GSK-3 signaling in developing cortical neurons is essential for radial migration and dendritic orientation

We have thoroughly revised our manuscript based on helpful comments of the reviewers. In particular, a weakness of the prior version that diminished novelty was inadequate verification of several of the lines of mutant mice on which we reported negative findings. As instructed by the editors, we have carefully verified protein reduction after Neurod6-Cre mediated recombination in each of the six lines and provided quantification of cortical neuronal migration. Most of these results are shown in a new Figure 6 where new confocal images of Cux1 expressing neurons in mutants and controls are shown together with Western blots documenting strongly reduced levels of proteins in the mutants.

Quantification is shown in supplemental material linked to this new data in Figure 6—figure supplement figure 1.

We have also carefully commented on all of the points raised by each reviewer and revised wording in the text in accordance with reviewer suggestions.

Reviewer #1: […] 1) The authors demonstrate a robust and clearly-defined deficit in the migration of newborn Cux1-positive cortical neurons as soon as they are born. In contrast, they see that the deeper-layer Tbr1-positive neurons are unaffected. Since these neurons are born quite a bit earlier, are the authors confident that their driver line Nex-Cre actually causes recombination early enough to affect these earlier-born neurons? Also, most Tbr1-positive neurons are generated directly by radial glia, while the Cux1-positive neurons are largely generated via intermediate progenitors. Since intermediate progenitors are targeted by their Nex-Cre construct, might this provide an explanation for these differences?

While it is not essential for the authors to address these issues experimentally here, they should be more cautious in their interpretation of these differences.

We are in complete agreement with the reviewer on this point. In fact we did not perform extensive analysis of the Tbr1-positive cells precisely because we did not think we could draw a strong conclusion about GSK-3 regulation of this population.

In order to make it clear that the strong GSK-3 regulation of migration may not apply to early born Tbr1-positive neurons, we modified the text in the Discussion as follows: “Deeper layer neurons expressing Tbr1 were not strongly influenced, but we cannot be certain that GSK-3 protein was fully depleted in these earlier born neurons at the time migration was occurring.”

2) In Figure 2A,B, using mice floxed for both GSK3A and B, it looks like there are many more Cux1-positive neurons in addition to their deficit in migration. Is it possible that inducible deletion of GSK3 in intermediate progenitors causes an expansion of cortical neurons which in turn affects their migration indirectly? The authors need to address this possibility by quantifying the relative proportions of Cux1-positive neurons in these floxed mice relative to the controls.

We think this unlikely because our cell autonomous recombination studies (Figure 4A’,B’) where recombination is targeted to postmitotic neurons with Neurod1-Cre show that a heavy majority of cells did not migrate to the upper layers. Also time lapse imaging of cells that are clearly neurons show that migration is strikingly delayed for almost all neurons imaged (Figure 4E).

In addition, as suggested by the reviewer, we quantified numbers of Cux1 positive cells in relation to total cell number in both Gsk3 mutants and controls. The number of Cux1 neurons in mutants was very similar to the number in controls.

3) In Figure 2D, the authors show statistical significance, but they only have an n=2 for controls. How can this be the case? I realize that some authors count each section as an n, but this isn't really valid for these types of analyses.

N = 2 for controls and n = 3 for the experimental refers to number of animals studied. These were littermate controls obtained from 2 separate pregnant females, multiple separate coronal sections of the electroporated cortex was studied in each animal (on average, 300 neurons per section). We do not necessarily get appropriate genotypes with matched controls from every litter electroporated. Since the n is low and the effects are relatively modest, we moved the overexpression image to Figure 2—figure supplement 1 and emphasize the point that overexpression of Gsk3 does not delay migration (based on n = 3 animals) as would have been predicted from prior studies.

4) One of the main conclusions of this manuscript is that GSK3 is essential for radial but not tangential migration, based upon their analysis of cortical excitatory versus MGE-generated inhibitory interneurons. This conclusion isn't really warranted by the data they show. In particular, while Figure 3A–A' does show that inhibitory interneurons make it into the cortex, these data need to be quantified and it needs to be shown that GSK3 has actually been deleted in these neurons to draw this conclusion.

We have addressed both of these important points. A Western blot of MGE lystates has been added to supplementary material, Figure 3—figure supplement 1, which demonstrates strong reduction GSK-3 protein in precursors and newly generated inhibitory neurons after recombination with Dlx5/6-Cre. Quantification showing only modest if any effects on migration is included in this supplement. We have also added the caveat to the text that we assessed only gross measures of tangential migration as we did not asses migration of interneuron subtypes in this study: “These results are not meant to imply that migration of interneurons was normal in every respect as we did not assess migration of specific interneuron subsets.”

5) Similarly, the authors conclusions about direct effects of GSK3 ablation on dendritic orientation are also difficult to draw based upon their experimental design. In particular, while I agree that their data convincingly show inappropriate orientation of the dendrites of Cux1-positive neurons in layers 2-3, the authors cannot rule out the possibility that these neurons were delayed or perturbed in their migration to this location, and that the dendritic deficits are a secondary consequence of any such perturbations. The authors need to deemphasize their conclusions that these are direct effects unless they have data dissociating migration from later events such as dendritic development.

The reviewer makes an interesting point. However, we are not in complete agreement. We specifically chose the few neurons in normal locations in Layer 2-3 for analysis. Figure 4A,A’ shows that some Gsk3 deleted neurons are normally positioned by E19 and Figure 1—figure supplement 1B shows a few Cux 1 neurons normally positioned as early as E16. It seems fair to assume that these neurons did not undergo a prolonged migration delay, presumably because GSK-3 levels were not drastically reduced in these few neurons at these early stages. As the reviewer acknowledges, the abnormal orientation of the dendrites is striking. We think the most plausible explanation is that dendrites of Gsk3 deleted neurons do not respond normally to directional cues that are present.

We agree that we can’t absolutely prove that dendritic polarization is directly related to GSK-3 abrogating signaling as opposed to an abnormality acquired during migration. We have therefore modified the title to: “GSK-3 signaling in developing cortical neurons is essential for radial migration and dendritic orientation” which is certainly true and avoids the terms “regulate” and “polarization”. We have also modified our text to reflect reviewer concerns on this point as follows: A) We changed the section heading from: “GSK-3 regulates dendritic orientation” to “Abnormal dendritic orientation in Gsk3 mutants”

B) We added text: “Images at E16 (Figure 1—figure supplement 1B) and E19 (Figure 4A’) show that a few Gsk3 deleted neurons are normally positioned and suggest that these normally positioned neurons did not undergo a substantial delay in migration.”

6) Data in Figures 6 and 7 showing negative results vis a vis migration with multiple signaling proteins are a nice addition to the manuscript. However, to really substantiate these negative data, it would be important to (i) show that the Cre-mediated recombination really did occur as predicted, and (ii) show some kind of quantification to rule out the possibility that there are indeed neuronal deficits, but that they are just more subtle than with GSK3 knockout.

Again, a very helpful point. As noted above, we have carefully verified protein reduction and provided quantification of migration in five of the six lines where we report negative results, Figure 6 and Figure 6—figure supplement figure 1. For comparison we show quantification of the strikingly abnormal quantification in Gsk3 mutants in Figure 2, new panels A and C. For Dvl123:Neurod6 we did not obtain enough 6 allele mutant mice for both Westerns and detailed quantification. We do show a representative image based on inspection of multiple sections from n = 3 of the 6 allele mutant mice. We hope we can keep the image in the paper as it will be of substantial interest to researchers in this field.

7) I'm not sure the small amount of microarray data shown in Figure 6B really adds anything to the manuscript. I would either add all of the microarray data or remove this.

We have now removed the Table shown in former Figure 6B and simply mention results in the text. The full array has been deposited in GEO, NCBI for reference (GEO accession number GSE58727) and will be released to the public shortly.

8) The data on Crmp-2 are important and interesting. The Dcx data are potentially also very interesting, but the authors don't show that there is no change in total Dcx levels. In the absence of such data, it is difficult to know whether relative phosphorylation of Dcx is actually decreased.

We probed for total DCX in these blots as directed. The reduction in pDCX remains very striking. The new results are shown in a revised Figure 7A.

Reviewer #2: […] The data are mostly well-documented and convincing. Indeed, as authors claim, to my knowledge, this is the first genetic demonstration that GSK-3's are required for specific aspects of neuronal migration. My main concern is with the novelty of the findings, relative to the J Nsci study by Asada and Sanada which also showed a requirement for GSK-3 in migration, albeit through an acute, non-genetic manipulation. Other studies, including Shi et al. Curr Biol 14, 2025-2032 (2004) have also implicated GSK-3 in dendrite vs. axon polarity.

We addressed this point in an exchange with the editors: our findings are in direct conflict with the Asada study. We did not verify that inhibition of GSK-3 via an LKB1→GSK-3→APC (dephosphorylation) signaling cascade is required for neuronal migration as was published in that paper. Analysis of GSK-3 in the Asada paper relied exclusively on activation of GSK-3 via the GSK-3β S9A construct and the demonstrated effects on migration were transient. We understand that GSK-3 inhibition due to ser9/ser21 phosphorylation and subsequent dephosphorylation of key cytoskeletal regulators has previously been implicated in vitro in elaboration of axons and migration of non-neuronal cells. However, we and others (Courchet et al. Cell 2013) do not find evidence that LKB1 signaling regulates migration in vivo. Finally, published work that we cite on GSK-3 mutant knock-ins that are viable and exhibit grossly normal brain development, suggest it is unlikely that ser9/ser21 phosphorylation of GSK-3 strongly regulates cortical neuron migration in vivo. Rather, our study shows that GSK-3 activity is required for cortical neuronal migration in vivo. Our evidence suggests that GSK-3 phosphorylation of key downstream cytoskeletal regulators (rather than dephosphorylation) is crucial for migration.

Additionally, while I appreciate the lack of migration phenotypes in other KOs examined here, contrasted with Asada's manipulations of LKB1, I am concerned about the extent of the loss of function induced. The fact that in the b-cat KO, the authors fail to detect any effect on Wnt signalling targets makes me wary of their strong conclusions. Especially, since the effect of KO on b-catenin expression is barely more than a two-fold change. How do we know that the induction of those KOs is efficient enough to warrant the strong conclusion that LKB1, Pten, b-catenin etc, are not required for migration?

To clarify, we did not assess Wnt target genes in Ctnnb1^Ex3:Neurod6 mice. Rather, we assessed Wnt target genes in Gsk3:Neurod6 mice. Reviewer 1 felt that this aspect should be given less emphasis and we have changed the presentation of the data as directed (see response to point 7 above)

We now show strikingly reduced protein levels after Neurod6-Cre mediated recombination for all of signaling mediators studied (new Figure 6 and—figure supplement 1) as noted in point 6 above. These strong protein reductions demonstrate that Neurod6-Cre was highly effective in generating recombination of all of the floxed alleles used in this study.

Reviewer 3: […] My main recommendation would be to revise the Abstract to better highlight the interest of the study. In particular, the first sentence is difficult to understand (“strongly regulates progenitors”?) and a bit discouraging as well since one of the virtue of the deleter line used here is to leave aside these previously investigated aspects of GSK3. They were also somewhat less interesting from a developmental neurobiology viewpoint since they dealt in essence with proliferation as opposed to migration. We have revised the Abstract and removed the first sentence as recommended.

In a similar vein, it is unfortunate that the Abstract fails to mention that GSK does not seem to be involved in the migration of cortical interneurons.

We have added a new sentence to the Abstract: “In contrast, tangential migration is not grossly impaired after Gsk3 deletion in interneuron precursors”. This point is also now mentioned in the Introduction.

Finally, the role of GSK in the Semaphorin pathway would be worth emphasizing in the Abstract and perhaps better explained in the Introduction. The important Chen et al. 2008 study is briefly mentioned there as well, but it would be useful if it would be better emphasize at the expense of some of the Wnt pathway discussion that could be kept for the Discussion.

Because of the word limitations of the Abstract, we cannot emphasize the Sema pathway here. However, we have revised the Introduction to include information on the Sema pathway as instructed.