being present only at 0.7% of the RINGs.
We further sub-classified RINGs on the basis of their chelating residues  and looked at the class-specific preference at IR6-4, if any (Figure 1D). Results reveal that the conservation of the tryptophan varies between different subclasses. In the canonical HC RINGs, tryptophan and leucine are equally prevalent (Figure 1D). In general, the IR6-4 residue is less conserved in RINGs belonging to this class. In contrast, H2 and PHD RINGs depict an overwhelming preference for the tryptophan (∼75%). A similar exercise on the adjacent positions, IR6-2 through IR6-6, confirms that this tryptophan preference does not merely reflect general sequence conservation in H2 RINGs (Supplementary Figure S1A). The preference of hydrophobic residue at IR6-5, on the other hand, is justified as this side chain buries itself in the core of the RING domain away from the surface (Supplementary Figure S1B).
Presence of tryptophan correlates with the oligomeric state of the RING domain
Closer look at all the RING E3:E2 complex structures reveal that the IR6-4 residue is present in the E2-binding surface (Figure 1E). Justifiably, mutation of the tryptophan led to drastic reductions in the E2 affinities of FANCLW341A  and ArkadiaW972A  compared with their wild-type versions due to the contribution of tryptophan in E2 binding via hydrophobic interactions. We, therefore, hypothesized that the presence of the tryptophan can be correlated with E3:E2 affinity. Although a search through the literature revealed that the presence of Trp is often associated with higher E3:E2 affinity, paucity of quantitative binding data and differences in the experimental set-ups prevented us from drawing any meaningful conclusion. Serendipitously, literature surveys revealed that IR6-4 Trp is practically invariant in monomeric E3s or multi-subunit E3s with one RING subunit. Exceptions to this rule are RNF25 and RNF146, both of which contain non-RING elements to bind their E2 partner. Furthermore, nearly all well-characterized homo- and heterodimeric E3s do not contain tryptophan at IR6-4, RNF8 being the sole exception. This unusual correlation between the oligomeric status and conservation of a residue, not directly involved in oligomerization, made us look through the published literature for more E3s that have been characterized in terms of their quaternary organization (Supplementary Table S1). Again, none of the dimeric RINGs contain the tryptophan while it is nearly invariant across E3 ligases that contain only one RING subunit (Figure 1F). Thus, Rbx1 and ANAPC11 have tryptophan, whereas none of the RING subunits of heterodimeric RNF40/RNF20 or BRCA1/BARD1 contains this residue. This unusual observation not only suggested that the importance of tryptophan is monomer specific but also indicated it to be rather disadvantageous for dimeric E3s.
Tryptophan is essential for the ligase activity of monomeric RINGs
To experimentally re-validate the importance of tryptophan in monomeric RINGs, we first compared the RING domains of RNF13 (henceforth referred as RNF13) and RNF38 with their corresponding tryptophan mutants (RNF13W270A and RNF38W493A) in multi-turnover ubiquitination assays. In these assays, we monitored E3 auto-ubiquitination in the presence of Ube2D2 (also known as UbcH5B). Results reveal that W270A substitution essentially abrogated RNF13 auto-ubiquitination (Figure 2A). In fact, we barely observed ubiquitinated E3 bands even after increasing the concentration of RNF13W270A all the way to 10 µM (Supplementary Figure S2A). RNF38W493A also failed to show any E3 ubiquitination unlike the robust modification observed for the wild-type E3 (Supplementary Figure S2B).
IR6-4 tryptophan is essential for the activity of monomeric E3s like RNF13 and ZNRF1.
To rule out the effect of tryptophan substitution to be Ube2D2-specific, we mutated the IR6-4 Trp (W214) in the C-terminal domain of ZNRF1 (residues 139–227 containing the zinc-finger and the RING domains) and probed its ability to enhance Lys63-linked free chain synthesis by Ube2N/Ube2V1 (also known as Ubc13/Uev1). Indeed, the activity of ZNRF1W214A is significantly attenuated in contrast with that of wtZNRF1 (Figure 2B and Supplementary Figure S2C). This attenuated ligase activity of ZNRF1W214A rules out the effect of tryptophan substitution to be Ube2D2-specific.
Comprehensive analyses of all RING sequences indicated that alanine is not prevalent in E3s unlike hydrophobic residues such as leucine and tyrosine (Figure 1C). However, none of the previous studies explored the effect of other amino acid substitutions in IR6-4 except alanine. We, therefore, generated a series of RNF13 mutants by replacing the tryptophan with a subset of five residues (Tyr, Phe, Leu, Ser, and Asp) chosen on the basis of our sequence analyses. Leu and Tyr were selected as they are the second and third most prevalent residue (see Figure 1C). Ser was chosen as it had the highest propensity among the uncharged polar residues, whereas Asp had the least among the charged ones. Multi-turnover assays revealed that all of the mutants, except RNF13W270Y, are essentially inactive at all E3 concentrations from 1 to 10 µM (Figure 2C and Supplementary Figure S2D). The RNF13W270Y mutant, on the other hand, is moderately active, albeit not as good as the wild-type.
In corroboration with the multi-turnover assays, pulse-chase Ube2D2∼Ub thioester release assays show wtRNF13 to be most efficient followed by RNF13W270Y (Figure 2D,E). On the other hand, RNF13W270F failed to enhance the basal rate of E2∼Ub dissociation observed in the absence of any E3. Based on all these data, we conclude that tryptophan is indeed indispensable for the activity of monomeric RNF13 and no other amino acid residue including aromatic or hydrophobic ones can functionally substitute tryptophan.
Mutating tryptophan decreases the E2 affinity of monomeric RINGs
We next monitored the binding of wtRNF13 and its W270Y and W270A mutants with Ube2D2 using isothermal titration calorimetry (ITC). Compared with the dissociation constant (Kd) of ∼1.1 µM obtained for the wtRNF13–Ube2D2 pair, we find that RNF13W270Y binds Ube2D2 with a Kd of ∼24 µM under identical experimental conditions (Figure 3A,B and Table 1). In comparison, binding of RNF13W270A to the E2 is significantly weaker than either the wild-type or RNF13W270Y (best estimated Kd of ∼124 µM) (Figure 3C and Table 1). Evaluation of the thermodynamic parameters reveals that the ∼25-fold reduction in the E2 affinity observed upon W270Y substitution primarily occurred due to a decrease in ΔHbinding from that of the wild type. W270A substitution, on the other hand, results in an endothermic binding leading to an unfavorable ΔHbinding at 25°C and interacts with the E2 solely due to a favorable change in entropy upon complexation. We verified our ITC data using fluorescence anisotropy-based binding experiments by titrating fluorescein-labeled Ube2D2 with the E3s. The obtained Kds of ∼0.8 µM, ∼17 µM, and ∼169 µM for wtRNF13, RNF13W270Y, and RNF13W270A are in excellent agreement with those from the ITC experiments (Figure 3D).
Substitution of IR6-4 tryptophan in monomeric E3s reduces their E2 affinities.
|Cell||Syringe||Kd (µM)*||ΔGbinding (kcal mol−1)||ΔHbinding (kcal mol−1)||ΔSbinding (cal mol−1 K−1)||n||Reference|
|ZNRF1||Ube2N||3.881 × 10−2||−10.12||−10.52||−1.384||–|||
|RNF4Y189A, S166W||Ube2D2||5.748||−7.149||−2.392||15.96||1.083||This study|
|RNF4Δ3C, S166W||Ube2D2||8.292||−6.932||−2.322||15.46||1.023||This study|
|Cell||Syringe||Kd (µM)*||ΔGbinding (kcal mol−1)||ΔHbinding (kcal mol−1)||ΔSbinding (cal mol−1 K−1)||n||Reference|
|ZNRF1||Ube2N||3.881 × 10−2||−10.12||−10.52||−1.384||–|||
|RNF4Y189A, S166W||Ube2D2||5.748||−7.149||−2.392||15.96||1.083||This study|
|RNF4Δ3C, S166W||Ube2D2||8.292||−6.932||−2.322||15.46||1.023||This study|
ITC experiments were carried out at 25°C in 25 mM sodium phosphate (pH 8.0), 150 mM NaCl. Raw data were fit to single site model using NANOANALYZE (TA Instruments, U.S.A.).
n was kept fixed during fitting due to weak binding.
To generalize the effect of tryptophan substitution on the E2 affinity, we carried out ITC experiments to monitor the binding of ZNRF1W214A to Ube2N and Ube2D2. We recently reported that at 25°C ZNRF1 binds Ube2N with an unusually high Kd of 39 nM unforeseen for any RING:E2 pair but displays an affinity of ∼1 µM for Ube2D2 (see Table 1) . Under identical conditions, ZNRF1W214A displays markedly reduced affinities towards both of the E2s yielding Kds of 7 µM and ∼15 µM for Ube2N and Ube2D2, respectively (Figure 3E,F and Table 1). We also note that the W214A substitution significantly reduced ΔHbinding for ZNRF1:E2 interactions. This substantial loss of binding enthalpy is partially compensated by an increase in the binding entropy.
Nonetheless, the observed decrease in the dissociation constants upon mutating a single binding surface residue underscores the crucial importance of the tryptophan side chain in dictating the E2 affinity of monomeric E3s. In fact, we note that none of the Ube2N or ZNRF1 point mutants previously explored by us exhibited such a drastic effect on their binding . We, therefore, conclude that the tryptophan plays a critical role in determining the E2 affinity across monomeric RINGs and replacing it with any other amino acid residue severely restricts their E3 ligase activity via reduction in E2 affinities.
Tryptophan is dispensable but enhances the activity of the dimeric E3s
RNF8 is the sole homodimer that contains tryptophan (Trp430) (see Supplementary Table S1 and Figure 1F). We, therefore, compared the ability of RNF8 and its tryptophan mutant, RNF8W430A, by monitoring Ube2N/Ube2V1-mediated Lys63-linked chain synthesis. Assays reveal that, unlike the monomers, RNF8W430A retains substantial ligase activity at 200 nM (Figure 4A). In fact, lowering the E3 all the way to 50 nM could not completely abolish the activity of RNF8W430A (Supplementary Figure S3A). On the other hand, increasing the E3 concentration above 200 nM progressively reduced the difference in the activities of the wild-type RNF8 and the W430A mutant (Supplementary Figure S3B).
Tryptophan is not essential for the activity of dimeric E3 ligases.
To rule out the presence of an unidentified non-RING element in full-length RNF8 that functionally compensated the effect of W430A mutation, we cloned and expressed the RING domain of RNF8 (Figure 4B). However, RNF8RING is quite unstable in solution and possesses negligible E3 ligase activity (see Supplementary Figure S3C). Moreover, in agreement with the previously published data , size exclusion chromatography (SEC) revealed that RNF8RING exists only as a monomer (Supplementary Figure S3D). We also failed to express RNF8RING, W430A as a soluble protein, suggesting that RNF8 RING domain is unstable in isolation and deletion of the extended helix most likely hindered its native folding. We, therefore, cloned and purified RNF8RC that included the extended helical region (Figure 4B). Unlike the RING, RNF8RC is stable and exists as a dimer in solution (Supplementary Figure S3D). Activity assays reveal that RNF8RC is equally proficient in enhancing Lys63-linked chain synthesis as the full-length (Figure 4C). RNF8RC, W430A could also enhance Lys63-linked chain synthesis similar to RNF8W430A ruling against the involvement of any non-RING element in E3 activity.
We further analyzed RNF8, RNF8RC, and their tryptophan mutants by employing a FRET-based di-Ub formation assay. In this assay, UbK63C-Dylight 594 and His6-UbG75C-Dylight 488 were used as the Ub-donor and the Ub-acceptor, respectively (Figure 4D and experimental procedures). The assay confirms that RNF8W430A and RNF8RC, W430A are equally competent to carry out di-Ub synthesis and display reduced di-Ub synthesis rates compared with the wild-type proteins (Figure 4E,F and Supplementary Figure S4). As expected, RNF8RING minimally enhanced di-Ub synthesis over the no-E3 control. All these biochemical assays establish that though the tryptophan is required for the optimal E3 activity of wtRNF8, W430A mutation is not as deleterious for this dimeric ligase unlike the analogous mutations in monomeric ones.
We next introduced tryptophan in another dimeric E3, RNF4, by mutating the IR6-4 serine (Ser166). Multi-turnover ubiquitination assays with Ube2D2 reveal that S166W substitution significantly enhanced the enzymatic activity of RNF4 over the wild-type as evident from the intensity of the ubiquitinated smear as well as the amount of leftover Ub (Figure 4G). Thioester release assays comparing wtRNF4 and RNF4S166W also reveal the markedly enhanced activity of the tryptophan-containing version (Figure 4H,I). Nonetheless, the substantial activity of the wtRNF4 in both multi-turnover and thioester release assays led us to conclude that the tryptophan is dispensable for the activity for dimeric E3 ligases.
IR6-4 tryptophan enhances the E2 affinity of dimeric RINGs
To gain insights into the E2 affinities of dimeric E3s in the presence and absence of tryptophan, we carried out ITC for RNF8RC and its corresponding W430A mutant. Data reveal that RNF8RC binds Ube2N with the Kd of 3.35 µM at 25°C (Figure 5A and Table 1) whereas under identical conditions, RNF8RC, W430A displays a ∼15-fold lower affinity for Ube2N (Kd ≈ 51.4 µM) (Figure 5B and Table 1). RNF8RING failed to show any interaction with the E2 providing further evidence of its folding defects (see Supplementary Figure S5A). Similarly, Ube2D2 binds RNF4S166W with a ∼30-fold enhanced affinity (Kd ≈ 6.15 µM) compared to the wild-type E3 (Kd ≈ ∼179 µM) (Figure 5C, Table 1 and Supplementary Figure S5B). To extend our studies, we introduced tryptophan in a third dimeric RING E3, Rad18, that binds Ube2B (also known as Rad6B) with a Kd of ∼41 µM at 25°C (see Table 1) . Previous studies carried out by Sixma and co-workers  also showed that Rad18RING binds Ube2B with a Kd of ∼35 µM. Rad18RING, F53W under identical conditions yields a Kd of ∼2.83 µM for Ube2B (Figure 5D and Table 1). Based on these binding data, we conclude that the presence of tryptophan enhances the E2 affinity across all RINGs irrespective of their oligomeric status in an E2-independent manner.
IR6-4 tryptophan enhances the E2 affinities of dimeric RING E3 ligases.
Tryptophan alleviates the necessity of dimerization for RNF4 activity
Binding studies revealed that the E2 affinities of monomeric and dimeric RING E3s are similarly affected by the IR6-4 tryptophan. The consequence of tryptophan substitution in their catalytic activities, however, is strikingly different. Monomeric E3s essentially lose their ligase activity in the absence of the tryptophan, while dimeric E3s could ubiquitinate even without this residue. These differences in the ubiquitination activities, though rationalized the non-essentiality of tryptophan for dimeric RINGs as opposed to monomers, fail to justify the near complete absence of this residue across all such E3s. Experiments on RNF4S166W, however, indicated that the presence of tryptophan could result in hyper-activate dimeric RINGs and it is, therefore, likely that such hyperactivity is deleterious for most dimers. To probe into this, we chose to utilize RNF4 that has been extensively characterized in terms of its structure and ubiquitination mechanism. RNF4 exists in equilibrium of inactive monomers and active dimeric ligases in vivo and in vitro. More importantly, the substrate, polySUMO2, promotes RNF4 dimerization and hence up-regulates its activity [25,26]. Furthermore, in the absence of the substrate RNF4 appears to dimerize only via its RING domain, unlike RNF8 or many other E3s, making it easier to carry out experiments.
To begin with, we prepared two well-characterized monomeric RNF4 mutants, RNF4Y189A and RNF4ΔC3, along with their S166W versions [26,27]. As expected, RNF4Y189A and RNF4ΔC3 display substantially attenuated ubiquitination activities in multi-turnover assays (Figure 6A,B compare the left and the middle lanes). In contrast with the single mutants, enzymatic activities of RNF4Y189A, S166W and RNF4ΔC3, S166W are comparable to those of the wild-type RNF4 (Figure 6A,B, right lanes). We also carried out Ube2D2∼Ub thioester release assays to get a semi-quantitative estimate for the ligase activities of all these four mutants (Figure 6C). As expected, both the S166W mutants are equally efficient in thioester discharge similar to the wild-type RNF4 (Figure 6D). On the other hand, RNF4Y189A and RNF4ΔC3 could barely release the activated Ub from the thioester. To confirm that S166W mutation did not cause any change in the oligomeric state of the mutants, though unlikely, we subjected RNF4RING,Y189A and RNF4RING,Y189A,S166W along with wtRNF4RING and RNF4RING,S166W to SEC analysis. These RING constructs were necessary as the full-length RNF4 displays an unusual hydrodynamic radius due to its flexible N-terminal region (see next section). The data clearly show that RNF4RING,Y189A and RNF4RING,Y189A,S166W exist as monomers, while the wtRNF4RING and its S166W version elute as dimers (Figure 6E). We also prepared two other monomeric RNF4 mutants, by mutating Ser151 and Gly155, to Arg and Glu, respectively (Supplementary Figure S6A). SEC analysis shows RNF4RING,G155E,S166W also exists as a monomer in solution (Supplementary Figure S6B). In corroboration with the previous mutants, RNF4S151R, S166W and RNF4G155E, S166W display substantial ubiquitination activity compared to that of the wild-type RNF4, while the single mutants failed to show ligase activity (Supplementary Figure S6C,D).
Introduction of IR6-4 tryptophan restores the activity of monomeric RNF4 mutants.
We also carried out ITC experiments to determine the E2-binding affinities of RNF4Y189A and RNF4Y189A,S166W (Figure 6F,G and Table 1). In agreement with all our previous observations, RNF4Y189A binds Ube2D2 with an estimated Kd of ∼85 µM at 25°C, whereas titrations with RNF4Y189A,S166W yield a Kd of ∼5.8 µM under identical conditions. Akin to RNF4Y189A,S166W, RNF4ΔC3,S166W also engages Ube2D2 with a Kd (∼8.3 µM at 25°C) (Figure 6H). All these binding data ascertain that the introduction of the tryptophan enhanced the E2 affinity of monomeric RNF4 mutants restoring their E3 ligase activity.
Avidity for E2∼Ub dictates the enzymatic activity of RNF4 and its mutants
The molecular mechanism that allowed dimeric E3s to retain activity without the tryptophan is far from being obvious from the results described thus far. The crystal structure of RNF4 bound to E2∼Ub revealed that the dimerization of E3s such as RNF4 is crucial for their activity as the second RING domain provides with an additional interaction surface to hold the activated Ub molecule in closed conformation . It is likely that the observed increase in the affinity for the E2 moiety upon introduction of the tryptophan reduced the requirement of Ub binding and, hence, reinstated the activity of monomeric RNF4 mutants. We, therefore, wanted to examine if the tryptophan in the monomeric RNF4 mutants entirely alleviated the necessity of Ub binding in the canonical closed conformation.
Formation of the closed complex involves extensive interactions between the activated Ub moiety with the carrier E2 and the RING domain. These interactions involve three conserved Ub residues, namely Ile44, Ile36, and Gln40 [6,7]. Consequently, mutation of either of these residues has been observed to mitigate the ubiquitination activity of RING E3s. We, therefore, carried out multi-turnover ubiquitination assays for RNF4Y189A,S166W as well as RNF4ΔC3,S166W using UbI36A, UbQ40A, and UbI44A. Results reveal that mutation of either of these Ub residues diminished the poly-Ub smear for both the RNF4 mutants (Figure 7A,B). Thioester release assays also confirm that these three Ub mutations attenuate the ligase activities of RNF4Y189A,S166W and RNF4ΔC3,S166W compared with the ones obtained with wtUb (Figure 7C–F and Supplementary Figure S7A–C). The least effect was observed for UbQ40A; whereas mutation of Ile44, implicated in mediating E2:Ub interaction, abolished ubiquitination.
E3:Ub interaction remains essential for active RNF4 monomers containing IR6-4 tryptophan.
All these findings establish that the interaction between the activated Ub and the RING domain is required for the ligase activity of monomeric RNF4 mutants akin to the dimeric wild type. Thus, it appears that a stronger Ub binding relieved the necessity of a strong E3:E2 interaction in the case of dimers. In other words, the ‘non-catalytic’ RING domain from the second protomer compensates for the low E2 affinity of the primary RING domain via its interaction with the Ub moiety from the E2∼Ub conjugate in dimeric E3s. To verify this proposition, we intended to determine the affinity of dimeric wtRNF4 and its monomeric mutants towards Ube2D2∼Ub conjugate. We prepared catalytically incompetent Ube2D2C85K∼Ub conjugate and subjected it to ITC with wtRNF4. Result shows that indeed wtRNF4 binds Ube2D2C85K∼Ub to Kd of ∼9 µM at 25°C (Figure 7G and Table 1). In contrast with E2 alone, binding of the conjugate to wtRNF4 is endothermic in nature. This unfavorable change in ΔHbinding is compensated by a highly favorable ΔSbinding. Intriguingly, we failed to observe any heat change in the ITC for the binding of Ube2D2C85K∼Ub to monomeric RNF4ΔC3,S166W (data not shown). However, SEC analysis of stoichiometric mixtures of RNF4ΔC3,S166W, or RNF4Y189A,S166W, or RNF4S166W with Ube2D2C85K∼Ub indicates formation of stable E3:E2∼Ub complexes in all the three cases indicating a high-affinity binding (Figure 7H). To solve this apparent anomaly, we first performed ITC to monitor binding of RNF4S166W and Ube2D2C85K∼Ub (Figure 7I). Interestingly, in contrast with wtRNF4, RNF4S166W binds Ube2D2C85K∼Ub with a favorable ΔHbinding. This contrast between the wild-type and the S166W mutants suggested that the binding of RNF4ΔC3,S166W to E2∼Ub may be driven entirely by the entropy and therefore cannot be probed with ITC. We, therefore, prepared a second Ube2D2C85K∼Ub complex using UbY59C that could be labeled with fluorescein for probing the binding using fluorescence anisotropy. We generated this particular mutant Ub over the already existing UbS20C as the fluorescein moiety conjugated to the latter retained substantial mobility restricting its use in anisotropy experiments. Indeed, as expected on the basis of SEC analyses, RNF4ΔC3,S166W could bind E2∼Ub with a Kd of ∼2 µM at 25°C (Figure 7J). In an identical experiment with the wtRNF4, we obtained a Kd of ∼6.2 µM that corroborated with the ITC data. Based on these studies, we conclude that indeed the avidity for the E2∼Ub complex confers enzymatic activity to RING E3s.
RING E3s are most prevalent among all the four E3 classes and constitute one of the largest protein families in eukaryotic proteomes comparable to that of the kinases . Enzymatic activities of these proteins depend on their competency to bind both the E2 and the Ub moieties in E2∼Ub thioester conjugates in proper orientation. Though the importance of E3:Ub [6,7] and E3:E2 interactions in RING ligase activity has been rigorously investigated, the existence of any correlation between these two binding events has not been thoroughly investigated. Biochemical studies on quite a few E3s beginning with c-CBL led to the identification of a tryptophan residue within the RING domain that was critical for E2 binding and hence the ligase activity . But, the importance of the tryptophan moiety across all RING E3s remained contentious likely due to its absence in quite a few active E3s. We show that tryptophan conservation and the oligomeric status of the RING domains are remarkably correlated. Systematic analyses, showing that not just alanine but also hydrophobic residues like leucine and phenylalanine cannot restore the enzymatic activity of monomeric RNF13, establish the indispensability of tryptophan in monomeric RINGs. Tyrosine, however, could partially compensate for the tryptophan in RNF13 as it led to a smaller change in the E2 affinity of RNF13W270Y compared with the wild type. This retention of activity by RNF13W270Y mutant suitably justifies its ranking among all the amino acids in the IR6-4 position. Furthermore, the difference between the phenylalanine and tyrosine points towards a role of the hydroxyl group of the latter in engaging the E2. It is likely that the hydroxyl group of tyrosine acts as an H-bond donor to interact with one of the main chain carboxyl moieties from the E2. However, we could not find any characterized monomeric E3 with the tyrosine residue but note that all the RBR RINGs contain tyrosine at their IR6-4.
Our studies further revealed that the dimeric RNF8, unlike the monomeric E3s, does not essentially require the IR6-4 tryptophan. The presence of the tryptophan, however, allows the wtRNF8 to be highly active. In an analogous manner, the introduction of the tryptophan enhances the activity of dimeric RNF4 (RNF4S166W) as observed by us and also of heterodimeric BRCA1/BARD1 observed by Klevit and co-workers . In agreement with this activity enhancement, we find IR6-4 tryptophan enhances the E2-binding affinity for all RINGs irrespective of their oligomeric state. These observations clearly suggest that the presence of tryptophan is not detrimental for the enzymatic activity of dimeric E3s per se. Why do most dimeric E3 ligases lack tryptophan then? The answer to this question lies in our observation that S166W mutation could rescue the enzymatic activity of monomeric RNF4 mutants. Substrate-induced dimerization has been shown to act as the regulatory mechanism for this E3 in vivo . Thus at low levels and in the absence of the substrate, wild-type RNF4 is essentially inactive. But RNF4S166W can be expected to be active even at low concentrations of the E3/substrate carrying out abnormal modifications inside the cell including enhanced ubiquitination of the E3 itself. Similar behavior is also likely in the case of other dimeric E3s upon the introduction of the tryptophan moiety. Such aberrant and undesired ubiquitination can be detrimental for cellular physiology. In fact, evidence in support of this proposition has been reported in the literature. For example, studies on BCA2 and MARCH5 show that E3 ubiquitination acts as a homeostatic mechanism to regulate their in vivo concentration and is critical for cellular health [30,31]. In yet another study, BRCA1/BARD1 auto-ubiquitination was reported to act as a signaling mechanism rather than causing proteasomal degradation of the E3 . Thus, hyperactivity of these E3s upon introduction of the tryptophan is expected to adversely alter the spatiotemporal regulation of their function. In fact, we note that RNF8, the only exception, has been implicated in a large number of biological pathways including DNA repair  and its malfunctioning has been linked to the progression of cancer. We, therefore, tend to propose that the presence of tryptophan in dimeric RNF8 emphasizes the importance of this E3 ligase in cells.
Dimeric E3 ligases utilize their second RING domain only to engage the activated Ub moiety [6,7]. We show that the dimeric wtRNF4 and its monomeric versions containing the tryptophan bind E2∼Ub with a comparable affinity despite showing >10-fold difference in their E2 affinities. Thus the introduction of the tryptophan could restore the enzymatic activity in RNF4 monomers via enhanced E2 binding despite the lack of the second RING moiety to interact with the activated Ub. On the other hand, most monomeric E3s lose their enzymatic activity upon mutating the tryptophan to alanine due to their weak affinity towards the activated Ub. This observation confirms our notion that the catalytic efficiency of RING E3s depends on their ‘avidity’ for the E2∼Ub conjugate rather than the individual affinities towards the E2 or the Ub. We also note that all the three monomeric RINGs, namely RNF25 (also known as A07), RNF146, and RNF125, that do not contain the IR6-4 tryptophan contain diverse secondary elements to assist the RING in engaging the E2 [13,34,35]. This observation further supports our avidity-based model for RING E3 activity.
In summary, this study underscores the importance of a single amino acid residue in E3 ligase activity. It also provides the first-ever evidence of an unusual, yet functionally relevant, correlation between the oligomeric status of a protein domain and the presence a specific amino acid residue away from the oligomeric interface. Furthermore, all outliers to this correlation, such as A07, are found to employ novel mechanisms to carry out their biological functions. Thus characterization of outliers to this general theme of E3 activity control may assist identification of novel regulatory mechanisms employed by RING E3 ligases.
A.B.D. along with other authors designed the experiments. S.S. carried out the experiments with RNF13, RNF4, RNF8, Rad18, and their mutants. A.P.B. carried out experiments with ZNRF1, Rad18, and their mutants. P.B. analyzed all the RING sequences and tabulated the data. P.A.B. prepared the UbS20C mutant and standardized the labeling procedure. A.B.D. carried out ITC experiments, created the database, analyzed the data and wrote the paper with S.S. and A.P.B.
This research work was partially supported by the Wellcome Trust-DBT India Alliance through their intermediate fellowship grant to A.B.D. [grant no. 500241/Z/11/Z] and by the intramural grant from Bose Institute. S.S., A.P.B., and P.A.B. are recipients of NET fellowship from Council of Scientific and Industrial Research, India.
The authors would like to thank Mr. Pritam Naskar and Dr. Shreyasi Dutta for helpful discussions. The authors also thank Mr. Mrityunjoy Kundu and Mr. Mrinal K. Das for their help with Typhoon Imager and DNA sequencing.
The Authors declare that there are no competing interests associated with the manuscript.
These two authors contributed equally.