This work extends previous theory on solid-effect DNP in viscous liquids by one of the authors to the more complicated situation where g anisotropy is partially averaged by slow tumbling. The derivations provide the basis for a computationally affordable numerical treatment. Field profiles of DNP enhancements (“DNP spectra”) fitted on the basis of this approach are in rather good agreement with experiments. The fitted parameters take physically reasonable values, despite the fact that the model for spin label dynamics is strongly simplified with respect to expected dynamics in the experimentally studied systems. The results constitute a substantial advance in a field of current interest. The manuscript is mostly clear and well written. I recommend publication after minor revision that takes into account the following issues.

1. The treatment of the experimental spectra assumes isotropic rotational diffusion (stated, a bit late, in line 602). For 10-Doxyl-PC in DOPC bilayers, this motional model is a strong simplification. Even for 16-Doxyl-PC, which is closer to the chain end and experiences less of an orienting potential, I am not sure how good an isotropic-motion model is. This, and the worsened agreement upon including nitrogen hyperfine coupling/anisotropy for 10-Doxyl-PC (Figure 4A) may indicate error compensation between deficiencies of the motion model and the neglect of ¹⁴N hyperfine coupling. In my opinion, this possibility should be mentioned. The difference in parameters between 10-Doxyl-PC and 16-Doxyl-PC (Table 2) might partially be due to the motional model being worse for 10-Doxyl-PC than for 16-Doxyl-PC.
2. Motion of spin-labelled lipids in lipid bilayers is a rather well researched topic (see, e.g., Livshits, V.A., Kurad, D., Marsh, D. Multifrequency simulations of the EPR spectra of lipid spin labels in membranes. J. Magn. Reson. 2006, 180, 63-71 and references therein). I believe that the motional model of 16-Doxyl-PC and 10-Doxyl-PC in DOPC bilayers could be substantially improved by measuring CW EPR spectra at a few more frequencies and fitting them by the microscopic-order macroscopic-disorder model (MOMD) developed by Freed et al. While this is clearly beyond the scope of the present study, I would like to see some comment whether the approach introduced in this manuscript could be used with a MOMD model or at least with a model of anisotropic rotational diffusion (axial rotational diffusion tensor).
3. I wonder how difficult it would really be to extend the approach to including ¹⁴N hyperfine anisotropy. I would assume that computation time increases only by a factor of three, because the subspaces with different nuclear magnetic quantum number of ¹⁴N could be treated independently. The g and hyperfine tensor being coaxial, each subspace could be computed with an effective tensor. The authors may want to consider whether this would be feasible and comment on it. Again, it may be beyond the scope of this study, but may also be of interest, as neglect of ¹⁴N hyperfine coupling does make a difference (Figures A4, A5).
4. Work such as this one is notoriously hard to reproduce without having access to the programs used for the computations. The authors may want to consider putting the program on GitHub.

Details:  

1. Around line 110: “cw-EPR spectrum deviates from a Lorentzian line”. I would have written “cw-EPR lineshape is not Lorentzian”, because spectra often consist of several Lorentzian lines, not only one.
2. When first discussing the deviation in Fig. 1(d) between experimental data and the decomposition into OE and SE contribution, it would be helpful to tell the reader that the paper addresses this issue.
3. Line 201, “When generalizing the Bloch equations to non-isotropic g tensor” I would have written “When generalizing the Bloch equations to an anisotropic g tensor”
4. Line 220/221: “For the same analysis to apply to solids, spin diffusion should be much faster”. I do not understand how spin diffusion enters the picture here. In the previous sentence, you mention the correlation time of the (supposedly electron-nuclear) dipolar interaction. There is no electron-nuclear spin diffusion that could be an analogous phenomenon in solids.
5. Line 224/225: “correlation time of the dipolar interaction is infinitely long (but still much shorter than the nuclear T1” Please replace “infinitely long” with “much longer than…”. Probably you compare to relaxation on zero-quantum and double-quantum transitions in the absence of fluctuating electron-nuclear dipolar interaction?
6. Figure A4: The simulation agrees better with experiment without considering the hyperfine interaction that certainly does exist. Some error compensation must be at play here. This needs to be mentioned.

The paper is a continuation of the series of theoretical works on the Solid Effect. The paper considers the effect of g-factor anisotropy on the DNP spectra. I believe that this theoretical work deserves to be published in the “Magnetic Resonance”.  The work is rather bulk, both in volume and in the cases and approximations considered, and, in my opinion, it would be better to divide it into two publications, for instance, in one to consider the theory and in the second to compare the theory with experiment. But this is at the discretion of the authors. I assume that specialists working in the field of DNP will still be able to understand this paper in detail.

I suggest to accept this paper as it is with the following minor remarks:

Already in the abstract states that "the dipolar interaction between the polarized nuclear spins and the polarizing electrons is not completely averaged out

by molecular diffusion." But in the case of isotropic rotation, which is considered in the paper, the mean value of the dipole interaction is zero, whether the rotation is slow or fast. It is necessary to explain in what sense the dipole interaction is not completely averaged out. Also I did not find (perhaps I did not notice) the criterion of "slowness" of rotation or tumbling. Perhaps dipole interaction is not averaged to zero in nitroxide-labeled lipids in fluid lipid bilayers, where there is preferred direction in the bilayer, but the paper considers, as I understood, isotropic rotation. Further, if the rotation is slow in the sense of adiabatic mode (the criterion of adiabaticity is known), it may be easier to consider SE in adiabatic approximation with solid angle \Omega as an adiabatic parameter.

The second minor remark. I was confused by the statement:

     "In (59), the sum over l mixes only expansion coefficients whose values l differ by two from L". Does this mean that l=L+2 or l=L-2, rather than the correct condition |L-2| <=l <=|L+2| ?

The manuscript describes detailed work on simulating DNP features originating from the solid effect, both in solids, and in liquids. The liquids case is the most interesting one, whereby asymmetry in the g-tensor leads to sizable solid-effect features. The work is truly impressive, providing simplified equations that allow very efficient computation of lineshapes, and thereby also providing an efficient data fitting modality. The work is probably easily accessible to the expert, but the broader readership would appreciate if further helpful comments were included. For example, in each section where a certain calculation is performed, it would be very helpful if at the beginning it were stated what the final goal of this portion is. I also feel that it may be suitable to cite the review article by Atsarkin. Some symbols are introduced in an ad hoc fashion which makes it harder to follow the thought process.  For example, the provenance of the i \omega_I terms in Eq. 19 is not explained, and one is left to guess. So some additional guiding statements would greatly improve the article.