Gimap5-dependent inactivation of GSK3β is required for CD4+ T cell homeostasis and prevention of immune pathology

GTPase of immunity-associated protein 5 (Gimap5) is linked with lymphocyte survival, autoimmunity, and colitis, but its mechanisms of action are unclear. Here, we show that Gimap5 is essential for the inactivation of glycogen synthase kinase-3β (GSK3β) following T cell activation. In the absence of Gimap5, constitutive GSK3β activity constrains c-Myc induction and NFATc1 nuclear import, thereby limiting productive CD4+ T cell proliferation. Additionally, Gimap5 facilitates Ser389 phosphorylation and nuclear translocation of GSK3β, thereby limiting DNA damage in CD4+ T cells. Importantly, pharmacological inhibition and genetic targeting of GSK3β can override Gimap5 deficiency in CD4+ T cells and ameliorates immunopathology in mice. Finally, we show that a human patient with a GIMAP5 loss-of-function mutation has lymphopenia and impaired T cell proliferation in vitro that can be rescued with GSK3 inhibitors. Given that the expression of Gimap5 is lymphocyte-restricted, we propose that its control of GSK3β is an important checkpoint in lymphocyte proliferation.

The Gimap GTPases are a set of septin/dynamin-related proteins which are expressed prominently in the haematopoietic lineages of mammals (and probably of other vertebrates), in particular in T lymphocytes. Mutations in rats and/or mice of two members of the gene family, Gimap5 and Gimap1, lead to substantial lymphopenias but the basis/bases of these detrimental effects are yet to be established. This paper addresses Gimap5 and proposes its involvement in the regulation of the activity of glycogen synthase kinase 3β (GSK3β). It proposes that Gimap5 is a negative regulator of GSK3β activity and helps to constrain the degree to which this Ser/Thr kinase promotes the degradation of some factors crucial to aspects of T cell proliferation (notably c-Myc). It goes on to propose that in the absence of Gimap5 the inhibition of GSK3β is impaired, with deleterious consequences for this proliferation and for cell survival, with increased levels of DNA damage. The manuscript contains some compelling data in support of the authors' hypothesis. At the wholeanimal level, the ameliorative effects on the sph/sph deficiency phenotypes of GSK3β ablation and GSK3β inhibitors are impressive. The narrative put together to develop a hypothesis for the potential molecular mechanism of Gimap5 function is very interesting. At the same time, however, the manuscript suffers from some substantial deficiencies in respect of (1) the quality and presentation of some of the data and (2) the writing of the manuscript, in particular the Results section, which suffers from a rather incoherent style and from inaccuracies in the proper description of the data (inaccurate scientific language). (The Discussion section is better written). The poor quality of presentation should have been evident to at least some of the long list of coauthors and I am surprised at the carelessness of submitting a paper to this journal without due attention. Major points 1. The argument and data presented surround the effects of Gimap5 deficiency on events following conventional T cell activation via the TCR. What is not clear is how this relates to the primary aspect of the phenotype associated with Gimap5 deficiency, namely the in vivo T cell lymphopenia. Gimap5-deficient animals fail to maintain a population of naïve, resting T cells. Part of this phenotype probably corresponds to a reduced thymic output of T cells and another part to a failure to maintain T cells in the resting condition (this point has been explicitly demonstrated in experiments on Gimap1-related lymphopenia). Neither of these stages corresponds obviously to explicit antigen-driven T cell responses. Late SP thymocytes and resting T cells express notably high levels of Gimap5. The authors should address this issue. 2. The authors have conducted the majority of their studies using T lymphocytes from 3-wk old mice, when the mutant mice possess a phenotypically fairly normal T cell population, in contrast to later time points. The assumption that the T lymphocytes at this stage are 'normal', however, is not without risk. For instance, Gimap5-deficient SP thymocytes have a survival deficiency and this may extend into the cell stage they are studying. In the same vein, calcium responses of Gimap5deficient SP thymocytes are reported to be down relative to WT (Ilangumaran et al. Mol Immunol 2009). A system in which Gimap5 could be electively deleted in lymphocytes would have been preferable. 3. Mcl-1. As a target of GSK3β, Mcl-1 is of interest to this study given its known role in T and B cell survival and specific published results vis-à-vis Gimap5 (Chen et al. J Exp Med, 2011). The authors' own data renders them a bit equivocal over engagement with this topic (see lines 88-92). At present, this reader feels he is being led in and out of a cul-de-sac on this subject. Some sharper writing/surgery is needed here (and in relation to what may be best addressed about this in the Discussion) in order to hold to the main message of the paper. 4. The molecular proposal for Gimap5 associated with vesicles such as MVBs sequestering and restraining GSK3β is an interesting idea with origins within the GSK3 field (as reviewed in quoted ref 24). The data presented, however, only begin to address the idea. Issues such as whether the two molecules interact directly with each other and of their disposition with respect to the membranes of the MVBs, or other vesicles with which they are associated, need to be addressed. In connection with this, the authors draw attention to the potentially linked issue of the requirements for Ser389 phosphorylation in this system. Other points (list not exhaustive). Line 21, Intro. As far as I know, human GIMAP5 Fig 4A,B. Percentage of T cells within the spleen is a problematic measure to consider in this context. The proliferative phenotype associated with lymphopenia-associated colitis is accompanied by substantial changes in the spleen including large increases in myeloid cell populations. These inevitably distort the percentages of lymphocytes relative to WT controls. Numbers of lymphocytes in spleen is a more informative measure. Fig 5A. As above, numbers of cells in spleen would be a more meaningful measure. The colour-coding isn't good either, such that I can't figure out which curve is which. Are all groups tamoxifen-treated? Couldn't it be important that the sph/sph, GSK3wt cells are showing the highest levels of GSK3β? Fig S6B. A plot of number would be better than percentage; as in C. Fig S7. We have the appearance of the Gimap5+/-genotype. Line 237 and following. The human data is interesting but it should be pointed out that the pedigree data does not yet seem sufficient to associate the phenotype definitively with Gimap5. Line 387-388. 'All mice used were generated on a C57BL/6 background and confirmed by whole genome SNP analysis'. Is this true? Or do you mean 'All strains/lines of mice'? Line 400 et al. Check that origins of centrally important antibody reagents (e.g. for human and mouse Gimap5, for GSK3 etc.) are given clearly, with refs. If appropriate. Real time PCR. Is the level of L32 RNA in mouse lymphocytes unaffected by T cell activation?
Reviewer #3 (Remarks to the Author): Patterson and co-workers focus on mice deficient for Gimap5 which have reduced CD4T cells ad B cells due to combined effects on survival and proliferation. Gimap 5 is needed for TCR induced inactivation of GSK3b which has b-catenin and NFAT signalling as downstream effectors. Interestingly pharmacological inhibition of GSK3 in Gimap5 deficient cells can restore proliferation, restore Treg numbers and overcome pathology in Gimap5 deficient mice. Authors report on one patient case with GIMAP5 deficiency where GSK3 inhibition also partially restored proliferation defects In general the work is well done, but it remains unclear how GSK3 mediates its effects. GSK3b has many targets, including metabolic targets, C-Myc, Wnt-b-catenin and NFATc. A few simple Wnt and NFAT reporter gene experiments can address these important mechanistic experiments that are now lacking. In addition, it is unclear if the effects observed are exclusively mediated by GSK3b or also by GSK3a. The title also is ambiguous in this respect. Comparing GSK3b kinase activity, how much is kinase activity inhibited after TCR/CD3/CD28 ligation vs. Pharmacological inhibition, which usually (at least with LiCl) is not complete In Fig 4, the improvement of pathology (preventing colitis)largely depends on the number pf Tregs. Therefore Treg numbers need to be included in panel 4A The human patient does not have a marked CD4 T cell defect (273) but CD8 cells (30) are much more reduced. This is different from the mouse model. Explanation? Are CD8 T cell responses and proliferation also rescued by targeting GSK3b?
The discussion on GSK3 and Wnt signalling in TH17 and Tregs is incomplete. Several papers published in top journals showing that Wnt signalling activates Th17 responses and inhibits Treg responses are not cited nor discussed.

Specific responses to the reviewers' comments:
We thank the reviewers for their overall positive and constructive reviews. We believe that their insight and comments have helped to further improve the overall quality of our manuscript. Please below find our responses to their specific concerns. All textual changes in the revised manuscript are marked in red. We have additionally added a model figure (Fig. 9) to summarize our findings regarding the effects of Gimap5-deficiency in relation to GSK3β regulation in the discussion (page 24).
Reviewer #1: 1. They convincingly demonstrate that GSK3β and Gimap5 colocalize but it remains unclear how this facilitates phosphorylation. Do the authors think that Gimap5 functions as a scaffolding molecule and facilitates GSK3β phosphorylation through p38. At least some discussion would be helpful.
Although outside of the scope of the current manuscript, we agree with reviewer that this is an important question that remains to be answered. Interestingly, our new studies further refined the localization of phosphorylated GSK3β and show that the phosphorylation of GSK3 at Ser 389 is predominantly present in the nucleus in line with previous observations. Importantly, overall GSK3β nuclear translocation as well as P-Ser 389 GSK3β is markedly reduced in Gimap5-deficient mice. These observations point to a critical role for Gimap5 in the sequestration/trafficking of GSK3β ultimately facilitating phosphorylation and translocation of GSK3β in the nucleus. Importantly, in the absence of Gimap5, reduced P-Ser 389 GSK3β can still be observed in vesicle-like structures that are predominantly outside of the nucleus (Fig.4E), suggesting the GSK3 Ser 389 phosphorylation at day 2 may occur in unique intracellular compartments. Thus, we propose a model in which Gimap5 facilitates GSK3β Ser389 phosphorylation and nuclear localization that is required for the late stage DNA damage response required for CD4 + T cell undergoing cycling. The exact molecular underpinnings of these observations, however remain to be defined in detail. We have added the data in the manuscript (new figure 3E-I) and expanded on this model in the discussion.

More background information on how unopposed GSK3β activity induces DNA damage and T cell loss would be helpful.
While different studies indicate that pharmacological inhibition of GSK3 or shRNA-specific knockdown of We agree with the reviewer that the overall levels of β-catenin, as assessed by immunoblot analysis, are low and therefore any change in β-catenin levels are difficult to interpret. In our revised manuscript, we have excluded the human catenin data and softened the overall conclusion on β-catenin while the relevant mouse data has been moved to Supplemental We have corrected this mistake in the revised version.
Reviewer #2 (Remarks to the Author): Major points

The argument and data presented surround the effects of Gimap5 deficiency on events following conventional T cell activation via the TCR. What is not clear is how this relates to the primary aspect of the phenotype associated with Gimap5 deficiency, namely the in vivo T cell lymphopenia. Gimap5-deficient animals fail to maintain a population of naïve, resting T cells. Part of this phenotype probably corresponds to a reduced thymic output of T cells and another part to a failure to maintain T cells in the resting condition (this point has been explicitly demonstrated in experiments on Gimap1-related lymphopenia). Neither of these stages corresponds obviously to explicit antigen-driven T cell responses. Late SP thymocytes and resting T cells express notably high levels of Gimap5. The authors should a5ddress this issue.
Our data depicted in manuscript Fig.1 indicates that CD4 + T cells levels in Gimap5-deficient mice are largely maintained in the absence of cognate antigen (Gimap5 sph/sph ; rag2 -/-; OT-II mice given water without ovalbumin); suggesting that activation-induced cell death (particularly post weaning in which the animals adapt to changing/new microbial flora) serves as an important contributor of the CD4 + T cell-specific lymphopenia. But the reviewer is correct in that the CD4 + T cell levels do not entirely reach WT levels (Manuscript Fig.1a) and thus, thymic output or reduced survival of peripheral resting CD4 + T cells as suggested by the reviewer could present a contributing factor of lymphopenia. We have now included studies to directly assess whether thymic output/survival or peripheral survival of CD4 + T cells are affected in Gimap5-deficient mice. Importantly, we observed a normal survival of thymic SP CD4 + T cells when cultured in the presence of IL-7 ex vivo (new Fig.S1A), while quantification the number of recent thymic emigrants (CD4 + , CD24 hi ) in the spleen of 3-week old WT and Gimap5 sph/sph mice, revealed no differences in thymic output between WT and Gimap5 sph/sph mice (new Fig.S1B, C). This is also in line with the GIMAP5 patient, who exhibits a normal proportion of CD31 + T cells that are indicative of recent thymic emigrants (see supplemental patient description). In contrast, peripheral CD4 + T cells showed a slight reduction in survival when cultured in the presence of IL-7 ex vivo (new Fig.1D). Together these studies suggest that in line with our previous studies (Barnes et al., JI, 2009), thymic development and survival of CD4 + T cells is largely unaffected. In contrast however, our studies do indicate a slight impediment in the survival of peripheral CD4 + T cells in the presence of IL-7. These studies are described on page 6 and 7 of our revised manuscript.

The authors have conducted the majority of their studies using T lymphocytes from 3-wk old mice, when the mutant mice possess a phenotypically fairly normal T cell population, in contrast to later time points.
The assumption that the T lymphocytes at this stage are 'normal', however, is not without risk. For instance, Gimap5-deficient SP thymocytes have a survival deficiency and this may extend into the cell stage they are studying. In the same vein, calcium responses of Gimap5-deficient SP thymocytes are reported to be down relative to WT (Ilangumaran et al. Mol Immunol 2009). A system in which Gimap5 could be electively deleted in lymphocytes would have been preferable.
See comments above (1): Our new studies indicate that Gimap5 sph/sph mice exhibit no apparent deficiency in thymic survival while displaying a normal thymic output of SP CD4 + thymocytes (see rebuttal Fig. 3). In addition to these studies, we tested calcium responses in SP CD4 + thymocytes upon activation with CD3/CD28 or ionomycin. Consistent with the normal survival and thymic output of SP CD4 + thymocytes, we observed no differences between the release of intracellular calcium between WT and Gimap5 sph/sph SP CD4 + thymocytes (rebuttal Fig. 3D; not included in the manuscript). These findings deviate from the observations in the lyp/lyp rat model. Thus, while the use of conditional Gimap5 KO would be ideal, we believe that the 3-week old Gimap5 sph/sph mice offer a reliable model to study peripheral CD4 + T cells at a relative naïve stage.

The molecular proposal for Gimap5 associated with vesicles such as MVBs sequestering and restraining
GSK3β is an interesting idea with origins within the GSK3 field (as reviewed in quoted ref 24). The data presented, however, only begin to address the idea. Issues such as whether the two molecules interact directly with each other and of their disposition with respect to the membranes of the MVBs, or other vesicles with which they are associated, need to be addressed. In connection with this, the authors draw attention to the potentially linked issue of the requirements for Ser389 phosphorylation in this system.
The vesicle association of GSK3 (colocalized with Gimap5) is indeed a surprising and novel observation and as a mechanism of GSK3-regulation. We have now further defined the GSK3β + /Gimap5 + doublepositive vesicles in CD4 + T cells using ImageStream analysis. Interestingly, our studies revealed limited expression of Rab5, Rab7 and Lamp1 + in Gimap5 + /GSK3 + vesicles. As far as a potential mechanism, our initial experiments performing co-immunoprecipitation studies, yielded no evidence for a direct interaction between Gimap5 and GSK3β through studies in resting or TCR-activated CD4 + T cells (results not shown). The timing and technical limitations associated with these Co-IP experiments, however, does not (yet) exclude this possibility. Importantly, further definition by ImageStream analysis reveals the P-Ser 389 GSK3β to be almost exclusively present in the nucleus of WT CD4 + T cells, while Gimap5 sph/sph CD4 + T cells exhibit markedly reduced P-Ser 389 GSK3β levels (see new manuscript Fig.3E-I and comments above for reviewer 1). These findings suggest that Gimap5 facilitates the P-Ser 389 and distribution of GSK3β in the nucleus that is required for the DNA damage response occurring during T cell cycling. The exact molecular mechanisms underlying these observations are beyond the scope of the current manuscript and will be a key focus in future studies in our laboratory. We apologize for this oversight; this citation has been added (page 3, line 25).
Line 53-56. Sentence should indicate that this refers to B lymphocytes.
We have modified this sentence to: "GSK3 by phosphorylation of Ser 389 is essential for lymphocyte viability upon double-stranded DNA breaks (DSBs) observed during Variable, Diversity, and Joining (V(D)J) recombination during thymic T cell development or in B cells undergoing immunoglobulin class switch recombination (CSR)." Line 53-56.

Lines 75-6. Lifespan etc. Does this refer to previously published data? I see none here.
This is based on our personal experiences keeping Gimap5 sph/sph ; Rag2 -/-OT-II breeders. Mice were kept generally up to 6 months and showed similar survival compared to Gimap5 WT Rag2 -/-, OT-II mice. We observed no spontaneous colitis in these mice as assessed by histology. Since the reference of the lifespan observation is anecdotal, and not compiled in a data plot, we removed this statement from the text.

'improved'. Relative to what? Cell numbers aren't necessarily the same as 'survival'. The cell numbers in the absence of Ova are still not as high as WT.
The text is now modified to "Moreover, their T cell compartment is largely maintained in the absence of antigen and CD4 + T cells remain predominantly naïve (CD44 lo ;CD62 hi ) (Fig.1A,B). On the other hand, compared to the Gimap5-sufficient controls, Gimap5 sph/sph ;Rag2 -/-;OT-II mice exposed to ovalbumin in drinking water, exhibited a reduced CD4 + T cell population with an increased proportion of remaining CD4 + T cells displaying a memory-like phenotype (CD44 hi ;CD62L lo ) (Fig.1A,B)" (line 93-99). The issue with peripheral survival is now also addressed with added studies assessing ex vivo survival of peripheral CD4 T cells with IL-7 (see above, reviewer 2, comment 2).   The immunoblot indeed represents different areas of the same gel. Nonetheless, to avoid any ambiguity, a new immunoblot showing all samples in a single gel is now included in Figure 2A. See above; Fig.2a has been replaced with a new immunoblot and shows consistent c-Myc data compared to 2C. In the previous Immunoblot, the induction of c-Myc was less robust (also in WT) and therefore the blots were longer exposed, explaining the differences in c-Myc at baseline (unstimulated conditions) between 2A and 2C.

Fig 2C data panel. How have the data been normalised? Is this pairwise relative to five different WT controls? Fairly pointless as a panel rather than just stating the relative increase in sph/sph mice (with stats).
We thank the reviewer on suggesting an alternative method to present the data, which are pairwise. Data points were generated by measuring the fold increase in the proportion of c-Myc phosphorylation in Gimap5 sph/sph CD4 + T cells compared to WT CD4 + T cells. The five data points correspond to five independent experiments; each comparison was performed on samples of a single blot to avoid introducing error from transfer efficiency, labeling efficiency, exposure, and stripping efficiency. Notably, in every experiment/blot we ran, the proportion of phosphorylated c-Myc in Gimap5 sph/sph CD4 + T cells was significantly higher than in WT CD4 + T cells. This is indicative of increased GSK3-mediated phosphorylation. The data is representative of 3 independent experiments included, (n of 2 per experiment, for a total of n=6). This is now further clarified in the legend of Fig.2. The similarity dilate represents the log transformed Pearson's Correlation Coefficient and is a measure of the degree to which two images are linearly correlated within a masked region. In this case both NFATc1 and DAPI were masked individually in live CD4 + T cells (>500 cells per sample for all experiments). We added the explanation to the material and methods (Page 33, Line 529-533). While mRNA stability and other factors may also affect the total mRNA level, this data shows that the failure to accumulate cMyc protein was not due to a failure to transcribe the myc gene. We have modified the language.
Line 127/ Fig 2C.  The suggested reference is now also included here.
Line 160-1, Fig 3A. The data as displayed/presented don't convince me that I am looking at 'high colocalisation'. Since the sph/sph mice are Gimap5-negative then the data should reflect zero vesicles positive for this.
The reviewer is correct and this is indeed an overstatement. In fact it is important to note that not all GSK3β + vesicles are Gimap5 positive. We believe that prior to localization to Gimap5, GSK3β likely must progress through a stepwise process through several different vesicular compartments (early endosomes, etc.). The vesicular distribution of GSK3β varies between cells and is dependent upon activation state. We have clarified this statement and in new Fig.S5, we further define of Gimap5 + /GSK3β + DP vesicles and their coexpression with endosomal/lysosomal vesicle markers. We also now quantified the % of Gimap5 + vesicles in human and mouse CD4 + T cells that are positive for GSK3 in the text (Line 197 and Line 322).

The 'Bright detail parameter is in need of explanation.
The Bright Detail Similarity is a standard statistical analysis performed by IDEAS software that represents the log transformed Pearson's correlation coefficient of the localized bright spots with a radius of 3 pixels or less within the masked area in the two input images. Our studies in this particular experiment were masked on Gimap5 spots and GSK3β spots. In general values <1 suggest no colocalization. This is now described in the material and methods on page 33, line 540-543.  The figure is meant to convey a higher resolution Image, while the Image stream provides the quantitative analysis of Gimap5/GSK3+ spots. The lower panel represent a control for the specificity of the Gimap5 antibody as one would expect to include in this type of experiment.

Line 182-3. There is a hint that S389 phosphorylation is increased at day 1. Is this a consistent observation?
GSK3β phosphorylation at S389 at day 1 appears similar between WT and Gimap5 sph/sph CD4 + T cells in the figure shown (and other blots) and appears unchanged to Gimap5 sph/sph levels in resting conditions or at day 2.
Lines 184-185. p38 MAPK. Data not shown. Should this be presented? p38 expression and phosphorylation is shown in Fig. 3D, blots 4-5 as well as Fig.S3B and Fig.S7C. We have added the Figure notations to clarify this.

Fig 3F. Was there no variance in WT data?
While there was variance in the WT data, error bars would be shorter than the height of the symbol. In these cases, Prism simply does not draw the error bars.
Line 189-191/ Fig 3H. "DNA damage…..was reduced to control levels…..". It was reduced, but whether that is to control levels is doubtful.
This has been modified and now simply states "reduced".

Line 193. To what does 'these data' refer?
The data shared in this section, as an aggregate.
Lines 201-2/ Fig 4A,B. Percentage of T cells within the spleen is a problematic measure to consider in this context. The proliferative phenotype associated with lymphopenia-associated colitis is accompanied by substantial changes in the spleen including large increases in myeloid cell populations. These inevitably distort the percentages of lymphocytes relative to WT controls. Numbers of lymphocytes in spleen is a more informative measure.
Gimap5 sph/sph mice are consistently smaller than WT littermates, even pre-weaning (Barnes, JI, 2010). This corresponds to an overall numeric reduction in cellularity within lymphoid organs. Furthermore, the LiCl treatment affects weight of mice (WT and Gimap5 sph/sph ) -something that cannot be accounted for in the untreated controls. We therefore believe that showing frequencies of CD4 + T cells and B cells for this experiment is a more accurate measurement. Given the low β-catenin levels detected in T cells and the associated difficulty to draw conclusive data from these WBs, we moved the β-catenin immunoblot to the supplemental data and are careful in our interpretation of these data. The individual data points present the mean values of >500 cells per sample (now clarified in the legend). Moreover, the data are now depicted as individual data points representing the average for each sample. The idea was to maximize histology panels so that the details in tissues actually showing pathology remain clear. Nonetheless, we have added these back in (current) Figure 5 in the text. The sixth curve was a second Gimap5 sph/sph Gsk3b fl/fl mouse. It has been removed. When comparing Gsk3b wt/wt cells, the increased GSK3β MFI in Gimap5 sph/sph CD4 + T cells is reflective of progressive shift of Gimap5-deficient CD4 + T cells to a memory like phenotype (CD44 hi CD62L lo ). Memory cells in general have higher expression of GSK3β independent of Gimap5 expression; GSK3β levels in WT and Gimap5-deficient CD4 + T cells are equivalent in CD44 hi CD62L lo CD4 + T cells. All groups are tamoxifen treated.

Fig S6B. A plot of number would be better than percentage; as in C.
We have included numbers instead of frequencies in the B cell graph (Fig.S9B). Correct, the GIMAP5 +/genotype (parent) was also used in Fig.7 and Fig.8H-J. (Current) Line 308 also clearly states that "we compared T cell proliferation from the patient with his heterozygous mother." Line 237 and following. The human data is interesting but it should be pointed out that the pedigree data does not yet seem sufficient to associate the phenotype definitively with Gimap5.
Given 1) the complete loss of GIMAP5 protein, 2) the observed similarity in molecular defects with Gimap5-deficient mice, 3) the absence of other candidate variants that could explain the deficiency, and 4) the use of GIMAP5 heterozygous parents that serve as ideal genetic controls, we believe we don't overreach with regard to our statements. More importantly, we believe we have been very careful in our overall conclusions in the manuscript by stating "we show that a human patient carrying a GIMAP5 loss-of-function mutation exhibits similar deficiencies, including lymphopenia, recurrent thrombocytopenia, and impaired T cell function. T cells of this patient exhibit reduced T cell expansion and c-Myc expression that can be corrected in vitro with GSK3 inhibitors." Line 387-388. 'All mice used were generated on a C57BL/6 background and confirmed by whole genome SNP analysis'. Is this true? Or do you mean 'All strains/lines of mice'?
This has been modified and instead of mice, now states "all mouse strains".

Reviewer #3 (Remarks to the Author):
In general, the work is well done, but it remains unclear how GSK3 mediates its effects. GSK3b has many targets, including metabolic targets, C-Myc, Wnt-b-catenin and NFATc. A few simple Wnt and NFAT reporter gene experiments can address these important mechanistic experiments that are now lacking.
We thank the reviewer for his constructive feedback. The reporter gene experiments would indeed be an elegant way to complement our findings with regard to the GSK3 targets c-Myc, NFATc1 and/or β-catenin. However, based on our experience using lentiviral vectors as well as transfection approaches in Gimap5deficient CD4 + T cells, we observed that the inherent fragility of these cells (likely due to their sensitivity to DNA damage) has made it impossible to reliably introduce exogenous vectors and, for instance, perform gain of function studies for specific transcription factors. Nonetheless, blocking NFATc1 dephosphorization (calcineurin inhibitor) in T cells causes a major loss of T cell proliferation (manuscript Fig.S4D), confirming the critical role of NFAT in T cell proliferation. We have also performed studies to address whether GSK3β vesicular sequestration involved Wnt signaling. Specifically, we incubated WT CD4 + T cells directly with Wnt3a and/or tested the effect of blocking Wnt secretion by incubating CD4 + T cells with CD3/CD28 in the presence/absence of IWP-2 (a Wnt secretion inhibitor). Subsequently, we assessed GSK3β sequestration by Imagestream and determined the ability of WT CD4 + T cells to proliferate. Neither Wnt3a nor IWP-2 affected the vesicular sequestration in WT CD4 + T cells or impacted their proliferation (Fig.S6A-D). Moreover, no significant differences in Wnt3a expression was observed between WT and Gimap5 sph/sph CD4 T cells at resting or activated conditions. These results are included in (Fig.S6) and discussed on page 15, line 208-215.
In addition, it is unclear if the effects observed are exclusively mediated by GSK3b or also by GSK3a. The title also is ambiguous in this respect.
With regard to the specific role of GSK3α and GSK3β, it is predicted that a majority of their functions are redundant. However, our data shows that genetic deletion of GSK3β is sufficient to mitigate CD4 + T cell loss and development of colitis in Gimap5 sph/sph mice. While this does not rule out the possibility that GSK3α is also dysregulated in the absence of Gimap5, these data point to a key role of GSK3β in the underlying T cell-dependent pathology. This specificity could be related to the DNA damage response which in the current literature is predominantly associated with GSK3β.
Other remarks:

In Fig 2, is PMA/Ionomycin treatment capable of bypassing the proliferation and c-Myc induction defects?
PMA/Ionomycin results in similar molecular and cellular defects, including impaired proliferation, survival, and induction of c-Myc. Given the normal calcium flux in response to Ionomycin or αCD3/CD28 (Fig.S4F) and the activation of ERK, JNK IkB, AKT and p38 (Fig.S3B) upon PMA/Ionomycin stimulation and AKT, p38, and GSK3β (S9) in Fig.S7A-D upon αCD3/αCD28 stimulation, it is likely that and Gimap5-associated defects occur further downstream. Rebuttal Fig.6 (not included in the manuscript): WT and Gimap5 sph/sph CD4 + T cells stimulated with PMA (50 ng/mL) and Ionomycin (500 ng/mL) for 24h prior to lysis.

In Fig 3C the individual Gimap5 and GSK3b spots do not always co-localize, what percentage does (quantification)?
The reviewer is correct and likely, prior to localization to Gimap5, GSK3β must progress through a stepwise process through several different vesicular compartments (early endosomes, etc.). The vesicular distribution of GSK3β varies between cells and is dependent upon activation state. In Fig. S5, we further define the Gimap5 + GSK3β + spots (Fig.S5) at 24h post stimulation and we also assess nuclear translocation of GSK3β which is primarily observed at a late stage (48h post stimulation) (Fig.3). In general we see a ~66% of Gimap5+ vesicles expressing GSK3β at 24 hours of activation for human CD4+ T cells, while in mice, the percentage of Gimap5 + vesicles expressing GSK3β is ~54%. These values have been included in the text (Line 197 and Line 322).

Comparing GSK3b kinase activity, how much is kinase activity inhibited after TCR/CD3/CD28 ligation vs. Pharmacological inhibition, which usually (at least with LiCl) is not complete
We show that GSK3 inhibitors LiCl and BIO both significantly reduce phosphorylation of cMyc. Shown to in the figure below is the level of p-cMyc (T58) in αCD3/αCD28 stimulated WT and Gimap5 sph/sph CD4 + T cells treated with either 2.5 mM LiCl or 100 nM BIO relative to untreated WT or Gimap5 sph/sph CD4 + T cells.
Using cMyc phosphorylation at T58 as a readout of GSK3 activity, these data show that GSK3 activity is reduced by ~60% by 2.5 mM LiCl and 40% by 100 nM BIO (n=3 and n=10, respectively).
In Fig 4, the improvement of pathology (preventing colitis) largely depends on the number of Tregs. Therefore Treg numbers need to be included in panel 4A.
The frequency of Treg cells in the spleen of vehicle-and LiCl-treated Gimap5 sph/sph mice remains unaffected and is not different from WT mice in those conditions (new Fig.S8C). Instead, Fig. 5D demonstrates that the therapeutic effect predominantly works on their suppressive capacity. As published previously (Aksoylar et al., JI, 2012) and shown here, while Gimap5 sph/sph have similar frequencies of regulatory T cells, they have a markedly impaired suppressive capacity. We show that 4 week treatment with LiCl restores Gimap5 sph/sph Treg function. This is consistent with previous reports of GSK3 inhibition potentiating Treg activity.

The human patient does not have a marked CD4+ T cell defect (273) but CD8 cells (30) are much more reduced. This is different from the mouse model. Explanation? Are CD8 T cell responses and proliferation also rescued by targeting GSK3b?
Patient CD4 + T cells are significantly reduced (246) compared to reference range for his age (610-1446). Expansion of both CD4 + and CD8 + T cells was rescued upon GSK3 inhibition with LiCl. This is actually remarkably similar to the Gimap5 sph/sph mouse. The mouse lacks peripheral CD8 + T cells, while CD4 + T cells are significantly reduced, but not absent. Given this phenotype, as well as the more severe reduction of CD8 + T cells in the GIMAP5 patient, it is reasonable to conclude that CD8 + T cells are more susceptible to Gimap5-deficiency.

The discussion on GSK3 and Wnt signaling in TH17 and Tregs is incomplete. Several papers published in top journals showing that Wnt signaling activates Th17 responses and inhibits Treg responses are not cited nor discussed.
We have expanded the discussion and added additional references discussing the roles of Wnt signaling, GSK3, and β-catenin in CD4 + T cell polarization, a topic that remains somewhat controversial. Loosdregt et al. report that addition of Wnt and inhibition of GSK3 in vitro inhibits Treg suppressive capacity (Immunity, 2013). Lee et al., demonstrate that inhibition of Wnt signaling promotes Th17 activity, while addition of exogenous Wnt or GSK3 inhibition reduce IL-17 production (Euro J Immuno., 2012). Conversely, expression of constitutively active β-catenin promotes Th17 polarization (Keerthivasan, et al., Sci Trans Med., 2014). Other studies suggest that inhibition of GSK3 promotes Treg activity (references 58,61,63). While the literature is ambiguous, our data support the idea that uncontrolled GSK3β activity impairs Treg function while at the same time an increased Th17 development is observed (Aksoylar et al., JI 2012). These additional papers are now referenced listed in the discussion (Line 422-424). Fig.8 (manuscript Fig.S8C): (A) Frequency of regulatory T cells (CD4 + CD25 hi Foxp3 + ) in the splenic CD4 + T cell compartment of vehicle-and LiCl-treated WT and Gimap5 sph/sph mice (n=11; mean ± SEM). Statistical significance is determined by ANOVA followed by Sidak's multiple comparisons test.

Rebuttal
The authors have appropriately addressed the issues that I have raised in my previous review.
Reviewer #2 (Remarks to the Author): I welcome the additional experiments that have been performed and the improvements that have been made to this manuscript. The study by Patterson and colleagues now makes a good case for a strong regulatory influence of Gimap5 on the actions of GSK3b in T cells. It proposes that Gimap5 has effects both on the physical sequestration of GSK3b (in MVBs) -an inhibiting effect on GSK3b -and on the nuclear translocation of GSK3b -an effect that somehow may promote a productive T cell response (see lines 387-89). I don't think the precise linkage of these effects to S389 phosphorylation of GSK3b has been fully delineated at this stage. In my opinion the manuscript could still do with some revision to correct inaccuracies, errors and improve presentation. Below I present my commentary which is in roughly line order. Items 1, 17, 18 raise points of clarification about the biology. I think the provision of a schematic diagram (Fig. 9) is helpful. This version, however, isn't entirely limpid for me in its explanatory power. I have asked questions relevant to this (items 17-19). 3. Line 150 -there is reference to resting T cells but Fig. S4A does not appear to show data for unstimulated cells, so this sentence should be modified. In the same sentence I think 'further indicating' is too strong and should be toned down to 'consistent with'. And similarly in line 153, the wording misleads when it says 'To test if GSK3 activity was elevated….' They are not 'testing' GSK3 activity but taking surrogate measurements. It might be better to start this paragraph -'To look for further evidence that GSK3 activity was elevated...'. 4. Lines 160-165. My reading of the data in Figure 2E is that the sph/sph CD4+ cells (without inhibitors) appear to be able to go through the same number of divisions as the WT cells after anti CD3+CD28 stimulation but have poor survival. 5. Line 167 and Fig. S4D. I think Fig. S4D should be omitted. I see no reason to re-visit the action of cyclosporin A when literature references cover the point being made. 6. Fig. 2F-H and lines 171-173. I welcome the additional explanation of 'similarity dilate' in the Methods section. I still suggest some improvement, however, on behalf of the reader. Panel 2F is essentially an illustration of the method being applied: one can stare at it for hours without it conveying much in terms of the Gimap5-related observation we are being told about. It is the