The authors studied a very interesting system where the interactions between two intrinsically disordered proteins are highly dynamic. As far as I know, currently there is little knowledge on IDP-IDP interactions. In fact, it has been unclear that whether strong interactions could naturally occur between two IDPs for molecular recognition. This study provides new insights into such IDP-IDP interaction system, which is of importance in the whole field.

Experiments were carefully designed and multiple experimental tools were employed, including NMR spectroscopy, MD simulation, and single-molecule FRET. Data are in general clearly presented, except for a few cases specified below. The model proposed by the authors can explain the experimental data well. In summary, it should be published after additional revision. Here are my comments on the current manuscript.

1. It would be interesting to see a titration experiment of 15N-NuMA with U-4.1G. The chemical shift changes and peak broadening should validate the authors’ model, if it is proper. Additional information, such as the residues on NuMA that are in close contact with 4.1G, may be obtained from this titration experiment.

2. Provide amino acid sequence information of 4.1G and NuMA in SI.

3. Is His6-tag cleaved in all the constructs for NMR measurements and smFRET experiments? If so, how does this exogenous tag affect the results compared to native constructs? Is this tag included in MD simulation? Is this a possible resource that causes the discrepancy between the smFRET and MD distance distribution data (Figure 4)?

4. Please rationalize the temperature used in ITC experiment (18˚C instead of, for example, 25˚C). Also, at what temperature did smFRET perform?

5. Measuring NMR relaxation parameters, such as backbone 15N R1, R2, and hetNOE, of the free 4.1G is beneficial. Based on Figure S5 orange spectrum, it seems promising to measure relaxation parameters for this 4.1G-NuMA (1:0.5) sample, which would be a great experimental input to further validate MD simulations.

6. Figure S6 may not be sufficient to support the authors’ statement that the construct is an IDP. The total number of peaks is apparently less than the expected one (~160), indicating many of the peaks are actually invisible. The corresponding residues of these peaks could form stable structure. Measuring relaxation parameters could help to resolve this issue to some degree. Otherwise, just remove this irrelevant (and even possibly misleading) figure.

7. Why MD simulation was performed with CHARMM 27 force field? The newer version CHARMM 36 seems have a better performance for IDPs.

Other Minor Points:

1. Figure 5 caption: d) “4.1G (C)” should read as “4.1G (C)/NuMA”, clearly a typo.

2. Figure S5, I would encourage the authors to show a superimposed 1D spectra at a representative cross-section (e.g. 15N 123.5 ppm). Contour plot of 2D NMR spectra has inherent limitation, and can be easily biased if contour is not properly selected. 1D cross-sections are good complements and should be shown whenever possible for superimposed 2D spectra. It also better shows the peak broadening effect.