In this manuscript, McAlpine and Zhu et al. report on discovering specific Ubiquitin Variants (UbVs) to inhibit the activity of E2 enzymes of the Ube2d family. They successfully identify 6 UbVs with varying specificity for Ube2d family and employ a suite of techniques (enzymatic assays, X-ray crystallography, and solutions biophysics) to characterize the interactions between these UbVs and Ube2d2. Interestingly, the crystal structure of UbV.1 bound to Ube2d2 shows that two UbV.1 molecules bind to one Ube2d2, at the Ube2d2 backside and Ring interacting sites. The authors found the UbV1 that is interacting with the back-side ubiquitin site on Ube2d2 enhances its inhibition of E2 activity. Further, the authors determine the structure of UbV.3 bound to Ube2d2. While these crystal structures are of high quality and well presented, the biophysics data testing the UbV.1 binding stoichiometry shows some inconsistencies and additional controls are required. The observation that UbV.1 and UbV.3 shows specificity for Ube2d2 and within the Ube2d family would be of interest. However, the specificity among all E2s is not as strong as claimed by the authors and prominently stated in the title (see details below). Further a previous study has already identified a Ube2d1-specific UbV with much higher potency than the UbVs reported here (ref. 17), taking away a lot of the novelty of the current study that identification of specific UbVs for single members of the Ube2d family is possible.

In summary, this study may be of interest to some researchers in the E2 field but needs some additional experiments and controls as outlined below.

Major points:
1) The biophysical data supporting the 2:1 binding stoichiometry of UbV.1 to Ube2d2 seems weak. While there are differences in the positions of the peaks in the SEC and AUC, the ITC experiments indicate a 1:1 stoichiometry. Importantly, the accuracy of the SEC method to determine MWs/stoichiometry is limited as the elution volume of a protein complex in SEC is not only dependent on the molecular weight. In line with this, the gels of the SEC experiments in Fig 5A and B show a large fraction of unbound UbV for both UbV.1 and UbV.3 (maybe even more for UbV.1 than UbV.3; but see point 11 below) even though a 2:1 ration of UbV and Ube2d2 was used in both cases. For the AUC experiments, the authors identify the molecular weight of the individual proteins but only report the Svedberg units for the complexes (main text p. 11). What are the calculated MWs from the reported Svedberg units and how to these compare to the expected MWs? How do the authors explain the discrepancies between ITC and the other AUC (and SEC) experiments?
2) In addition to the point above, the effect of the S22R mutant appears inconsistent between different experiments. The ITC data for UbV.1 shows a >4-fold stronger binding for Ube2d2 S22R (2.3 µM) compared to Ube2d2 WT (10.8 µM), with the same N = ~1 for both Ube2d2 variants. It is difficult for this reviewer to match these data with the SEC and AUC data, where the S22R mutant seems to show a loss of binding as the peaks are shifted to lower MW (shifted right in SEC, shifted left in AUC) compared to UbV.1 + Ube2d2 WT.
3) The SEC data is difficult to interpret as the calibration standards for the two experiments in Fig 4A and 4B have very different elution volumes. The authors conclude that UbV.1 with Ube2d2-S22R elutes between a 2:1 and 1:1 ratio and that for the UbV.3 complex with Ube2d2-S22R they report a 1:1 ratio as with Ube2d2-WT. However, in Fig 4A, Ube2d2-UbV.3 elutes left of the 29 kDa calibration standard, whereas Ube2d2-S22R-UbV.3 elutes later than the 29 kDa standard in Fig 4B. How do the authors explain this difference? This difference for the UbV.3 complexes makes it difficult to interpret the differences observed with the UbV.1 complexes. It would be good to repeat all these experiments under identical conditions and also include the individual UbVs alone as additional controls.
4) The authors claim that UbV.1 and UbV.3 are specific for Ube2d2 and within the Ube2d2 family. First, the authors should make it much clearer throughout the manuscript that a Ube2d1-specific UbV with much higher potency has previously been reported (ref. 17). Second, the characterised UbVs are not as specific as the authors claim in the title and elsewhere. The data in Fig 6 show that UbV.1 binds 2-3 out 4 Ube2d family members. While UbV.3 seems more specific within the Ube2d family (Fig. 6), it shows strong binding for other E2s, Ube2b, Ube2i and Ube2u (Fig 1D). The authors should more carefully word the sections on specificity (incl. the title) considering these observation. In addition, the authors test UbV specificity against different Ube2d family members using qualitative GST-pulldown assays. The authors should include more quantitative assays such as titrations in ELISA or ITC and should consider including other E2s (Ube2b, Ube2i and Ube2u) for UbV.3.
5) Many of the experiments rely on visualisation of fluorescently labelled ubiquitin (Fig 1B,C, Fig 3E,F, Fig 6C). The authors need to show Coomassie-stained gels for these experiments to validate even loading of enzymes and ubiquitin in each assay.
6) It is unclear how many times most of the biochemical and biophysical experiments were performed. The authors present single gels for all the biochemical assays without quantification. Are these the only experiments performed or are these representative gels of multiple experiments? The authors should specify this and repeat experiments if necessary to increase rigor.
7) The manuscript contains some inconsistencies in the experimental descriptions, e.g. between the main text and methods and the methods and presented data. Examples include the E2 charging assays. The main text (p. 9, l. 6) states that these experiments were performed at 30°C for 10 min. Fig 3A only shows 1 and 2 min time points. The methods section states that the experiment was performed at 28°C (p. 19, l. 8).
8) The ITC experiments also show inconsistencies and unexpected features. The methods sections states that ~200 µM ligand was titrated into ~20 µM E2 for all experiments. However, the titration in Fig 4 and S3B show very different final molar ratios (~2, >2.5 and >4). The experiment in Fig S3B shows very high DP even though no binding is detected. Can the authors explain and clarify these observations? If different experimental setups were used, the authors need to include these details.
9) Fig 2: It appears that Ube2d2 and used for the crystal structures contains additional N-terminal residues (S0, G-1 L-2, P-3). The same seems to be the case for UbV.1 and UbV.3. While the latter is not such a problem as these are synthetic proteins anyway, the fact that Ube2d2 contains additional residues that are part of the interaction surface with the UbVs is concerning. Did the authors use the same Ube2d2 constructs for their biophysical analysis? The authors need to show that the UbVs can bind Ube2d2 with a native N-terminus to confirm that the UbVs bind endogenous Ube2d2 without these additional residues.
10) Fig 3A–D: The authors show the interaction surface of multiple proteins of the ubiquitin cascade that interact with Ube2d2 to highlight the competition of the UbVs with these other ligands. In addition to just showing the binding interface, the authors should show (e.g. in a supplemental figure) the full structures, as regions outside the core interaction surface may clash with the UbVs. The authors also need to indicate the PDBs (and their references) used for these comparisons.
11) The gels in Fig 5A and B are of very poor quality, making it almost impossible to see some of the bands. The authors should repeat these experiments and produce better gels. Why does the Ube2d2-S22R run much lower in the middle panel in Fig 5B compared to all other gels?

Minor points:
i) How were the ELISA data in Fig 1D normalised. Can the authors provide more details and better specify the range used. “High” and “Low” is not very informative.
ii) Additional references should be included, e.g. for the PDB structures used for comparisons in Fig 3 and molecular replacement models, rather than just the PDB codes, and for the software used to analyse ITC data (only one of the 3 programs is referenced).
iii) Many of the figures would be easier to follow if additional labels were added. Examples include:
a. Fig 1B/C: label the gels with Ube2d2 WT and Ube2d2 S22R respectively.
b. Fig 2B,C,F: What do the dotted lines represent?
c. Fig 6B: Label the bands
d. Fig 3F, 6C: label the UbV bands
iv) The authors should include more information on the AUC experiments. How long were the protein complexes preequilibrated for? How was buffer matching done. What was the pathlength of the sector and targeted total absorbance? At what resolution was the data processed with. Is frictional ratio free floating or fixed? Does the residual bitmap and the fit of experimental data look good?
v) For the AUC experiments, can the authors include a table in supplement of sedimentation coefficient, the integrated absorbance values, and molecular mass?

In their revised manuscript, the authors have tried to address many of our previous points and have improved their manuscript by repeating some of the experiments and rephrasing the main text. The specificity of the UbVs is now more adequately presented throughout the manuscript. We recognise the authors’ efforts to refocus their analysis on the ITC and AUC experiments and provide additional AUC data analysis details in new Table 2.

However, in our view the data on the different stoichiometry of UbV1 and UbV3, specifically the binding of UbV1 to Ube2d2, still does not convincingly support a strong 2:1 binding model of a UbV1 dimer. In our view, the presented data is most consistent with one strong UbV1 binding site on Ube2d2 (UbV1a) and a second much weaker binding site (UbV1b). This interpretation is consistent with the most robust biophysical data (ITC and AUC) where the authors observe a stoichiometry of just over 1:1 under their conditions used. It is also consistent with the X-ray structure, where the authors observe 3/3 UbV1a binding sites occupied but only 2/3 UbV1b binding sites. While the authors could test the affinity of the 2nd binding site, e.g. using much higher UbV concentrations in ITC or AUC to achieve full saturation of the 2nd site or using Ube2d2 mutants that disrupt only the 1st binding site, we acknowledge that this may be very difficult due to the weak affinity, and we therefore do not require these experiments for this manuscript. However, the authors need to rephrase the relevant statements about stoichiometry and UbV dimer in the abstract, end of introduction, results and discussion.

Other issues the authors should address are:

1) Discussion p. 12 line 21. Here, the authors write that UbV.1 binds with an intermediate stoichiometry to the Ube2d2 S22R mutant. What do the authors mean by ‘intermediate stoichiometry’? The presented data, e.g. the AUC data in Table 2 (fitted MW = 29.5 kDa, calculated MW = 28.3 kDa), is most consistent with a 1:1 binding model of UbV.1 and Ube2d2.
2) The authors now put more focus on the ITC and AUC biophysical data rather than the SEC data but removed some of the ITC data analysis. They need to include at least the specific cell and syringe concentrations rather than a range, calculated Kd, statistics, and stoichiometry values for the ITC data presented in Fig 6; either for each ITC curve in the figure or as a separate table.
3) The authors replaced some of the original ITC experiments with new experiments and for some obtained different affinities between the different experiments. How do the authors justify only showing the new results? In our view, it would be preferred to treat these different experiments as multiple repeats and fit the data globally or at least report the confidence interval via SEDPHAT or averages of the multiple experiments.
4) In response to our comment in the first round of reviews (Major point 9), the authors write that they used different protein constructs for crystallography and biophysical studies, and, importantly, used E2 enzymes with native N-termini for the biophysical studies. This is not reflected in the methods section where only Ube2d2 constructs with an N-terminal His-tag for the phage display and a GST-fusion Ube2d2 are described. The GST-fusion Ube2d2 would presumably still contain additional N-terminal residues even after 3C cleavage of the tag. Which of these proteins was used for crystallography, which for the biophysical assays? The authors should clearly describe which constructs are used for which experiments.
5) In the first round of reviews, we referred to labelling UbV bands in the presented fluorescent gels (point iii-d.) and the authors responded that UbV bands would not be detectable in the fluorescent images, which we agree with. Nevertheless, in Fig. 8D prominent additional bands are visible between di- and tri-Ub only when the UbVs are added. Could these bands be ubiquitinated UbVs?
6) What do the authors mean by somewhere between 1 and 2 molecules bind UbV2 (in new text added to p.10)? This can’t physically be true. Instead, this suggests that there are multiple binding sites with different affinities that are occupied at different levels depending on the UbV concentration. The authors should rephrase this also considering our main point raised above.
7) Some of the descriptions in the Methods should include additional details:
a. It would be easier if the authors split the ITC and GST pull-down experiments into separate sections rather than a single "Binding experiments” section.
b. For the ITC, the authors should include the stirring speed and number of injections as well as the model used in SEDPHAT for processing.
c. For the GST-pulldown assay, what type of GSH resin was used? What amount/concentration of bait and candidate proteins were used? What buffer conditions were used?
d. For the analytical SEC please report flowrate and injection volume.

The authors now have addressed all remaining comments and concerns.